SlideSpeech: A Large-Scale Slide-Enriched Audio-Visual Corpus

TL;DR

SlideSpeech is a large-scale audio-visual dataset with synchronized slides, enabling new research in multi-modal speech recognition by leveraging slide text and images to improve accuracy.

Contribution

This paper introduces SlideSpeech, a novel large-scale corpus with synchronized slides and transcripts, and proposes baseline methods to incorporate slide text into speech recognition.

Findings

Incorporating slide text improves speech recognition accuracy.

The corpus contains over 1,000 hours of transcribed speech with synchronized slides.

Baseline methods demonstrate the potential of multi-modal integration.

Abstract

Multi-Modal automatic speech recognition (ASR) techniques aim to leverage additional modalities to improve the performance of speech recognition systems. While existing approaches primarily focus on video or contextual information, the utilization of extra supplementary textual information has been overlooked. Recognizing the abundance of online conference videos with slides, which provide rich domain-specific information in the form of text and images, we release SlideSpeech, a large-scale audio-visual corpus enriched with slides. The corpus contains 1,705 videos, 1,000+ hours, with 473 hours of high-quality transcribed speech. Moreover, the corpus contains a significant amount of real-time synchronized slides. In this work, we present the pipeline for constructing the corpus and propose baseline methods for utilizing text information in the visual slide context. Through the application of keyword extraction and contextual ASR methods in the benchmark system, we demonstrate the potential of improving speech recognition performance by incorporating textual information from supplementary video slides.