MUTASK

Multimodal Translation and Adaptation of Scientific Knowledge for Global Accessibility

About MUTASK

MUTASK, Multimodal Translation and Adaptation of Scientific Knowledge for Global Accessibility, is a European research project developing AI-driven methods to translate, adapt and communicate scholarly content.

The project addresses the CHIST-ERA Science in Your Own Language vision. In MUTASK, "own language" means more than a language code. It also includes a person's prior knowledge, discipline, goals, doubts, vocabulary and preferred way of learning.

The Challenge

Scholarly knowledge often fails to reach the people who could use it because:

  1. researchers, students and wider audiences face language barriers;
  2. translated scientific content can remain too dense or too specialized;
  3. adapted outputs are rarely connected to a reliable workflow that reaches the intended communities.

Research Direction

MUTASK integrates machine translation, automated summarization, semantic indexing, audience adaptation and video-based storytelling into one workflow. The aim is to reshape complex academic documents, including research articles, conference papers and outreach texts, into outputs such as:

The project is not a generic summarizer, a generic video generator or a one-size-fits-all outreach tool. It studies how scientific meaning survives translation, transposition to an audience and transformation into media.

The MUTASK Lifecycle

The current technical model follows one paper through the whole chain:

Ingest Structure Understand Adapt Produce Review

AGH focuses on ingestion, parsing, extraction, document structure and media production. LORIA focuses on NLP, semantic indexing, translation, retrieval and personalization. PHZH contributes audience research, pedagogical transposition, dissemination and feedback design. Review and release decisions are shared across the consortium.

Audience And Evaluation

The human side of the project is central. MUTASK plans to involve target groups such as students, researchers, journalists, nonprofits, teachers, policy actors and citizens in testing adapted outputs.

Quality is evaluated across several dimensions:

Open Science

MUTASK aims to deliver practical methods that fit current publishing and research infrastructures while supporting Open Science standards, reproducibility and reusable data formats.