Edinburgh Research Explorer

Mr. Andreas Grivas

Research Assistant

Research Interests

I am interested in knowledge extraction from unstructured text using neural networks. My current focus is on fine-grained named entity recognition and relation extraction from biomedical text. More specifically, my work focusses on tagging electronic health records (EHRs) with disease information (phenotyping). Disease mentions in such records have a varying level of granularity. For example, "acute ischaemic stroke" is more precise/granular than a mention of "ischaemic stroke". However, depending on the application, a different level of tag granularity is required. Can we learn to encode mentions using neural networks in a way that maintains the subset/is-a relationship between tags? Can we use information from disease ontologies towards this goal?

I have previously worked on dependency parsing for languages with rich morphology and named entity recognition in news text.

Qualifications

BSc Informatics, Harokopeio University of Athens 2014.
MSc Artificial Intelligence, University of Edinburgh 2017.

Research outputs

  1. What do character-level models learn about morphology? The case of dependency parsing

    Research output: Chapter in Book/Report/Conference proceedingConference contribution

  2. PaloPro: a platform for knowledge extraction from big social data and the news

    Research output: Contribution to journalArticle

  3. Author Profiling using Stylometric and Structural Feature Groupings

    Research output: Chapter in Book/Report/Conference proceedingConference contribution

View all (3) »

ID: 81960065