B&Anet: Combining bidirectional LSTM and self-attention for end-to-end learning of task-oriented dialogue system

Abstract

Building dialogue systems plays an important role in modern life. amongst them, task-oriented dialogue for resolving problems in actual life is most worth exploring. Motivated by the development of end-to-end approaches, a task-oriented dialogue model based on bidirectional LSTM and self-attention mechanism is proposed. It not only makes good use of context and effectively solves the long-term dependency, but also identifies the relationship between sentences, optimizes feature vectors, and has good parallelism. In our method, the dialogue state tracker(DST) is firstly improved. Our dialogue state tracker can identify the multiple slot key-value pairs involved in the utterance without manual label. In addition, we apply the data augmentation of merging machine translation and bilingual dictionary to create more diversified data sets. Finally, in the experimental part, the enhanced data and the results of DST are fed into the proposed B&Anet (bidirectional long and short memory network and self-attention mechanism network) model. The evaluation results on DSTC2 (dialogue state tracking chanllenge2) show that our method achieves competitive performance.