---
license: cc-by-nc-sa-4.0
language:
- en
library_name: transformers
pipeline_tag: text-classification
datasets: 
  - ClaimRev
widget:
- text: "Teachers are likely to educate children better than parents."
  context: "Homeschooling should be banned."
---

# Model
This model was obtained by fine-tuning `microsoft/deberta-base` on the extended ClaimRev dataset.

Paper: [To Revise or Not to Revise: Learning to Detect Improvable Claims for Argumentative Writing Support](https://arxiv.org/abs/2305.16799)

Authors: Gabriella Skitalinskaya and Henning Wachsmuth

# Suboptimal Claim Detection
We cast this task as a binary classification task, where the objective is, given an argumentative claim and some contextual information (in this case, the **main thesis** of the debate), to decide whether it is in need of further revision or can be considered to be phrased more or less optimally.

# Usage

```python
from transformers import AutoTokenizer, AutoModelForSequenceClassification
import torch

tokenizer = AutoTokenizer.from_pretrained("gabski/deberta-suboptimal-claim-detection-with-thesis-context")
model = AutoModelForSequenceClassification.from_pretrained("gabski/deberta-suboptimal-claim-detection-with-thesis-context")
claim = 'Teachers are likely to educate children better than parents.'
thesis = 'Homeschooling should be banned.'
model_input = tokenizer(claim, thesis, return_tensors='pt')
model_outputs = model(**model_input)

outputs = torch.nn.functional.softmax(model_outputs.logits, dim = -1)
print(outputs)
```