#macroanalysis | Explore Tumblr posts and blogs

am--f · 2 years ago

Text

The cloudy humanities

Matthew Jockers grounds his argument for a data-scientific approach to literary studies in a familiar observation from twentieth-century cultural theory: it is not possible, using traditional literary-critical means, to exhaustively process the massive corpus of literary history. The humanities have been restricted to “microanalysis”—the analysis of a “random” selection of things from a single “representative” work or a small set of “representative” works (Jockers, 8)—perhaps in part due to technical constraints. Even the most ambitious formalisms and structuralisms could not classify and graph empirical cultural data at scale; in fact, they often explicitly rejected empirical approaches for this very reason, opting instead to develop elegant and minimal codes by which any given text (or “message,” to extend the information-theoretical analogy) could be analyzed (by human analysts). But by 2013, both text digitization (the accumulation of literary-textual data) and digital algorithms (tools for automatically processing this data into human-interpretable information) had become accessible enough for scholars like Jockers to mount a convincing macroanalytic challenge that would quantify the different features of global literary production, enabling researchers to answer literary-historical questions in ways that were previously impossible.

With the neoclassical synthesis in economics as a model (a questionable choice!), Jockers suggests that literary macroanalysis is the successor to the literary theory initiated by visionary twentieth-century critics like Alexander and Aleksey Veselovsky, Victor Shklovsky, and Vladimir Propp. The semiotic squares of A.-J. Greimas and the tables of Levi-Strauss give way to Jockers’s graphs and word clouds, derived from machine readings. In a chapter on theme, Jockers demonstrates the importance of topic modeling as a prototypical macroanalytic operation. Through natural language processing software like the open-source Machine Learning for Language Toolkit (MALLET), texts can be discretized into frames or “chunks”—a kind of quantization of the signifying space that Roland Barthes called “lexia”—which are then scanned for clusters of words (collocations of collocations) by an algorithm trained to recognize them; the findings are correlated across a large corpus or corpus subset to demonstrate trends that can be associated with national and historical contexts, a way of quantifying the kinds of claims that critics, especially historicist ones, tend to make about literary objects (that some text is or is not part of a pattern of similar texts conditioned by the circumstances of their production, that some text was or was not influenced by a historically proximate mode or manner of writing, etc.). Jockers’s method, as he seems to recognize, but as many others have also pointed out repeatedly, hinges on the question of whether this “distant reading,” as Franco Moretti named it, yields anything novel or “unconscious” that couldn’t have been adequately obtained otherwise. But rather than rehearse the debate between evangelists of qualitative and quantitative methods, I would like to briefly address features of the field that Macroanalysis tries to bolster, and to assess their bearing on some tentative, qualitative suppositions about digital culture. I am interested in the cloudy aesthetics of data visualization, which might be linked to a media archaeology of information display, as well as to the tendency to emphasize theme over narrative structures or story templates. Following from this, I want to begin pose the question of whether it is possible or useful to make historical claims about shifts away from narrativity itself in the broader cultural context.

The word cloud is Jockers’s preferred form. Nebulous, fuzzy, the word cloud (or tag cloud) gives an impression of a probability field by arranging and scanning its terms in two-dimensional space. The size and sometimes color of a term is proportional to its frequency. Position may be random, or not: higher-frequency terms may form the center of the cloud, or the terms may be distributed along axes. Like Stephan Mallarmé’s Un coup de dés jamais n��abolira le hasard (1897), the word cloud deals with the noise in the channels of literary work, the many throws of dice that make up literary output, and the recognition of patterns in these performances of language.* One account credits its invention, at least in the scientific context, to social psychologist Stanley Milgram, who, in a 1967 study, created a word cloud about Paris from a set of survey responses (Viégas and Wattenberg 2008). Such clouds saturated the Internet in the early years of Web 2.0, becoming prevalent on blogging platforms (where they would graph a user’s posts), and were notably a major feature of the photo hosting site Flickr. The form has since fallen out of favor in user experience design, but it remains prominent in popular visual culture. It often works, interestingly, in a decorative mode, taking on architectural dimensions. I have, for instance, encountered countless interiors or storefronts adorned with more or less nonsensical word clouds that attempt to somehow index the experience of the product or the space. This usually produces a comic effect: it is absurd to have my enjoyment of a beverage from a juice bar mediated by a cloud of words like “nature,” “relaxation,” “kale delight,” “freshness,” “serenity,” “wellness,” and “NYC” in various sizes and arranged at various angles on the wall, probably because the descriptive efforts of these (possibly fabricated) n-grams are so incongruous with what is actually happening. In this case of acute semiosis, there is something like an overemphasis on or overexposure of the branding process, the “theming” of the activity of consumption laid bare. The migration of digital infographics onto fast-casual wallpaper blurs the boundary between the commodity and the consumer analytics used in the process of designing it. There is a wonderful cheapness or laziness to this kind of marketing: it is as though the keywords used in the marketing brainstorm session simply ended up composing the marketing itself.

It sometimes seems to me that the production of word clouds is analogous to the way that consumers increasingly are trained to process everyday cultural objects: by scanning (not reading) a deluge of media for trends, moods, and vibes. In this sense, the aesthetics of literary analytics—the way in which the scientific program of someone like Jockers makes choices about presentation that are not exempt from the system of fashion—corresponds to the focus on theme and topic modeling in our contemporary media environments. There is a cloudiness to the cultural logic of digital networks, in which algorithmic immersion (on social media) yields vagueness and diffusion. We consume alongside our recommendation algorithms; we desire in collaboration with statistical models of taste (on Spotify, TikTok, YouTube). And when it comes to narrative entertainment, it seems that we are (as consumers) often less interested in following stories than enjoying characters, affects, and styles. The media archaeologist Wolfgang Ernst, drawing on the etymological relation between erzählen and zählen, recounting and counting, has suggested that culture today emphasizes counting over recounting in precisely this way. It might be too much to assert the “end” of narrative, just as it might be too much for certain modes of scientism or pragmatism to assert the “end” of theory, but it does seem possible to make the case that specifically theme-obsessed digital humanities research is the course of research appropriate to a broader cultural ambience of near-universally distant, algorithmic, computational reading. The digital humanists of the future may of course be able to demonstrate this correspondence with more quantified certainty.

*Quentin Meillassoux has produced an interesting numerological interpretation of the Coup de dés, taking Mallarmé’s text quite literally as a poem about number and measure; here, the Symbolist work and the n-gram cloud are both performances of counting (Meillassoux 2012).

References

Barthes, Roland. (1973) 1974. S/Z, translated by Richard Miller. New York: Hill and Wang. Ernst, Wolfgang. 2013. Digital Memory and the Archive. Minneapolis: University of Minnesota Press. Jockers, Matthew. 2013. Macroanalysis. Champaign, IL: University of Illinois Press. Meillassoux, Quentin. 2011. The Number and the Siren. Falmouth, UK: Urbanomic. Viégas, Fernanda B., and Martin Wattenberg. 2008. “Tag Clouds and the Case for Vernacular Visualization.” ACM (Association for Computing Machinery) Interactions: 49–52.

#writing #narrative theory

1 note · View note

omcik-blog · 8 years ago

Text

New Post has been published on OmCik

New Post has been published on http://omcik.com/analyst-on-reaction-to-bank-earnings-its-a-stock-pickers-market/

Analyst on reaction to bank earnings: It's a stock-pickers market

Goldman Sachs surprised investors after reporting disappointing first-quarter earnings, a rare occurrence for a bank that normally beats Wall Street projections and estimates.

Meanwhile, JP Morgan and Bank of America posted strong earnings for the quarter, leaving investors wondering whether Goldman’s disappointing results could mean anything for financials long term.

BMO Capital Markets chief investment strategist Brian Belski emphasized the importance of a stock-picking market.

“We learned today that stock picking matters. Because, again, look at what happened to Goldman and then look what happened with Bank of America and Citigroup, right?” Belski told CNBC’s “Power Lunch” on Tuesday. “We’re in a stock-pickers market, so you have to be in the right areas.”

Belski said there’s a variety of factors contributing to market volatility.

“There’s a tug of war going on with respect to fundamental analysis, macroanalysis, policy analysis and clearly geopolitical,” Belski said. “And until the end of the day, until we start to see some things happen in Washington and until we see real earnings hit these companies, we’re going to have this volatility.”

But Lindsey Group chief market analyst Peter Boockvar said despite the interesting context surrounding the post-election rally and quarter, there’s another reason earnings are slowing down.

“The fundamental basis of banking, of taking deposits and lending them out, is slowing down, and I think that’s going to be the bottom-line driver for the banking sector at least for the next couple of quarters,” Boockvar said. “It’s clear that the economy remains extremely mediocre, and now we’re seeing signs of parts of the economy rolling over, particularly the auto sector, at the same time the Fed is raising interest rates.”

Watch: Market down over 100 points today

0 notes

topinforma · 8 years ago

Text

New Post has been published on Mortgage News

New Post has been published on http://bit.ly/2pkZVw3

Analyst on market reaction to bank earnings: We're in a stock-pickers market

Goldman Sachs surprised investors after reporting disappointing first-quarter earnings, a rare occurrence for a bank that normally beats Wall Street projections and estimates.

Meanwhile, JP Morgan and Bank of America posted strong earnings for the quarter, leaving investors wondering whether Goldman’s disappointing results could mean anything for financials long term.

BMO Capital Markets chief investment strategist Brian Belski emphasized the importance of a stock-picking market.

Belski said there’s a variety of factors contributing to market volatility.

But Lindsey Group chief market analyst Peter Boockvar said despite the interesting context surrounding the post-election rally and quarter, there’s another reason earnings are slowing down.

Watch: Market down over 100 points today

0 notes