TermInformer

TermInformer extracts biomedical terms from literature and analyzes their semantic relationships using unsupervised methods and word vector training to produce reusable term embeddings for downstream biomedical NLP applications.

Key Features:

• Unsupervised Term Mining: Extracts biomedical terms from unlabelled corpora without relying on external resources such as labeled datasets or predefined dictionaries.
• Semantic Relationship Analysis: Uses natural language processing (NLP) techniques to reveal semantic relationships between terms.
• Reusable Term Embeddings: Trains word vector models to generate term embeddings that can be applied across downstream NLP applications.
• General Applicability: Operates without extensive prior knowledge and is applicable to diverse document data, in contrast to named entity recognition models that require extensive training and limited entity-type specification.

Scientific Applications:

• Lung cancer: Applied to identify potential factors and treatments for lung cancer by mining terms and analyzing their semantic relationships.
• Breast cancer: Applied to identify potential factors and treatments for breast cancer through term mining and semantic analysis.
• Coronavirus: Applied to identify potential factors and treatments for coronavirus by extracting terms and mapping semantic relationships.

Methodology:

Applies an unsupervised approach that calculates term meanings in an open fashion, employs word vector training algorithms to generate term embeddings (compared to existing word embeddings), and uses NLP techniques to analyze semantic relationships while operating without external labeled datasets or predefined dictionaries.