Research On Intelligent Retrieval Method Of Economic Crime Based On Topic Model

With the expansion of public security economic investigation business,the increasing number of historical data and the extensive use of artificial intelligence technology,the intelligent retrieval mode for economic crime information has become an urgent demand in the field of economic crime investigation.At present,most search engines mostly use keyword matching mode,without considering the inherent semantics of the text,resulting in low accuracy and recall of search results.In order to solve this problem,topic model has been widely used in information retrieval because of its advantage of mining hidden concepts in documents.However,the following problems still exist: First,its unsupervised learning form leads to unclear explanation of generated topics;Second,the topic model is more suitable for processing long text and is not suitable for query statements with fewer features.In order to solve the above problems,starting from the semantic characteristics of intelligent retrieval,this paper proposes an intelligent retrieval method based on topic model.This method researches from the two angles of document retrieval model and user's question,and uses topic model combined with clustering algorithm to mine the potential semantics of documents and construct document index.The ontology technology is applied to deal with the user's question sentences,and the retrieval formula is standardized.Firstly,the expert knowledge in the field of economic crime is selected as the prior knowledge of the theme model,which improves the defect that the theme of the unsupervised theme model deviates from the actual semantics,realizes the semi-supervised theme modeling,and generates the document theme label conforming to the characteristics of economic crime.Secondly,the clustering method of documents is studied,and the traditional k-means algorithm is improved by combining Bagging's integrated learning idea to solve the problem of sparse topic distribution of text generated by topic model training of original corpus.Thirdly,the domain ontology of economic crimes is constructed to describe the logical relationships among entities and to complete the persistence of the ontology.The semantic and logical relationships of query sentences are identified in the retrieval stage,the retrieval conditions of users are reconstructed,and the real retrieval requirements of users are discovered.Finally,according to the comprehensive matching rate between the query and the document,the result list is sorted according to the similarity to improve the retrieval accuracy and recall rate.On the basis of intelligent retrieval scheme based on topic model,an intelligent retrieval system for economic crimes is developed.Taking the data of an economic crime field as anexample,the effect and value of the improved topic model and intelligent retrieval scheme in this paper are verified.