Research On Technical Expert Recommendation System Based On Semantics

Universities and colleges and other scientific research institutions in our country have a large number of experts,but due to information "asymmetry" and lack of demand-expert accurate matching services,companies lack effective ways to quickly and accurately match specific technical requirements.With the rise of big data and artificial intelligence technologies,the development and utilization of scientific and technological achievements and other data resources will effectively solve these problems and have important implications for effectively promoting the transformation of scientific and technological achievements.This paper mainly studies the semantic representation of scientific and technological achievements document vector and scientific and technical experts,as well as the expert's precise recommendation method and system for specific technical requirements.The main research work of this paper includes:1)Proposed an improved document vector representation model based on PV-DM to solve the problem of extracting semantic feature information from different locations in semi-structured documents.The model is trained through the documents of academic research,invention patents,and scientific research projects that represent experts' scientific research capabilities.Semantic vector representations of scientific documents are generated and vector libraries are formed.2)Proposed a technical requirements-oriented expert scoring model.Using the related scientific documents to model the characteristics of experts;taking the technical requirement semantic vector as the center,calculating the radial basis function value of the corresponding document vector as the weight of the result;considering the quality of the document,the publish time,the relevance between the document and requirement,and expert contribution of a variety of features to calculate the score of how much the expert matches the technical requirement;weighted sum of all the scores to obtain the final score;3)Based on the above models,proposes an expert recommendation method for specific technical requirements.This method uses the typical two-stage recommendation in the recommendation system to solve the massive expertrecommendation problem.Firstly,a vector database of scientific documents is constructed based on the aforementioned document vector model.Based on the specific technical requirements,related scientific and technical documents are obtained by using semantic similarity and K nearest neighbor search,and the recommended candidate ranges are demarcated according to the experts corresponding to these documents;then,based on the above-mentioned expert scoring model,the calculation is performed.Get the score of each candidate expert,rank the candidate experts according to the score,and take Top N to complete the expert recommendation.The above research results have been verified experimentally,and we developed a semantic-based scientific and technical expert recommendation system.Real-time recommendation of appropriate experts according to the demand text submitted by the company has been conducted,so that efficient and accurate online matching can be achieved and the efficiency of scientific and technological achievements conversion can be effectively improved.