Research On Robot Question Answering System For Freshmen Register In Colleges And Universities

During the register work of freshmen in colleges and universities,there are plenty of i.e.departments and procedures involved,so it is extremely difficult to achieve a unified interpretation of the correlative register affairs.It is urgent that provide a synchronous,real-time and interactive register services for freshmen to improve their resister works efficiently.In this paper,the robot question answering system is constructed to provide a more intelligent new way to answer their questions,which is based on the application of colleges and universities register work and aim to solving the problem.These services can not only improve the quality and efficiency of freshmen's register work but also have positive significance for similar needs and related issues like government offices,transportation and tourism distribution centers.Generally,the robot question answering system combines natural language processing technology with information retrieval technology,according to the openness of the question and answer content,it can be divided to two modules: open-domain type and closed-domain type.This paper studies the robot question answering system of closed-domain type,and focuses on improving the related intelligent algorithms at the same time.The question answering system of university welcome robots constructed in this paper mainly includes four modules: question answer library construction?problem understanding?problem retrieval and answer extraction,and the key technology of this robot development includes text pretreatment?word vector?convolution neural network?long short-term memory and Flask.Among this modules,focus on understanding their meanings.The most important is further abstracted as how to effectively improve the classification accuracy of the original question documents and how to effectively improve the matching rate between the new questions and the questions in the question answering database.Therefore,the author's main job is to improve and implement problem classification model based on two-channel convolution neural network and question similarity computational model based on feature fusion,at the same time,the author complete construction of the welcome robot's question and answer system,in Chongqing University of Technology.The problem classification model based on two-channel convolution neural network mainly uses word vector technology and convolution neural network technology.Amongthem,the word vector models are Skip-Gram and CBOW models,which proposed by Google,the optimization goal of convolution neural networks is to improve the classification accuracy.This paper proposes a two-channel convolution neural network model,which combines contribution degree of part-of-speech with word vector model application scenarios.By introducing part-of-speech probability and word vector weight factors,the problem of sparse and inaccurate features of short text representation in traditional methods is improved.At the same time,the two-channel input mode is selected to enrich the network input information and achieve the optimal effect of classification.The question similarity computational model based on feature fusion is a model,which uses convolution neural networks to extract part features,and uses bi-directional long short-term memory networks to extract global features,then combines the two features for computing similarity between questions.The model complementary combines features of different granularity to solve the problem that the convolution neural network does not consider the relations of words in context,and avoid the problem of gradient disappearance or gradient explosion of traditional recurrent neural network.Thereby improving the accuracy of the model similarity calculation.Experiments show that the retrieval accuracy of the robot question answering system described in this paper is 11% higher than that of the traditional question answering system,which verifies the feasibility of the improved algorithm.