Skip to content

manikandan-ravikiran/Source-Free-Domain-Adaptation

Repository files navigation

Source-free-domain-adaptation

Data sharing restrictions are common in NLP datasets. For example, Twitter policies do not allow sharing of tweet text, though tweet IDs may be shared. The situation is even more common in clinical NLP, where patient health information must be protected, and annotations over health text, when released at all, often require the signing of complex data use agreements

The goal is to develop an accurate system for a target domain when annotations exist for a related domain but cannot be distributed. Instead of annotated training data, participants are given a model trained on the annotations.

Currently baseline models and data are provided inline with https://github.com/Machine-Learning-for-Medical-Language/source-free-domain-adaptation.git

About

Source free domain adaptation in clinical temporal expression extraction

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published