Update README.md
Browse files
README.md
ADDED
|
@@ -0,0 +1,83 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
---
|
| 2 |
+
language: "en"
|
| 3 |
+
tags:
|
| 4 |
+
- gpt2
|
| 5 |
+
- exbert
|
| 6 |
+
- commonsense
|
| 7 |
+
- semeval2020
|
| 8 |
+
- comve
|
| 9 |
+
license: "mit"
|
| 10 |
+
datasets:
|
| 11 |
+
- ComVE
|
| 12 |
+
metrics:
|
| 13 |
+
- bleu
|
| 14 |
+
widget:
|
| 15 |
+
- text: "Chicken can swim in water. <|continue|>"
|
| 16 |
+
---
|
| 17 |
+
|
| 18 |
+
# ComVE-gpt2-medium
|
| 19 |
+
|
| 20 |
+
## Model description
|
| 21 |
+
|
| 22 |
+
Finetuned model on Commonsense Validation and Explanation (ComVE) dataset introduced in [SemEval2020 Task4](https://competitions.codalab.org/competitions/21080) using a causal language modeling (CLM) objective.
|
| 23 |
+
The model is able to generate a reason why a given natural language statement is against commonsense.
|
| 24 |
+
|
| 25 |
+
## Intended uses & limitations
|
| 26 |
+
|
| 27 |
+
You can use the raw model for text generation to generate reasons why natural language statements are against commonsense.
|
| 28 |
+
|
| 29 |
+
#### How to use
|
| 30 |
+
|
| 31 |
+
You can use this model directly to generate reasons why the given statement is against commonsense using [`generate.sh`](https://github.com/AliOsm/SemEval2020-Task4-ComVE/tree/master/TaskC-Generation) script.
|
| 32 |
+
|
| 33 |
+
*Note:* make sure that you are using version `2.4.1` of `transformers` package. Newer versions has some issue in text generation and the model repeats the last token generated again and again.
|
| 34 |
+
|
| 35 |
+
#### Limitations and bias
|
| 36 |
+
|
| 37 |
+
The model biased to negate the entered sentence usually instead of producing a factual reason.
|
| 38 |
+
|
| 39 |
+
## Training data
|
| 40 |
+
|
| 41 |
+
The model is initialized from the [gpt2-medium](https://github.com/huggingface/transformers/blob/master/model_cards/gpt2-README.md) model and finetuned using [ComVE](https://github.com/wangcunxiang/SemEval2020-Task4-Commonsense-Validation-and-Explanation) dataset which contains 10K against commonsense sentences, each of them is paired with three reference reasons.
|
| 42 |
+
|
| 43 |
+
## Training procedure
|
| 44 |
+
|
| 45 |
+
Each natural language statement that against commonsense is concatenated with its reference reason with `<|continue|>` as a separator, then the model finetuned using CLM objective.
|
| 46 |
+
The model trained on Nvidia Tesla P100 GPU from Google Colab platform with 5e-5 learning rate, 5 epochs, 128 maximum sequence length and 64 batch size.
|
| 47 |
+
|
| 48 |
+
<center>
|
| 49 |
+
<img src="https://i.imgur.com/xKbrwBC.png">
|
| 50 |
+
</center>
|
| 51 |
+
|
| 52 |
+
## Eval results
|
| 53 |
+
|
| 54 |
+
The model achieved fifth place with 16.7153/16.1187 BLEU scores and third place with 1.94 Human Evaluation score on SemEval2020 Task4: Commonsense Validation and Explanation development and testing dataset.
|
| 55 |
+
|
| 56 |
+
These are some examples generated by the model:
|
| 57 |
+
|
| 58 |
+
| Against Commonsense Statement | Generated Reason |
|
| 59 |
+
|:-----------------------------------------------------:|:--------------------------------------------:|
|
| 60 |
+
| Chicken can swim in water. | Chicken can't swim. |
|
| 61 |
+
| shoes can fly | Shoes are not able to fly. |
|
| 62 |
+
| Chocolate can be used to make a coffee pot | Chocolate is not used to make coffee pots. |
|
| 63 |
+
| you can also buy tickets online with an identity card | You can't buy tickets with an identity card. |
|
| 64 |
+
| a ball is square and can roll | A ball is round and cannot roll. |
|
| 65 |
+
| You can use detergent to dye your hair. | Detergent is used to wash clothes. |
|
| 66 |
+
| you can eat mercury | mercury is poisonous |
|
| 67 |
+
| A gardener can follow a suspect | gardener is not a police officer |
|
| 68 |
+
| cars can float in the ocean just like a boat | Cars are too heavy to float in the ocean. |
|
| 69 |
+
| I am going to work so I can lose money. | Working is not a way to lose money. |
|
| 70 |
+
|
| 71 |
+
### BibTeX entry and citation info
|
| 72 |
+
|
| 73 |
+
```bibtex
|
| 74 |
+
@article{fadel2020justers,
|
| 75 |
+
title={JUSTers at SemEval-2020 Task 4: Evaluating Transformer Models Against Commonsense Validation and Explanation},
|
| 76 |
+
author={Fadel, Ali and Al-Ayyoub, Mahmoud and Cambria, Erik},
|
| 77 |
+
year={2020}
|
| 78 |
+
}
|
| 79 |
+
```
|
| 80 |
+
|
| 81 |
+
<a href="https://huggingface.co/exbert/?model=aliosm/ComVE-gpt2-medium">
|
| 82 |
+
<img width="300px" src="https://hf-dinosaur.huggingface.co/exbert/button.png">
|
| 83 |
+
</a>
|