LASIGE’s PhD student Márcia Barros is presenting her research work at the second PhD in Informatics Seminar of 2019/2020, on this Thursday, May 14 at 12h.
The PhD in Informatics Seminars are open to PhD students, faculty, MSc and BSc students and are an excellent opportunity for cross-disciplinary development through presentation and discussion of edge research going on in Ciências ULisboa Departments of Informatics.
Title: Recommender systems for scientific items: a Hybrid semantic recommender algorithm for chemical compounds
Speaker: Márcia Barros (LASIGE/DI)
When: May 14 (Thursday) at 12:00
Where: Zoom videoconference (for more information access: https://moodle.ciencias.ulisboa.pt/course/view.php?id=2228#section-2)
Abstract:
Recommender systems have been widely used in fields such as movies, music and online stores, however, they are poorly applied to scientific fields. We identified as one of the main challenges for using recommender systems in scientific fields, the lack of datasets suitable for evaluating recommender algorithms. In previous work, we overcame this challenge by developing a new methodology called LIBRETTI (LIterature Based RecommEndaTion of scienTific Items). LIBRETTI allows the creation of standard datasets of implicit feedback with the format of <user, item, rating>, where the users are authors from research papers, the items are scientific items, such as chemical compounds, genes, or phenotypes, and the ratings are the number of papers an author wrote about an item. We have now as case studies a dataset in the field of Astronomy, and a dataset in the field of Chemistry. With these dataset available, it is now possible to start testing and developing new recommender algorithms. Thus, this seminar will present a Hybrid recommender model for recommending Chemical Compounds. The Hybrid model integrates state-of-the-art collaborative-filtering algorithms for implicit feedback and a new content-based algorithm, based on the semantic similarity between the Chemical Compounds in the ChEBI ontology. This model was evaluated using the dataset of implicit feedback of Chemical Compounds, CheRM, created through LIBRETTI.