#3-d semantic segmentation | Explore Tumblr posts and blogs

gtssidata4 · 2 years ago

Text

Image Annotation Helping The Feature Of ADAS 2023

Annotation for Advanced Driver Assistance Systems (ADAS) in Computer Vision. The most advanced driver assist systems (ADAS) provide cars and drivers with the most advanced technology and details to aid them in being aware of their surroundings and handling potential situations with greater ease making use of semi-automation. AI together along with ADAS Annotation aids in developing applications that can detect various scenarios and objects and make quick and accurate choices to ensure safety while driving.

What is the reason? ADAS is important for Safe and Controlled Driving

ADAS is like self-driving vehicles and utilizes similar technologies like radar, vision, and the combination of various sensors such as LIDAR to automate dynamic driving tasks such as braking, steering or acceleration to make sure that the safety of drivers and a controlled, safe driving.

In order to incorporate this technology, ADAS demands labelled information to help the algorithm recognize the driver's movements and the objects. Annotating images is a well-known service that creates such data for computer vision training.

What is the difference between ADAS and Self-Driving Cars?

In autonomous or self-driving automobiles, the driver enjoys full control over its steering, the brakes, and many other aspects. There is no need for a driver as it can move in a certain direction and avoid all types of obstacles without the intervention of a human.

This assistance is all included in ADAS to assist or warn drivers if they do not see the situation. In the absence of a driver's attention all systems operate semi-autonomously and take the proper step for safety as well as the comfort of driving.

We employ ADAS Annotation to detect the driver's various bodies and other objects. Image Annotation Service is a popular tool to generate computer-generated training data.

Annotation ADAS for Traffic Detection

We utilize the ground-truth-labelling method to label recorded sensor data in line with the expected ADAS state. Pattern recognition, learning is a method of extract tracker, 3-D Vision as well as other Computer Vision techniques are used in ADAS traffic labelling.

AI Workforce is a well-known company that has developed a driver assistance system which provides high-quality information about traffic detection that can help create a live algorithm capable of recognizing patterns in traffic in the near future ADAS technology.

Annotation of Driver Monitoring for ADAS

Drivers who are tired and sleepy, or distracted, can be monitored by ADAS The driving monitor. ADAS detects signs that indicate the drivers mental strain, his conduct and the car's surroundings. AI Workforce annotates ADAS systems by using frames to assist ADAS by monitoring the motorist's facial expression, behavior and body movements.

Annotation Segmentation (AS) in ADAS is the process of labelling and indexing objects in frames. If there are multiple objects that are labeled with a unique colour code, each object is identified with a distinct colour code , without background noise. It is essential to eliminate background noise in order to ensure that the object will identify the item's edges.

We offer semantic segmentation of images, which is required to identify fixed and essential objects. To solve high-level vision issues that arise in computer vision like image understating and parsing scenes, the image segmentation can help computer vision applications overcome problems with low-level vision including 3D motion reconstruction and motion estimation and reconstruction.

What are the reasons you would want outsourcing your ADAS annotation tasks?

The most important asset for training autonomous vehicles and develop. A huge amount of varied and rich tag data is used to confirm. This involves gathering information about an area in order to link image data with actual conditions on the ground.

It is able to make use of the annotated data to develop and test recognition algorithms as well as predictive models in a systematic manner. Ground truth labels can help autonomous vehicles to recognize and identify moving objects through the identification of urban surroundings such as road markings, highway signs and even weather conditions.

There are many reasons why one ought to consider using Cogito to outsource ADAS annotation and other functions. Here are some of the most well-known reasons:

Excellent Quality Services

It is evident that cost is a major aspect to take into account when outsourcing the annotation process. Companies such as Cogito and Analytics can provide high-quality data Anotation Service with reasonable costs.

Infrastructure and technologies at their best

Companies that provide data annotation are innovative and cutting-edge. They make use of the most recent Artificial Intelligence, Machine Learning and robotics technologies.

Clients will get the most current and up-to-date software and customer support to help you with data annotation.

Services for images annotation

The success of an Artificial Machine Learning (AI/ML) implementation needs a high-quality learning data model. However, in addition to high-quality, the quality of AI/ML training will be defined by the size, speed and speed of annotations, security of data and bias reduction. Making sure the annotations of images are precise for projects that involve Machine Learning/AI as well including incorporating all these elements, will assist in creating the proper dataset for the project.

Without professional annotation experts, businesses usually face one or two of the following problems:

Understanding the meaning behind any image

Attention to detail, and a keen understanding

Recognition of faces, and following analysis (identifying gender, categorizing emotions, etc.)

A vast database is analyzed and analyzed with the goal of preserving the accuracy.

The use of classifiers is to organize every image.

Data security compliance

The consistency in the subjectivity of data sets

It's an arduous process

The process of consuming takes more effort and time than it is acceptable to try by yourself. This is why it's cheaper outsourcing annotation of images to a reputable company.

Perception models that have been trained with the 2D bounding boxes datasets can enhance your model's ability to search visually by recognizing different objects, including the most intricate images. Our annotation tools use 2D bounding boxes as well as 3D bounding boxes to make annotations that can be used to develop projects across various industries, including e-commerce autonomous vehicles, traffic control, and other projects that require training data to teach autonomous vehicles to identify pedestrians and cyclists pedestrians, traffic lights, footpaths,

Teach students about the ecommerce and retail model to recognize furniture and accessories, clothing food items, other items that can speed up the checkout process or create an income

The billing process is automated.

Develop computer vision models that find damage to objects such as cars and buildings to determine the amount of help needed to settle insurance claims as well as other things of this kind.

Recognize objects, people and tracks in satellite and drone images

We have designed our bounding box annotation workflow to meet the specific needs of you to find the objects that are of interest by using precise image labelling services.

0 notes

spamzineglasgow · 6 years ago

Text

(REVIEW) Not your minute turns from the blueprint: Body Work, by Tom Betteridge (SAD Press, 2018)

Denise Bonetti <[email protected]> Mon, 10 Dec 2018, 20:21 to maria.spamzine Hey Maria, Hope life’s good :) I’m just writing to see you if you’ve read Tom B’s new Body Work? There’s a paper I should be writing, but have been reading and rereading that pamphlet instead, or more like dipping in and out really, cause it's so beautifully layered and expanding that you can only take so much at a time. Don’t you think Tom’s poetry has this strange earworm quality to it? (I think he’d like the annelid comparison.) The way he choreographs words (I don’t want to use the word 'images’) makes its way into my brain and never wants to leave. He draws these, like, lateral paths of meaning so clearly that the weeds never grow back.Tom Raworth has this bit in 'Writers / Riders / Rioters' that goes:

the present is surrounded with the ringing of ings which words have moss on the northside

like, words naturally arrange themselves into a system of semantic habit, which is so stable and stale that makes them grow moss, but also so rich and vibrant when it's exploited productively. Obviously this is Raworth so it probably also means the opposite of this and so much more, but it kinda makes me want to say that the present (poem) makes the ringing of ings deviate so well that the moss can never grow again. I’d say that his poems behave like sophisticated lines in the sand, but they're more like brutal carvings on a rock. He had a couple poems in Blackbox Manifold ages ago (I think) and there was this one bit

‘nerve truffled plume lead pickled breast’

I think about all the time (especially when I cook). It’s so smooth. Why can I not stop thinking about it. It’s cause it’s so shameless, it wants it all - the feather-light and the corpse-heavy, never perturbed, so lucid. It plays at tasting good, but it tastes of an unrealistically blank texture. A-ha! Anyway the new pamphlet is gr8, if you haven’t read it yet look at the first poem pls - ‘OCCAM OCEAN’ (sounds like an anagram or palindrome)

It all dwells in systematic abstraction but flies so close to materiality, like a mosquito buzzing around the ear ('Not your minute turns from the blueprint'). I love that ‘plate’ in the first phrase, too: it behaves like an adjective but feels nothing like it. I can't help but think it's the subject of the sentence in a parallel universe that's created by scrambling syntax. It makes me think this is the way language should always work, and that we're the fkn idiots living in the parallel universe in which syntax is scrambled in ordered to be as boring as possible. Idk - it's late and I need to go back to writing boring essay syntax 'bound to decision and the pursuit of what follows'. Lemme know ur thoughts you smart queen D xxx

Maria Sledmere <[email protected]> Wed, 12 Dec 2018, 17:30 to denise.spamzine

Dearest Denise,

Life is good thanks. I'm sitting in the attic of the law building and I can hear the construction work going on and every time I leave I look out at the skyline and slivers of infrastructural alteration. I was walking along the road earlier because the pavement was closed off for construction and cba crossing and the high-vis guy was like, 'you'll not see Christmas walking on the road like that', but I guess I misheard him saying something else because I was really engrossed in this old Slowdive album, so I just smiled sweetly. Anyway, that got me thinking back to the question of erasure, which I think is pretty vital to Body Work.Have been carrying this pamphlet in my bag for so long that the cover has started to peel and revealing speckles of white underneath, like the text itself is ready to reveal itself, and yes I was thinking Barthes and strip-tease and paratextual enticement.

So I had to google the word annelid and now can't get the phrase 'segmented worm' out of my ear/head/throat (gross!). I was thinking about what sort of stains are on the cover of this book, you know, with Hrafnhildur Halldórsdóttir's gouache/pencil work. A stain is one thing stuck to another. Gouache is a funny kinda substance that is half watery gauze, half binding, thick, gummy akin to acrylic. Wet, it will easily smudge. My thumb keeps bleeding where the skin thins, hardens, peels and sloughs off. Tom's poems are kinda mucilaginous in parts (v. insecty, molluscky, sap emission; but also they have a crispness and precision, like cuts of leaf). Things that smudge or fall in flakes. Bodies are maybe whatever we leave behind. I didn't want to mention Hookworms the band because of the singer's sexual abuse scandal BUT it seems significant that a group named after an earworm/type of parasitic larvae would have a song called 'Negative Space'. Like what we eat into in the process of making, existing, remixing. Is language a rash, the result of these parasitic inf(l)ections?

I've been to a couple of reading groups on microbiomes lately, and we were thinking through this idea of how acknowledging your bodily composition in terms of myriad genes and organisms challenges conventional, bounded notions of 'self'. What kinds of affects does this produce? There's a weirdness to that, in Mark Fisher's sense of the weird as 'that which does not belong', that which 'brings to the familiar something which ordinarily lies beyond it, and which cannot be reconciled with the "homely" (even as its negation)'. Fisher suggests that the form most conducive to rendering the weird might be montage. So I was thinking about how montage involves splicings, gaps, juxtapositions, cuts and suddenness. I mean you open the very first poem of Body Work, 'Occam Ocean', and see that its prose-poetic paragraph compacting is split in the middle by the juncture of line break and indent. And ofc the title recalling Occam's razor, the philosophic principle by which in the case of two explanations for an occurrence, it's best to go with the one that requires least speculation. Razor things down and erase speculation? What are we left with, more of the Rreal? Lately I've been hankering for cleaner prose, crisp lines, simpler solutions. The Anthropocene is all of a goddamn tangle. Do I want to follow the myriad threads or just cut cut cut -- who gets to do that?

Did you ever cut a worm in two as a kid?

Okay so I love how 'Occam Ocean' might promise, title-wise, this clean prosaic expanse (like the oxymoron of expanding ocean and occam's, which requires surely a condensing), but what we get is clustering, motion, shiver, containment. The sensual trash magickk of P. Manson! The little syncope of this thing or that, the 'maple neck', vibrating canes, 'chambered breath bowed into the driest soundboard'. These aren't like 'Latour litanies' because they are not like concrete OOO segments of things; idk, they are more about processes and mutable assemblages, emphasis on action and change, sometimes transmission, things inside things. Lynn Margulis and symbiogenesis. How things interact, communicate up close; all of a mutable, prose-poetic swallow. It's actually an incredible intimate pamphlet, don't you think? I feel inside a thing inside a thing inside a thing. I feel a vague ecological sorrow, which gnaws at introspective tendency. The clue to that, you might see, is the cutaway phrase, 'emo Chord' in 'String Growth'.

'Collapse all tears allowing echo retreat'; these lyric lines of 'Glissando' expression, smoothing and shimmering over cuts and junctures: a slide between notes. I used to play trombone and I wish I cld articulate linguistically what kinds of lip vibration occur when you attempt a glissando and feel it slide down your arm muscle but then also through your chest as you try to sustain a sound. It's maybe the way you glide through a scattered poem, with your eye, which is different ofc to the spikier way you'd have to read it aloud, stuck on the vowels. Stuttering. I would love to hear Nat Raha perform these poems, because she does such wonderful things with punctuation and bodily performance, a kind of grammar of breath and cough and click. Reading over the more field erasure poems like 'O--NE' and 'String Growth', it's easy to say something like ~vibrant materialism~ here, but as usual I reach for Steven Connor on noise. Return to the ear, which is 'vulnerable' and 'resembles the skin in being the organ of exposure and reception'. I love what Connor says about Levinas' perspective on 'the awareness of the vacant anonymity of being, of an abstract, encompassing sense that "there is"' being 'an experience of noise'. <3 Acknowledging that breath in the void, that is not nothing but a sparkling something, entails a sense of noise. I am here in the attic of the law building, listening to construction, the type of my fingers on the keys. Someone is murmuring of their distress. What is the difference between living and existing, and being and nothing?

Karen Barad:

'Suppose we had a finely tuned, ultrasensitive instrument that we could use to zoom in on and tune in to the nuances and subtleties of nothingness. But what would it mean to zoom in on nothingness, to look and listen with ever-increasing sensitivity and acuity, to move to finer and finer scales of detail of...?'

When she asks, 'What is the measure of nothingness?' I think surely it is a bodily measure, as everything is: 'bound', as Tom puts it, 'to decision and the pursuit of what follows'. Of course 'what follows' recalls Derrida's 'The Animal That Therefore I Am (More to Follow)', where he's just out the shower and watching his cat watching his phallus, etc. What kinds of dislocation occur. But I mean in all that grandeur of encounter, there's still anthropocentricism. Tom gives you this cinematic CUT, like the instructive 'keen SNAP' that occurs in 'Occam Ocean' to dramatise 'Still, pondsnails whir and blindly source [...] a working leaf shutter'. Soever the language enacts the slurs of the snails up close. We look for the answer to the question of ellipsis, the more to follow [what follows]: inevitable, a question. Sometimes Tom is writing about silence ('then silence confronts an earful underhand') but the music of his language does all the noise, so we just can't have nothingness: there is always a vibrational residue that speaks of something in miniature, atomic, happening.

Ofc with the ear again I am thinking of the ear at the start of Lynch's Blue Velvet and how it's covered in rasping wee insects whose hum is a sort of white noise of trauma that runs through Lumberton's suburban idyllicism.

And so what happens next is I flip open to the following page of Body Work and there is 'String Growth', one of Tom's sprawling erasure poems, which for more than a split-second resembles hundreds of crawling, shimmering ants. I actually think my earliest childhood memory is of looking down at my bare feet on the patio of our old house in Hertfordshire and seeing red ants run over my toes. Then looking up to a greying, English sky. Constantly struck by the cinematic image of that, its splicing out of time: the vividness of insects on human flesh, then milky smog of skyward nothingness. 'String Growth', the accompanying notes to Body Work tell me, is an erasure poem of the Chordoma Foundation's 'Understanding Chordoma' information page.

Erasure can expose sequences of nightmare at work in the lexis and syntax of the text on which it parasitically feeds. I am scared to go on the Chordoma Foundation's website, for fear that just reading or saying the word 'tumour' will activate some kind of malignant growth in my body. And so something of the word chord as a sonorous relation between materials (bodily, textual; textural, cellular). Chordomas are tumours often located in the spine and so I find myself looking for the undulating shape of a spine in the scatter-text of Tom's poem. My eyes cascade down textual spines. Why is it sometimes I otherwise latch upon a 'keystone' word which then extends with adjacent resonance? Musical abnormalities accumulate. Thought swells.

And yea I wonder how this fits into what you say about the poems being 'so smooth'. Like Lynch's waxen silicone ear. Because even though fragmentation makes me think of bits and jaggedness etc, there's this sheen of aestheticism to Tom's work that makes me think of gloopiness, fullness, thereness but also the glaze of potential nothingness. Like in Barad's sense, or miniaturised ecological window shopping - a la Morton's Romantic consumerism? Or do we get into the things themselves? What are your thoughts on the question of recalcitrance? Maybe cos he named a previous pamphlet Pedicure I've just got varnish on my mind. Things an insect might stick to, and be amberized in. Mm.

'[...] Phosphorus crystals may be white, red, burgundyor alight as urine passes'

I keep a stone of citrine under my pillow sometimes. It is supposed to alleviate nightmares and 'manifest abundance'. It is the colour of a rich, dehydrated piss and sometimes when I come back to bed after peeing in the night I think it's some kind of organ lain on my bedsheets, hopped out of my body, and I have to stop my heart and breathe. Is that syncope?

On the <topic of piss>, isn't there a sort of caustic quality, even to the smoothness? Like it is working at making a brittleness of its sheen? And that is what poetry is, cracking the veneer of language or something? Punctuational insects dwelling in splits and fissures? It is nice and cool in Tom's poetry, a place for thinning the self and dwelling. Even though the lexis is so rich and dense, it still seems slender somehow; there's a suppleness. Tease threads of your silk(worm).

Was thinking about what Lisa Robertson says about 'commodiousness' in poetry and what kinds of space there are for the reader here, because I don't think there is much space at all, in the conventional humanly readerly sense. Maybe what I mean by (straw man: Romantic) lyric, which requires something of declarative expansiveness? The density and clutter of specialist language in Body Work makes me feel like a worm, trying to hook my way lusciously into a line: 'espalier's / strains unfinished by the scarp trellis' ('Body Work'), 'rooted to a middle-ground / no more than motion defibrillates'. And I become a parasite on the body of the text, which is a parasite on the body, which is made up of millions of (para)sites. Para ofc meaning side by side, which made me think of Haraway's sympoeisis (making-with) but also, admittedly, Limmy's madeup psychic show, Paraside (lalalol what you were saying about the scrambling parallel universe maybe, is that a lalaLacian Real which necessitates ululation, stammering? Complex remixing musicality of language throughout Body Work as summoning?). Going back to my incidental Slowdive reference earlier, maybe there's a shoegaze thing here, like setting up these 'noise-worlds' which shimmer indiscriminately behind/inside/through the semiotic oscillations of lyric? Is shoegaze a form of sonic gouache? Well it is certainly an ontic form of seduction, where I can't pick out the instruments of expression but I look for them hungrily in the haze. And the idea that transmission between worlds (the living/dead, human/nonhuman) might require a strain of humour (like haha but also meant in the sense of bodily humours?). For instance, shoegaze is decidedly not a humorous genre, but it sort of works on bodily humours, sometimes giving me the bends, or the blurry spaced-out feeling of having one's pleasure receptor's caressed by sound. Was wondering how YOU experienced the space and physicality of the poems -- was there anything u found FUNNY or sufficiently sultry as to produce a long and gorgeous sigh?

Mm and aren't there these tasty, cute moments of wow like 'tropic glut' ('O--NE') and 'prism arousal' ('Body Work') and 'clamour to emboss' ('Sapling').

Come to think of it, there are quite a lot of trees in Body Work, at the very least between 'Sapling' and 'Copsing', but also resonance in 'Awning', 'Annual' (which mentions 'yield', 'Thicket', 'sky-light muddle' etc) and 'Georgel' (georgics, idk?). Something about sprawl and thread: like the action of branches as arboreal mirror for threads of viruses, threads of code?

Side note: Can a person in a crowd of people experience canopy-shyness? Emily Berry has this lovely poem about crying and canopies and language.

Ways to dwell in inertia, violence, suspense ('Poem for July') as a 'clearing' within the pamphlet? Body Work as a title seems to combine two distinct fields: car repair and alternative medicine (hence mention of plants, cancers and crystals). The question of holistic approach, therapeutics, restoration. The sheen of metal, the sheen of health. O wise one of la letteratura del contemporaneo, pray tell your thoughts on possible Ballardian comparisons? Like obv v. different but I was struck by something to do with the cut-up structures of The Atrocity Exhibition and the way erasure works in Tom's work (probably in a more precise, attentive way, like the specialist's collage of tiny skins and digits, as opposed to grander themes of mediation that explode all over Ballard's work? -- generalising for the sake of interest obv).

Longing for a 'carvery [of] / uncommoning / rave'. Some kind of party you'd give up your skin for (is skin mere synecdoche of identity here?). Maybe the rave is what you were saying about scrambling.

Anyway, I hope your essay is going well. I must go read Hillis Miller's thoughts on Ariadne's thread, maybe make a tea. I've been getting these headaches lately, dawn to dusk & beyond, like the kind you get after being swimming (chlorine headache) or after crying (hormone headache). Pressurisation. I wonder if I have a parasite in my brain? So tonite I will probs lie awake, sleepless, listening for tinnitus :(

With warmth, Maria xxxxxxx

p.s.

Of course, by the time I get to the end of rereading I realise that it's only white marks being revealed underneath because literal holes have appeared in the Body Work cover, like some kind of fungus has been eating away at the book, performing another erasure.

Denise Bonetti <[email protected]> Fri, 1 Feb, 12:15 to maria.spamzine Dear Maria, Once again my legendary inability to reply to personal emails within a reasonable 1-month window manifests itself. Invoke my Scorpio moon (?) etc. I love it when people are like 'RIGHT - enough Facebook for me, email me if you want to talk etc', because that sounds like a nightmare to me. Long live IM! Long live the short form! (quite rich coming from someone whose job at the moment, I guess, is to churn out a dissertation?)

But then you know what I was thinking - someone like Clark Coolidge, for example, can get away with long form, intense long form. Not only get away, but own that long form. That long form I'm into. Clark Coolidge drown me in words and I'm fine with it, because he never dilutes, there's never any stagnation, you know what I mean, he just goes and goes and goes and you're like !!? YES!! HOW ARE YOU DOING THIS!? He just never runs out of steam. And I'm thinking of Coolidge because you mentioned crystals as agents in Tom B. (cf. The Crystal Text, how would the crystal speak etc), and of course because both Tom B. and Clark C. are just doing mad things with language, bold things and exciting things... They are like scientists you're friends with but who (maybe) don't like to talk about their work, then one day they decide to let into their basement lab where they've been secretly working on the most complex, organic, project for years, and they're like, don't freak out, here it is:

(Sorry for terrible quality [#postinternet] First is Tom B., second is Clark C.) Body Work looks quite controlled in form, visually speaking, with its vaguely justified lines, BIG symmetrical margins.. even the scattered pages look orderly! Like the bit of 'STRING WORTH' you sent. Which going back to your erasure thing, it makes me feel like Tom B. is giving us an OXO cube of his writing, all concentrated and delicious. But then my response is - show us more!!! Which rarely happens because I am scared of long form. And email. And dissertations. I also LOVE what you said about how Body Work combines car repair and alternative medicine!! That is absolutely spot on. Like the material, pragmatic tinkering motions of his writing, the referral to structures and the intention of like, see how far we can bend them and push them, but then it is never as dry as that! Very sweet motor oil. It's very kind poetry... generous! (A word that my friend Phoebe used to describe a certain type of poetry at a party last week and I thought, very interesting). Linguistically generous because it offers so many networks of reading, but then also.. approachable? As approachable as experimental poetry of this kind can be. I'm sure like, DANIEL would not think this is approachable lol (#COYBIG #romance). Which, fair enough. But if you're a nerd for this kind of poetry, then yeah. Like this bit from 'ANNUAL'??

*Cries!!! This is like you said, healing!! I feel looked after! 'Stomach prefers sound to day-to-day camphor'!!! Honestly what is this! So touching, so simple!

(Btw, I started experimenting with aromatherapy in my tiny room lol, do you know how to stop the water from boiling in the oil burner??)

Thanks for sending such interesting ideas over, I have to shoot to a seminar ! PPS: I saw Steven Connor in the English library yesterday (Oliver pointed at him silently like !!!!!!!!!!) so I kind of followed him to see what his approach to book browsing is.. very natural-looking and orderly? Surprised. Love the guy. *bubble sounds*

Lots of love Maria ‘let’s see where the spirits take us’ ur the best

xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx

Maria Sledmere <[email protected]> 6 Feb 2019, 23:19 to denise.spamzine Dear Denise,

Makes so much sense to map your message sensibilities onto your taste in poetry! I am so torn between the percolated richness of the email, its classic deferral (omg hun I owe you a million emails!) through a sort of quantum dimension of procrastination, and in opposition the sugar rush nowness of IM. I am such a frantic typist that often I send the wrong messages or cross my wires or just gush too much, so email is probably a safe option for me. There is all too much blue in my Messenger windows...Temptation of x's and endless emojis. But such a beauty to IM and texts out of context, like I wonder how many people read your probs too late 4 a snog now :'((( as micro-fictions, versus poems. I have a whole folder of screenshots on my computer from things that happened on Facebook that I have no memory of. Something about the Romantic fragment, accumulating ruin.

(btw) I feel like these extracts also shed light upon Body Work somehow. Biodegradables versus hard minerals and synthetic matter. Inner/outer. Flush. Tbh I think the middle one was from you?

Yes the dissertation, the dissertation as labour; it's like you have to find your scaffold first. Sometimes I feel like the scaffold in that wonderful Sophie Collins poem, 'Healers', and I write and don't notice myself and suddenly I'm so there, but the scaffold is secretly taking her bolt pins out the more I write around her. You can only be so respectful to your scaffold when she's so in the way. Gemini problems?Duality; structure/content. What is it Tom McCarthy says: 'structure is content, geometry is everything'. I want to be a wee fractal in a sequence of massive refraction. Is that how it works? Back to scaffolds, maybe we need to find the kindest mode of dismantling, and that's when you work into a form or something. And then also the more organic structures! So for CC it's the whole crystal thing, and working out of crystal logic. And then you just go and go and it's wonderful, much extravagant fractality, almost like poetry as virus, replicate replicate, grow, change. Mm, it's so good. My friend Kirsty did this mad poem about a tree, I couldn't tell if it was a story or poem, it was just branching out in a way that seemed hungry, necessary, spreading its roots. She said she wrote it in a rush! As if trees could rush! I like to think she inhabited a concentrated moment of becoming-tree, like she was a myriad in the roots or leaves. I don't think it could have happened without the tree, you know? But also the tree was almost entirely absent, it was like a ghost of form. Maybe I forgot how it goes. The lines looked like branches or something. Can you have long-form concrete? Concrete I guess by necessity is long-form. It takes a lot of energy to make. People are building houses out of mycelium instead, which is rad. Talking of roots and that, I just wrote 26k words on ecopoetics & t h r e a d s over the past fortnight and it was kind of that process, like letting a sort of tapestry take hold and I was maybe just one more thread, I was hardly doing the weaving, everything was moving around me and I wanted to wriggle into more and more gaps. Becoming-thread, perhaps. The next step is to slack and cut, which is exciting. Where to even start?

Your description of the complex, organic project is so gorgeous. Poems slow-cooked in a lab with tender organic care. My two scientist PhD pals are always gramming these beautiful pictures of crystals they're growing or mad wave patterns on screens. And we go for lunch and I'm like what you doing this afternoon and they're like, Oh just shooting photons. And is that much different from spending your afternoon writing poems? (Yes, they'd groan). I'm just chasing bits of light. Reading Tom B's work it's this whole precision thing, the actual inhabitation of process as such, so you see the energy buzz between things. I don't mean to say simply this is atomic poetry or poetry as tool analysis. It's more a betweening.

Isn't it super difficult to write non-anthropocentrically and really inhabit micro-relationality and also sound interesting and sexy in the way Claire Colebrook (she has that great essay in Tom Cohen's Telemorphosis) describes as 'sexual indifference', i.e. that threat to heteronormative reproduction that 'has always been warded off precisely because it opens the human organism to mutation, production, lines of descent and annihilation beyond that of its own intentionality'? Well anyway Body Work really works this way for me, it's like a poetics of sexual indifference that is nevertheless charged with desire you can't really predict, it's something in the frisson between objects and lines and coils of form. I think of crystal charge, iPhone battery (mine's always dying, Gemini trait 100%), engines. Neat miniaturisations of entropy, surge, spike and flux. When the 'I' comes in I'm like hey, what flow are you? It's actually so satisfying to quote these poems as fragments btw, they can do so much on their own as much as in poems and pamphlets, I wonder if that goes back to the accessibility thing. Like the absolute charm of a line as auto-affection:

This bit is from 'Temper' and I go back to my point about the flush/fluster! Globe of air/your *bubble sounds*. Isn't everything held so neatly, and yet it never feels neat, it just feels sharp and sparking, this 'technical glossy finish' like a really nice car, the body paint of a poem, its prosody so tightly held it feels more surface, a selection of hues and textures. And the erotics of the text or at the very least its pleasure is the shift between bodies, synecdoche, yes you could say bodies without organs but things in themselves are also important. Maybe another poet who does this is Sylvia Legris, she writes these apparently impersonal poems filled to the seams with specialist lexis (you have to have like twelve tabs open per poem to get it), but there's an affirmative humour and energy that feels v much a personal sensibility, a deliberated skewing of world that splices the poet's agency among items, artefacts, language. I mean how nice are these poemsshe published in Granta. I feel like I want cutlery to read them with, if that makes sense. Maybe a scalpel, for the succulence. The appearance of an ear again!

And then the beautiful metallurgy of this line from Tom, like somebody pierced my ears with perfect silver and it let all the demons out:

I am worried about what a certain seizure would look like. When we talk about vitality is it a willing naivety towards matter qua matter, as if we could just step out of correlationism? Such thoughts for 11pm of a Wednesday night. I can't help but think of the body image that Elizabeth Grosz describes in Volatile Bodies, kind of riffing off Paul Schilder: 'What psychoanalytic theory makes clear is that the body is literally written on, inscribed, by desire and signification, at the anatomical, physiological, and neurological levels'. And yeah, cool, what about the nonhuman body also? Has anyone done a really good psychoanalysis of the object. Parsed its psychic striations (traumatic or pleasurable residues of every microbial, huh?). In fact, what about the psychodynamic model of actual icebergs? Time we started literalising the matter of metaphors, absenting 'real cultural / medium' and filling with meltwater, fire and flow. Maybe it comes down to a bead of ink, the 'intimate concentrate' which is Lucozade, hangover piss, sick pH levels. So yeah, Body Work for me is this totally seductiveintersubjective space which actually works out pretty visceral states, sometimes disembodying me into a more fractal, mineral or bacterial being. I could start talking Kathryn Yusoff and geomorphism too, but maybe enough strata for one email? Plus I'm mixing my metaphors, I'm sure, mostly because I'm still morphing, dissolving inside those lines. I think I ate too many OXO cubes.

As for your oil burner boiling, sounds like you have an overactive candle? Maybe try a cooler tealight, nestle it to the back a little to redirect the strength of the flame? I like rosemary oil for remembrance, cranberry for comfort, ginger for energy. That line about resin is so nice. I was in Crianlarich at the weekend and my friend Patrick found this massive log and he carried it for so long that you could smell the resin on his skin, it was amazing. I keep thinking about the word 'pitch' and lush tree-ness, and the Log Lady in Twin Peaks and poetry you can chew like new molasses, prior to melt. Is that how it works?

Somebody is smashing glass into a bin in my garden and probably I should just close the pamphlet...

...but it's like a delicious pdf that gives infinities...

Yours in multiples & cherryish flusters,

Maria xoxoxoxooxoxoxoxoxooxoxoxox

#review #tom betteridge #body work #Sad Press #reviews

1 note · View note

drmikewatts · 1 month ago

Text

Complex & Intelligent Systems, Volume 11, Issue 5, May 2025

1) Predicting trajectories of coastal area vessels with a lightweight Slice-Diff self attention

Author(s): Jinxu Zhang, Jin Liu, Junxiang Wang

2) Micro-expression spotting based on multi-modal hierarchical semantic guided deep fusion and optical flow driven feature integration

Author(s): Haolin Chang, Zhihua Xie, Fan Yang

3) Wavelet attention-based implicit multi-granularity super-resolution network

Author(s): Chen Boying, Shi Jie

4) Gaitformer: a spatial-temporal attention-enhanced network without softmax for Parkinson’s disease early detection

Author(s): Shupei Jiao, Hua Huo, Dongfang Li

5) A two-stage algorithm based on greedy ant colony optimization for travelling thief problem

Author(s): Zheng Zhang, Xiao-Yun Xia, Jun Zhang

6) Graph-based adaptive feature fusion neural network model for person-job fit

Author(s): Xia Xue, Feilong Wang, Baoli Wang

7) Fractals in Sb-metric spaces

Author(s): Fahim Ud Din, Sheeza Nawaz, Fairouz Tchier

8) Cooperative path planning optimization for ship-drone delivery in maritime supply operations

Author(s): Xiang Li, Hongguang Zhang

9) Reducing hallucinations of large language models via hierarchical semantic piece

Author(s): Yanyi Liu, Qingwen Yang, Yingyou Wen

10) A surrogate-assisted differential evolution algorithm with a dual-space-driven selection strategy for expensive optimization problems

Author(s): Hanqing Liu, Zhigang Ren, Wenhao Du

11) Knowledge graph-based entity alignment with unified representation for auditing

Author(s): Youhua Zhou, Xueming Yan, Fangqing Liu

12) A parallel large-scale multiobjective evolutionary algorithm based on two-space decomposition

Author(s): Feng Yin, Bin Cao

13) A study of enhanced visual perception of marine biology images based on diffusion-GAN

Author(s): Feifan Yao, Huiying Zhang, Pan Xiao

14) Research on knowledge tracing based on learner fatigue state

Author(s): Haoyu Wang, Qianxi Wu, Guohui Zhou

15) An exploration-enhanced hybrid algorithm based on regularity evolution for multi-objective multi-UAV 3-D path planning

Author(s): Zhenzu Bai, Haiyin Zhou, Jiongqi Wang

16) Correction to: Edge-centric optimization: a novel strategy for minimizing information loss in graph-to-text generation

Author(s): Yao Zheng, Jingyuan Li, Yuanzhuo Wang

17) A reliability centred maintenance-oriented framework for modelling, evaluating, and optimising complex repairable flow networks

Author(s): Nicholas Kaliszewski, Romeo Marian, Javaan Chahl

18) Enhancing implicit sentiment analysis via knowledge enhancement and context information

Author(s): Yanying Mao, Qun Liu, Yu Zhang

19) The opinion dynamics model for group decision making with probabilistic uncertain linguistic information

Author(s): Jianping Fan, Zhuxuan Jin, Meiqin Wu

20) Co-evolutionary algorithm with a region-based diversity enhancement strategy

Author(s): Kangshun Li, RuoLin RuanHui, Wang

21) SLPOD: superclass learning on point cloud object detection

Author(s): Xiaokang Yang, Kai Zhang, Zhiheng Zhang

22) Transformer-based multiple instance learning network with 2D positional encoding for histopathology image classification

Author(s): Bin Yang, Lei Ding, Bo Liu

23) Traffic signal optimization control method based on attention mechanism updated weights double deep Q network

Author(s): Huizhen Zhang, Zhenwei Fang, Xinyan Zeng

24) Enhancing cyber defense strategies with discrete multi-dimensional Z-numbers: a multi-attribute decision-making approach

Author(s): Aiting Yao, Huang Chen, Xuejun Li

25) A lightweight vision transformer with weighted global average pooling: implications for IoMT applications

Author(s): Huiyao Dong, Igor Kotenko, Shimin Dong

26) Self-attention-based graph transformation learning for anomaly detection in multivariate time series

Author(s): Qiushi Wang, Yueming Zhu, Yunbin Ma

27) TransRNetFuse: a highly accurate and precise boundary FCN-transformer feature integration for medical image segmentation

Author(s): Baotian Li, Jing Zhou, Jia Wu

28) A generative model-based coevolutionary training framework for noise-tolerant softsensors in wastewater treatment processes

Author(s): Yu Peng, Erchao Li

29) Mcaaco: a multi-objective strategy heuristic search algorithm for solving capacitated vehicle routing problems

Author(s): Yanling Chen, Jingyi Wei, Jie Zhou

30) A heuristic-assisted deep reinforcement learning algorithm for flexible job shop scheduling with transport constraints

Author(s): Xiaoting Dong, Guangxi Wan, Peng Zeng

#ai #artificial intelligence

0 notes

randomblogger-stuff · 4 years ago

Text

10 PROVEN CONTENT WRITING TIPS FOR BEGINNERS - STEP BY STEP GUIDE

INTRODUCTION:

Content writing is an art & creativity, still don't you know, How to write perfect Content? Easy-peasy, I give some fantastic content writing tips for beginners to improve their writing skills. Content writing is up to educate the audience; by writing eye-catchy attention-grabbing tense access the reader's mind to take immediate action; Content communicates with your audience via writing excellent sentences in your blogs or on the social networking description. Follow the upcoming content writing formats, accomplish and nurture your content writing skills, and it will help you to add more value to your article.

1.DEFINING YOUR NICHE

· Before you start writing the content fix with your Niche, based on your Niche, you need to register your Content that should satisfy your Niche and be related to your audience. This is a very base content writing tip for a beginner.

2.KEYWORDS FILTER

· Come up with the proper sense of keywords relevant to your Niche. Before start writing the Content, search for the phrases user's most searched for; choose the low competition keywords with high demand. Pick 5 to 10 keywords integrate those keywords in your articles.

3.DECIDE THE CONVEYING TONE

Ø Narrate your Content in a different tone

FIRST-PERSON NARRATIVE

· The first-person narrative is about describing yourself, sharing your feedback, real-time experience, and giving reviews for what you had experienced in the past. Use (I, WE) terms to write the article in a first-person narrative tone.

Ø Example: Bloggers and Reviewers use the First-person communicative tone.

SECOND PERSON NARRATIVE

· Second Person narrative describes the products or services you're selling, Detailing the advantages and disadvantages of the products, educating the audience about your products or services. Use (You) term to write the article in a second-person narrative tone.

Ø Example: Business owners or service providers uses a second-person narrative tone to up-sell their products.

THIRD-PERSON NARRATIVE

· The third-person narrative speaks about others' achievements, drawbacks, and gossip about whatever went viral on the market. Whether it's a good or lousy contrivance, we can talk about that activity. Use (He, She, It, They) terms in your article to write the article in a third-person narrative tone.

Ø Example: News articles, Media articles, Movie reviewers.

4.DECIDE COMMUNICATION STYLE FOR YOUR BLOG

FORMAL:

· If you're Writing any technical blog for industries or writing any new articles for your business or services, your approach and writing style should be Formal. This ethical way of approach helps to convey and compromise your target audience to take immediate action.

UNFORMAL:

· If you're writing any non-techie blogs apart from the ROI mindset, you can write any new articles up to your mindset or the Niche that you're trusting very confident and unbreakable. That gives more engagement to your blogs and grabs more unique visitors to your site.

5.TITLE FRAMING:

· The title is the crucial factor for all blogs to increase engagement for our page. Title framing plays a hugely significant role in CTR. Play with your title before you publish a blog post. Every reader should have an eye on the title so every content writing beginner be much focused on title framing; Your title must have these significant elements,

Ø Magical words + Keywords + Powerful/Emotional words.

6.SKETCH AN OUTLINE:

· Before we start the content writing, Frame an outline for the article to keep engaging the audience throughout the article, The below elements need to consider in the blog article writing,

Ø INTRODUCTION

Ø BODY

Ø CONCLUSION

7.INTRODUCTION

· Make one good attention-grabbing introduction before start explaining the article in detail. Once you grab the attention and exaggerate the audience's problem in sentences, then give a solution to their problems. Make one clear CTA at the end of the content description. The universal formula for writing perfect descriptions is called an AIDA. Implement the AIDA strategy in the introduction for your blog posts and integrate the AIDA in social media descriptions for your business or services. One of the best-proven writing strategies followed by the entire universe is called the AIDA method. Make one perfect good description using AIDA is up to your creativity and content lineup tactic.

8.BODY

· The body of the article is like furnishing the aesthetics of the entire article. Break the paragraphs, use subheadings to the side topics, use more images as many possible this creates the desire in the accessor's mind to read the entire article. The body is the main element to reduce the bounce rate on your page.

Ø NOTE: The requisite element for every content writing beginner needs to look at these above content writing formats.

9.CONCLUSION

· A decision will summarize what the reader learned from your writing and how they benefitted from your blog. If the reader not gets satisfied with your Content, you can rearrange some aspects of the article according to the reader's review perspective or come back with some other good blog article.

Example:

Ø Ask a question to the reader to leave a comment or react to your post.

10.CTA

· Give clear CTA at the end of every article that increases your social score and share links to your own referring page that gets more visibility to your re-directing page also for your social media profiles, This drives more engagement to your referring page and social networks, must needed element every content writing beginner need to consider it in blog article writing format.

STEPS NEED TO BE FOLLOWED IN CONTENT WRITING

11.FORMATTING THE ARTICLE

· Format your article with the following upcoming sections. Formatting helps us to candle up the article more for reader convenience without any obstacles,

a) BREAK PARAGRAPHS

· Break your paragraphs. Don't make longer paragraphs in writing that create lazy dopamine on the reader's mind to finish the entire section; Segment each paragraph section-wise, which can easily occupy the reader to look after the entire article.

b) AXING

· Ax your article after completing your Content. Proofread your article twice or thrice to find mistakes and irrelevant lines accumulated on your writing, remove all the unrelated sentences and unwanted phrases. Also, make sure your essay needs to perfect with the extracted sentences, and prioritize keywords frequently while axing the entire article.

c) KEYWORD FREQUENCY

· Frequently split your keywords; don't stuff your keywords inside the article that makes some annoying over do's while reading the article. Split up your keywords according to the words count for 200 words. One keyword is suggested. Put LSI keywords in your writing to overcome the Keyword stuffing.

d) IMAGES

· Images play a significant role in blog articles. We can explain most sections without words by using the images, in return we receive more engagement rate to our blog post. Use images in your blog articles that are related to your Content. If you're not good at creating images, outsource the ideas for designing or if you're looted the other creativity, give the credits to the original creator.

e) BULLET POINTS AND NUMBERING

· Step down your paragraphs using Bullet points and Numbers, line down the essential points put numbers for your subheadings. That creates a beautiful appearance for your article. It's the most innovative way to make the reader engage with your blogs.

f) CONSISTENCY:

· Maintain consistency throughout the article; Content detailing which you're implemented on the first segment of the blog needs to be in the remaining every segment of the article. Every single detail needs to maintain consistency throughout the entire article. Don't make an uneven segment that gives more bounce rate to your page.

Ø Follow the metrics and standards; these were proven and standardized blog article writing format; also, it has been the verified tactics to acquire and engage your audience via content writing.

FAQ:

1.WHAT IS CONTENT WRITING?

· Content writing is an art; it accesses the reader's mind to take immediate action. The content writing creativity is up to educate the audience and conveys your message via writing unique Content.

2.WHAT IS A KEYWORD?

· Any word you searched on the internet to get a specific answer on the search engine is called a keyword.

3. WHAT ARE ALL THE TYPES OF KEYWORDS?

· SHORT-TAIL KEYWORD

· MID-TAIL KEYWORD

· LONG-TAIL KEYWORD

· LSI KEYWORD

SHORT TAIL KEYWORD

Ø The word that contains a single term or a single phrase is called the short tail keyword. This keyword gives many broad search results to the users to discover lot more results according to their query.

EXAMPLE: Non-Veg food

MID TAIL KEYWORD

Ø The word that contains more than one single term or single phrase is called the mid-tail keyword. This type of keyword gives slight broad search results to the user's query.

EXAMPLE: Non-Veg food Restaurants.

LONG-TAIL KEYWORD

Ø The word that contains a long phrase in the search query is called the Longtail keyword. This keyword gives exact search results to the users to get specific information about the searcher query.

EXAMPLE: Non-Veg food Restaurants in Chennai.

LSI KEYWORDS

Ø LSI Keywords stand for LSI- (Latent Semantic Indexing).

Ø The words which are related to the primary keywords are called the LSI keywords.

EXAMPLE: NON-VEG FOODS ( PRIMARY KEYWORD )

Ø Chicken, Mutton, Prawn, Crab

0 notes

compneuropapers · 6 years ago

Text

Interesting Reviews for Week 41, 2019

What and How the Deaf Brain Sees. Alencar, C. D. C., Butler, B. E., & Lomber, S. G. (2019). Journal of Cognitive Neuroscience, 31(8), 1091–1109.

Neural correlates of sparse coding and dimensionality reduction. Beyeler, M., Rounds, E. L., Carlson, K. D., Dutt, N., & Krichmar, J. L. (2019). PLOS Computational Biology, 15(6), e1006908.

A review of neurobiological factors underlying the selective enhancement of memory at encoding, consolidation, and retrieval. Crowley, R., Bendor, D., & Javadi, A.-H. (2019). Progress in Neurobiology, 179, 101615.

Recent progress in semantic image segmentation. Liu, X., Deng, Z., & Yang, Y. (2019). Artificial Intelligence Review, 52(2), 1089–1106.

The role of the opioid system in decision making and cognitive control: A review. van Steenbergen, H., Eikemo, M., & Leknes, S. (2019). Cognitive, Affective, & Behavioral Neuroscience, 19(3), 435–458.

#science #Neuroscience #computational neuroscience #Brain science #research #scientific publications #reviews #machine learning

4 notes · View notes

senzacaponecoda · 5 years ago

Text

pAA?

(Decided to reduce the names to E, O and S because this isn’t something that should be treated seriously or like there’s anything in particular respectable or responsible about it, but might be helpful to someone like me trying to find something out for personal reasons.) So since I finally got my hands on a copy of O and S’s reconstruction of PAA, I thought I’d go through and pick out what I thought were their best etymons against what I think are E’s best etymons. As I’ve said before both reconstructions seem to be basically mass comparison reconstructions.

The wiki page on Afroasiatic claims they agree on basically nothing, and out of 1000ish and 2600ish I still got less than 100 shared roots, so that’s probably accurate. The page and what I pulled out didn’t exactly line up though, but also there were some roots especially on E’s end (due to a need to reconstruct verbs to nominalize for his PS theories) that I outright changed the meaning of because they actually made his individual arguments stronger; like, say, if you had 4 roots that mean knife in each sub-branch, why assume the etymon means cut?

I compared the roots in case there was something that was an obvious oversight, and while I don’t think their methods were all that good (although O and S’s are a lot better), I didn’t want to discount everything in case there was something like “well their vowel systems are different so I’m not counting roots where even V[os] = V[e]” or like “O/S posit (g,x) < *q while E has (g,x) < *ÿ” being overlooked and miraculously they agreed on like 300 more roots than the page gives them credit. There’s also issues like, well, “Is this particular author doing such a bad job but the critics aren’t seeing that, that he/she/they are dragging the other author/s down by undermining their ok etymons?” etc.

So, like, I’ve messed with the data in a way that’s irrepairable and shit. Mind that. This isn’t scholarly work and isn’t intended to be. And I messed with some of the data - even if I think it’s more representative of their works at a level, from any kind of responsible position I need to make it absolutely clear that it’s not really reflective of their work, but a reflection of my amateur, shitty thoughts on their concordances.

So, for the curious:

So when putting this list together I accidentally kept mixing which side I wrote whose roots together. E tends to have labiovelars where OS have o, E has long vowels, OS don’t, E has tones, OS don’t, E has the voiced velar fricative I transcribed as ÿ and ŋ, OS have q and q’ vs k’ and k. E has p’, the rest I guess you’ll have to guess.

I also marked a few words with * for what might be imitative or baby speak words. *? like on horn is “I’m on the fence of it being imitative”

Vowels are pretty much ignored, but otherwise definite correlates:

kol-f bark (n) dam blood k'os~k'as bone naf breath ben~bin build di3, da3, du3 call dumn~deman cloud k'at' cut* 2ab father* pir fly*? ba2 go sim hear lib heart k'ar horn*? inkwal~2ankol kidney 2er~2âr know lVk'~lak lick* tir liver sum,sim~süm,sim name wan open bu~baw place dak,duk~dik pound k'u2 rise tuf spin ra3-raa3 sun dab,dib-dub tail les,lis tongue ma2 water

Close matches, differing by not a lot of phonological space, but the correlations seem unsystematic or quasi-systematic (E *k = OS *x? or the other way). A lot of this might be due to obfuscation of the subgroups; it’s known beyond these reconstructions that Semitic b, p, f correspond to Egyptian b, p, f but there doesn’t seem to be a pattern to the correspondences beyond [+labial][-nasal]=[+labial][-nasal]... except some roots where b = m, mostly Egyptian to Semitic (iirc it’s believed that snb = slm).

! maaw~mawut die ! har~heraw day ! t(l)'ok' beat ! bak~bax burn ! k'al~k'ar burn ! (t)san brother ! kor~kw'al angry #related to kidney? ! pak~pax break apart* ! yar~3er burn #n.b. OS *e = E *ya pretty consistently ! ka2(up) cover ! t'ub~duf drip* ! g'arub~ÿar(b) dusk ! 2et~iit eat*? ! gur~guud enclose ! 2ir~2il eye #2il is also given for both by the wiki author ! 2aakw~2ax fire ! ŋiiwr~gir flames ! pur~fir flower 1 k'ur~kâr go around ! ĉa3ar~ła2r hair ! qafV3~gâf hold ! qam~kam hold ! fil~bul hole ! ĉa(2i)d~gwi/ad land ! ne2ul~ñaw moist ! bakr~bar morning ! âf~2ap mouth* ! ĉer~sar,sir root ! cab~sVp sew ! sur~tsur sing ! tsoon~soon smell ! bak'~p'ak' split* ! da2~daw walk

Really kind of stretched correspondences, usually requiring twice the amount of special pleading as above:

!! rip~2erib sew !! dabn~zab hair !! yawr~yabil bull !! büł~fil skin !! 3ir~raw sky

A root that’s probably wrong but I didn’t delete it:

!!! c'eyg~c'a3ek shout

Roots on the wiki article it says are part of “the fragile consensus” but either escaped my standards for good etymologies from E and OS or were lost when I changed E or OS’s proposed definitions to fit the data they presented for etymologization:

>> (2a)bVr bull >> (2a)dVm land >> 2igar~kw'ar enclosure >> 3ayn eye #I think for both this was only found in Sem and Egn, which is proto-Sem-Egn, not PAA >> bar son >> gamm mane, beard >> gVn cheek, chin >> gwar3~(gora3?) throat #These words weren’t in E’s but look like it and I guessed on what the OS reconstruction would look like >> gwina3~(gona3?) hand >> kVn cowife >> kwaly??? kidney #that this isn’t VnkwVl damns something >> k'awal, qwar say, call >> sin tooth >> siwan know >> zwr seed >> łVr root >> šun sleep, dream

These I sort of gathered based on a decent scholarly Egyptologist’s work

** (n)i,ku,nak(u) 1sc ** nVn 1pc ** kumV, kV 2sm ** kimV, ki 2sf ** suwa, sV 3sm ** si(y/t)a, si 3sf **-ú~-aw nom (earlier an ergative) **-á abs **Egyptian had an outright ergative participle so like yeah Some AA assumptions $$ m- participle, possible 3 or 4 denoting instruments, resultatives, passive and active meanings, and more. $$ s- causative (likely su-) $$ -í nisba = adjectivizer $$ -át/-út ending for abstract, mass, diminutive, augmentative, female nouns $$ top two combine frequently $$ maybe k became a masculine? maybe u/w too? $$ perhaps an -n- passive? -t- passive? reciprocal voice? (kick each other) $$ emphatic coronal series $$ prefixing conjugation {2 t t y t n t t y t} on imperfect-type stems $$ core verb stems seem to be perfect vs imperfect $$ a lot of sources report *-r has spreading out semantics at the end of a Semitic verb and I believe it. $$ suffixing possessive pronouns $$ probably all of pAA derives its verb system from those same possessive pronouns suffixed to participles, which probably helped the ergative to nominative alignment change. Again, Egn had an ergative participle at one point. It actually used a slightly different form of the possessed agreement system for that though, maybe reflective of a dead case system. $$ personally I think egn broke off early, sem and amz were geographically between it and ethiopia, the cushitic languages broke early but stayed in contact, chadic is cladistically cushitic that broke off in the middle of the ethiopianization of cushitic in the post-islamic era, since it seems to be in cultural memory. beja wasn’t as ethiopianized and is slightly closer to semitic, amz, and in many ways egn surprisingly, omotic was probably adopted into the family, ongota is too creolized to know anything, tones, vowel rebalancing, probably influences of Central/Eastern Sudanic. I have no idea why languages with ATR (i.e. pharyngealization) distinctions would make pharyngealized languages switch to ejectives, so if sem-amz is a thing you just have to postulate why they switched, since egyptian is no help. for all i know egyptian always had an aspiration distinction (which would explain some insane correlations) and then they fell together and split again for coptic. Don’t think this can be projected to the proto-lang: $$ ablauting everywhere, in nouns and verbs. masdars and singular nouns probably reflect original vowel patterns. !! E and OS reconstruct consonant systems similar to PS and idk about that. Both give 5 vowel core systems but for different reasons and different results. E has length and the central short vowels fall together in all but Omo. Some shenanigans happen to get Cu uu, ii, aa, ə, and ä. Ch ends up with a i u ə, and the other three a i u. OS has e yield ya and o yield wa, ü for u~i results, and does overall less movement.

These are numbers from my number project:

## tsin~tsar two ## kwrad three ## fVd'w four ## magw ten

It’s interesting that many of these shared roots are in fact Swadesh roots and that that mostly happened on accident. E reconsructed 1000 roots, OS 2600, and just using judgements like “a root needs to mean the same thing in two (standard I held E to) or three (standard I used for OS) subbranches” to sift through them brought me down to these. But like most of these roots are fundamentally two segments long. There were even ~false matches in some of them with the phonaesthetic shape of their English glosses - horn would go back to something like *karn in PIE, which looks like the root here. Maybe horns make a kar sound, but with 2iit~2et, idk, I don’t think of eating as making an “eat-eat” sound; I actually threw away a number of roots that had a “kwa-l-kwa-l” (whence Ar. 2akulu) sound which at least kind of sounds like swallowing.

0 notes

happycoincidences · 8 years ago

Text

Neuroscience, linguistics and neurolinguistics

@epiphanicwiring - I made a list of sources (with some publications I found interesting that hopefully are interesting to you too):

Journal of neuroscience sometimes has free pulications e.g. Stimulating Multiple-Demand Cortex Enhances Vocabulary Learning (Magdalena W. Sliwinska et al., 9 August 2017).

eNeuro is an open-access journal of the Society for Neuroscience.

NeuroscienceNews.com is an independent science news website focusing mainly on neuroscience and other cognitive sciences.

Another one is Science but with restricted access.

Nature.com: an article I found interesting, and I think you might find interesting as well is How to map the circuits that define us (Kerri Smith, 9 August 2017). Of course this is not about the human brain. With today’s technology it would take too long to completely map a human brain, a whole fruit-fly brain is already too big to reconstruct, a map of only that would even take hundreds years to complete.

Pubmed: it compromises citations for biomedical literature, life science journals, and online books about science in general, and thus also neuroscience:

Sensitivity to syntax in visual cortex (Suzanne Dikker et al., 3 January 2009). Language is among the most complex of human cognitive systems, yet its processing is extremely automated and fast. It is about syntactic prediction and the violations of syntactic predictions in the brain.

Prediction Signatures in the Brain: Semantic Pre-Activation during Language Comprehension (Burkhard Maess et al., 15 November 2016): another one about word and sentence prediction.

The interplay of discourse congruence and lexical association during sentence processing: Evidence from ERPs and eye tracking (C. Christine Camblin et al., 7 January 2007). Subjects read sentences that were locally congruent but whose discourse congruence was manipulated.

Statistical learning of tone sequences by human infants and adults (Jenny R. Saffran et al., March 1999), that was published to Pubmed (but you have to buy it), luckily I have found a free source. The paper shows that preverbal infants could compute transitional probabilities (here they used them for word segmentation).

Springer provides scientific articles for all disciplines and they sometimes do a sale, last year I downloaded many linguistics research publications for free. This is the link for all the neuroscience articles . I do have to warn you that most of the articles are restricted (payment) access, but perhaps they will do a sale again!

Some selected neuroscience publications from The Neuroscience Institute (1998-2013! Possible outdated! But perhaps still useful though! I have read publications from the early 20th century that were still interesting and useful today, but when you write your scientific theses these sources do not count, for you always have to use the most recent one of that topic of course).

There is also Eos (maandblad over wetenschap) a scientific Belgian magazine in Dutch. And Wetenschap in beeld, also in Dutch (unfortunately). I buy both of these magazines from time to time, especially when there is a special publictation on astronomy, neuroscience, linguistics, etc.

Wait but Why is a blog about all things science and more:

Neuralink and the Brain’s Magical Future (Tim Urban, 20 April 2017).

The AI Revolution: The Road to Superintelligence (Tim Urban, 22 January 2015), this is about artificial intelligence and what possible future effect it can bring, so not completely related to neuroscience, but a very interesting read nonetheless!

These are more purely linguistics (with some neurolinguistcs in there too) than neuroscience:

Language Project Muse: some are open and some are restricted access.

Another popular one is Language Log

Speculative Grammarian : satirical linguistics.

ConLangery: a podcast about linguistics.

Phonology; International Phonetic Alphabet (IPA) sources:

web.uvic.ca : a whole chart

Seeing Speech: The International Phonetic Alphabet (revised to 2005)

Sounds of the World’s Languages : the two last websites do not only provide a chart but also sound and video. To not only hear but also see the sound pronounced in the mouth and throat is very entertaining to me (I have a big fascination for phonology, so I have these websites even bookmarked).

If you are interested in morphosyntax Sabine Iatridou's work would probably be interesting.

Lastly, I think this might be the most interesting article about linguistics to you: An empirical comparison of rhythm in language and music (Aniruddh D. Patel, Joseph R. Daniele, 3 September 2002; accepted 1 October 2002): it shows that English and French musical themes are signiﬁcantly different in a measure of rhythm, which also differentiates the rhythm of spoken English and French, suggesting that spoken prosody influences a culture's music.

#thought of giving you a few links but got carried away #hope you like it #neuroscience #linguistics #language #neurolinguistics

14 notes · View notes

recosense-blog · 6 years ago

Text

Top Trends in Cognitive Computing

As we have arrived in the third decade of computing –the cognitive decade–and it will bring on a fundamental change in the manner humans work with machines. There will be a breakthrough in the way the computers used to function earlier, as with the onset of cognitive computing, various software developers and engineers are bridging the gap between the way the computer functions and the way a human brain thinks. The result is intuitive and interactive however which improves over time and can help humans in expanding their limits of capabilities, increasing knowledge and making well-informed decisions.

Cognitive Computing is a new technology that permits people to communicate with computers using a natural language. Earlier, people used to code or format their text in a manner for the system to understand. For example: if they wish to search something, they used to input the keywords. Natural language processing permits the users to ask questions or speak in sentences while communicating with their systems in the same manner as we talk to another person.

How Cognitive computing works?

Cognitive computing systems are a consistent flow of learning techniques. These techniques employ different processes such as data mining, recognition of patterns and language processing. Although, cognitive computing is witnessing a dramatic change as companies are now engaged in developing Artificial Intelligence(AI) and are competing in the market to get a foothold. Cognitive computing can assist us in realising the accurate AI as it helps in imitating the activity of the human brain. These systems can be useful in developing an automated system for solving complex problems by itself.

IBM is the forerunner of Cognitive computing technology. The company has made a billion-dollar investment in big data analytics and is now on the verge to spend near to one-third of its R&D budget in the development of cognitive computing technology.

Pillars of Cognitive Computing

The three main pillars of Cognitive Computing are:

Machine Learning

Big Data Analytics

Cloud Computing

Machine Learning

Machine Learning is making the algorithms put to use. The objective of the algorithm is to examine the data and estimate the trends. Normally, there is a set of training data that feeds the information.

You are required to examine this data on several data parameters to determine its efficiency. However, a self-learning feature is necessary for Cognitive Computing. From there itself, Machine Learning comes into the picture.

Big Data Analytics

The processing of huge amounts of information is the function of the human brain. Understanding the relation of any question is essential. Machines can make it happen faster once you feed the data in it.

But there are two forms of data – Organized and Unorganized. Here, Big Data Analytics plays a role in processing the huge data for you and finding the relevant one you need.

Cloud Computing

If you have huge data to process, you need to have high computing power. The demand is continuously changing in cognitive computing. However, scalability becomes a big factor.

From there Cloud Computing comes in the limelight which helps in providing you with scalable resources. This type of environment works great for Cognitive Computing.

Phenomenal Trends Seen in Cognitive Computing

Chatbots

The simulation of human behaviour by computer programs are known as Chatbots. It employs the feature of Natural Language Processing (NLP).

NLP is used for taking inputs from a human. The chatbots are the intelligent feature used by Cognitive Computing. It can analyse the situation and answer you accordingly.

For example, Chatbots such as Google Assistant, Alexa, Siri are all intelligent bots using cloud-based natural language processing. The cloud capacity permits the business to store user based relevant information and offer tailor-made solutions, products, and messages subject to the behaviour and user preferences. These chatbots optimize the cloud computing potential to offer personalised context significant customer experiences.

Sentiment Analysis

Sentiment Analysis is associated with determining various forms of emotion. Emotions are a normal thing in a human which comes out naturally. But you need to train a machine by feeding a set of data conversations.

Sentimental analysis can be put to use by many companies for splitting up the social media conversation. You can examine individual posts, likes, replies, comments, etc. very spontaneously.

Face Detection

The image processing that detects your upper face is known as face detection. The cognitive system is put to use to analyse the eye colours, lips, contours, etc. to scan a face. Once a facial image has been generated, one can easily recognize the face.

Normally, 2D images are used for face detection. However, it can also be performed for 3D models. Primarily, the face detection is helpful in any security surveillance system.

Risk Management

Examining the possibility of unforeseen risks via current trends prediction is all about risk management. This feature does analyze data, instincts, behaviour, etc. It is a mixture of both art and science.

Cognitive Computing supports you to amass data and trends. Depending on this collection, it can create actionable insights. Moreover, an analyst would be a better person to contact for any further details in this regard.

Fraud Detection

Identifying any abnormal transaction results in fraud detection. In order to study the frauds happened, you need to go through the historical data for identifying the reason. From there, cognitive computing plays a vital role.

Trends by Industry

The top cognitive computing trends are clearly witnessed in the following segments:

(1) Cognitive Computing in OTT

Cognitive Computing capabilities are also harnessed to produce Over-the-Top-Content for delivering video-based content in the form of web series and movies. The cognitive technologies allow the learning and classifying the information from different organized and unstructured data may at times be cumbersome. But the substance administration can potentially be benefitted from AI innovation as feeling, image and voice recognition has involved media innovation buyers to increasingly depend on AI apparatuses to cluster and search for substance files. So that, AI can provide support in substance labelling (which is a costly affair). With the fundamentals of metadata, go through the substance administration structure which ends up having more precision in this way which empowers monetization avenues. Artificial intelligence, coined as “cognitive computing intelligence” in media industry claims to track the user client habits of utilization.

Now, OTT is no more confined with Android-backed Smartphone, instead, manufacturers are developing Android TV sets along with apps to necessarily assist the users in delivering Omnichannel experiences throughout the devices.

(2) Cognitive Computing in News Industry

To allow natural human interaction to flow with cognitive computing tools, the IBM Watson platform has assemblage various capabilities such as:

Speech and Image recognition

In-depth text understanding

Detecting emotion and expression

Translation tools

Relation extraction

When a user wants to read, listen or hear news – with the already embedded cognitive abilities in the device can present the news in the form of text, image or voice according to the user preferences.

The latest news update, forums, weather forecast, sports, politics, business, finance –are all under the cognitive controls backed by four API categories:

Language: An API combination, but not restricted to, categorizing natural language text, entity extraction, conversations, converting documents, extracting semantic concept, language translation, retrieval of passage and ranking, relationship extraction, tone analysis and much more.

Speech: An API combination for speech to text conversion and text to speech conversion which includes the potential to train with your own language models.

Vision: APIs to search new perspectives, gather relevant value and take actionable insights from images.

Data Insights: News is a pre-enriched content that comes with a natural language processing which allows the highly-targeted search and trend analysis.

(3) Cognitive Computing in Retail

Cognitive computing permits enterprises to increase human expertise vigorously. The cognitive computing technologies enable consumers to assist the retail industry to discover new opportunities and make rational shopping decisions.

The technologies also provide interactive, adaptive and contextual systems for making and permit retailers for creating quality tailor-made experiences for individual customers. The following are some rapid trends that are transforming the retail industry:

IBM Watson Commerce

IBM Watson Commerce is a software program for managing the business processes of both retail and online merchandisers. It also focuses on enhancing the customer involvement initiatives. The software program put Insights Assistant to use which help merchandisers to recognize abnormal situations in their marketplaces and advocate suitable actions.

Predictive Analysis

Predictive analysis helps online retailers anticipate what the customers would like to buy in advance. This would help them to predict the customers buying choices such as knowing the highest amount for a certain item the customer is willing to pay. It assists them in optimizing their customer service. Their analytical tools search at the several determinants that can assist in generating the desired involvement from the client such as clicking an online product promotion or subscribing to a newsletter.

Voice Assistants

It is amazing to know the engagement of activated voice digital assistants revolutionising the retail sector. These devices permit the customers to make orders of their preferred product choices and get them delivered within a few hours at their homes.

Direct Purchase via Social Media

Social media platforms such as Facebook, Twitter, Instagram and Pinterest are also navigating the online shopping sales by offering several purchase options. For example: when you see a product suggestion on your social media platform, upon clicking on the product’s image, an application combined with cognitive capabilities will navigate you to the product page. This will help to drive instant traffic on an e-commerce website and enhance the chances of a customer buying a specific product.

Personalized Recommendations

Online retailers use cognitive technology to offer their clients instant personalized recommendations in the form of mobile notifications or e-mail newsletters regarding products they should consider to purchase. This will help the retailers to develop complex insights into their product and generating customer base for a particular product.

Demand Forecasting

To sustain the immediate demand from their customers, online retailers make estimations regarding future demands with the help of demand forecasting. There are a couple of demand forecasting tools that are based on cognitive technology and machine learning will help the online retailers facilitate the precision of deriving probabilities of future demand.

Price Optimization

It becomes difficult for online retailers to accurately determine the price fluctuations responsiveness in their respective markets. This is due to factors such as competitor pricing, supply constraints and complex predictions of customer demands. To face this situation, online retailers can put price optimization tools to use subject to the cognitive technologies for making predictions on how their clients will respond to the price fluctuations of different products or services. These tools help the online retail industry to fulfil their objective of maximizing profit.

#cognitivecomputing #cloudcomputing #artificial intelligence #trends

0 notes

drmikewatts · 3 months ago

Text

IEEE Transactions on Cognitive and Developmental Systems, Volume 17, Issue 2, April 2025

1) Programmable Bionic Control Circuit Based on Central Pattern Generator

Author(s): Qinghui Hong, Qing Li, Jia Li, Jingru Sun, Sichun Du

Pages: 233 - 246

2) Biomimetic Spiking Neural Network Based on Monolayer 2-D Synapse With Short-Term Plasticity for Auditory Brainstem Processing

Author(s): Jieun Kim, Peng Zhou, Unbok Wi, Bomin Joo, Donguk Choi, Myeong-Lok Seol, Sravya Pulavarthi, Linfeng Sun, Heejun Yang, Woo Jong Yu, Jin-Woo Han, Sung-Mo Kang, Bai-Sun Kong

Pages: 247 - 258

3) UAV Coverage Path Planning of Multiple Disconnected Regions Based on Cooperative Optimization Algorithms

Author(s): Yang Lyu, Shuyue Wang, Tianmi Hu, Quan Pan

Pages: 259 - 270

4) Multimodal Emotion Fusion Mechanism and Empathetic Responses in Companion Robots

Author(s): Xiaofeng Liu, Qincheng Lv, Jie Li, Siyang Song, Angelo Cangelosi

Pages: 271 - 286

5) Location Reasoning of Target Objects Based on Human Common Sense and Robot Experiences

Author(s): Yueguang Ge, Yinghao Cai, Shuo Wang, Shaolin Zhang, Tao Lu, Haitao Wang, Junhang Wei

Pages: 287 - 302

6) Brain Compensatory Mechanisms During the Prolonged Cognitive Task: fNIRS and Eye-Tracking Study

Author(s): A. A. Badarin, V. M. Antipov, V. V. Grubov, A. V. Andreev, E. N. Pitsik, S. A. Kurkin, A. E. Hramov

Pages: 303 - 314

7) Pretrained Dynamics Learning of Numerous Heterogeneous Robots and Gen2Real Transfer

Author(s): Dengpeng Xing, Yiming Yang, Jiale Li

Pages: 315 - 327

8) Progressive-Learning-Based Assist-as-Needed Control for Ankle Rehabilitation

Author(s): Kun Qian, Zhenhong Li, Yihui Zhao, Jie Zhang, Xianwen Kong, Samit Chakrabarty, Zhiqiang Zhang, Sheng Quan Xie

Pages: 328 - 339

9) Identifying Longitudinal Intermediate Phenotypes Between Genotypes and Clinical Score via Exclusive Relationship-Induced Association Analysis in Alzheimer's Disease

Author(s): Meiling Wang, Wei Shao, Shuo Huang, Daoqiang Zhang

Pages: 340 - 351

10) A Behavioral Decision-Making Model of Learning and Memory for Mobile Robot Triggered by Curiosity

Author(s): Dongshu Wang, Qi Liu, Xulin Gao, Lei Liu

Pages: 352 - 365

11) CCANet: Cross-Modality Comprehensive Feature Aggregation Network for Indoor Scene Semantic Segmentation

Author(s): Zhang Zihao, Yang Yale, Hou Huifang, Meng Fanman, Zhang Fan, Xie Kangzhan, Zhuang Chunsheng

Pages: 366 - 378

12) Neighborhood-Curiosity-Based Exploration in Multiagent Reinforcement Learning

Author(s): Shike Yang, Ziming He, Jingchen Li, Haobin Shi, Qingbing Ji, Kao-Shing Hwang, Xianshan Li

Pages: 379 - 389

13) Edge-Centric Functional-Connectivity-Based Cofluctuation-Guided Subcortical Connectivity Network Construction

Author(s): Qinrui Ling, Aiping Liu, Taomian Mi, Piu Chan, Xun Chen

Pages: 390 - 399

14) The Methodology of Quantitative Social Intention Evaluation and Robot Gaze Behavior Control in Multiobjects Scenario

Author(s): Haoyu Zhu, Xiaorui Liu, Hang Su, Wei Wang, Jinpeng Yu

Pages: 400 - 409

15) Fatigue State Recognition System for Miners Based on a Multimodal Feature Extraction and Fusion Framework

Author(s): Hongguang Pan, Shiyu Tong, Xuqiang Wei, Bingyang Teng

Pages: 410 - 420

16) Touch Gesture Recognition-Based Physical Human–Robot Interaction for Collaborative Tasks

Author(s): Dawoon Jung, Chengyan Gu, Junmin Park, Joono Cheong

Pages: 421 - 435

17) Graph-Laplacian-Processing-Based Multimodal Localization Backend for Robots and Autonomous Systems

Author(s): Nikos Piperigkos, Christos Anagnostopoulos, Aris S. Lalos, Petros Kapsalas, Duong Van Nguyen

Pages: 436 - 453

#ai #artificial intelligence

0 notes

wonbindatascience · 6 years ago

Text

LDA (Latent Dirichlet Allocation)

1. Concept

각 문서에 어떤 주제들이 존재하는지에 대한 확률모형

문서가 생성되는 과정을 확률모형으로 모델링한 것

LDA의 inference는 실제 관찰가능한 문서 내 단어를 가지고 우리가 알고 싶은 (1)토픽의 단어분포, (2)문서의 토픽분포를 추정하는 과정

In natural language processing, latent Dirichlet allocation (LDA) is a generative statistical model that allows sets of observations to be explained by unobserved groups that explain why some parts of the data are similar. LDA is an example of a topic model.

2. Input

문서 내 단어

3. Output

각 topic을 이루고 있는 단어들이 무엇인지

한 topic 내에서 각 단어들이 차지하는 가중치는 얼만큼인지

4. Architecture

LDA는 (1)토픽의 단어분포와 (2)문서의 토픽분포의 결합으로 문서 내 단어들이 생성된다고 가정

LDA의 아키텍처, 즉 LDA가 가정하는 문서생성과정:

사용자가 지정한 hyperparameter α,β와 우리가 corpus로부터 관찰가능한 w_d,n을 제외한 모든 변수가 미지수다. (# of topics, K 도 사용자가 지정해야 하는 hyperparameter)

5. LDA 모델의 변수

ϕ_k

각 단어가 k번째 topic에 속할 ‘확률’로 이루어진 topic_k의 벡터

k = # of topics

이 벡터는 전체 corpus의 단어 개수만큼의 길이를 가졌다

즉 ϕ_k 벡터의 각 요소값은 해당 단어가 k번째 토픽에서 차지하는 ‘비중’을 나타낸다.

ϕ_k의 각 요소는 확률이므로 모든 요소의 합은 1이 된다. (아래 표 기준으로는 각 column의 합이 1)

e.g.)

Architecture를 보면 ϕ_k는 hyperparameter β에 영향을 받는 걸 알 수 있다. 이는 ϕ_k이 Dirichlet distribution 따른다는 가정을 취하기 때문

Dirichlet distribution이��? : k차원의 실수 벡터 중 벡터의 요소가 양수이며 모든 요소를 더한 값이 1인 경우에 확률값이 정의되는 연속확률분포

θ_d

d번째 문서가 각 topic에 속할 ‘확률’로 이루어진 document_d의 벡터

d = # of documents

이 벡터는 전체 topic의 개수 K 만큼의 길이를 가졌다

즉 θ_d 벡터의 각 요소값은 해당 topic이 d번째 문서에서 차지하는 ‘비중’을 나타낸다.

θ_d의 각 요소는 확률이므로 모든 요소의 합은 1이 된다. (아래 표 기준으로는 각 row의 합이 1)

e.g.)

Architecture를 보면 θ_d는 hyperparameter α에 영향을 받는 걸 알 수 있다. 이는 θ_d이 Dirichlet distribution 따른다는 가정을 취하기 때문

z_d,n

z_d,n은 d 번째 문서의 n번째 단어가 어떤 topic에 해당하는지 할당해주는 역할을 한다.

e.g.)

세번째 문서의 첫번째 단어는 ‘Topic2’일 가능성이 높겠네요. Topic1과 2가 뽑힐 확률이 각각 0.375, 0.625이거든요.

w_d,n

문서에 등장하는 단어를 할당해주는 역할

ϕ_k 와 z_d,n에 동시에 영향을 받는다.

e.g.)

직전 예시에서 z3,1은 실제로 Topic2에 할당됐다고 칩시다. 이제 ϕ_2를 봅시다. 그러면 w3,1은 Money가 될 가능성이 높겠네요. Topic2의 단어 분포 가운데 Money가 0.313으로 가장 높거든요.

6. 과정 설명

여기에서 LDA가 가정하는 문서생성과정이 합리적이라면 해당 확률과정이 우리가 갖고 있는 말뭉치를 제대로 설명할 수 있을 것. 바꿔 말해 토픽의 단어분포와 문서의 토픽분포의 결합확률이 커지도록 해야 한다는 이야기

우리는 p(z,ϕ,θ|w)를 최대로 만드는 z,ϕ,θ를 찾아야 한다. 이것이 LDA의 inference

그런데 분모에 해당하는 p(w)를 단번에 구할 수 없기 때문에 gibbs sampling 같은 기법을 사용하게 된다.

Gibbs sampling?

LDA에서는 나머지 변수는 고정시킨 채 한 변수만을 변화시키되, 불필요한 일부 변수를 샘플링에서 제외하는 collapsed gibbs sampling 기법을 쓴다. LDA에서는 p(z,ϕ,θ|w)를 구할 때 ϕ,θ를 계산에서 생략한다. 어쨌든 LDA의 깁스 샘플링 과정을 나타낸 수식은 다음과 같다. p(z_i=j|z_−i,w) 위 식의 의미는 이렇다. corpus가 주어졌기 때문에 w는 우리가 이미 알고 있는 값이다. z는 ‘각 단어가 어떤 토픽에 할당돼 있는지’를 나타내는 변수인데, z_−i는 ‘i번째 단어의 토픽 정보를 제외한 모든 단어의 토픽 정보’를 가리킨다. 이 정보를 토대로 해당 단어가 어떤 토픽에 속할지 할당하는 것이 LDA의 깁스 샘플링 과정이다. 식 전체적으로는 w와 z_−i가 주어졌을 때 문서의 i번째 단어의 토픽이 j일 확률을 뜻한다.

이렇게 z가 update 되고 나면 차례로 θ와 ϕ도 update가 된다. 이런 방식으로모든 문서, 모든 단어에 대해 깁스 샘플링을 수행하면 모든 단어마다 토픽을 할당해줄 수가 있게 된다. 이 과정에서 ϕ,θ 또한 자연스럽게 구할 수 있다. 보통 1000회~1만회 반복 수행하면 그 결과가 수렴한다고 한다.

각 단어마다의 토픽 zi는 초기엔 랜덤하게 할당

Dirichlet parameter α,β의 역할?

‘d번째 문서에 i번째로 등장하는 단어의 토픽이 j번째일 확률’을 다시 쓰면 다음과 같습니다.

이전 예시에서 첫번째 문서에 Topic2에 할당된 단어들이 하나도 없었다(n1,2=0). 원래대로라면 첫번째 문서가 Topic2와 맺고 있는 연관성 강도, 즉 A는 0이어야 할 것이다. 이렇게 되면 z_d,i가 Topic2가 될 확률 또한 0이게 됩니다. 하지만 사용자가 지정하는 hyperparameter α 존재 덕분에 A가 아예 0으로 되는 일을 막을 수 있게 된다. 일종의 smoothing 역할을 한다는 것이다. 따라서 α가 클수록 토픽들의 분포가 비슷해지고, 작을 수록 특정 토픽이 크게 나타나게 된다. 이는 β가 B에서 차지하는 역할과 대동소이하다. 디리클레 파라메터 변화에 따른 토픽 분포의 변화를 직관적으로 나타내면 다음과 같다.

7. How to find optimal K which is # of topics

Unfortunately, there is no hard science yielding the correct answer to this. 고로 최적 topic 개수 K는 여러 실험을 통해 구해야 하는 미지수이다.

Overall, when choosing the number of topics and evaluating the interpretability of your topic model, it is important to look at both qualitative and quantitative factors.

Some preferable links

https://medium.com/square-corner-blog/topic-modeling-optimizing-for-human-interpretability-48a81f6ce0ed

https://www.quora.com/What-are-good-ways-of-evaluating-the-topics-generated-by-running-LDA-on-a-corpus

Topic Coherence

https://coredottoday.github.io/2018/09/17/%EB%AA%A8%EB%8D%B8-%ED%8C%8C%EB%9D%BC%EB%AF%B8%ED%84%B0-%ED%8A%9C%EB%8B%9D/

추출된 각 주제에 있는 상위 단어들 사이의 의미론적 유사도를 측정한다. <=> 유사도가 높으면 일관성 있게 주제가 잘 추출된 모델.

따라서 의미론적 유사도를 측정하기 위해서 외부 Corpus가 필요로 된다.

Choosing a ‘k’ that marks the end of a rapid growth of topic coherence usually offers meaningful and interpretable topics.

Picking an even higher value can sometimes provide more granular sub-topics.

계산 과정:

First, the word set t is segmented into a set of pairs of word subsets S. Second, word probabilities P are computed based on a given reference corpus. Both, the set of word subsets S as well as the computed probabilities P are consumed by the confirmation measure to calculate the agreements ϕ of pairs of S. Last, those values are aggregated to a single coherence value c.

e.g.)

The above plot shows that coherence score increases with the number of topics, with a decline between 15 to 20. Now, choosing the number of topics still depends on your requirement because topic around 33 have good coherence scores but may have repeated keywords in the topic. Topic coherence gives you a good picture so that you can take better decision.

(https://rare-technologies.com/what-is-topic-coherence/) (https://datascienceplus.com/evaluation-of-topic-modeling-topic-coherence/)

Hierarchical Dirichlet Process (HDP)

HDP is an extension of LDA, designed to address the case where the number of mixture components (the number of topics in document-modeling terms) is not known a priori.

“As far as pros and cons, HDP has the advantage that the maximum number of topics can be unbounded and learned from the data rather than specified in advance. I suppose though it is more complicated to implement, and unnecessary in the case where a bounded number of topics is acceptable.”

Topics extracted via HDP could be too granular to derive generalizable meaning from for the use case at hand.

“I've gotten much better results by running a few iterations of regular LDA, manually inspecting the topics it produced. ‘Hierarchical LDA’ doesn't know what you actually intend to use the topic modeling for.”

‘Perplexity’

From a quantitative perspective, some data practitioners use perplexity or predictive likelihood to help determine the optimal number of topics and then evaluate the model fit. Perplexity is calculated by taking the log likelihood of unseen text documents given the topics defined by a topic model. A good model will have a high likelihood and resultantly low perplexity. Usually you would plot these measures over a spectrum of topics and choose the topic that best optimizes for your measure of choice. But sometimes these metrics are not correlated with human interpretability of the model, which can be impractical in a business setting.

Perplexity might not be the best measure to evaluate topic models because it doesn’t consider the context and semantic associations between words.

Another quantitative solution you can use to evaluate a topic model that has better human interpretability is called topic coherence.

LDA는 문서가 생성되는 과정을 확률모형으로 가정한다. LDA로부터 추정된 토픽 정보(z)를 활용해 계산한 각 단어의 발생 확률이 클수록 학습 corpus가 생성되는 과정을 제대로 설명하는 것이라는 얘기다. 최적 토픽 수를 찾기 위한 방법도 이 아이디어를 차용한다.

깁스 샘플링으로 구한 ϕ,θ를 활용해 전체 문서, 모든 단어의 발생 확률 p(w)를 정의하고, p(w)가 클수록 좋은 inference라는 것에 입각하여 토픽 수 K를 바꿔가면서 실험한다.

p(w) 정의

그리고 이 값을 활용한 Perplexity 지표는 다음과 같다.

p(w)는 클수록 좋은 inference이므로 exp(−log(p(w)))는 작을수록 좋다. 아래 그림처럼 토픽 수 K를 바꿔가면서 Perplexity를 구한 뒤 가장 작은 값을 내는 K를 최적의 토픽수로 삼으면 된다.

(https://www.machinelearningplus.com/nlp/topic-modeling-python-sklearn-examples/) (https://stackoverflow.com/questions/17421887/how-to-determine-the-number-of-topics-for-lda) (https://datascience.stackexchange.com/questions/128/latent-dirichlet-allocation-vs-hierarchical-dirichlet-process) (https://medium.com/square-corner-blog/topic-modeling-optimizing-for-human-interpretability-48a81f6ce0ed) (https://www.quora.com/What-are-good-ways-of-evaluating-the-topics-generated-by-running-LDA-on-a-corpus)

8. How to tune hyperparameters

num_topics

The most important tuning parameter for LDA models is num_topics

alpha, eta

affect sparsity of the topics

According to the Gensim, both defaults to 1.0/num_topics

chunksize

is the # of documents to be used in each training chunk

update_every

determines how often the model parameters should be updated

passes

is the total number of training passes

9. Tips for LDA

the quality of the final model (which is to say the coherence and usefulness of the topics) is largely dependent upon preprocessing.

(http://www.matthewjockers.net/2013/04/12/secret-recipe-for-topic-modeling-themes/)

(https://ratsgo.github.io/from%20frequency%20to%20semantics/2017/06/01/LDA/) (https://ratsgo.github.io/statistics/2017/05/28/binomial/) (https://en.wikipedia.org/wiki/Latent_Dirichlet_allocation)

0 notes

toothextract · 6 years ago

Text

14 SEO Predictions for 2019 & Beyond, as Told by Mozzers

Posted by TheMozTeam

With the new year in full swing and an already busy first quarter, our 2019 predictions for SEO in the new year are hopping onto the scene a little late — but fashionably so, we hope. From an explosion of SERP features to increased monetization to the key drivers of search this year, our SEO experts have consulted their crystal balls (read: access to mountains of data and in-depth analyses) and made their predictions. Read on for an exhaustive list of fourteen things to watch out for in search from our very own Dr. Pete, Britney Muller, Rob Bucci, Russ Jones, and Miriam Ellis!

1. Answers will drive search

People Also Ask boxes exploded in 2018, and featured snippets have expanded into both multifaceted and multi-snippet versions. Google wants to answer questions, it wants to answer them across as many devices as possible, and it will reward sites with succinct, well-structured answers. Focus on answers that naturally leave visitors wanting more and establish your brand and credibility. [Dr. Peter J. Meyers]

Further reading:

Content for Answers: The Inverted Pyramid – Whiteboard Friday

We Dipped Our Toes Into Double Featured Snippets

Desktop, Mobile, or Voice? (D) All of the Above – Whiteboard Friday

2. Voice search will continue to be utterly useless for optimization

Optimizing for voice search will still be no more than optimizing for featured snippets, and conversions from voice will remain a dark box. [Russ Jones]

Further reading:

The Influence of Voice Search on Featured Snippets

Lessons from 1,000 Voice Searches (on Google Home)

How to Discover Featured Snippet Opportunities – Whiteboard Friday

How to Target Featured Snippet Opportunities – Whiteboard Friday

3. Mobile is table stakes

This is barely a prediction. If your 2019 plan is to finally figure out mobile, you’re already too late. Almost all Google features are designed with mobile-first in mind, and the mobile-first index has expanded rapidly in the past few months. Get your mobile house (not to be confused with your mobile home) in order as soon as you can. [Dr. Peter J. Meyers]

Further reading:

How Does Mobile-First Indexing Work, and How Does It Impact SEO?

How and Why to Do a Mobile/Desktop Parity Audit

Internal Linking & Mobile First: Large Site Crawl Paths in 2018 & Beyond

How Mobile-First Indexing Disrupts the Link Graph

4. Further SERP feature intrusions in organic search

Expect Google to find more and more ways to replace organic with solutions that keep users on Google’s property. This includes interactive SERP features that replace, slowly but surely, many website offerings in the same way that live scores, weather, and flights have. [Russ Jones]

Further reading:

Zero-Result SERPs: Welcome to the Future We Should’ve Known Was Coming

What Do You Do When You Lose Organic Traffic to Google SERP Features?

Google’s Walled Garden: Are We Being Pushed Out of Our Own Digital Backyards?

5. Video will dominate niches

Featured Videos, Video Carousels, and Suggested Clips (where Google targets specific content in a video) are taking over the how-to spaces. As Google tests search appliances with screens, including Home Hub, expect video to dominate instructional and DIY niches. [Dr. Peter J. Meyers]

Further reading:

YouTube SEO: Top Factors to Invest In – Whiteboard Friday

A Step-by-Step Guide to Setting Up and Growing Your YouTube Presence

Beyond YouTube: Video Hosting, Marketing, and Monetization Platforms, Compared

6. SERPs will become more interactive

We’ve seen the start of interactive SERPs with People Also Ask Boxes. Depending on which question you expand, two to three new questions will generate below that directly pertain to your expanded question. This real-time engagement keeps people on the SERP longer and helps Google better understand what a user is seeking. [Britney Muller]

Further reading:

Infinite “People Also Ask” Boxes: Research and SEO Opportunities

7. Local SEO: Google will continue getting up in your business — literally

Google will continue asking more and more intimate questions about your business to your customers. Does this business have gender-neutral bathrooms? Is this business accessible? What is the atmosphere like? How clean is it? What kind of lighting do they have? And so on. If Google can acquire accurate, real-world information about your business (your percentage of repeat customers via geocaching, price via transaction history, etc.) they can rely less heavily on website signals and provide more accurate results to searchers. [Britney Muller]

Further reading:

The Ultimate Cheat Sheet for Taking Full Control of Your Google Knowledge Panels

How to Optimize Your Google My Business Listing

8. Business proximity-to-searcher will remain a top local ranking factor

In Moz’s recent State of Local SEO report, the majority of respondents agreed that Google’s focus on the proximity of a searcher to local businesses frequently emphasizes distance over quality in the local SERPs. I predict that we’ll continue to see this heavily weighting the results in 2019. On the one hand, hyper-localized results can be positive, as they allow a diversity of businesses to shine for a given search. On the other hand, with the exception of urgent situations, most people would prefer to see best options rather than just closest ones. [Miriam Ellis]

Further reading:

The State of Local SEO Industry Report

Local Search Ranking Factors 2018: Local Today, Key Takeaways, and the Future – Whiteboard Friday

9. Local SEO: Google is going to increase monetization

Look to see more of the local and maps space monetized uniquely by Google both through Adwords and potentially new lead-gen models. This space will become more and more competitive. [Russ Jones]

Further reading:

New Research: 35% of Competitive Local Keywords Have Local Pack Ads

What Do SEOs Do When Google Removes Organic Search Traffic? – Whiteboard Friday

10. Monetization tests for voice

Google and Amazon have been moving towards voice-supported displays in hopes of better monetizing voice. It will be interesting to see their efforts to get displays in homes and how they integrate the display advertising. Bold prediction: Amazon will provide sleep-mode display ads similar to how Kindle currently displays them today. [Britney Muller]

11. Marketers will place a greater focus on the SERPs

I expect we’ll see a greater focus on the analysis of SERPs as Google does more to give people answers without them having to leave the search results. We’re seeing more and more vertical search engines like Google Jobs, Google Flights, Google Hotels, Google Shopping. We’re also seeing more in-depth content make it onto the SERP than ever in the form of featured snippets, People Also Ask boxes, and more. With these new developments, marketers are increasingly going to want to report on their general brand visibility within the SERPs, not just their website ranking. It’s going to be more important than ever for people to be measuring all the elements within a SERP, not just their own ranking. [Rob Bucci]

Further reading:

Mapping the Overlap of SERP Feature Suggestions

Make Sense of Your Data with These Essential Keyword Segments

12. Targeting topics will be more productive than targeting queries

2019 is going to be another year in which we see the emphasis on individual search queries start to decline, as people focus more on clusters of queries around topics. People Also Ask queries have made the importance of topics much more obvious to the SEO industry. With PAAs, Google is clearly illustrating that they think about searcher experience in terms of a searcher’s satisfaction across an entire topic, not just a specific search query. With this in mind, we can expect SEOs to more and more want to see their search queries clustered into topics so they can measure their visibility and the competitive landscape across these clusters. [Rob Bucci]

Further reading:

Build a Search Intent Dashboard to Unlock Better Opportunities

It’s Time to Stop Doing On-Page SEO Like It’s 2012

Using Related Topics and Semantically Connected Keywords in Your SEO – Whiteboard Friday

How to Feed a Hummingbird: Improve Your On-Page SEO with Related Topics in Moz Pro

13. Linked unstructured citations will receive increasing focus

I recently conducted a small study in which there was a 75% correlation between organic and local pack rank. Linked unstructured citations (the mention of partial or complete business information + a link on any type of relevant website) are a means of improving organic rankings which underpin local rankings. They can also serve as a non-Google dependent means of driving traffic and leads. Anything you’re not having to pay Google for will become increasingly precious. Structured citations on key local business listing platforms will remain table stakes, but competitive local businesses will need to focus on unstructured data to move the needle. [Miriam Ellis]

Further reading:

The Guide to Building Linked Unstructured Citations for Local SEO

Why Local Businesses Will Need Websites More than Ever in 2019

14. Reviews will remain a competitive difference-maker

A Google rep recently stated that about one-third of local searches are made with the intent of reading reviews. This is huge. Local businesses that acquire and maintain a good and interactive reputation on the web will have a critical advantage over brands that ignore reviews as fundamental to customer service. Competitive local businesses will earn, monitor, respond to, and analyze the sentiment of their review corpus. [Miriam Ellis]

Further reading:

Time to Act: Review Responses Just Evolved from “Extra” to “Expected”

How to Respond to the 5 Types of Google Reviews

Location Data + Reviews: The 1–2 Punch of Local SEO

See more local SEO predictions for 2019 by Miriam in our Q&A!

We’ve heard from Mozzers, and now we want to hear from you. What have you seen so far in 2019 that’s got your SEO Spidey senses tingling? What trends are you capitalizing on and planning for? Let us know in the comments below (and brag to friends and colleagues when your prediction comes true in the next 6–10 months).

Sign up for The Moz Top 10, a semimonthly mailer updating you on the top ten hottest pieces of SEO news, tips, and rad links uncovered by the Moz team. Think of it as your exclusive digest of stuff you don’t have time to hunt down but want to read!

from https://dentistry01.wordpress.com/2019/02/27/14-seo-predictions-for-2019-beyond-as-told-by-mozzers/

0 notes