Model description

This repo contains the model and the notebook for fine-tuning BERT model on SNLI Corpus for Semantic Similarity. Semantic Similarity with BERT .

Full credits go to Mohamad Merchant

Reproduced by Vu Minh Chien

Motivation: Semantic Similarity determines how similar two sentences are, in terms of their meaning. In this tutorial, we can fine-tune BERT model and use it to predict the similarity score for two sentences.

Training and evaluation data

This example demonstrates the use of the Stanford Natural Language Inference (SNLI) Corpus to predict semantic sentence similarity with Transformers.

• Total train samples: 100000

• Total validation samples: 10000

• Total test samples: 10000

Here are the "similarity" label values in SNLI dataset:

• Contradiction: The sentences share no similarity.

• Entailment: The sentences have a similar meaning.

• Neutral: The sentences are neutral.

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

Model Plot