Empowering Knowledge Extraction to Empower Learners

Type of project: European  |  Start date: 01/12/2023  |  End date: 30/11/2025

Artificial intelligence and Natural Language Processing are playing a key role in the digital transformation of society towards a growing ability of artificial agents to process data, understand meaning and exploit the learnt knowledge to support people in a wide variety of activities.

The project aim to improve the current state of the art in autonomous understanding of contents from video lessons and to exploit the extracted knowledge to provide advanced services for learning support, in a way that makes the service provision sustainable and the knowledge shareable and reusable.

The model will be applied in the context of higher education and training and is aimed to have an impact on empowerment in accordance to the Digital Education Action Plan 2021-27, geared towards the development of high-performing digital education ecosystem.

More in detail, the project is focused on knowledge extraction from educational videos in terms of concepts explained and dependency relations among them, in particular prerequisite relations (PR), which express the knowledge required to understand another concept.

Video lessons have seen an exponential growth in the last years for a variety of educational purposes, formal/informal, academic/vocational. The scope of EKEEL is to improve PR extraction from educational videos by proposing a model of PR Knowledge Graph aimed to exploit such amount of digital resources to produce new value, through knowledge transformation and reuse.

For PR extraction, EKEEL will exploit Language Models, Deep Learning for text, image and video processing, and AI techniques for Natural Language Processing and temporal reasoning in order to identify the concepts and their changing role along the video. The scientific contribution in terms of knowledge extraction is an innovative audio-visual approach for PR extraction that takes into account the analysis of the concepts’ dynamics throughout the video stream, whose results will provide, in addition, novel features for training deep learning models.


Funding programme:
PRIN 2022 PNRR - P20227PEPK

Funding body:
European Union

Grant agreement:
CUP B53D23026030001


CNR-ILC role:

Project coordinator:
Ilaria Torre (University of Genoa)

CNR-ILC Research Unit Chair:
Felice Dell'Orletta

Dominique Pierina Brunato
Giulia Venturi
Noemi Terreni