Redlining Culture by Richard Jean So

Author:Richard Jean So
Language: eng
Format: epub
Publisher: Columbia University Press

RACE AND LITERARY DISTINCTION

Let me begin with a simple hypothesis: based on an analysis of textual features of novels—namely, diction, syntax, style, and narrative—there exists measurable differences between literary categories of U.S. bestselling novels, U.S. prizewinning novels, and novels written by black authors published between 1950 and 2000. And if this is true, let me add a brief corollary: the difference that exists is greater between novels by black authors and both bestselling and prizewinning novels than that between bestselling and prizewinning novels. And if this is true, one final corollary: this difference—what we might think of as a boundary line that divides our three categories—is unchanging over time.

To test this hypothesis, we’ll need a language model. Our model includes four types of features. First, diction: my model simply counts the frequencies by which different words appear in a text. Word frequencies can only tell so much, and they cannot capture things like metaphor or irony, but they provide a baseline to glimpsing the content or story of a novel. Second, syntax: my model also counts the frequencies by which certain parts of speech, such as “nouns” or “adverbs,” appear in a text. Here, too, simply counting the frequency of parts of speech in a text can only tell us so much and cannot tell us about paragraph-level syntactical ambitions, but it can still tell us much about the sentence-level syntactical habits of the text.19 Third, style: my language model computes the amount of repetition and the lexical density of the text. How often does a text repeat the same words per sentence or paragraph? How difficult is the text to read based on a measurement of lexical diversity per sentence or paragraph? Again, this approach will not be able to pick up on such finely defined tendencies like a “Jamesian style,” but it provides a useful report on the broader stylistic habits of the author. Last, narrative: the model will compute statistics relevant to the represented world of the story. These include the ratio of dialogue to narration, the number of characters that appear in the story, the average amount of attention paid to each character, the number of manmade and natural objects that appear in the story on average, the number of locations or settings that appear in the story on average. In total, I refer to these features as “narration.” Once more, these narrative features cannot capture the full complexity of the concept of narrative as asserted by literary theorists. But they are useful, albeit coarse, accounts of how novels employ the building blocks of narration: characters, dialogue, setting, objects, and so on.20

The value of the language model is that it allows us to identify concretely and specifically, at the level of words, style, form, and narrative, how two corpora of literary texts are similar or different. Traditional close reading, of course, can spot such differences in single texts, but the machine can tell us how consistent and regular these differences or similarities are across a large number of literary works.

Download

Copyright Disclaimer:
This site does not store any files on its server. We only index and link to content provided by other sites. Please contact the content providers to delete copyright contents if any and email us, we'll remove relevant links or contents immediately.