Dramatic Conversation Disentanglement

Kent K. Chang,Danica Chen,David Bamman

DOI: https://doi.org/10.48550/arXiv.2305.16648

2023-05-26

Abstract:We present a new dataset for studying conversation disentanglement in movies and TV series. While previous work has focused on conversation disentanglement in IRC chatroom dialogues, movies and TV shows provide a space for studying complex pragmatic patterns of floor and topic change in face-to-face multi-party interactions. In this work, we draw on theoretical research in sociolinguistics, sociology, and film studies to operationalize a conversational thread (including the notion of a floor change) in dramatic texts, and use that definition to annotate a dataset of 10,033 dialogue turns (comprising 2,209 threads) from 831 movies. We compare the performance of several disentanglement models on this dramatic dataset, and apply the best-performing model to disentangle 808 movies. We see that, contrary to expectation, average thread lengths do not decrease significantly over the past 40 years, and characters portrayed by actors who are women, while underrepresented, initiate more new conversational threads relative to their speaking time.

Computation and Language

What problem does this paper attempt to address?

The problem that this paper attempts to solve is conversation disentanglement in movies and TV dramas. Specifically, the authors focus on how to identify different conversational threads in complex multi - character dialogue scenes. These threads may occur simultaneously, and each thread has its own topic and focus. This conversation disentanglement is of great significance for understanding cultural representations on the screen, dialogue patterns, and power relationships between characters. For example, by analyzing issues such as who initiates new conversational threads and how long conversations usually last, we can enhance our understanding of cultural representations on the screen.

To study this problem, the authors constructed a new dataset, which contains 10,033 dialogue turns (comprising 2,209 threads) extracted from 831 movies. They also compared the performance of several conversation disentanglement models on this new dataset and applied the best model to analyze 808 movies, exploring the relationship between historical thread length and enhanced continuous style, as well as the relationship between gender and the struggle for the right to speak in conversations. The study found that, contrary to expectations, the average thread length has not decreased significantly in the past 40 years, and although characters played by female actors are under - represented, they initiate new conversational threads more frequently relative to their speaking time. These findings contribute to a deeper understanding of dialogue structures and their social impacts.