对话管理中基于槽特征有限状态自动机的方法研究

对话系统的研究已经成为人机交互技术发展的新热点。而对话管理则是其中最重要的组成部分．该文在当前对话管理的各种实现方法的基础上，提出了一种基于槽特征的自动机设计方法，其中应用了状态压缩和状态集、动作集的子空间划分。并着重以确认过程为例，阐述了确认策略控制函数及其对对话过程的影响．文中还提出了一种树形的意图分层结构，并将这种分层结构应用于主题检测与主题切换，成功解决了多主题对话系统的主题切换问题．最后，实验表明该文提出的设计方案在策略控制、主题检测与主题切换等方面具有较好性能，同时也具有一定扩展性．

A Finite State Automata Approach Based on Slot-Feature for Dialogue Management in Spoken Dialogue System

Spoken dialogue systems have become a new hotspot of human-machine interaction, in which dialogue management is the most important. In this paper, we propose a new method for dialogue management. The scheme uses finite state automata based on slot-features to establish the whole structure. In the automata, a two-dimensional state feature is proposed and state reduction is applied. State set and action set are divided into subsets; hence it becomes easier to control. The most distinctive aspect is that two strategy control functions are used to implement the strategy control in the verification as well as in the request. Furthermore, an intention-layered structure is proposed and applied to the topic detection and switch, which work for multi-topics applications. The structure is portable and feasible to be extended. At last, experiments show that our method has good performance.