Research And Application Of Topic Expansion Technology For Intelligence Service

Topic expansion of intelligence service is to expand the customer's intelligence needs, and use the expanded topic words as a supplement and explanation of the original topic. It is one of the key technologies to ensure the comprehensive intelligence acquisition. Therefore, after in-depth analysis of the service scene, this paper constructs Word Embedding from the aspects of document co-occurrence words and semantically similar words to improve the comprehensiveness and accuracy of topic expansion. To improve the recommendation effect by the user's feedback for recommendation words, make the recommendation results gradually approach approximate the users' needs and improve artificial expansion's efficiency, this paper uses interactive expansion method to help the users expand the topic. The main work of this paper includes the following aspects:In this paper, we propose a topic expansion technology based on Word Embedding for intelligence service. This paper gets two different kinds of Word Embedding models by using vector space model and skip-gram model from the document co-occurrence words and semantically similar words, and constructs two topic expansion methods according to the two kinds of Word Embedding models, and linear combination is used to combine these two topic expansion methods. We compute the divergence between the similarity of candidate words with the related words and with the unrelated words, filter the candidate words by the similarity divergence, and update related words set and unrelated words set by the user's confirmation for recommendation words. The experimental results show that the proposed method achieves good performance on topic expansion.In this paper, we analyze the application scene of topic expansion, and study the corresponding application program for the problem of application scene. Firstly, we analyze the topic expansion process and the artificial expansion mode, and summarize the problems of the expansion process and the law of the expansion model. Then we analyze the case and the mode of the artificial expansion. By summarizing the case, the topic expansion technology we proposed is better in the intelligence service of consumption evaluation and public opinion information. Finally, by using expansion technology based on Word Embedding, we improve the topic expansion process according to the problems and the law in the application scenarios.This paper designs and implements the topic expansion system. The document co-occurrence words, semantically similar words and the results of linear combination are provided to users in this system, and the system recommends candidate words interactively by the user's feedback.