17 April 2012 by William

Changing how research is done is a very big task, and we can’t do it alone. We’re particularly appreciative of our development partners who are working with us to chip away at the problems hindering research efficiency today. One problem is sifting through the volume of search results to find the most important and timely results. Jason Priem of Total Impact is working on this problem at the School for Information Science at the University of North Carolina. He and his colleagues are doing a study to determine if scholarly search can be improved by personalizing search results based on the previous reading history of the scholar — that’s where you come in. If you’re willing to share your academic search and paper reading history to improve science, sign up for his study!
Here’s how Jason describes the study:

Transforming scholarly search with Mendeley: participants needed

Once upon a time, scholars mostly accessed the literature by reading select journals. Today, though, we increasingly turn to searches in expansive, flat indexes like Google Scholar, and that’s a win. But these searches could be better!

For one thing, academic search engines make heavy use of citation as ranking signal. This is effective, but citations can take years to accumulate, and so rankings can be stale. Academic search engines also deliver the same results to everyone, which can present a lot of irrelevant results for an individual user. For instance, a Google Scholar search for “status updates” gets you articles about social media, wireless sensor networks, and Belizean bird species. You probably just want one of these.

One potential answer to both these problems is personalizing scholarly search results using Mendeley libraries. Looking at someone’s Mendeley collection, you could take a pretty good guess as to whether they’d prefer to see an article about the bird species Forster’s Tern or one about gateway fault tolerance. As an added bonus, Mendeley collections tend to be pretty up-to-date compared to publication or citation histories.

Personalizing search has a long and checkered history. But using reference libraries this way is, as far as we know, pretty novel. It’s also exciting — since the domain (scholarship) is relatively narrow, we might see better performance than general-purpose personalization.

Given this potential, we are planning a study to see if we can in fact improve scholarly search rankings by looking at users’ personal Mendeley libraries. To do this, we need some participants. Specifically, we’re looking for folks who will let us record their academic searches (and only the academic searches) for a month, and let us have a look at their Mendeley libraries via the Mendeley API. In return, we’re offering $50 USD and the satisfaction of helping science!

To participate, please send an email to priem@email.unc.edu. We will follow up with instructions on how to proceed. To learn more about the study, visit the study website.

Thank you and watch this spot for our results in the fall!

Tags: , , , ,

One Response to “Transforming Scholarly Search with Mendeley: Your help needed!”

  1. Steven Beuford Says:

    I agree that Google should keep up what they’re doing, and give the majority of searchers what they want for a keyword. If a person wants status updates they probably DO mean Facebook/Twitter status updates, and Google should give exactly that.