PTVD: A Large-Scale Plot-Oriented Multimodal Dataset Based on Television Dramas

TL;DR

PTVD is a comprehensive, large-scale multimodal dataset based on TV dramas, designed to improve plot coherence understanding, with extensive annotations, comments, and a baseline model for multimodal learning research.

Contribution

This paper introduces PTVD, the first plot-oriented multimodal TV drama dataset, including long-form video, detailed annotations, and a strong baseline model for multimodal tasks.

Findings

PTVD enables new multimodal research in TV dramas.

Extensive experiments validate the dataset's usefulness.

Counter-intuitive insights from cognitive-inspired tasks.

Abstract

Art forms such as movies and television (TV) dramas are reflections of the real world, which have attracted much attention from the multimodal learning community recently. However, existing corpora in this domain share three limitations: (1) annotated in a scene-oriented fashion, they ignore the coherence within plots; (2) their text lacks empathy and seldom mentions situational context; (3) their video clips fail to cover long-form relationship due to short duration. To address these fundamental issues, using 1,106 TV drama episodes and 24,875 informative plot-focused sentences written by professionals, with the help of 449 human annotators, we constructed PTVD, the first plot-oriented multimodal dataset in the TV domain. It is also the first non-English dataset of its kind. Additionally, PTVD contains more than 26 million bullet screen comments (BSCs), powering large-scale pre-training. Next, aiming to open-source a strong baseline for follow-up works, we developed the multimodal algorithm that attacks different cinema/TV modelling problems with a unified architecture. Extensive experiments on three cognitive-inspired tasks yielded a number of novel observations (some of them being quite counter-intuition), further validating the value of PTVD in promoting multimodal research. The dataset and codes are released at \url{https://ptvd.github.io/}.