Text Mining Provider

Back to Home

Brief Overview

The Text Mining Provider KG contains subject-predicate-object assertions derived from the application of natural language processing (NLP) algorithms to the PubMed Central open-access collection of publications plus additional titles and abstracts from PubMed.

Caveat: Text-mined assertions must be interpreted with caution, as NLP algorithms may introduce false assertions.

Example Edge:

License/Restrictions: None.

URL: http://smart-api.info/registry?q=978fe380a147a8641caf72320862697b; http://smart-api.info/registry?q=71fa2e0f0f1fe1ec67f4ddb719db5ef3

Detailed Description

The Text Mining Provider aims to provide an up-to-date, Biolink-compatible, knowledge graphs (KGs) composed of assertions mined from the available biomedical literature. Two flavors of KGs are provided:

1. A concept cooccurrence KG where nodes are ontology concepts and links between nodes reflect the cooccurrence of the concepts in text, e.g. in the same sentence or abstract. Edges are scored using the Normalized Google Distance metric.
2. A KG composed of text-mined assertions where the nodes are ontology concepts and the edges represent explicitly defined BioLink relations between the two concepts.

Team Contact:

Bill Baumgartner

Interfaces

See Available Knowledge Graphs.

Knowledge Sources

• Public biomedical literature archives (like PubMed).
• See also Concept Recognition.

Source Code

See Associated Code Repositories.

External Documentation

See the for details on the development status and implementation plans for the NCATS Translator Text Mining Provider. The referenced repository includes, among others:

• a project board to monitor progress of the milestones from the initial Text Mining Provider proposal
• facility to request new text mining targets and to report errors
• the performances of component tools being used by the Text Mining Provider
• the status of ongoing work related to the Text Mining Provider