collaboration between Nenad Macesic and Kernyu Park
Aiming at an ontology-based approach to tackle antibiotic resistance.
- Ontology Learning process (semi-automated)
- Topic-modeling methods using tf-idf weighting of terms
- annotation of biomedical texts using Metamap
- network analysis (clustering and community detection)
-
obtain MEDLINE corpus of abstracts and MeSH terms
-
process abstracts using tf-idf weighting of terms and constrain to an adjacency matrix of co-occurrences between all terms occurring in all abstract texts (obtain "tokens", or "terms")
-
map to UMLS using Metamap to identify concepts in the term cloud (obtain "concepts")
-
combine with matrix of MeSH terms to identify hierarchical as well as modifier relationships between terms (define relationships)
-
expert validation of the established concepts and their relationships
-
construct an ontology (ontology generation)
-
expert validation of the obtained ontology for further improvement