Helping researchers to find new articles and opportunities
Benj Pettit works at Mendeley and works on text and data mining tools that help researchers to find new articles, collaborators etc. One of the special things about the Mendeley catalogue is that it is formed in a crowdsourced way.
“I am Benj Pettit, and I work on Mendeley, at Elsevier. And my interest in text and data mining is for making tools that help researchers discover articles, collaborators, funding opportunities and other entities that are out there, to help them with their research.
So, we’re making an online ecosystem for this kind of discovery and networking. And I’m specifically working on recommender systems for that. So, somebody may be has been using Mendeley as a reference manager, and we can harness the power of all of those researcher libraries to give them recommendations for articles to fill in the gaps in their knowledge. Or may be suggest something new and serendipitous.
One of the things about the Mendeley catalogue is that it actually arises in a crowdsourced way from people grabbing articles by using the browser plugin or uploading pdfs and skimming metadata from those. So it’s actually quite a messy dataset compared to say the core datasets that Elsevier maintains as a publisher. And for that reason, we tend to rely on the user-article interaction data as a strength, but not necessarily trust all the content of the text fields. So that’s why we use a lot of collaborative filtering approaches, which are very popular in recommender systems, and don’t necessarily depend on the text fields themselves. That also lets us make interdisciplinary recommendations, whereas a very strict text classifier may draw a line between a researcher and an article that’s of interest to them in a different discipline.”