SimuLingua consortium reviews progress on materials data, AI tools, and project communication

A central theme was collaboration: making diverse datasets consistent, verifiable, and useful requires close interaction between materials researchers and AI specialists. The meeting helped clarify priorities for the coming months and the contributions needed from partners.

Collecting and harmonising materials data

A substantial part of the meeting focused on collecting data from public materials databases and scientific publications. KTH presented progress in gathering computational datasets, including structural, thermodynamic, and selected mechanical properties, with particular attention to refractory high-entropy alloys. Partners discussed the need to harmonise units, conventions, and metadata so that information from different sources can be used consistently across the project. They also explored opportunities to enrich existing datasets with additional property estimates and assessments of high-temperature phase stability.

Developing traceable, AI-assisted data extraction

The Data Factory presentation demonstrated how AI-assisted tools can extract materials-property data from the text, tables, and plots of scientific papers. A key feature is traceability: extracted measurements can be linked to their location in the original publication and checked by a researcher before being confirmed or rejected. The tool also supports data search, export, and access through AI assistants.

Further development will focus on improving extraction completeness and accuracy, benchmarking against manually extracted data, making extraction requirements easier to define, and preparing data for compatibility with knowledge graphs.

Connecting materials expertise and AI research

Partners highlighted the importance of close collaboration between materials researchers and AI specialists in designing the project’s knowledge graph and the scientific foundation model architecture.

The University of Sheffield shared progress towards a literature-based dataset for piezoelectric electroceramics. A curated collection of relevant publications is being assembled to support materials descriptor research and provide a future test case for the Data Factory.

Coordinating communication and the next steps

The consortium reviewed the website and LinkedIn activity, newsletter preparation, and contributions to the dissemination, exploitation, and communication plan. Partners were encouraged to share research updates, report conference participation, and contribute information for public-facing content.

The meeting concluded with reminders about upcoming deliverables and partner inputs. Future monthly meetings will include a brief update from each organisation, supporting a clearer overview of progress and stronger coordination across the consortium.