Data and code related to our recent [EMNLP'18 paper] (https://arxiv.org/abs/1808.10012) will be released here soon...
Reasoning about Actions and State Changes by Injecting Commonsense Knowledge, Niket Tandon, Bhavana Dalvi Mishra, Joel Grus, Wen-tau Yih, Antoine Bosselut, Peter Clark, EMNLP 2018
A repository of the state change prediction models used for evaluation in the Tracking State Changes in Procedural Text: A Challenge Dataset and Models for Process Paragraph Comprehension paper accepted to NAACL'18. It contains two models built using the PyTorch-based deep-learning NLP library, AllenNLP.
- ProLocal: A simple local model that takes a sentence and entity as input and predicts state changes happening to the entity.
- ProGlobal: A global model for state change prediction that takes entire paragraph and an entity as input and predicts the entity's state at every time-step in the paragraph.
These models can be trained and evaluated as described below.
- Create the
propara
environment using Anaconda
conda create -n propara python=3.6
- Activate the environment
source activate propara
- Install the requirements in the environment:
pip install -r requirements.txt
- Test installation
pytest -v
You can download the dataset used in the NAACL'18 paper from
http://data.allenai.org/propara/
Example command to run eval script:
python propara/eval/evalQA.py tests/fixtures/eval/para_id.test.txt tests/fixtures/eval/gold_labels.test.tsv tests/fixtures/eval/sample.model.test_predictions.tsv
If you find these models helpful in your work, please cite:
@inproceedings{proparNaacl2018,
Author = { {Bhavana Dalvi, Lifu Huang}, Niket Tandon, Wen-tau Yih, Peter Clark},
Booktitle = {NAACL},
Title = {Tracking State Changes in Procedural Text: A Challenge Dataset and Models for Process Paragraph Comprehension},
Year = {2018}
}
** Bhavana Dalvi and Lifu Huang contributed equally to this work.