A cloud-native web application for AI-powered queries using fine-tuning on pre-trained Natural Language Models for Named Entities Recognition - THESPIAN-NER

When researchers and professionals try to retrieve stored analysis raw data, they usually face three major issues: data availability, missing (meta)data standardisation, and difficult query creation for database fetching.

To ease these issues, in the framework of the Eosc-Pillar project, the cloud platform Tools for HEritage Science Processing, Integration, and ANalysis (THESPIAN) was developed.

The first web service of the THESPIAN platform, THESPIAN-Mask, was developed to resolve the first two issues; it is a web-app for FAIR storage of scientific analysis on Cultural Heritage, and it is based on an hoc developed CIDOC-CRM-compatible ontology, CRMhs, describing the metadata of scientific files.

To ease the third issue, and also to help the metadata generation process, an additional cloud-native tool was developed: THESPIAN-NER.

It is a tool based on a deep neural network for Named Entity Recognition, enabling users to upload their Italian-written report files and obtain labelled named entities usable as keywords, either for (semi)automatically compose custom queries to the database, or fill (part of) the metadata form describing the file to be uploaded.

Learn more about EOSC-Pillar use case 9