Cross-lingual Approaches for Task-specific Dialogue Act Recognition
This work addresses the problem of data scarcity in dialogue act recognition for specific tasks and languages, though it is incremental in nature.
The paper tackles dialogue act recognition for specific tasks with limited annotations by using cross-lingual models, achieving significant performance improvements over related approaches.
In this paper we exploit cross-lingual models to enable dialogue act recognition for specific tasks with a small number of annotations. We design a transfer learning approach for dialogue act recognition and validate it on two different target languages and domains. We compute dialogue turn embeddings with both a CNN and multi-head self-attention model and show that the best results are obtained by combining all sources of transferred information. We further demonstrate that the proposed methods significantly outperform related cross-lingual DA recognition approaches.