Research On Paper-Reviewer Recommendation Based On Research Field Label

In academic research,more and more researchers are submitting academic manuscripts to journals and conferences.For journal or conference organizations,how to efficiently and accurately recommend appropriate reviewers to review papers is very important.The traditional reviewer recommendation is generally decided by the academic journal editors or academic conference organizes according to the research field of the reviewer and the relevant information of the paper.Obviously,this is unscientific,and it is very time consuming and requires a lot of manual operations.However,many studies only use the content information of the papers or reviewers as research objects,ignoring their research field label information.Research field label information can guide the learning process of paper-reviewer recommendation.Thus,this dissertation mainly studies an automatic,efficient and accurate paper-review recommendation method.Since the reviewer information and the paper information are basically composed of text content,in order to explore the relationship between the reviewer and the manuscript paper,it is necessary to study the relationship between the text content of the two.Obviously,discrete text information contains complex semantic information.There is a complex semantic gap between different text content.In order to obtain the relevance of the research field between the two,it is necessary to eliminate the semantic gap between the two and explore its semantic relevance.In order to ensure the quality of the paper review,it is necessary to ensure that the reviewer has a high correlation with the research field of the manuscript paper.Therefore,it is very important to obtain the research correlation between the two.On the basis of analyzing the textual information of reviewers and papers,more useful information is introduced to explore deeper domain relevance,which helps to accurately recommend appropriate reviewers for manuscript papers.This dissertation uses the research field label information of the paper and the reviewer to supervise and guide the process of the reviewer recommendation,and explore the complex semantic relationship between the reviewer and the paper to obtain accurate research relevance.The main research work of this dissertation includes:1.It firstly introduces the background knowledge recommended by the reviewer of the paper,and fully studies the research status and basic method theory of the paper-reviewer recommendation.In addition,the dissertation mainly analyzes the difficulties and challenges of this research,and proposes corresponding research methods for these challenges.In order to eliminate the semantic gap between reviewers and papers,and to make full use of the existing information to mine research field correlation between papers and reviewers.Two kinds of classification methods based on research field labels are proposed.2.Existing methods for obtaining research field relevance between reviewers and papers do not largely eliminate the semantic gap between the two,and they do not make full use of their field label information.In response to this problem,this study proposes a new classification model WMD-CCA based on the research field label,and transforms the paper-reviewer recommendation problem into a classification issue.The method first transforms the reviewer's important text information(keywords)into a word vector integrates the semantic information.Based on the distance between the word vectors,the minimum distance between the reviewer and the paper is optimized,and the minimum distance is regarded as the research field relationship.Based on this relationship,combined with the research field label of the paper,the constructive covering algorithm is used to adaptively predict the reviewer field labels.Finally,experiments were performed on several datasets,and the results suggest that this study has certain applicability in the review of experts.3.Due to the emergence of interdisciplinary fields,many reviewers or papers involve multiple research field labels.Recommending the reviewers who involve in multiple fields increases the difficulty of recommendation than individual field.In addition,obtaining features that cover multiple research fields is important for the relevance of mining reviewers and papers.In response to the above problems,this study proposes a new multi-label classification method Hiepar-MLC,based on the research field label information,the paper-reviewer recommendation problem is transformed into multi-label classification problem.The method first introduces a hierarchical and transparent feature representation method,and it learns the inherent hierarchical structure information in the text.It uses the attention mechanism to focus on the components related to the research field.Based on this feature representation and research field label information,a multi-label classification method is used to predict multiple field labels of a paper.Finally,the reviewers and papers are obtained from ACM Digital Library,and the experiment shows that the multi-label method has certain applicability in the paper reviewer recommendation.