In this work we present a new dataset of literary events—events that are depicted as taking place within the imagined space of a novel. While previous work has focused on event detection in the domain of contemporary news, literature poses a number of complications for existing systems, including complex narration, the depiction of a broad array of mental states, and a strong emphasis on figurative language. We outline the annotation decisions of this new dataset and compare several models for predicting events; the best performing model, a bidirectional LSTM with BERT token representations, achieves an F1 score of 73.9. We then apply this model to a corpus of novels split across two dimensions—prestige and popularity—and demonstrate that there are statistically significant differences in the distribution of events for prestige.

Literary Event Detection

A history of literary prestige needs to study both works that achieved distinction and the mass of volumes from which they were distinguished. To understand how those patterns of preference changed across a century, we gathered two samples of English-language poetry from the period 1820–1919: one drawn from volumes reviewed in prominent periodicals and one selected at random from a large digital library (in which the majority of authors are relatively obscure). The stylistic differences associated with literary prominence turn out to be quite stable: a statistical model trained to distinguish reviewed from random volumes in any quarter of this century can make predictions almost as accurate about the rest of the period. The “poetic revolutions” described by many histories are not visible in this model; instead, there is a steady tendency for new volumes of poetry to change by slightly exaggerating certain features that defined prestige in the recent past.

/pdf/the-longue-duree-of-literary-prestige-1bwm0bgmos.pdf

The Longue Durée of Literary Prestige

To judge by debates over distant reading, literary scholars might have something in common with the unadaptable tutor in george Eliot's The Mill on the Floss, the Rev. Mr. Stelling, who applies the same educational method to any boy in his charge: he sees Tom Tulliver's mind as a field to be plowed with the classics, not as a stomach that can't digest them. But if we look more closely at the range of projects in literary studies, from digital humanities to the new formalist poetics, we find few hidebound Stellings. Applying one's favorite method to all sizes and shapes of data is an obvious genre of error, and scholars and critics of all sorts studiously avoid it. In a maxim now popular among practitioners of digital humanities, when you have a hammer, everything looks like a nail. The humorous warning not to become enamored of shiny tools and ply them regardless of object (a schoolboy, all schoolboys) can come from people who use MALLET (MAchine Learning for LanguagE Toolkit) or other software to produce statistical analyses of words or of white spaces in nineteenth-century newspapers. As researchers who are primed to question the prescriptive bias of our doxa, our samples, our models, and our tools, we nevertheless have incentives to overstate the predictable errors of competing approaches.

Mid-Range Reading: Not a Manifesto

This article is in two parts. The first seeks to set an agenda for the study of Victorian popular fiction by examining what the field comprises today in terms of texts studied, methodologies and affective engagement, and then thinking through the implications of studying such fiction today in a global and remediated context. I argue that Victorian sentimental popular fiction self-consciously models processes of relationship formation and exploration in its characters, its explicit scenes of reading, and above all in its plots, in order to mould and maintain readers’ relationships to it. “Sympathy” and its interrogation defines both the representation of characters’ relations to one another and readers’ relationship to that representation. It is a textual technique used by the fiction industry to create and maintain customer loyalty. Our affective responses to this technique constitute one reason we, as students of a still marginal field, continue to read it with energy and enthusiasm. What we need to do is self-consciously think through the implications of this energy’s rootedness in the commercial imperatives of the nineteenth-century publishing industry. Part 2 offers two case studies to examine that call, one very famous and the other virtually unknown, by American women writers whose work circulated globally: Susan Warner and E.D.E.N. Southworth. I ask what the ethical and methodological implications might be for the checking of our pleasure through what neuroaesthetics calls “cognitive elaboration.”

/pdf/victorian-popular-fictions-today-feel-these-words-as-mama-4qvnf6jaec.pdf

Victorian Popular Fictions Today: “feel these words as mama does!”

A long time ago now (maybe 1988?), Gerald (Gazdar) and I supervised Adam’s DPhil at the University of Sussex. Adam was my age, give or take a year, having come to academia a little late, and was my first doctoral student. Adam’s topic was polysemy, and I’m not really sure that much supervision was actually required, though I recall fun exchanges trying to model the subtleties of word meaning using symbolic knowledge representation techniques—an experience that was clearly enough to convince Adam later that this was a bad idea. In fact, Adam’s thesis title itself was Polysemy. Much as we encourage short thesis titles, pulling off the one-word title is a tall order, requiring a unique combination of focus and coverage, breadth and depth, and, most of all, authority. Adam completely nailed it, at least from the perspective of the pre-empirical Computational Linguistics of the early 1990s. Three years later, after a spell working for dictionary publishers, Adam joined me as a research fellow, now at the University of Brighton. I had a project to explore the automatic enrichment of lexical databases to support the latest trends in language analysis, and, in particular, task-specific lexical resources. I was really pleased and excited to recruit Adam—he had lost none of his intellectual independence, a quality I particularly valued. Within a few weeks he came to me with his own plan for the research—a “detour,” as he put it, from the original workplan. I still have the e-mail, dated 6 April 1995, in which he proposed that, instead of chasing a prescriptive notion of a single lexical resource that needed to be customized to each domain, we should let the domain determine the lexicon, providing lexicographic tools to explore words, and particularly word senses, that were significant for that domain. In that e-mail, Computational Lexicography at Brighton was born. Over the next eight years or so, Computational Lexicography became a key part of our group’s success, increasingly under Adam’s direct leadership. The key project, WASPS, developed the WASPbench—the direct precursor of the Sketch Engine, recruiting David (Tugwell) to the team. In addition, Adam was one of the founding organizers of SENSEVAL, an initiative to bring international teams of researchers together to work in friendly competition on a pre-determined word sense disambiguation task (and which has now transformed into SEMEVAL). Together we secured funding to support the first two rounds of SENSEVAL; each round required the preparation of standardized data sets, guided by Adam’s highly tuned intuitions about lexical data preparation and management. And we engaged somewhat in the European funding merry-go-round, most fondly in the CONCEDE project, working on dictionaries for Central European languages with amazing teams from the MULTEXT-EAST consortium, and with Georgian and German colleagues in the GREG project.

Adam kilgarriff

Of the novelties introduced by digitization in the study of literature, the size of the archive is probably the most dramatic: we used to work on a couple of hundred nineteenth-century novels, and now we can analyze thousands of them, tens of thousands, tomorrow hundreds of thousands. It's a moment of euphoria, for quantitative literary history: like having a telescope that makes you see entirely new galaxies. And it's a moment of truth: so, have the digital skies revealed anything that changes our knowledge of literature? This is not a rhetorical question. In the famous 1958 essay in which he hailed "the advent of a quantitative history" that would "break with the traditional form of nineteenth-century history", Fernand Braudel mentioned as its typical materials "demographic progressions, the movement of wages, the variations in interest rates [...] productivity [...] money supply and demand." These were all quantifiable entities, clearly enough; but they were also completely new objects compared to the study of legislation, military campaigns, political cabinets, diplomacy, and so on. It was this double shift that changed the practice of history; not quantification alone. In our case, though, there is no shift in materials: we may end up studying 200,000 novels instead of 200; but, they're all still novels. Where exactly is the novelty?

Canon/archive : large-scale dynamics in the literary field

Writing the Mind

This article describes a series of ongoing efforts at the Stanford Literary Lab to manage a large collection of literary corpora (~40 billion words). This work is marked by a tension between two competing requirements – the corpora need to be merged together into higher-order collections that can be analyzed as units; but, at the same time, it’s also necessary to preserve granular access to the original metadata and relational organization of each individual corpus. We describe a set of data management practices that try to accommodate both of these requirements – Apache Spark is used to index data as Parquet tables on an HPC cluster at Stanford. Crucially, the approach distinguishes between what we call “canonical” and “combined” corpora, a variation on the well-established notion of a “virtual corpus” (Kupietz et al., 2014; Jakubiek et al., 2014; van Uytvanck, 2010).

/pdf/organizing-corpora-at-the-stanford-literary-lab-balancing-4k80zjbroh.pdf

Organizing corpora at the Stanford Literary Lab. Balancing simplicity and flexibility in metadata management

In its ability to extract feature sets, relate texts within an abstract space, and semantically parse groups of texts, computational textual analysis has functioned primarily as a formalist intervention into literary study. When practitioners venture outside of the formal features of the texts themselves, it is author or date that serves as the point of contact between the text and its wider context. And yet, texts offer a rich history of reception: as different interpretive communities (Fish 1980) receive and reinterpret novels, poems or plays, they recontextualize the literary object to suit the particular socio-cultural goals of their period or nationality. Lacking detailed accounts of reading practices at large scales, even traditional practitioners of literary history have been unable to reconstruct the history of reception of even the most historically canonical texts. In this project, we leverage the ability of Digital Humanities to recover, at least provisionally, a large-scale history of textual reception by exploring the patterns of citation that reveal the attention paid to specific texts across their history as readerly objects. How are certain canonical texts cited over time and what can the attention paid to different segments of texts with a rich reception history tell us about the reading or social practices of different historical periods? How do different groups of readers (particularly authors and critics) quote text differently as they make use of passages in their own writing? And how do specialists and non-specialists cite the same text differently? By exploring the locus of attention within a canonical text, both across groups of readers and across history, we provisionally reconstruct a historically and socially contingent map of a text’s reception history. As Piper and Algee-Hewitt have argued in “The Werther Effect” (Piper and Algee-Hewitt 2014), practices of citation, the embedding of the language of a text within other works, can reveal patterns of reception even within a single author’s corpus. In this project, we expand this approach multi-dimensionally, identifying passages of canonical works quoted in other novels, in critical articles by specialists and nonfield specialists, and in a larger undifferentiated corpus of text. The scale of our analysis enables us to identify what parts of a text have received the most writerly attention overall and how that attention has been shaped over time. By moving from semantics to passages, we switch our attention from the intangible metrics of semantic similarity, to specific, quotationlevel instances of citation that demonstrate specific attention to identifiable parts of our target texts. Drawing on work in sequence alignment by David Smith et al (2013) and Richard So et al. (forthcoming), we will therefore be able to explore patterns of attention that have been paid to a text by identifiable groups of readers. We argue that these patterns of citationality serve as a proxy for the reception of a text: while necessarily limited to readers who themselves were authors (or critics), they nevertheless represent an important category of reception available to analysis. To extract the quotations, we used Python’s “difflib” module, wrapped up as a parallelized MPI program that runs on an HPC cluster. To compare any two individual texts for example, when checking for passages from Hamlet inside of a novel from the Gale American Fiction corpus the texts are first split into tokens and passed through a filter that removes a set of 200 stopwords. This speeds up the alignment algorithm (the high-frequency words that get pulled out make up a significant portion of the total words in any given text, producing shorter sequences) and also has the advantage of making the alignment process less sensitive to small changes in function words, which seem to get shuffled around or changed fairly frequently when a text is quoted. For example, a change from: And crook the pregnant hinges of the knee Where thrift may follow fawning

/pdf/reconstructing-readerly-attention-citational-practices-and-4jusdgdgj2.pdf

Hannah Walser

Papers

Canon/archive : large-scale dynamics in the literary field

Writing the Mind

Organizing corpora at the Stanford Literary Lab. Balancing simplicity and flexibility in metadata management

Reconstructing Readerly Attention: Citational Practices and the Canon, 1789-2016.

The Novel as Data