Research And Implementation Of Information Retrieval System Oriented To Automobile Field

With the development of Internet technology,how to quickly retrieve the desired content in the face of the huge amount of automobile data in the network is one of the important research directions of information retrieval technology.On the other hand,the application of automobile knowledge graph to automobile information retrieval technology is a hot spot in the current academic and industrial circles.Therefore,this thesis proposes a solution for the construction of automobile knowledge graph and applies it to the automobile information retrieval system to complete the design and implementation of the automobile information retrieval system.The research content mainly includes the following three aspects:(1)Research how to construct the knowledge graph in the automotive domain,including ontology construction,knowledge extraction,and knowledge storage.The Scrapy crawler framework is used to obtain the data in the automotive field,and the knowledge map mode layer is constructed according to the content of the automotive data.In order to obtain knowledge data from unstructured data,the MacBERT-BiLSTMCRF is used for car entity recognition and the Attention-BiLSTM is used to extract the relational content in the automotive field,and finally the construction of the knowledge map in the automotive field is completed.(2)Study how to construct the function of automobile information retrieval,and respectively realize two technical routes about automobile knowledge retrieval and automobile text retrieval.The entity link uses the name dictionary to filter the candidate entities,and combines the improved Jaccard coefficient to sort the candidate entities to obtain the query entity.Relation mapping uses TextCNN technology suitable for short text classification to convert user retrieval questions into specific query relations.For the retrieval of automobile text data,this paper uses TextRank to extract text summaries,and uses BM25 to calculate text similarity for content matching.Finally,the LambdaMART sorting learning method is used to present the obtained text result set to the user in a reasonable order.(3)Build an information retrieval platform in the automotive field.Based on the above research content,combine the data content in the automotive field and analyze the needs of different roles to achieve user login registration,authority authentication,Information management,as well as comprehensive query of car data and management of car text data are multiple functional modules.The availability of the system is verified by functional and performance tests.