Recommender Systems for Research Data Management

Zenodo (CERN European Organization for Nuclear Research)

Abstract

This bachelor thesis is about creating a recommender system for researchers. On the basis of a research project, other similar research projects will be proposed. The goal is to improve the collaboration between researchers. The data comes from Coscine, a research data management platform. Researchers can upload their files of a research project on Coscine. Thereby they can fill out a metadata schema. The schema depends on their research field. The recommendations are based on this explicit metadata of the files in a project. Because there is no empirical data to derive a relationship between the projects, a content based recommendation system is proposed. The unstructured data is processed with a pretrained natural language processing model, based on BERT. The similarity between projects is the distance between the centroids of all files of a project.

Authors 1

1. Markus Klein corresponding Aachen

   FH Aachen

   Affiliation as printed

   FH Aachen