Understanding The Semantic Intent Of Domain-Specific Natural Language Query

This paper mainly studies the method to understand the semantic intent of domain-specific natural language queries. Parsing natural language queries for the purpose of understanding user’s query intent is an essential factor to search engine and automatic question answering system. Traditional methods for understanding user’s query intent only take keyphrase query into consideration, so most proposed approaches are not suitable for natural language query. In this paper, we focus on understanding the semantic intent of natural language queries. To this end, we study an cascaded conditional fandom fields model for chinese address normalization and then a structured SVM model for understanding the semantic intent of natural language query.Because the complexity of natural language query from map search engine, it is essential to recognize the different constituents of query for purpose of under-standing the semantic intent of such query. But there exists a lot of problems in labeling such query, such as the complexity, the integrity and the ambiguity of that query. So, this paper first presents a method for Chinese address normaliza-tion based on the cascaded conditional fandom fields. The results of experiments indicated that the proposed Chinese address standardization method is effective and the F-score was 81% in open testing.Based on the first work, this paper then propose an approach for the purpose of understanding natural language query intent. Firstly, a hierarchical structure is introduced to represent the semantic intent of natural language query, and then propose a method to map natural language queries to the corresponding semantic intents based on a structured SVM classification model. Experimental results on a manually labeled corpus show that our method achieved a sufficiently high result in term of precision and Fl.