Exploring Representations for Singular and Multi-Concept Relations for Biomedical Named Entity Normalization.

Area of Science:

• Biomedical Informatics
• Natural Language Processing
• Computational Biology

Background:

• The COVID-19 pandemic increased biomedical literature queries with diverse naming conventions.
• Disambiguating chemical and disease mentions is crucial for accurate information retrieval.
• Current methods struggle with the variety of nomenclatures in biomedical texts.

Purpose of the Study:

• To develop and evaluate automated named entity normalization techniques for biomedical texts.
• To improve the relevance of search results in biomedical repositories.
• To address the challenge of varied naming conventions for chemicals and diseases.

Main Methods:

• Exploration of term embedding aggregation techniques.
• Evaluation of context's impact on normalization performance.
• Assessment of embedding approaches for normalizing single and multi-relation terms in unstructured text.

Main Results:

• Term embedding aggregation and contextual analysis significantly improve named entity normalization.
• The proposed methods effectively disambiguate diverse chemical and disease mentions.
• Enhanced accuracy in linking text mentions to biomedical knowledge base concepts.

Conclusions:

• Automated named entity normalization using term embeddings and context is effective for biomedical literature.
• This approach enhances data curation and researcher access to relevant scientific articles.
• The methods provide a scalable solution for managing the growing volume of biomedical data.