Augmentext 0.1.7

Text augmentation library for NLP with a focus on biomedical applications.

Augmentext

Augmentext is a text augmentation package for Natural Language Processing, with a focus on applications in the biomedical domain.

Augmentext is work in progress! Some features are functional, but it not yet in a usable state.

Features

• Auto-generated, randomised misspellings
• Dictionary-based thesaurus word replacement
• Auto-generated abbreviations
• More to come...

Biomedical Domain Specific Features

Although a general library, Augmentext has a special focus on biomedical text, such as

• Replacement of mm/g^2 with common mistakes, e.g. g/mm^2 etc.
• Conversion of units from metric to imperial/customary and vice versa
• Integration of SNOMED, ICD, MeSH, RxNorm and other text corpora in to the augmentation pipeline
• Synonym replacement using pre-trained models using GloVe, fasttext, and word2vec.

More Information

See the project's GitHub respository https://github.com/mdbloice/Augmentext

Help will be available here once the software has been made public on GitHub: https://augmentext.readthedocs.io