Predicting Editor Activity for Open-Access Mega-Journals Using Fixed-Size Abstract Embeddings

During, Joël (2022) Predicting Editor Activity for Open-Access Mega-Journals Using Fixed-Size Abstract Embeddings. Bachelor's Thesis, Artificial Intelligence.

	Text toestemming.pdf Restricted to Registered users only Download (125kB)
Preview	Text bAI_2022_DuringJDG.pdf Download (634kB) | Preview

Abstract

Open-access mega journals (OAMJs) can make thousands of publications per month. Due to their broad scope and large pool of editorial board members, finding a suitable editor for a new manuscript is a complex problem. This work examines editors' responses to invitations to edit new manuscripts for one OAMJ and this dataset's potential to be used in an editorial board member recommender system, which aims to automate the finding of suitable editors. The main challenge in this data is that only a few editors are invited for each manuscript, leading to a highly sparse dataset with limited data per editor. Different NLP techniques such as Word2Vec, SciBERT, and Doc2Vec are used to transform the abstracts of new manuscripts into fixed-size embeddings that various machine learning methods can use. We present an experiment to evaluate the performance of these methods in predicting an editor's response to invitations for new manuscripts using a baseline classifier or support vector classification. We find that the embedding methods presented here provide an improvement over random classification. Word2Vec produces the best results in our experiment, although its performance is similar to that of SciBERT and Doc2Vec. An additional experiment examines whether similar editors can be clustered to reduce the problem of limited data availability per editor. None of the clustering methods presented here provide an improvement over the method without clustering.

Item Type:	Thesis (Bachelor's Thesis)
Supervisor name:	Spenader, J.K.
Degree programme:	Artificial Intelligence
Thesis type:	Bachelor's Thesis
Language:	English
Date Deposited:	15 Jul 2022 11:36
Last Modified:	26 Jul 2022 07:09
URI:	https://fse.studenttheses.ub.rug.nl/id/eprint/27911

Actions (login required)

View Item