Work

Automating Peer Reviewer Recommendations for Journal Submissions Using a Deep Reviewer Assignment Pipeline

Year: 2023
Type: dissertation
Abstract: To address an exponentially increasing submission rate of scientific manuscripts and the subsequent increase in peer reviewer fatigue, this thesis aims to bring automation to the assignment of highly qualified reviewers to manuscripts submitted to journals. This is achieved by developing a deep learning reviewer assignment pipeline that uses transformer-type neural network models and Natural Language Processing techniques to match the semantics of a peer reviewer’s publication history with the semantics of a submitted manuscript. To do this, the pipeline processes manuscripts and reviewer publication histories, determines the keywords and sentence fragments that best represent the semantics of these corpora, and uses these to create reviewer matches. Currently, the pipeline generates a test dataset containing primarily biomedical literature derived from the open-access Semantic Scholar database but is ultimately data source agnostic. Of the four pre-trained models used in the pipeline, SPECTER was found to generate the greatest number of sufficient matches.
Language: English
Open Access status: green
OpenAlex ID: W7212346410
Cites: 26
Cited by:
Related to: 10