TRIPLE Deliverable: D5.6 Report on the Discovery System

In this deliverable, we present an innovative discovery system that brings together textual and visual interface components. This discovery system is based on Head Start, an open source framework that has been developed by Open Knowledge Maps over more than six years. Head Start provides interactive, web-based overviews of research topics, authors, and projects. The overviews are based related resources, for example, the most relevant publications on a topic, or the articles and datasets published by an author or project.
Head Start is one of the innovative services that is integrated with GoTriple in order to support discovery. In comparison to a pure list-based approach, Head Start provides a bird’s eye view of the contents by way of visualisation. Head Start highlights connections between research outputs using clustering. This helps users to sort the relevant from the irrelevant with respect to their research question or information need. Head Start also supports users to come to know the concepts related to a topic, and it enables the easy identification of open access content.

These features are very relevant in the SSH domain, as supported by the user research carried out in this project. Among the issues that Head Start alleviates is getting and keeping an overview of a field and obtaining a better understanding of the terminology related to a topic. Head Start gives users a starting point, especially those who are not yet experts in a field. In this way, TRIPLE can also attract new audiences beyond researchers, including students, practitioners, journalists, and citizens. Finally, using some of the machine learning algorithms included in Head Start, such as on-the-fly deduplication and missing subject inference, GoTriple can provide an engaging search experience even if the metadata is messy or incomplete.

With Head Start, GoTriple can provide a popular discovery option that has become a mainstay for users from around the world. Since 2016, a million users have used the visual search on openknowledgemaps.org and more than 400,000 knowledge maps have been created so far.

Head Start is a tightly integrated innovative service in GoTriple. The Head Start interfaces can be accessed by users as part of the core platform. The integration is modeled after popular web search engines that link to additional services and interfaces directly within their search results page of GoTriple. On the backend, Head Start has a direct connection to the ElasticSearch index of GoTriple, where GoTriple’s data is stored.

In this deliverable, we discuss Head Start, its architecture as well as its integration in GoTriple in detail. Furthermore, we describe the work done in the associated task T5.6, which includes a refactoring of the Head Start backend, the development of a standalone search flow package, the integration of Head Start into GoTriple, adapting the software to the GoTriple data, and improvements and adaptations to the machine learning pipeline.