Research On Intelligent Search For Domain

With the explosive growth of data in domains in the information age,the efficient knowledge search and discovery has been increasingly urgent needed.In many domain-oriented search scenes,the generic search engines are unhelpful because of data confidentiality,so the search functions are often based on conditional matching of database queries,and the semantic information of the query captured by them is limited.Therefore,there is great significance to obtain the semantics of the query and realize domain-oriented intelligent search.When faced to the query entity linking,which is the key problem in intelligent search,the traditional methods are usually to recognize the named entity mentions in the query first,and then they are linked to entities in knowledge graph.These methods require a large amount of work in data processing(like entity mentions tagging)and feature selection.Furthermore,they are easy to cause cumulative errors and reduce the linking effect.In order to address these issues,this thesis proposes an Attention Mechanism based model for entity linking.In this model,Long Short-Term Memory Network is used to encode the questions.Then through the decoder process by Attention Mechanism,entity mentions and disambiguation information are generated as outputs.Finally,these outputs are linked to the entities in knowledge base.The experiments are conducted on a dataset of questions and entities,which are about products in automotive domain.The results show that the proposed model only use rare contextual information,but can obtain good results.This idea provides a novel solution for entity linking.Based on the Knowledge Graph,Text Classification,Entity Linking,Representation Learning and other technologies in Natural Language Processing,this thesis proposes an intelligent search framework for domain after practices and summary.The framework consists of network layer,query classification layer,entity linking layer,service layer and database layer.Briefly introduction of each layer's functions and optional schemes are given,understanding the query's semantics can be customized and implemented on a certain degree by choosing different schemes at each layer.The schemes of query classification and result document sorting are also compared,and then a simple knowledge graph is constructed by the use of data obtained from the network.Based on the proposed query entity linking model and domain-oriented intelligent search framework,an intelligent search system for the automobile domain is implemented by using Web application technologies,it proves that the entity linking model and the domain-oriented intelligent search framework are feasible.