Improving Automatic Quotation Attribution in Literary Novels

TL;DR

This paper improves automatic quotation attribution in literary novels by breaking down the task into interconnected sub-tasks, benchmarking models on a large dataset, and demonstrating that simple models can achieve competitive accuracy.

Contribution

It introduces a modular approach to quotation attribution, benchmarks state-of-the-art models on a new dataset, and shows simple models can perform well in this complex task.

Findings

Sequential models achieve high accuracy in speaker attribution

Benchmark dataset enables comprehensive evaluation of sub-tasks

Modular approach improves flexibility in quotation attribution

Abstract

Current models for quotation attribution in literary novels assume varying levels of available information in their training and test data, which poses a challenge for in-the-wild inference. Here, we approach quotation attribution as a set of four interconnected sub-tasks: character identification, coreference resolution, quotation identification, and speaker attribution. We benchmark state-of-the-art models on each of these sub-tasks independently, using a large dataset of annotated coreferences and quotations in literary novels (the Project Dialogism Novel Corpus). We also train and evaluate models for the speaker attribution task in particular, showing that a simple sequential prediction model achieves accuracy scores on par with state-of-the-art models.