Research On Expert Retrieval Based On Topic Model

At present,the method of expert retrieval can not analyze the semantic structure information in the text.Mining the relationship between author and word in the document by the author topic model,establishing the relationship between document,author,topic and word,breaking the assumption of the independence of expert and query in document model.according to expert retrieval of the author topic model cluster result,using an array of three elements to represent each word in a new document,According to the relationship between expert and document build candidate supporting document,in support of the document,the author calculates the co-occurrence of experts and words and the relationship between the author and the document.According to the experts and the relevance of the query to return to the list of experts,the use of JS distance method to find similar experts with an expert.choose 1557 papers from the ACL and EMNLP CONLL meeting as experimental data sets,through comparison between expert retrieval based on author topic model and document model in precision,recall rate,MAP value and retrieval time,found that expert retrieval based on author topic model can effectively improve the performance of expert retrieval.