Hierarchical3D Adapters for Long Video-to-text Summarization

In this paper, we focus on video-to-text summarization and investigate how to best utilize <PRE_TAG>multimodal information</POST_TAG> for summarizing long inputs (e.g., an hour-long TV show) into long outputs (e.g., a multi-sentence summary). We extend SummScreen (Chen et al., 2021), a dialogue summarization dataset consisting of transcripts of TV episodes with reference summaries, and create a multimodal variant by collecting corresponding full-length videos. We incorporate <PRE_TAG>multimodal information</POST_TAG> into a pre-trained textual summarizer efficiently using adapter modules augmented with a hierarchical structure while tuning only 3.8\% of model parameters. Our experiments demonstrate that <PRE_TAG>multimodal information</POST_TAG> offers superior performance over more memory-heavy and fully fine-tuned textual summarization methods.