统计与规则结合的问句理解

APPROACH FOR QUESTION UNDERSTANDING BASED ON COMBINATION OF STATISTICS AND RULE

提出了一个结合统计和规则的口语理解方法。首先,用统计分类器对输入进行主题分类,然后用语义规则提取主题相关的语义槽。该方法在主题分类和语义槽提取方面都具有较低的错误率,同时具有很好的鲁棒性,并在图书馆查询系统的查询需求理解中取得了很好的结果。

In this paper we propose an approach for spoken language understanding based on the combination of statistics and rule. Our approach first categorizes the topic of an input utterance using statistical classifier, and then extracts the semantic slots associated with the topic by semantic rules. The method has both quite low failure rate and good robustness in topic classification and semantic slot extraction, and performed well in experiments of query requirement understanding in library query system.