Bert Base Cased Qa Evaluator

🚀 BERT-base-cased-qa-evaluator

This model takes a question-answer pair as input and outputs a value indicating its prediction of whether the input is a valid question-answer pair. It's based on the pretrained BERT-base-cased with a sequence classification head.

🚀 Quick Start

This model accepts a question-answer pair as input and predicts whether it's valid. It's built upon the pre - trained BERT-base-cased with a sequence classification head.

✨ Features

Intended Uses

The QA evaluator was initially designed to work with the t5-base-question-generator to assess the quality of generated questions. The input for the QA evaluator adheres to the BertForSequenceClassification format, using the question and answer as two sequences. Inputs should follow this format:

[CLS] <question> [SEP] <answer [SEP]

Limitations and Bias

The model is trained to evaluate semantic relationships between questions and answers but can't determine the actual truth or correctness of an answer.

📦 Installation

No installation steps were provided in the original README, so this section is skipped.

💻 Usage Examples

No code examples were provided in the original README, so this section is skipped.

📚 Documentation

Training Data

The training data consists of question-answer pairs from the following datasets:

• SQuAD
• RACE
• CoQA
• MSMARCO

Training Procedure

In 50% of cases, the question and answer were concatenated. In the other 50%, a corruption operation was carried out (either replacing the answer with an unrelated one or copying part of the question into the answer). The model was then trained to predict whether the input sequence was an original QA pair or a corrupted input.

📄 License

No license information was provided in the original README, so this section is skipped.