| Social Question Answering(SQA)websites is a social platform where people share knowledge and disseminate information based on communications.In SQA websites,Q&A is the first requirement.Users can publish their own puzzles and ask questions based on their actual situation.By the mean time,they can also serve as knowledge carriers and share their experience and wisdom with others in the form of answering,thus forming a social platform with interactive attributes.However,existing SQA websites also have problems such as insufficient quality of answers,users receiving excessive information,and long Q&A cycles.Therefore,in order to accurately connect matching questions and answer users,and provide high-quality services,it is necessary to improve the Q&A quality of SQA websites,which requires accurate judgment and analysis of concerns and areas of competence of each user.The purpose of this thesis is to accurately determine the areas of concern and competence of the users in the SQA websites.The question and answer user match more accurately,which needs to find points of interest and the ability of users simultaneously,the more accurate the recommendation of answering questions to the users who have both interests and expertise to provide high quality answers will be.In response to this point,the following researches are carried out:Firstly,in view of the fact that the existing research is generally an independent study of interest and ability,an NLP-based user interest and capacity analysis method N-UICAM is proposed.This method is based on NLP's semantic analysis of texts,eliminating the need to mine large amounts of data as a training set.It also adds data pre-processing modules for possible short-text and garbage-recovery situations,through which can filter the answers and combine question and answer lists according to given rules.Using the degree value calculation method to extract features of the user's short-term interest,long-term interest and ability,and finally the analysis results are presented as a coarse-grained secondary classification tree and fine-grained label item vectors,describing interest and capacity characteristics from a richer perspective.Secondly,the algorithm for calculating the degree of feature words of different benchmarks is proposed.Based on the user's short-term interest,long-term interest,and capability characteristics,the values are mapped to a linear scale in combination with a normalized function.Then,according to their respective data characteristics,the degree values are calculated based on the time decay,time window,and vote approval number,respectively.User dynamic update method of short-term and long-term interest feature items is also introduced.Finally,through the user data of the social question and answer website,the accuracy of the results of the algorithm itself is evaluated by using the own labels of question as the verification standard,including the coincidence degree of the classification tree and the accuracy of the label item vectors.Then compare it with the existing algorithms,and calculate the recall rate and average absolute error of the classification tree,and the performance index calculated by using the tag item as the recommendation ranking,verifying that the method proposed in this thesis has better accuracy and performance advantages in user interest and capability analysis. |
