ABNER

ABNER performs named entity recognition in biomedical text to extract entities such as proteins and cell lines for downstream bioinformatics and computational biology analyses.

Key Features:

• Named entity types: Two distinct tagging modules target biomedical entities, including proteins and cell lines.
• Conditional Random Fields (CRFs): Employs CRFs as the core statistical model for sequence labeling.
• Feature representation: Incorporates orthographic and contextual features to improve entity recognition accuracy.
• Training corpora: Modules are trained on standard corpora to establish model parameters.
• Java API integration and training: Provides a Java API to integrate ABNER into workflows and to train new models on user-defined corpora.

Scientific Applications:

• Entity extraction: Automates extraction of proteins and cell lines from scientific literature for data extraction.
• Downstream analyses: Supplies recognized entities for bioinformatics and computational biology analyses.
• Literature review acceleration: Facilitates rapid identification of relevant entities across large text collections.
• Data curation: Supports curation efforts by providing structured entity annotations.
• Hypothesis generation: Helps identify novel relationships between biological entities by extracting entity mentions from text.

Methodology:

ABNER applies machine learning via conditional random fields (CRFs) using orthographic and contextual features; its modules are trained on standard corpora and can be retrained on user-defined corpora via a Java API.