Game-Based Video-Context Dialogue
This work addresses the challenge of building dialogue systems that handle dynamic visual contexts and multiple speakers, which is incremental by extending existing methods to a new dataset.
The authors tackled the problem of multimodal dialogue by introducing a new dataset based on soccer game videos and chats, enabling models to generate relevant language from dynamic visual contexts and chat history. They developed and evaluated baseline models, achieving results measured through retrieval ranking-recall, automatic metrics, and human studies.
Current dialogue systems focus more on textual and speech context knowledge and are usually based on two speakers. Some recent work has investigated static image-based dialogue. However, several real-world human interactions also involve dynamic visual context (similar to videos) as well as dialogue exchanges among multiple speakers. To move closer towards such multimodal conversational skills and visually-situated applications, we introduce a new video-context, many-speaker dialogue dataset based on live-broadcast soccer game videos and chats from Twitch.tv. This challenging testbed allows us to develop visually-grounded dialogue models that should generate relevant temporal and spatial event language from the live video, while also being relevant to the chat history. For strong baselines, we also present several discriminative and generative models, e.g., based on tridirectional attention flow (TriDAF). We evaluate these models via retrieval ranking-recall, automatic phrase-matching metrics, as well as human evaluation studies. We also present dataset analyses, model ablations, and visualizations to understand the contribution of different modalities and model components.