摘要
现有许多对话系统都是面向任务的 ,但很少考虑对话过程涉及的多主题、主题切换、主题间的信息共享 ,以及不同信息重要程度不同等问题 .该文提出基于主题森林结构的对话管理模型 ,较好地处理了上述问题 ,能实现对话过程的混合主导 .该模型能保证交互过程中对话内容的一致性 ,对上下文的理解、对话控制和应答决策灵活自然 .其推理机应用了多种与领域无关的策略 。
There are many task-oriented dialog systems, but few take into account the issues on multiple topics, topic shifts, information sharing among topics, and difference in importance for information items. A dialog management model based on the topic forest structure is presented, which can well handle those issues with mixed-initiative dialog control. This model is consistent during the dialog transaction, so that it can understand the user according to the context. Its dialog control and response decision are both natural and flexible. With several domain-independent strategies, its reasoning engine is portable too.
出处
《自动化学报》
EI
CSCD
北大核心
2003年第2期275-283,共9页
Acta Automatica Sinica
基金
国家自然科学基金 (198710 4 5 )
清华基础研究基金 (JC19990 4 2 )资助~~
关键词
主题森林结构
对话管理模型
自然语言理解
人机口语对话技术
计算机
信息处理
Data structures
Indexing (of information)
Inference engines
Management
Mathematical models
Speech intelligibility
Trees (mathematics)