Nora: A Vocabulary Discovery Tool for Concept Extraction.

Area of Science:

• Medical Informatics
• Natural Language Processing
• Computational Linguistics

Background:

• Controlled medical vocabularies have limited coverage of domain-specific terms.
• Augmenting these terminologies is a resource-intensive process.
• Existing natural language processing (NLP) concept extraction methods can be hindered by incomplete lexicons.

Purpose of the Study:

• To develop an interactive tool, Nora, for discovering and augmenting domain-specific terminology from text corpora.
• To enhance the performance of NLP-based concept extraction tasks through improved terminology resources.

Main Methods:

• Nora visualizes word chains derived from text corpus word frequency indexes.
• Domain experts curate these chains to identify relevant terms and lexical variants.
• The tool facilitates interactive exploration and refinement of terminology.

Main Results:

• Nora demonstrated a 38% increase in the domain lexicon for homelessness and related psychosocial factors.
• This expansion in terminology led to an additional 10% in extracted concepts.
• The tool's interactive approach proved effective in identifying and curating lexical variants.

Conclusions:

• Nora is an effective interactive tool for discovering and expanding domain-specific terminologies.
• The tool enhances NLP concept extraction by enriching medical vocabularies.
• This approach offers a resource-efficient method for terminology augmentation in specialized domains.