You are here:
Development of an Ontology for Occupational Exposure
Citation:
Hubbard, H., A. Varghese, P. Egeghy, AND D. Vallero. Development of an Ontology for Occupational Exposure. ISES 2016, Utrecht, NETHERLANDS, October 09 - 13, 2016.
Impact/Purpose:
Presented at the ISES 2016 Conference.
Description:
When discussing a scientific domain, the use of a common language is required, particularly when communicating across disciplines. This common language, or ontology, is a prescribed vocabulary and a web of contextual relationships within the vocabulary that describe the given domain with a view to organizing information. This presentation describes a methodology to ontology development that uses machine learning and natural language processing algorithms, including vector space language models, lexical relation extraction, and topic discovery algorithms, to define an ontology for describing occupational exposures. By applying these automated processes to support expert judgment, a much larger body of literature can be considered than if an individual was required to evaluate each document. Additionally, computer-generated synonym lists can work as an aid to researchers by suggesting keywords that may not otherwise be considered. In order to use the automated tools, publicly available scientific abstracts from PubMed were gathered using keywords related to “occupational exposure”. The titles and abstracts from each study were combined into a single text field. This textual vector was analyzed using ICF’s DoCTER (Document Classification and Topic Extraction Resource) tool to determine clusters and inter-cluster distances that were used to suggest taxonomies. Then, ICF’s L-Rex (Lexical Relationship Extractor) and ToxSyn (Toxicologic Semantic Similarity Discovery) tools were used to propose ontology rules, by finding synonyms, antonyms, hyponyms and hypernyms, discovering range-domain relationships, and assessing term similarity queries. These results were used to develop a semantic model or visual representation of the ontology pattern suggested for describing occupational exposures.
