Reference Resolution and Context Change in Multimodal Situated Dialogue for Exploring Data Visualizations

TL;DR

This paper investigates reference resolution in multimodal dialogue involving visualizations, proposing a context-aware pipeline and comparing traditional and deep learning models, highlighting transfer learning benefits and generalization issues.

Contribution

It introduces a new annotation scheme and reference resolution pipeline for multimodal visualization dialogue, and evaluates models including CRF and transformer-based methods.

Findings

Transfer learning improves deep learning model performance.

CRF outperforms deep learning models on low-resource data.

Contextual information enhances reference resolution accuracy.

Abstract

Reference resolution, which aims to identify entities being referred to by a speaker, is more complex in real world settings: new referents may be created by processes the agents engage in and/or be salient only because they belong to the shared physical setting. Our focus is on resolving references to visualizations on a large screen display in multimodal dialogue; crucially, reference resolution is directly involved in the process of creating new visualizations. We describe our annotations for user references to visualizations appearing on a large screen via language and hand gesture and also new entity establishment, which results from executing the user request to create a new visualization. We also describe our reference resolution pipeline which relies on an information-state architecture to maintain dialogue context. We report results on detecting and resolving references, effectiveness of contextual information on the model, and under-specified requests for creating visualizations. We also experiment with conventional CRF and deep learning / transformer models (BiLSTM-CRF and BERT-CRF) for tagging references in user utterance text. Our results show that transfer learning significantly boost performance of the deep learning methods, although CRF still out-performs them, suggesting that conventional methods may generalize better for low resource data.