»In TILDE, we aim to support the general public in their health information by providing trustworthy and balanced search responses – relevant information is thus made accessible to a population with diverse information literacy, reading and language skills.«

Dr. Mark Kröll
Key Researcher, Group Leader “Natural Language Processing” Know-Center, Austria

Project TILDE – Trustworthy Access to Knowledge from the Indexed Web

The TILDE project introduces an advanced AI-driven component for enhancing the accessibility, efficiency and inclusivity of web search technologies. By leveraging recent advancements in NLP and LLMs, we aim to redefine how concepts, relationships, and data insights are aggregated, prototypically demonstrated by a use case in the health domain. Our methodologies, particularly around the development of a Knowledge Graph and the application of RAG techniques, will serve as a benchmark for future vertical web search technologies. This approach contributes to increasing the accuracy and trustworthiness of search results while ensuring a reduction in biases, promoting a more balanced representation of information. The creation of an interactive web application for data exploration, analysis and reporting aims to showcase how AI technologies (i) can facilitate access to information in an objective, fact-based way as well (ii) allow for more intuitive and insightful search experiences.

Our work aligns with the OpenWebSearch.eu goal to foster a European ecosystem for web search infrastructure, emphasizing transparency, trustworthiness, and user empowerment. By incorporating trustworthiness aspects, we tackle key challenges related to bias and fairness, directly contributing to the project’s ambition to deliver ethical, unbiased search solutions.

The organization behind it:

Know-Center GmbH (KC) is Austria’s leading research center for trustworthy data-driven Artificial Intelligence with over 20 years of experience in top research and more than 700 applied scientific projects. Its dedicated Natural Language Processing group – led by Dr. Mark Kröll – provides state-of-the-art knowledge in (i) information / relation extraction (ii) text aggregation, (iii) text generation, as well as (iv) deep learning based architectures. The Fair-AI Group investigates aspects of fairness, diversity and non-discrimination, its measurements and mitigation in AI. Dr. Simone Kopeinik and Tomislav Đuričić will bring in their expertise to measure and better understand biases in LLMs. The Human-AI Interaction area – co-led by Dr. Vedran Sabol – brings in expertise in visual analytics, interactive machine learning, eXplainable AI, UI design and usability engineering, with 2 decades of project and publication track record in applications for textual data.