Research On Expert Recommendation System Based On Text Analysis

The online Q&A community has become an important platform for people to solve problems and gain knowledge.However,when users ask questions in the community,they are limited by their own language habits,which leads to inaccurate description of the problem,and it is often difficult to obtain high-quality answers.Therefore,this paper proposes an expert recommendation system based on text analysis.The system can model the problems raised by users,and the improved Labeled-LDA model and BERT model can be used to model some experts who are most suitable to answer this question.In the expert text classification part,considering the excessive useless word interference of the original Labeled-LDA model,it is easy to cause the defects of the theme.This paper combines the TF-IDF algorithm,the chi-square test algorithm and the special noun library of the specific domain on the original model structure.Improve,increase the weight of feature words that can represent a topic,and improve the classification performance of the model.In the expert recommendation section,the paper first classifies the questions to be answered through the improved Labeled-LDA model to know the domain category to which the problem belongs,and then extracts the expert information text to be matched from the list of experts under the category.Finally,the BERT language model is used to model the problem to be answered and the expert information to be matched,and the similarity between the two vectors is calculated.Experts with higher similarity coefficient are recommended to the questioner.This paper aims to use the above method to implement an expert recommendation system for a childcare network user.The experiment part first obtains the user's question and answer information of the childcare network in the past ten years through the Scrapy-Redis distributed crawler system,and obtains the historical answer set of each expert user after preprocessing.Then use the improved Labeled-LDA model to model the text data of all experts,get the probability distribution of each expert's field of interest and the lexical probability distribution in each field to form a list of experts.Finally,through BERT,the problem to be answered is matched with the expert information,and the expert recommendation is completed.The experimental results show that the recommended accuracy of the system is significantly improved compared with the recommended accuracy of other algorithms,which proves the feasibility of the system and provides a new idea for the expert recommendation field.