Periodic Reporting for period 1 - ARTIFI (Accelerating Research Through Individualized Funding Intelligence)
Período documentado: 2020-09-01 hasta 2021-09-30
Researchers are under ever increasing pressure to compete for research funding. More and more time is spent on writing lengthy proposals with dwindling success rates. IDfuse is a Dutch SME that wants to help researchers identify specialised, less well known funding opportunities. In 2016, IDfuse has launched impacter.eu a successful online tool that runs automated analyses on drafted grant applications, allowing academic researchers to strengthen their grant proposals. IDfuse envisions to go a step further in its service by providing researchers with alternative funding opportunities based on the reviewed proposal and the scientific profile of the researcher, thereby accelerating research through individualized funding intelligence (ARTIFI). The envisioned product should essentially consist of a machine learning engine that is able to actively learn from interacting with the individual researcher. Moreover, in order to be able to present researchers with opportunities that go beyond what they would typically think of, IDfuse envisions that the engine should be able to expand the researcher profile with insights that can be extracted from open source research materials, including abstracts and open access scientific publications.
Work in this project has led to the development of a new Active Learning system, now both implemented in the impacter.eu tool as delivered directly in the context of our partners' application, ResearchConnect. The Active Learning technology uses smart matchmaking algorithms to determine a first set of relevant funding opportunities, and then learns from the user evaluation of the attractiveness of the funding opportunities. The results of the matchmaking exercise can be saved as a profile, and used as a basis for alerts when new matching funding opportunities emerge.
The application of the algorithm combines the state-of-the-art technology for machine text comprehension with intuitive usage and design to arrive at a quick and efficient way of navigating through a forest of funding opportunities. The application of Active Learning on top of smart document vector representations presents a novel way of using AI methodology as a decision support tool. With this approach: the researcher stays in control of the process, with the machine merely providing the suggestions. A better fit between academic and funding opportunity will improve the effectiveness of both the researchers writing the application as well as the funding instrument, getting the most relevant expertise for the stated policy goal.
In addition, there's substantial potential to apply this technology to other matchmaking questions. An example is journalism, where news feeds could be linked to expert profiles in order to characterize and reflect on important societal developments.
In addition, there's substantial potential to apply this technology to other matchmaking questions. An example is journalism, where news feeds could be linked to expert profiles in order to characterize and reflect on important societal developments.