Speaker Naming in Movies

TL;DR

This paper introduces a multimodal model for speaker naming in movies that combines visual, textual, and acoustic data, demonstrating superior performance on a new dataset and achieving state-of-the-art results in subtitle-based question answering.

Contribution

The paper presents a novel unified optimization framework for multimodal speaker naming and an end-to-end memory network that advances the state-of-the-art in movie subtitle understanding.

Findings

Significant improvement over baselines in speaker naming accuracy

New dataset with TV show and movie episodes for evaluation

State-of-the-art results on MovieQA 2017 Challenge subtitles task

Abstract

We propose a new model for speaker naming in movies that leverages visual, textual, and acoustic modalities in an unified optimization framework. To evaluate the performance of our model, we introduce a new dataset consisting of six episodes of the Big Bang Theory TV show and eighteen full movies covering different genres. Our experiments show that our multimodal model significantly outperforms several competitive baselines on the average weighted F-score metric. To demonstrate the effectiveness of our framework, we design an end-to-end memory network model that leverages our speaker naming model and achieves state-of-the-art results on the subtitles task of the MovieQA 2017 Challenge.