Using Models of Lexical Style to Quantify Free Indirect Discourse in Modernist Fiction

Brooke, Julian; Hammond, Adam; Hirst, Graeme

doi:10.1093/llc/fqv072

Cited by 13 publications

(14 citation statements)

References 15 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…A number of recent contributions show that CL techniques can be expanded to construct analysis systems for literary texts. They for instance support the extraction of Social networks among literary characters (Elson et al 2010), an analysis of the textinternal dynamics of inter-character relationships (Chaturvedi et al 2016;Iyyer et al 2016) or aspects of plot structure (Goyal et al 2010), they induce types of characters from large text collections (e.g., Bamman et al 2014) or help understand the stylistic characterization of certain character types (Brooke et al 2017). Over the past few years, small communities of researchers pushing targeted computational modeling techniques have evolved in several field-specific branches of DH.…”

Section: Preliminariesmentioning

confidence: 99%

“…Linguistic indicators (explicit attribution of speech or thought, deictic elements, adverbial modifications, reference to sensory perception, etc.) and contextual build-up, including certain patterns of character references, are included in the feature set, and so are style indicators (as Brooke et al 2017 showed in a detailed analysis of free indirect discourse in the writings of Virginia Woolf and James Joyce).…”

Section: Modeling Subjective Categorizations: Another Place For Rapidmentioning

confidence: 99%

“…39 37 The experiments, which were run on the English translation of the Schnitzler texts, were based on the output oft he Stanford CoreNLP tools (https://stanfordnlp.github.io/CoreNLP/). In addition, the verb classification from VerbNet (https://verbs.colorado.edu/verb-index/), and the stylistic profile vectors from Brooke et al (2017) were used, which are included in the Gutentag tools suite (www.cs.toronto.edu/*jbrooke/gutentag/; the style profile lexicon is available as built_in_lexicons/ sixstyleplus.txt; thanks to Julian Brooke). 38 What we call ''focal'' here is operationalized as evoking the reader's interpretation of the character being the present exponent of subjective point of view.…”

Section: Modeling Subjective Categorizations: Another Place For Rapidmentioning

confidence: 99%

See 2 more Smart Citations

Computational text analysis within the Humanities: How to combine working practices from the contributing fields?

Kuhn

2019

Lang Resources & Evaluation

View full text Add to dashboard Cite

This position paper is based on a keynote presentation at the COLING 2016 Workshop on Language Technology for Digital Humanities in Osaka, Japan. It departs from observations about working practices in Humanities disciplines following a hermeneutic tradition of text interpretation versus the method-oriented research strategies in Computational Linguistics (CL). The respective praxeological traditions are quite different. Yet more and more researchers are willing to open up Keywords Digital humanities Á Adaptation of NLP tool chains Á Interdisciplinary working practice 2 For Computational History, see for instance the contributions in Bozic et al. (2016); for Computational Literary Studies, many researchers came together for the 2017 International DFG Symposion in Literary Studies at Villa Vigoni, Italy 9-13 October 2017, which was dedicated to Digital Literary Studies. A volume of the contributions is in preparation (Jannidis, in preparation). The development of the subfield of Computational Literary Studies in Germany is quite dynamic. In 2018, the Deutsche Forschungsgemeinschaft (DFG), established a priority program Computational Literary Studies (SPP 2207) with a runtime of 6 years. 3 This is not to say that digital editions and computational support for corpus search and exploration have not been widely adopted. Indeed, fields like Literary Studies have changed with the omnipresence of computational tools and resources, hence the observation of a ''computational turn'' (Berry 2011). Most influential have arguably been the contributions by Moretti (2007) and Jockers (2013). The emphasis in this article is however on the integration of formal/algorithmic models in the core argumentation of scholarly research in the Humanities disciplines, which up until today are still in its early stages. 4 Our focus is not on Philosophical Hermeneutics (associated with Friedrich Schleiermacher, Wilhelm Dilthey, Hans-Georg Gadamer a.o.), but on hermeneutics as a broadly adopted practical method in Literary Studies and other disciplines targeting the interpretation of works of art. As Newton (1989) points out, the term, which has long been central in the German tradition, gained prominence in the Anglo-American tradition with the work by Hirsch (1967). ''The central concern of hermeneutics is the problem created by the fact that texts written in the past continue to exist while their authors and the historical context which produced them pass away in time. Reading such texts therefore becomes inseparable from the question of interpretation.'' (Newton 1989: 116). 5 See also the discussion of the aims of text interpretation in Mantzavinos (2016).

show abstract

Section: Preliminariesmentioning

confidence: 99%

Section: Modeling Subjective Categorizations: Another Place For Rapidmentioning

confidence: 99%

Section: Modeling Subjective Categorizations: Another Place For Rapidmentioning

confidence: 99%

See 1 more Smart Citation

Computational text analysis within the Humanities: How to combine working practices from the contributing fields?

Kuhn

2019

Lang Resources & Evaluation

View full text Add to dashboard Cite

show abstract

“…We started our investigation by looking at free indirect discourse (FID), an important formal vehicle of dialogism in modernist literature, through which authors deliberately blur the boundaries between the words of narrators and characters. Our initial experiment was simply to test whether readers (some 480 undergraduate students) were able to consistently identify spans of FID in Woolf's To the Lighthouse and Joyce's “The Dead” (Brooke, Hammond, & Hirst, ). Validating these results proved challenging, given the expectation in computational linguistics that readerly consensus must reach levels where disagreements can be reasonably dismissed as noise—an expectation that was unreasonable in the context of our experiment (Hammond, Brooke, & Hirst, ).…”

Section: Collaboration or Interdisciplinary Adaptationmentioning

confidence: 99%

“…Our initial aim had been to use the reading data to train a machine learning model that could detect passages of FID in any literary text. Again, the computational linguists used their disciplinary expertise to determine that this was not practicable; Woolf's and Joyce's employments of the device were simply too peculiar to be generalizable (Hammond, Brooke, & Hirst, ). As a way forward, I came up with the idea of producing digital editions of To the Lighthouse and “The Dead” that would visualize students' vastly differing interpretations of FID in the words—“readers' maps” to show how great an interpretive challenge dialogic modernist works present to their readers .…”

Section: Collaboration or Interdisciplinary Adaptationmentioning

confidence: 99%

The double bind of validation: distant reading and the digital humanities' “trough of disillusionment”

Hammond

2017

Literature Compass

Self Cite

View full text Add to dashboard Cite

The digital humanities (DH) is currently in the phase of the “hype cycle” known as the “trough of disillusionment.” Franco Moretti, perhaps the most prominent practitioner of the most prominent discipline of DH—“distant reading,” the computational analysis of large quantities of literary texts—recently expressed his exasperation with the state of DH, reflecting “our work could have been better” and asking why, “considering the amount of energy, talent, and tools, going into [DH], that we have such difficulty producing great results.” Surveying leading recent work in distant reading by Moretti, Matthew L. Jockers, Laura Mandell, Ryan Heuser, Long Le‐Khac, and Joanna Swafford, this paper provides a twofold explanation to the field's failure to produce “great results.” Both explanations relate to “validation,” the process by which quantitative results are shown to be reliable and trustworthy. Many distant reading projects have produced disappointing results because they have been more interested in validating their tools—showing that their computational methods are able to confirm existing stereotypes—than in pursuing genuine discoveries. Many others, meanwhile, produce provocative results that cannot be meaningfully validated. Although the double bind of validation is real, I propose collaboration and “interdisciplinary adaptation” as promising solutions.

show abstract

Empirie – Beschreibung – Interpretation. Über den Platz von Computermodellen in den hermeneutisch-historisch orientierten Literaturwissenschaften

Kuhn¹

2022

Germanistische Symposien

View full text Add to dashboard Cite

ZusammenfassungAusgangspunkt für diesen Artikel sind die Vorannahmen, die für einen Text erfüllt sein müssen, damit datenbasierte Textanalysemodelle aus der Computerlinguistik methodisch zuverlässig auf ihn angewandt werden können. Der Artikel stellt die Frage, inwieweit die empirisch-korpusbasierte Methodik unter diesen Annahmen Befunde zu interpretationsrelevanten Eigenschaften literarischer Texte liefern und damit einen Beitrag zu Kernfragen einer hermeneutisch-historisch orientierten Literaturwissenschaft leisten kann. Von besonderer Bedeutung ist die methodische Vorannahme, dass die Trainingsdaten, auf deren Basis die Modellparameter abgeschätzt werden, in den relevanten Eigenschaften repräsentativ seien für die Texte, die im Zentrum der eigentlichen Untersuchung stehen. Dies scheint zunächst das Spektrum der erzielbaren Befunde auf deskriptive Untersuchungskategorien zu beschränken, für die eine robuste Operationalisierung möglich ist. Interpretationsrelevanten Texteigenschaften scheinen sich dem Zugang zu entziehen. Der Artikel regt jedoch an, eine alternative Praxis für die Anwendung von Computermodellen zu explorieren, bei der über deskriptive Standardkategorien hinaus untersuchungsspezifische Analysekategorien aus einer bestimmten Fragestellung heraus entwickelt und komputationell modelliert werden können. Basierend auf einem Vorverständnis im Sinne einer hermeneutisch geprägten Arbeitspraxis können so auch interpretationsrelevante Texteigenschaften in die Modellierung einbezogen werden. Zur Absicherung der methodischen Validität der Befunde muss das Forschungsteam bei diesem Vorgehen seine Modelle anhand geeigneter, sorgfältig annotierter Referenzdatensätze selbst kalibrieren und die zu erwartende Vorhersagegenauigkeit abschätzen.

show abstract

Using Models of Lexical Style to Quantify Free Indirect Discourse in Modernist Fiction

Cited by 13 publications

References 15 publications

Computational text analysis within the Humanities: How to combine working practices from the contributing fields?

Computational text analysis within the Humanities: How to combine working practices from the contributing fields?

The double bind of validation: distant reading and the digital humanities' “trough of disillusionment”

Empirie – Beschreibung – Interpretation. Über den Platz von Computermodellen in den hermeneutisch-historisch orientierten Literaturwissenschaften

Contact Info

Product

Resources

About