Research On College Student-oriented Education Secretary Question Answering System

In the modern society,the network has become a main consulting medium,because of the diffusive computer and explosive information.At present,college students usually consult somethings including study,life and play through spot,telephone and online etc.However,the amount of involved questions are always discrete,large,repeated.Therefore,How to use internet technology to establish a scientific and reasonable QA service system,it becomes significant.And it also avoids the teacher answer the cumbersome questions.This is the source of solving these thesis.The paper mainly research on college student-oriented education secretary QA system,it will achieve answer students questions partly,and provide students with a convenient and quick way to get answer.In order to improve the accurate rate,the paper takes the students' dairy management as the relatively closed case to study,and the case can be extended.In the secretary QA system,students can ask the education secretary questions through the ordinary text,and the teaching secretary will return questioner a more accurate answer or other ways to get answer,so it improves the efficiency of answering question.The hot technologies involved in the paper are mainly two fields: the natural language processing and knowledge base construction.And the study of natural language processing is the core concern,that is,how to improve the classify efficiency of the origin text data and how to improve the match rate between user's question and the one in the knowledge base.Therefore it is the focus of the study.And the paper shows 5 parts:(1)Introduce research status:Paper introduced the present research on QA system and QA robot,and it also introduced text representation and word semantic similarity algorithms based on hownet,and it also introduced text similarity algorithms.Besides,it analyzed the characteristics of the collected text data and data preprocessing ways.(2)Modifying text representation to improve the classify efficiency of the origin text data:Paper modified the idftf-algorithm based on Vector Space Mode.Considering the algorithm only considers the relationship between the term feature and the document,the paper absorbed the square factor and lexical factor to cover the shortage of origin algorithm ignoring the feature words classify well.And the improved algorithm is verified validity by using the origin text data set.(3)Modifying word semantic similarity algorithms to improve the match rate between user's question and the one in the knowledge base.The paper finds out that the current algorithm doesn't consider the different importance of distance and depth,and it will bring out negative effect.So the paper proposes constraint depth factor to improve the original algorithm,and it was verified validity from the theoretical and experimental aspects.(4)Further modifying word semantic similarity algorithms:the paper put forward a new word similarity calculation method of computing the combination terms of corresponding to the first basic meaning of the highest similarity to eliminate the low ones,Besides,it absorbs weighted sum to modify the original algorithm.And it is verified validity to improve the accuracy rate of return the accurate answer by using the student's test set and candidate question data set.(5)Research the education secretary QA system:Finally,using the prepared college management text date,the paper established the college management-oriented and narrow domain knowledge base based on hownet,and gives the example explanation.and the base apply to the education secretary QA system.