Introducing the EOSC Recommender System

1. Bridging the information gap

A major challenge facing researchers today is information overload. While navigating large amounts of data may come with the job, researchers can get inundated with too much of the same information or incomplete information.

Later this year, the EOSC Future project is releasing the EOSC Recommender System (RS). This AI-based tool will be capable of processing all of the content and resources currently available on the EOSC marketplace (e.g. publications, datasets, software, workflows, related information, etc.). The fully functional software system is being built entirely from scratch, with a view to bolstering the EOSC platform as a complete virtual space for experimentation and research.

2. A recommender system for EOSC

Recommender tools are a cornerstone of modern-day internet systems. These tools typically rely on complex algorithms, machine and deep learning to collect information to match users with a particular set of options that meets their preferences and particular needs. They are highly prevalent in our daily lives – shaping the way we shop, order food, watch movies and more.

Within the scope of the EOSC Future project, developing an EOSC RS will foster discovery and access to a variety of research data resources and support on the EOSC platform based on a user’s profile, previous platform interactions, domain expertise, professional connections/engagements and interests.

Key components

The EOSC RS dashboard is still in development. Nevertheless, we have a picture of how users will be able to engage with the system.

For researchers, the EOSC RS will offer 3 main paths for a personalised view of EOSC resources:

  • interactive workflows, in the form of a chatbot or another online AI/ML module, that appear at the resource request stage to guide and advise on the available options, constraints, solutions
  • smart search options, with links to related resources based on the keywords used to search for data, services or software
  • personalised recommendations for services and resources based on a user’s previous activity, connections and other profile features.

A comparable functionality for providers and other EOSC stakeholders will be based on information collected in the onboarding process, in addition to feedback from researchers using providers’ services via EOSC.

The EOSC RS will be able to:

  • learn about and match relevant content extracted from the EOSC platform (EOSC-Core: resource registry, marketplace, ordering system, etc.)
  • filter out information irrelevant to specific requests
  • drive service engagement based on a user’s specific interests, research area, previous searches, orders and other ‘behavioural’ data
  • collect and analyse content with user consent, in line with GDPR regulations.

A novel recommendation

For the user, the EOSC RS will help to speed up the identification (or ‘findability’) of publications, datasets and other resources available on EOSC. What is particularly distinctive about this system is the fact that it will not only tailor content available through the EOSC platform but also push users to provide input on this content. For each recommendation, the user will be able to indicate whether it was useful or not (in other words, user feedback on a simple ‘like-dislike’ scale).

User input is essential for the EOSC RS to be able to truly offer intelligent discovery and smart recommendations. The intelligent discovery of EOSC resources and smart recommendations, based on users' feedback and preferences, will significantly enhance user experience and engagement with the platform as a source of information, not to mention as a virtual community. 

3. Looking to the next phase

The EOSC Future project has already catalogued existing data resources, in addition to defining internal and technical interfaces in the EOSC marketplace.

With the above objectives in mind, the EOSC RS is on track to come out in late 2022. In parallel with the release, users will have the chance to review and test out the EOSC RS concept, offering critical insight and input into the EOSC RS interface and components. In turn, this input will improve the overall user experience and uptake of EOSC resources among different research communities.

Below is a visual representation of how the EOSC RS will feature in the EOSC platform.

24 March 2022