Dr. Prashanti Manda (Computer Science) received new funding from the National Science Foundation for the project “CAREER: Ontology-Powered Named Entity Recognition and Robust Semantic Similarity Metrics.”
Ontologies have been used for data representation in biology since 2006 and have powered an array of computational applications. However, the process of translating knowledge in literature to an ontology-based representation is still largely manual and does not scale well. This proposal will address challenges at the frontier of Natural Language Processing – automated curation of biological literature via Named Entity Recognition (NER) of ontology concepts.
The accurate evaluation of ontology-based NLP systems requires robust semantic similarity measures that can estimate degrees of partial relatedness between an NLP tool’s output and a gold standard. While a number of semantic similarity metrics do exist, their robustness and ability to identify small amounts of similarity is unclear. Here, the researcher proposes to develop novel deep learning architectures for ontology based NER and robust semantic similarity metrics that can accurately assess the performance of the above models.