Date of Graduation

7-2021

Document Type

Dissertation

Degree Name

Doctor of Philosophy in Computer Science (PhD)

Degree Level

Graduate

Department

Computer Science & Computer Engineering

Advisor/Mentor

Gauch, Susan E.

Committee Member

Andrews, David L.

Second Committee Member

Cronan, Timothy P.

Third Committee Member

Panda, Brajendra N.

Keywords

bioinformatics; consumer health vocabulary; laymen terms; ontology; text processing; word embeddings

Abstract

Clear language makes communication easier between any two parties. However, a layman may have difficulty communicating with a professional due to not understanding the specialized terms common to the domain. In healthcare, it is rare to find a layman knowledgeable in medical jargon, which can lead to poor understanding of their condition and/or treatment. To bridge this gap, several professional vocabularies and ontologies have been created to map laymen medical terms to professional medical terms and vice versa. Many of the presented vocabularies are built manually or semi-automatically requiring large investments of time and human effort and consequently the slow growth of these vocabularies. In this dissertation, we present an automatic method to enrich existing concepts in a medical ontology with additional laymen terms and also to expand the number of concepts in the ontology that do not have associated laymen terms. Our work has the benefit of being applicable to vocabularies in any domain.

Our entirely automatic approach uses machine learning, specifically Global Vectors for Word Embeddings (GloVe), on a corpus collected from a social media healthcare platform to extend and enhance consumer health vocabularies. We improve these vocabularies by incorporating synonyms and hyponyms from the WordNet ontology. By performing iterative feedback using GloVe’s candidate terms, we can boost the number of word occurrences in the co-occurrence matrix allowing our approach to work with a smaller training corpus.

Our novel algorithms and GloVe were evaluated using two laymen datasets from the National Library of Medicine (NLM), the Open-Access and Collaborative Consumer Health Vocabulary (OAC CHV) and the MedlinePlus Healthcare Vocabulary. For our first goal, enriching concepts, the results show that GloVe was able to find new laymen terms with an F-score of 48.44%. Our best algorithm enhanced the corpus with synonyms from WordNet, outperformed GloVe with an F-score relative improvement of 25%. For our second goal, expanding the number of concepts with related laymen’s terms, our synonym-enhanced GloVe outperformed GloVe with a relative F-score relative improvement of 63%.

The results of the system were in general promising and can be applied not only to enrich and expand laymen vocabularies for medicine but any ontology for a domain, given an appropriate corpus for the domain. Our approach is applicable to narrow domains that may not have the huge training corpora typically used with word embedding approaches. In essence, by incorporating an external source of linguistic information, WordNet, and expanding the training corpus, we are getting more out of our training corpus. Our system can help building an application for patients where they can read their physician's letters more understandably and clearly. Moreover, the output of this system can be used to improve the results of healthcare search engines, entity recognition systems, and many others.

Download

Share

COinS