#ocr answers

[ask-ocean-ocr]

Favorite Speech & Debate event?

...If we had the funds for a team. We're working on it.

Sincerely, Ocean OCR

[finnlongman]

i hope this doesn't sound patronising if you've already tried that route, but in case you haven't: if this is a text that has been given to you as is and been produced by a general OCR, it might be worth looking into whether ppl have already trained language-specific OCR models for your field of study and using that to transcribe the scanned text again. there's a lot of transcription solutions/software and different fields prefer different ones, and idk the standard for Celtic studies personally, but a site i use often (transkribus) has 2 Irish models whose related projects you might be able to use as a starting point for research at least. best of luck to you either way!

So there are several factors at work with the OCR problems with this text specifically.

The PDF of the text is from Archive. The library copy it was scanned from has various pencil markings and annotations that are interfering with the printed text -- it's not a clean scan. It's also not super high definition, so letters like "h" sometimes get misread as "li", even though they're totally readable to human eyes.

The edition uses frequent italics and brackets to show where abbrevations in the manuscript has been expanded. Individual italicised letters confuse the OCR, as do random square brackets in the middle of words.

It also has a lot of superscript numbers corresponding to manuscirpt variants in the footnotes. Sometimes these are in the middle of a word. This also confuses most OCR systems, even if it can tell that the footnotes are separate from the main text.

The language of the text is late Middle / Early Modern Irish, from two different manuscripts that have their own unique spelling quirks (for example, one of them loves to spell Cú Chulainn's name "Cú Cholain", which is a vibe).

In order to run the text through a more sophisticated OCR system that was equipped to cope with a) annotations, b) weird formatting and punctuation, c) incredibly frequent footnotes (variants), and d) non-standardised spelling (which throws off many language models), I would probably still need to have a reliable, clear, and high-definition scan of the text. Which would require re-digitising it from scratch.

So, the quickest and easiest way to get a version of the text that I personally can use is to sit here and type up 20,000 words into a document. This is 2-4 days' work, depending on how focused I am, and gives me the chance to go through the text in detail and spot things I might miss otherwise, so it's probably a whole lot less effort for more benefit than trying to adapt an entire language model that could read this terrible PDF. Especially as I have no experience of using these programmes so would have a steep learning curve.

Now, somebody absolutely should do that, so we could get proper searchable editions of more things. But honestly, if using transcription tools for medieval/early modern Irish I think there are higher priorities than things already available in printed form, so I doubt it's at the top of anyone's to-do list!

[ace-and-the-rpg-horrors]

First it was Shin now it's Rui. Which of the accusations are worse is the question.

Probably Shin.

DEFINITELY SHIN

what is it with me getting compared to tech boys my Computer Science GCSE caused me one of my worst mental breakdowns

[maplesynth]

your program is really cool and ofc would be more accurate than existing OCR bc of how u pattern match but i actually was able to use google lens to get the text from the image with an exact match. it didn't have the problems that u detailed in your post like uppercase vs lowercase or spelling correction. so existing tech is smarter than u think :)

i don't think any of you know what "exact" means

[astralscrivener]

🎵+ your favourite oc ship? 👀

oh shit. oh shit

i can't necessarily choose a favorite because that's like picking a favorite child but declan and mylor have been on the brain

the top song i think about when i think of them is we have it all by pim stones. i first heard it in the car with beatrice on a cloudy gray summer afternoon and i have not been normal about it since

i just need you to listen to this song and imagine two people who love each other more than anything else and who just want a quiet soft life together but circumstance keeps ripping them apart . and then you try being normal after that

SEND ME A ♫ + A CHARACTER’S NAME AND I WILL RESPOND WITH A SONG THAT REMINDS ME OF THEM. SEND A ♫ + A SHIP AND I WILL DO THE SAME.

[anistarrose]

I want to make my posts more accessible, but can't write IDs myself: a guide

[Plain text: "I want to make my posts more accessible, but can't write IDs myself: a guide." End plain text.]

While every image posted online should be accessible in an ideal world, we all know it 1) takes time to learn how to write image descriptions, and 2) is easy to run out of spoons with which to write IDs. And this says nothing of disabilities that make writing them more challenging, if not impossible — especially if you're a person who benefits from IDs yourself.

There are resources for learning how to write them (and if you already know the basics, I'd like to highlight this good advice for avoiding burnout) — but for anyone who cannot write IDs on their original posts at any current or future moment, for any reason, then there are two good options for posting on Tumblr.

1. Crowdsource IDs through the People's Accessibility Discord

[Plain text: "1. Crowdsource IDs through the People's Accessibility Discord". End plain text.]

The People's Accessibility Discord is a community that volunteers description-writing (and transcript-writing, translation, etc) for people who can't do so themselves, or feel overwhelmed trying to do so. Invite link here (please let me know if the link breaks!)

The way it works is simple: if you're planning to make an original post — posting art, for example — and don't know how to describe it, you can share the image there first with a request for a description, and someone will likely be able to volunteer one.

The clear upside here (other than being able to get multiple people's input, which is also nice) is that you can do this before making the Tumblr post. By having the description to include in your post from the start, you can guarantee that no inaccessible version of the post will be circulated.

You can also get opinions on whether a post needs to be tagged for flashing or eyestrain — just be able to spoiler tag the image or gif you're posting, if you think it might be a concern. (Also, refer here for info on how to word those tags.)

The server is very chill and focused on helping/answering questions, but if social anxiety is too much of a barrier to joining, or you can't use Discord for whatever reason, then you can instead do the following:

2. Ask for help on Tumblr, and update the post afterwards

[Plain text: "Ask for help on Tumblr, and update the post afterwards". End description.]

Myself and a lot of other people who describe posts on this site are extra happy to provide a description if OP asks for help with one! This does leave the post inaccessible at first, so to minimize the drawbacks, the best procedure for posting an image you can't fully describe would be as follows:

Create the tumblr post with the most bare-bones description you can manage, no matter how simple (something like "ID: fanart of X character from Y. End ID" or "ID: a watercolor painting. End ID," or literally whatever you can manage)

Use a tool like Google Lens or OCR to extract text if applicable and if you have the energy, even if the text isn't a full image description (ideally also double-check the transcriptions, because they're not always perfect)

Write in the body of the post that you'd appreciate a more detailed description in the notes!

Tag the post as "undescribed" and/or "no id" only if you feel your current, bare-bones description is missing out on a lot of important context

When you post it and someone provides an ID, edit the ID into the original post (don't use read mores, italics, or small text)

Remove the undescribed tag, if applicable. If you're posting original art, you can even replace it with a tag like "accessible art" for visibility!

And congrats! You now have a described post that more people will be able to appreciate, and you should certainly feel free to self-reblog to give a boost to the new version!

[notbecauseofvictories]

hello! longtime follower and current 1L - am starting the rounds of the law firms and there's always the "so what practice area are you interested in?" bit. i was curious what that is for you + what you do/don't like about it? iirc, you mentioned something abt being in-house, healthcare related, regulatory side of things? i'm interested in regulatory stuff but not sure if that takes the shape of litigation or an explicitly regulatory-focused practice. found your blog back in the star wars days, then got into silmarillion and the silt verses after seeing your posts abt it haha - hope you're doing well and staying warm!

Congrats! 1L summer is a fun, anxiety-ridden time that mostly involves telling people over and over, who you are and what you're interested in---which can be tricky if you're not sure what you're interested in.

Personally, I went in knowing that I liked healthcare (the field I worked in prior to law school) and that wanted to stay healthcare-adjacent.....but not much else.

I learned I was not destined for litigation pretty much the first time my Legal Writing professor handed back our appellate briefs. (Mine did not have the grade I wanted at the top.) This was compounded by our final project, where we presented in front of real live attorneys and I was a nervous, sweaty wreck. After that, I decided that becoming Atticus Finch was not in my future.

But there are still lots of other kinds of law to practice! I live in the healthcare regulatory space---and I work for a pretty under-resourced company, which means I have lots of contact with other areas like R&D, clinical research, data privacy, marketing and adtech, direct patient care, healthcare compliance, and negotiating between various international laws. Not to mention my scope is always expanding, which is...challenging, but I'm also the kind of person who enjoys spending a weekend reading about Brazilian law.

(One of my guiding stars through the whole law school/job search process was "I don't want to be bored." I am never, ever bored.)

And this wasn't even my first stop! When I was in law school, I spent my semesters interning/clerking at firms, consulting boutiques and government agencies; policy-focused clinics and hospitals and giant corporate behemoths. I've said before that observing all these different settings was valuable, that it gave me a better understanding of myself, how I work, and the kind of work I was looking for. While I won't ever claim that every experience I had was amazing (it was not) it did give me the opportunity to explore, in a way that most adult professionals simply can't.

I mean---look. If you're committed to the brass ring of OCR and a high-profile law firm, then you might have to make this decision now. (Or at least come up with a good answer for interviewers...) But I highly encourage you and everyone choosing that path to keep the other doors open, just a crack. There are interesting things that sneak through when you aren't looking.

[shallowseeker]

Nonny, you don't have to use AI to grab the episode snippets or transcripts. They're searchable thanks to the wonderful @butch--dean's search engine here.

And the scripts are here thanks to @spnscripthunt-inactive and the transcripts are available here.

But by far the search engine is the most helpful for cross-referencing themes for whatever you'd like to write about (or answer). For example, I hadn't seen Metatron's dog episode in awhile, but just typing "Toto" into the search engine was enough to pull up the exact scene.

The scripts can be a little harder to search unless you've run OCR on it to pull out the text to make it searchable. (I'm more of a script-girlie reader myself, so I tend to just... remember where things are.)

And although I haven't watched all of them yet, I’m finding that the SPN DVD box set is an absolute goldmine of commentary in the extras—on themes, set design, all of it. From there, you can sort of just… string the pearls together or criss-cross the red thread on your crazy investigation board—however you want to emphasize The Things!

But yeah, I'd say just shoot from the hip, and try not to worry about it. You don't need AI to ramble on, and it doesn't even matter if you're wrong, change your mind, or repeat yourself so long as you're having fun. Two of my fave books I've read that I think about a lot with respect to meta are Shot by Shot (volumes I & II). It just encourages you to look at things a bit differently.

And sometimes, I literally just see things that aren't there *cough* mouse trap ball YOU ARE AN EYE BECAUSE I SAID SO.

[girlfriendsofthegalaxy]

tuesday again 9/24/2024

you might be wondering “is my dear friend tumblr user girlfriendsofthegalaxy still unemployed?” the answer is yes. take this cat off my hands please i don’t think he’s causing the unemployment but he certainly isn’t helping

listening

via Wendy @dying-suffering-french-stalkers, Huoy Meas' ប្រគល់ក្ដីស្នេហ៍មកខ្ញុំវិញ. figuring out what this incredibly zippy Cambodian rock song is named and what it's about was really difficult bc spotify is a bane upon this earth and won't let you fucking copy-paste and OCR was not working on the Khmer script. i ended up listening to the first couple seconds of each of her songs on apple music, and finally figured out this roughly translates to Give Me Back My Love and is about begging a fuckboy for closure.

via the spotify discover weekly, Night Club's Pretty Girls Do Ugly Things. all Night Club's songs sound the same so if you like one, great news! i had this song on for a full gregorian hour bc, i am only a tiny bit ashamed to say, i was storyboarding a The Man With No Name fancam to this. i think it would go pretty hard.

Smoke you like a cigarette Choke you like a lariat Fatalistic tourniquet Do you want more?

-

reading

thank you mackintosh.

i did not Adore any of these comics from the library. i sort of enjoyed Night of the Ghoul, a one-volume TPB by Scott Snyder and Francesco Francavilla. i think ive blogged about this before but every once in a while i'll get a bee in my bonnet to read some horror comics even though i am a giant baby about horror movies.

Night of the Ghoul is about how you can't save your dad from PTSD but also about a lost horror film and also about the extremely dad behavior of tracking down every scrap of info about an auteur. there's also a monster.

the subtle art changes from present day to the remains of the film to the non-film flashbacks are well done, imo. the cover screams mignola but the inside pages are really fun pulp nonsense. i love a piece of genre writing that rolls around and delights in being a piece of genre writing.

im doing my level best not to get sucked into tiktok but i DO love watching this lady revive antique nail polish and look for dupes for shades from like the 20s. she found an almost exact dupe for a shade produced during wwii which is crazy insane to me!!!

-

watching

The Asphalt Jungle (1950, dir. Huston), it's a very painterly heist noir. i even like Sterling Hayden in one of the more prominent roles, even though i think he generally has the appeal of undercooked dough.

much like Fritz Lang's M, it presents the criminal element of the city as its own class with its own reputation and reference systems. it got in some trouble with the censors for having a VERY clearly laid out heist plan and execution. it's also got the babiest Marilyn Monroe in one of her earliest roles

this was such a gloriously messy movie. everyone is such a fucking mess. this woman only known as Doll is heartbreakingly, head over heels in love with Sterling Hayden's character. she's a little flighty and bumbling and silly, but determined! they're constantly orbiting the gravitational weight of her desire for this man and desire for a real life with this man. and that's just one subplot! she has maybe five minutes total screentime! she should have gotten a supporting actress oscar!!! everyone acted their fucking hearts out and it was so much fun to watch!

-

playing

monument valley is in the netflix games library this month (i don't actually know what their liscencing agreements would even look like, they and the studios they worked with were very tightlipped about that when they were rolling this out three years ago) but i assume it's going to be on the service for a while. i have never played this game, which makes me feel a little bit like a bad gamer. you can tell it's ten years old from some of the color and texture choices, but WOW did literally everyone take inspiration from this game.

this is the platonic ideal of a phone game. i get why everyone went insane about it and there was a brief boom of geometry-based puzzle mobile games. it is MUCH much harder now to get people to pay money to play a game that has a planned endpoint and planned number of levels, so netflix is a good home for it.

i was often frustrated but always delighted. the level below involves making something happening that made me genuinely gasp out loud in glee. well worth the annoyance of downloading the netflix app and scrolling through the poorly labeled and poorly sorted carousel of games.

great retrospective, a bit about how you need to have tiny teams go off and just kind of fuck around and bring weird stuff back, and a lot about how they actually designed the levels

The end result had a pixel-perfect axonometric aesthetic that not only went hard on its references to Dutch master artist and printmaker Maurits Cornelis Escher, but also dug deep into classic video game design, going right back to early arcade machines and 8-bit titles. Each of the ten levels is like a piece of fine furniture, built with invisible dovetail joints and inlaid with marquetry, stuffed with secret compartments and little design flourishes. Gray cites the world of theatre and stage design, as well as graphics, as important keystones in the way the levels were constructed. ‘Ken would always talk about flower arranging, and how you frame a silhouette of a level on the screen,’ he says.

-

making

update on the Phantom Menace fabric: pinked the raw edges and threw it in the laundry again with a very large quantity of vinegar. 50% poly was too high for it to really do anything, which is interesting. it didn’t lessen the seam edge effects either, which is a little annoying bc the seams were so gigantic and that’s a good chunk of fabric to lose. i am going to buy a camp shirt pattern at some point when i have money again but for now it goes in The Box

also! thrifted a pack of o-rings for jars for a dollar and finally put my grains etc in my pretty jars. they’re going to live in the pantry but today they live out on the countertop

[communist-ojou-sama]

Since nothing I'm doing is all that special and I figure some others might want to help out with it, I decided to make a shared folder: In it is the full text of "I've Found My Answers" by the martyr Bassel al-Araj, along with the book divided into pieces short enough for Deepseek's length limits (about 30 pages each of the original book), and a Word document that currently contains the raw OCR of the first 70 pages, with the first few having been proofread. If you can read Arabic and would like to help me proofread, help would be much appreciated! I think it'd be nice if after making a nice epub for this book (what I'm aiming to eventually do) we could do the same for the Thorn and the Carnation, and for other books that currently only exist as somewhat-blurry pdfs! by default this link only gives reading and commenting privileges because I don't want some zionist to come and grief my work. If you want editing privileges DM me.

[ask-ocean-ocr]

i think your hair colour is nice!! don't listen to the haters, live laugh love ginger ✨️

- 🦭 anon

Some of you all could take a page out of this book! Spread positivity.

Sincerely, Ocean OCR

[four-loose-screws]

Blog Update 2025.04.22

Hey ya'll! I've gotten an influx of translation asks so far this year, which have been a lot of fun to dive into! The only downside is the time it takes for me to research and write up an answer to each ask when I have a whole bunch at once, so I thought I'd take a moment to briefly mention some of the posts I am going to be slowly but surely working on for a while. This isn't the entire list on my ask backlog, just the big stuff.

Tharja personality post - I don't think she's different enough in localization to merit a full analysis like I usually do for my personality posts, but once I read and think through all of her supports, I will for sure be sharing my thoughts.

Soleil personality post - I forgot I never did an analysis for Soleil! Well, better late than never! She's one of the top characters who needs a full Japanese vs. localization deep dive for sure.

Folder with raw FE novel scans in Japanese - They'll be far from perfect since I don't edit my novel scans, I need them just good enough to run relatively accurately through an OCR to help with my translations, but I'll still share my novel scans for those interested.

[turtburglers]

i truly wish i could see the appeal in a "universal software" the way other people do. i just can't understand canva. i need to put together a program at the library explaining how to use canva. but i still can't answer the question of why you would use canva for myself. it's a photo editor, but it doesn't do any real photo editing. you can make pdf documents, but it's OCR is so terrible that you can't really edit something once it's made. or anything you've scanned. it's a slideshow designer. but... who seriously came to canva looking to make slideshows? its a video editor???? this program that most people use to make basic ass flyers is also a video editor???? and we aren't even talking about how frustrating it to use as a freemium platform. oh you don't have to pay for it UNLESS YOU WANT ANY OF THE PARTS YOU'D ACTUALLY COME TO CANVA FOR. the only thing it's actually good for is providing free to use stock photo, clip art, and templates inside of the program, but if you want anything that looks halfway decent you have to sign up for their ridiculous subscription.

i'm just. if you actually care about what you do 90% of the things canva does it's insufficient for. it's a bad photo editor, it's a bad document editor, it's a bad video editor, it's a bad presentation creator, and we haven't even gotten to how bad the AI tools are. i hate AI so much but Canva's AI isn't even worth attempting to use. you ask for a picture of gold leaf and it serves you gold soup. whatever gold soup is. but also. i guess it technically does all of those things. so if you're not actually interested in anything you make being good then. there you go.

[anpesir]

Master Julius Caesar Act 1: In-Depth Q&A Notes for ICSE, CBSE, IGCSE, GCSE, IB, AP & 30+ Global Curricula.

Dive into expertly crafted question-and-answer notes for Act 1 of Shakespeare’s Julius Caesar, aligned with ICSE, CBSE, Maharashtra & West Bengal State Boards, Cambridge IGCSE, GCSE (AQA, Edexcel, OCR), IB Diploma, AP English Literature, and over 30 international systems from Australia and Canada to UAE, Kenya, Nigeria and the European Baccalaureate. Explore Caesar’s triumphal return, the Soothsayer’s warning, Cassius’ envy, Brutus’ moral reflection, Portents and omens, and conspirators’ early plotting through detailed answers that highlight themes, character dynamics, language features and dramatic tension. Ideal for exam revision, classroom discussion or independent study to build analytical skills, deepen understanding and boost exam confidence.

Click the link below to increase your grade 👇:

[cuprohastes]

Trying to make AI Bros useful.

So I have an R1 from Rabbit.

If you don't know what an R1 is: It's a very nicely designed palm sized block of glossy orange plastic with a small indoor screen, camera, a scroll wheel and a button. It costs $200 and is a really basic Android phone where all the android phone parts are hidden. It uses Perplexity/ChatGTP to answer questions. There's also a LAM - Large Action Model, which doesn't exist, but is supposed to.

The idea is that the R1 will answer questions, find information and be able to take action on your behalf. Many people have done clickbait Youtube videos about how this is a scam.

It's only partially a scam.

The part where you activate the camera, which is unable to see anything until activated, by physically not allowing it to point out the front or back until it's rolled around, and ask it for information about what it sees, is fine.

That works.

The rest...? Bullet point time.

The Bad

The LAM is just a bunch of scripts for using four different websites, all of which are the ones you'd assume a Tech-Bro thinks are important. Uber, Door dash, Yelp. No other websites are supported, and every time one of these services changes their layout the "LAM" goes down for "maintenance".

The GPS is just garbage.

You have to tether it to a Hotspot or SIM that has voice and data. Data alone won't work. The R1 cannot make phone calls.

No app integration: You can't send a message, check your mail etc.

Screen is hilariously shit outside or in bright light

UI prioritises icon of a cute rabbit vs the results.

So WTF?

✨Use Case Scenario!!✨

OK: Imagine you're partially sighted. You pull this thing out of your pocket: It has one easy to find physical button that does nearly everything.

You turn it on with a click.

You double click to go into vision mode.

You hold the button down and say "What am I looking at?"

The R1 then uses a pretty good speech model to describe the scene!

You hold the button down and ask further questions about the image:

"How much is this?" "What's the weight?" "Is there a special offer?"

Or a bill:

"Who is this from?" "Wow much is it for?" "When do I have to pay by?"

Common Arguments

"This could be an app!"

Yes it could be an app and if you have absolutely fine vision and motor control and want to pull a phone out, unlock it, find the icon, then navigate around a UI, go for it. Perplexity is right there, you can use it now. It's fantastic at checking websites (Even ones it's explicitly banned from using) and summarising them.

But again: If you're having to run things by touch, a single button is way easier.

"It wasn't designed as an assistive product so..."

No. And a genuine assistive product is most likely an android phone of the cheapest off the shelf hardware, an OCR app, and a £1,200 price tag. Comes with the cutting edge synthesised voice tech of 1995. The rabbit is $200/£160, and is voice controlled and still somehow has fewer buttons than a device that's supposedly designed for people who can't see buttons.

"Ah but does it really..."

The hype is massive. But I physically have one. One moment. I am now holding it.

Did I run down to the supermarket and test it? Yes. Did it work? Yes. Was it minorly a pain in the ass? Yes.

"You ain't visumably Impaired!!"

Ah, no. Yes. Depends on how nitpicky you want to be.

However, I do have someone here whose vision is basically Noped out, who will be helping me test and either she hates it or I'll have to spend a few hours planning a heist to pry it out of her hands, so I can charge the battery.

[shanaocsmuseum]

๋࣭ ⭑⚝𝙎𝙪𝙥 𝙨𝙪𝙥. 𝙒𝙚𝙡𝙘𝙤𝙢𝙚 𝙩𝙤 𝙢𝙮 𝙜𝙤𝙤𝙛𝙮 𝙖𝙨𝙨 𝙗𝙡𝙤𝙜 𝙄 𝙜𝙪𝙚𝙨𝙨?✮⋆˙๋࣭ ⭑

╰┈➤𝙒𝙚𝙡𝙡, 𝙬𝙝𝙖𝙩 𝙙𝙤 𝙬𝙚 𝙝𝙖𝙫𝙚 𝙤𝙣 𝙩𝙝𝙞𝙨 𝙗𝙡𝙤𝙜, 𝙩𝙝𝙚𝙣?

ᯓ★Headcanons, moodboards, arts, fics and scenarios

ᯓ★Miscellaneous shit and posts

ᯓ★Yaps about my ocs

About the Ocs:

ᯓ★ 𝑱𝒆𝒎í𝒎𝒂 Á𝒎𝒃𝒂𝒓 𝒅𝒆𝒍 𝑴𝒂𝒓 𝑯𝒆𝒓𝒏á𝒏𝒅𝒆𝒛⭑.ᐟ - Call of Duty Ghosts OC tag: #cod ghosts oc: jemíma hernández >Looking For America, ongoing <TW!: strong language, angst, depressive thoughts, canon reality scene, mentions of sensitive topics

ᯓ★ Jackie Walker⭑.ᐟ | Call of Duty Ghosts

ᯓ★ Linda Reinhardt-Merrick⭑.ᐟ | Call of Duty Ghosts

ᯓ★ Nathaniel Sean Ocre⭑.ᐟ | Call of Duty Ghosts

ᯓ★Tags:

Art tag->#shani draws Asks and Q&A->#shani replies Rambles, yap and rest of the shi->#shani talkies Fanfics n' writings->#shani's literature corner Headcanons->#shani's thinking cap Creating scenarios or interactions->#shani's acting set Reblogging->#shani reblogs

Character interaction blog-> lt-hernandez-answers Dive into Teen!AU blog-> Nothing but the Beat! Explore the literature corner/Cod Ghosts related content-> Shanawrites