bert-base-uncased-finetuned-mrpc
This model is a fine-tuned version of bert-base-uncased on the GLUE MRPC (Microsoft Research Paraphrase Corpus) dataset.
It is designed to evaluate sentence pairs and determine if they are semantically equivalent (paraphrases of one another).
Model Description
- Developed by: manyanda
- Model type: BERT (Bidirectional Encoder Representations from Transformers)
- Language(s) (NLP): English
- License: Apache 2.0
- Finetuned from model: bert-base-uncased
Intended Uses & Limitations
Intended Use
This model can be used to check sentence pairs for semantic similarity. It is ideal for deduplicating text, matching user questions to FAQ databases, or validating paraphrased text segments.
Limitations
The model is fine-tuned specifically on the MRPC corpus, which primarily consists of news articles. It may perform less accurately on casual chat scripts, technical documentation, or highly stylized text domains.
Training and Evaluation Data
The model was trained on the MRPC split of the GLUE Benchmark. This dataset contains sentence pairs extracted from online news sources, each manually annotated to indicate whether they capture the same meaning.
- Training set: 3,668 pairs
- Validation set: 408 pairs
- Test set: 1,725 pairs
How to Use
You can easily use this model with the Hugging Face pipeline API for sentence-pair classification:
from transformers import pipeline
classifier = pipeline("text-classification", model="gwisumanyanda/hugging-pipeline-model")
result = classifier({
"text": "The weather today is completely sunny.",
"text_pair": "It is a bright and beautiful sunny day outside."
})
print(result)
# Expected Output: [{'label': 'LABEL_1', 'score': 0.99}]