|
--- |
|
language: en |
|
tags: |
|
- bert |
|
- sequence-classification |
|
- mrpc |
|
- paraphrase |
|
license: mit |
|
--- |
|
|
|
# Model description |
|
|
|
Fine-tuned version of `bert-base-uncased` on the Microsoft Research Paraphrase Corpus (MRPC) dataset for paraphrase detection using the MRPC dataset. |
|
|
|
## Intended uses & limitations |
|
|
|
This model is intended for paraphrase detection tasks, particularly those similar to the MRPC dataset. It may not perform well on substantially different datasets or tasks. |
|
|
|
## Training and evaluation data |
|
|
|
The model was trained on the MRPC dataset, which contains 5,801 sentence pairs extracted from news sources on the web. 3,900 pairs were labeled as paraphrases by human annotators. |
|
|
|
## Training procedure |
|
|
|
The model was fine-tuned using the Hugging Face Transformers library. We used a batch size of 16, learning rate of 2e-5, and trained for 3 epochs. |
|
|
|
## Evaluation results |
|
|
|
The model achieved the following results on the MRPC validation set: |
|
|
|
- Accuracy: 0.8480 |
|
- F1 Score: 0.8927 |
|
|
