The Study Of Q & A System Based On Maximum Entropy Model Semantic Parsing

Question answering system is a prerequisite for processing. This paper is aimed to study the problem understanding, make understanding the problems accompanying the robot used in the intelligent question answering system of the nursing home, with understanding of the problems elderly, intelligent answers of questions and the search results are back elderly.This paper describes the module in the various steps of understanding question, including the issue of sub-word segmentation, semantic annotation, identify the problem type, extract keywords and keyword expansion, and analyze the status of each stage of the study and introduced the maximum entropy model algorithm thinking of ways to build the models, how to select the feature set of the algorithm. This paper introduces the M-word segmentation algorithm and the C-K algorithm that describe the ideas, procedures, examples, and experiments. The proposed algorithm will be used to accompany the robot in automatic question answering system, and have a good effect.For existing ambiguity word segmentation and uncompleted words, we propose M-segmentation algorithm, the system is small question answering system, based on statistics and reasoning segmentation method is high, so we need improve the mechanical sub-word method, M-segmentation algorithm does not miss any chance of a word may be to ensure the accuracy of segmentation. After the using of M-segmentation algorithm, in order to determine the classification of types more accurately, we proposed C-K algorithm from the semantic and structural understanding of the sentence questions. This method can determine the sentence type classification of the trunk to analyze problems and extract key words. The extended maximum entropy model result as the characteristics of building is the best to build a maximum entropy model.After M-segmentation algorithm is being used, it can get segmentation sequence accurately, as the C-K algorithm to analyze the problem of classification. We can extract the keywords and improve the problem of understanding. In the actual project, we can get a more understanding of the elderly to ask questions accurately, and improve searching the accuracy of the answer and saving search time.