Text mining for neuroanatomy using WhiteText with an updated corpus and a new web application

Area of Science:

• Neuroscience
• Computational Linguistics
• Bioinformatics

Background:

• Extracting neuroanatomical connectivity from scientific literature is challenging.
• Existing methods require manual curation and lack scalability.

Purpose of the Study:

• To describe the WhiteText project's progress in automatically extracting neuroanatomical connectivity statements.
• To present an expanded manually curated corpus and a new web application for data access.

Main Methods:

• Developing and refining methods for brain region mention recognition and standardization.
• Implementing connectivity statement extraction algorithms.
• Manually curating a corpus of neuroanatomical connectivity statements.

Main Results:

• The project successfully extracts neuroanatomical connectivity statements from text.
• A new corpus version added 2,111 statements, resulting in a merged corpus of 5,208 statements.
• Cross-validation achieved 67% recall and 51% precision for connectivity statement extraction.

Conclusions:

• The WhiteText project provides a valuable resource for neuroscience research.
• The automated extraction and curated corpus can seed connectivity matrices and train machine learning models.
• A new web application facilitates data access and further curation.