面向任务口语对话系统中期待模型的实现算法

The Construction of Expectation Model in Task-Oriented Spoken Dialogue Systems and Its Realization Algorithms

该文研究如何从口语对话系统特定场景的期待内容中,抽象出期待所需要的一般的数据结构和处理方法,提出了基于任务模型的期待模型的构建算法。将该期待模型与对话上历史相结合,可以为系统构造出随对话过程动态改变的语境,使系统具备初步地利用语境推导用户意向的推理功能,从而大大提高语义分析的鲁棒性及对话成功率,最后给出了该期待模型在实验系统中应用的有效性结果。

This paper mainly studies how to extract the common structure and process methods of expectation from scene-specified expectation setting, then puts forth the algorithms to build expectation model suitable for dialogue systems based on the task model.When incorporated with dialog context, this model could create dynamic situation that varies with dialogue process, which endows the system with the preliminary ability to reason users' intentions by reference to this situation, so as to improve the robustness and precision of semantic analysis and the dialogue success rate. Finally, the experimental results prove the effectiveness and usability of this model.