UMLS-Interface and UMLS-Similarity : open source software for measuring paths and semantic similarity

Area of Science:

• Biomedical Natural Language Processing
• Computational Linguistics
• Bioinformatics

Background:

• Numerous computational methods exist for assessing semantic similarity between biomedical concepts.
• These methods often rely on diverse standards and programming environments, leading to fragmentation.
• A unified approach is needed for consistent development and validation.

Purpose of the Study:

• Introduce two novel open-source frameworks, UMLS-Similarity and UMLS-Interface, for calculating semantic similarity.
• Provide a common platform for developing, testing, and comparing semantic similarity measures based on the Unified Medical Language System (UMLS).
• Facilitate advancements in biomedical Natural Language Processing (NLP).

Main Methods:

• Developed UMLS-Interface to provide path information for UMLS concepts.
• Developed UMLS-Similarity to compute semantic similarity using existing measures and allow for extensions.
• Validated framework functionality by replicating results from prior research.

Main Results:

• Successfully created and validated two open-source frameworks for UMLS-based semantic similarity.
• Demonstrated the frameworks' ability to reproduce previous findings, confirming their accuracy and utility.
• Established a standardized platform for semantic similarity research.

Conclusions:

• The UMLS-Similarity and UMLS-Interface frameworks offer a significant contribution to biomedical NLP.
• These open-source tools provide a unified environment for the development and evaluation of semantic similarity measures.
• The frameworks will accelerate research and development in understanding relationships between biomedical concepts.