Research On Expert Recommendation Algorithm Oriented To Project Requirements

The expert recommendation algorithm oriented to project requirements studied in this paper can quickly and accurately provide expert recommendation strategies,which has important practical significance.Expert recommendation for project needs is mainly based on the degree of matching between their scientific research achievements and projects,and the scientific research achievements of experts mainly include two categories: academic papers and non-academic papers such as key projects,patents,and Project Fund.Compared with academic papers,Text of non-academic paper results have the characteristics of less vocabulary and sparse semantics,and belong to short text data.In view of the problem that the accuracy of expert recommendation decreases due to the dynamic change of the subject of expert academic papers and the sparse semantics of non-academic paper results,the traditional LDA long text topic model and BTM short text topic model cannot efficiently extract features.Based on this,This paper studies from the perspective of integrating time factors and integrating attention mechanism in these two topic models,and proposes two content-based expert recommendation algorithms.The specific contents are as follows:(1)An expert recommendation algorithm based on dual topic modelAiming at the problem of the decline of recommendation accuracy caused by the dynamic change of the topic of expert academic papers,this paper builds a new expert recommendation algorithm based on the dual topic model.First,the algorithm uses two topic models to extract topic features from expert paper data and project requirement document data respectively(S-TFATM),this model studies the influence of time factors on the weight changes of subject words in expert academic papers by introducing the time forgetting factor,and uses the keywords as the subject labels of the text,which solves the problem that the topic model is difficult to determine the number of topics;on the other hand,for the project requirement document data,the LDA topic model is used to extract the topic feature of the project.Second,use the KL distance to align the topics of experts and project requirements texts;thirdly,based on the topic distribution information of experts and project requirements after topic alignment,calculate and sort the matching degree of experts and projects,and generate a recommendation list according to the calculated value of the matching degree.Finally,the effectiveness of the algorithm is verified by text classification and expert recommendation experiments.(2)An expert recommendation algorithm for word-to-topic model with Selfattention mechanismAiming at the problem of the decrease of recommendation accuracy caused by the sparse semantics of expert non-academic paper results,this paper builds a new expert recommendation algorithm based on the word pair topic model fused with the selfattention mechanism.First,a word-to-topic model(SA-BTM)fused with self-attention mechanism is constructed to extract text features from expert short text data and project requirement document data.Based on the word pair co-occurrence mechanism of the BTM model,the SA-BTM model integrates the semantic similarity value between word pairs and the TF-IDF value of the word into the model as prior knowledge.At the same time,The Self-Attention mechanism is introduced into the SA-BTM model to obtain the contextual semantic information of words in the original text.After obtaining the text features of the expert’s short text and the project text,the two distribution information output by the SA-BTM model is used to calculate the matching degree between the expert and the project,and the recommendation is made according to the matching degree.Finally,the effectiveness of the algorithm is verified by text classification and expert recommendation experiments.