MDG-NLP

Multi-Dimensional Generalization

This Hugging Face organization hosts the datasets and models used in our study of multi-dimensional generalization in temporal extraction tasks.

Tasks

TIMEX Time Expression Extraction

EVENTX Event Expression Extraction

Generalization Dimensions

Domain — Domain transfer

Robustness — Adversarial perturbations

Compositional — Joint extraction

Structural — Sentence to document

Vocabulary — Lexical shift

Resources

This Hugging Face organization contains the datasets and models used in our experiments, while the experimental code is available on GitHub.

Datasets Models GitHub

Clinical Data Access

The clinical-domain datasets (T2C) are derived from the 2012 i2b2 Clinical Temporal Relations corpus and are not publicly redistributed because they are subject to the n2c2 data use agreement.

If you already have authorization to use the original i2b2/n2c2 dataset, please contact us. We can share the corresponding processed T2C data with authorized users after verifying their access.

Paper

Evaluating Multi-Dimensional Generalization of Large Language Models in Temporal Extraction Tasks

Fahmid Shahriar Iqbal · Ritam Dutt · Soumitra Das · Arnav Verma · Sagnik Ray Choudhury