Already for some time now, we closely collaborate with our partners from RESEARCHconnect for all fund search related offerings to our customers. This enabled universities like Utrecht, Coimbra and Maastricht to search for funding opportunities on the basis of the publication profile of their researchers, or simply by pasting an abstract/research summary in the search bar. However, a couple of months ago we started experimenting with personalized search algorithms.
We observed that in many cases people don’t have the time to create the perfect search query, no matter whether they are searching for funding opportunities or searching in Google. This observation, in combination with the tendency to only look at the first few results (again, how often do you look beyond page one of the Google Search results?), we started looking for different ways of searching.
We figured we wanted to let the feedback of the user influence the results of the matches the algorithm creates. In the above screenshot you see a potential relevant funding opportunity for the researchers and the opportunity to like or dislike this opportunity. All likes move into the ‘your matches’ list that you can download but also feed the algorithm with the characteristics of that specific funding opportunity that you apparently like. All the dislikes feed the algorithm with the characteristics of the call that you don’t like for your funding opportunities. All these likes and dislikes lead to the development of your ‘matching model’ that is unique to you. The model can be saved to continue another time, but…the model can also be used to search for you.
Once you have created you own personalized search algorithm, this algorithm can do a periodic search for you. Since it now knows what type of funding you like and dislike it can generate a list of potential matches, for instance every quarter, from the RESEARCHconnect database that match with your personal preferences!
This is fundamentally different from current e-mail services who keep you up to date about developments of a specific calltopic that you liked. Our e-mails actually perform a whole new search for you, based on your preferences. So it’s possible that funding opportunities are found that you weren’t even aware of, without spending any time on performing a search. And if you change your interest, research topic or something else? You can simply create a second, third or entirely new search algorithm!
Currently, we are testing this with the University Medical Hospital Groningen where participants in the pilot not only train the model and receive a periodic e-mail. In these e-mails, they will also tell the system which matches were good, and which were bad so the personalized algorithm keeps learning from their input.
Are you, just like the UMCG, also interested in the opportunity to create personalized search algorithms with the researchers in your institution? Please feel free to reach out to us to learn more!