Modelling social readers: novel tools for addressing reception from online book reviews.

Pavan Holur,Timothy R Tangherlini,Shadi Shahsavari,Vwani Roychowdhury,Ehsan Ebrahimzadeh

doi:10.1098/rsos.210797

Pavan Holur, Timothy R Tangherlini + Show 3 more

Open Access

https://doi.org/10.1098/rsos.210797

Copy DOI

Abstract

Social reading sites offer an opportunity to capture a segment of readers’ responses to literature, while data-driven analysis of these responses can provide new critical insight into how people ‘read’. Posts discussing an individual book on the social reading site, Goodreads, are referred to as ‘reviews’, and consist of summaries, opinions, quotes or some mixture of these. Computationally modelling these reviews allows one to discover the non-professional discussion space about a work, including an aggregated summary of the work’s plot, an implicit sequencing of various subplots and readers’ impressions of main characters. We develop a pipeline of interlocking computational tools to extract a representation of this reader-generated shared narrative model. Using a corpus of reviews of five popular novels, we discover readers’ distillation of the novels’ main storylines and their sequencing, as well as the readers’ varying impressions of characters in the novel. In so doing, we make three important contributions to the study of infinite-vocabulary networks: (i) an automatically derived narrative network that includes meta-actants; (ii) a sequencing algorithm, REV2SEQ, that generates a consensus sequence of events based on partial trajectories aggregated from reviews, and (iii) an ‘impressions’ algorithm, SENT2IMP, that provides multi-modal insight into readers’ opinions of characters.

Highlights

Online reader comments about works of literary fiction offer an intriguing window onto how people read
We introduced a pipeline addressing two important tasks that are instrumental in constructing this network representation: entity mention grouping (EMG) and inter-actant relationship clustering (IARC) [60], which we summarize below
Before computing the cosine similarity, we reduce the dimension of the BERT embeddings to four principal components using principal component analysis (PCA), having found that the resultant scores generalize well with this choice of principal components

Summary

Introduction

Online reader comments about works of literary fiction offer an intriguing window onto how people read. The reviews of a single novel provide a view onto the collective imagining of what is important in the novel, including aspects of plot, the interactions between various characters, and even the metadiscursive space of authors, genre, critics, film adaptations and movie stars. These reviews provide impetus for a data-driven analysis of readers’ responses to a work of literary fiction [4]. They help us understand how readers create an ‘imagined community’ of readers, an extension of Fish’s notion of ‘communities of interpretation’, engaged in the collective enterprise of literary analysis [5,6]

Objectives

Methods

Results

Conclusion