Domain Intelligent Q&A user intention recognition based on keyword separation

In view of the disadvantages of current intelligent Q&A user intention recognition technology, which can't make good use of the detailed features in user questions, this paper proposes a layered method to identify user intention. This method first identifies the keywords in the user problem as the user's first level intention, then distinguishes the sentence pattern of the user problem according to the number of keywords, and determines whether to identify the second level intention according to the sentence pattern of the problem. Then, the key word features are fused to identify the user's problem type, and the result of problem type recognition is regarded as the second layer of user's intention. Finally, the user's two layers of intentions are integrated and regarded as the user's overall intentions, and retrieved in the Knowledge Graph. Experiments show that the F1 value of user problem type recognition is increased by 6% after keyword features are added. The method of keyword separation is used to identify the user's intention in question and answer, which separates the two layers of user's intention. For the problem sentence pattern that is not necessary to recognize the second layer of user's intention, the judgment of user's intention is no longer performed, which reduces the complexity of user's intention recognition, and opens up space for the case that keywords need to be processed. In addition, keyword features are added to the second layer of user intention recognition model, which makes better use of the detailed features in user question statements and improves the effect of user intention recognition.