Update README.md
Browse files
README.md
CHANGED
@@ -1,3 +1,41 @@
|
|
1 |
---
|
2 |
license: mit
|
3 |
---
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
---
|
2 |
license: mit
|
3 |
---
|
4 |
+
# roberta-temporal-predictor
|
5 |
+
A RoBERTa-base model that is fine-tuned on the [The New York Times Annotated Corpus](https://catalog.ldc.upenn.edu/LDC2008T19)
|
6 |
+
to predict temporal precedence of two events. This is used as the ``temporality prediction'' component
|
7 |
+
in our ROCK framework for reasoning about commonsense causality. See our [paper](https://arxiv.org/abs/2202.00436) for more details.
|
8 |
+
|
9 |
+
# Usage
|
10 |
+
|
11 |
+
For simplicity, we implement the following TempPredictor class. Example usage using the ``TempPredictor`` class:
|
12 |
+
```python
|
13 |
+
from transformers import (RobertaForMaskedLM, RobertaTokenizer)
|
14 |
+
from src.temp_predictor import TempPredictor
|
15 |
+
|
16 |
+
TORCH_DEV = "cuda:0" # change as needed
|
17 |
+
|
18 |
+
tp_roberta_ft = src.TempPredictor(
|
19 |
+
model=RobertaForMaskedLM.from_pretrained("CogComp/roberta-temporal-predictor"),
|
20 |
+
tokenizer=RobertaTokenizer.from_pretrained("CogComp/roberta-temporal-predictor"),
|
21 |
+
device=TORCH_DEV
|
22 |
+
)
|
23 |
+
|
24 |
+
E1 = "The man turned on the faucet."
|
25 |
+
E2 = "Water flows out."
|
26 |
+
t12 = tp_roberta_ft(E1, E2, top_k=5)
|
27 |
+
print(f"P('{E1}' before '{E2}'): {t12}")
|
28 |
+
```
|
29 |
+
|
30 |
+
# BibTeX entry and citation info
|
31 |
+
|
32 |
+
```bib
|
33 |
+
@misc{zhang2022causal,
|
34 |
+
title={Causal Inference Principles for Reasoning about Commonsense Causality},
|
35 |
+
author={Jiayao Zhang and Hongming Zhang and Dan Roth and Weijie J. Su},
|
36 |
+
year={2022},
|
37 |
+
eprint={2202.00436},
|
38 |
+
archivePrefix={arXiv},
|
39 |
+
primaryClass={cs.CL}
|
40 |
+
}
|
41 |
+
```
|