Simple Question-Answering System with BERT

This repository contains a simple question-answering system built with the BERT model. The model is fine-tuned on the SQuAD dataset and can answer questions based on a given context.

Features

• Uses the bert-large-uncased-whole-word-masking-finetuned-squad model from Hugging Face’s transformers library.
• The system takes a context (a piece of text) and a question, and returns an answer based on the context.
• The answer is extracted directly from the context.

Usage

The main function in this repository is robot(context, question). Here’s how to use it:

from transformers import BertForQuestionAnswering, BertTokenizer
import torch

Load the pre-trained BERT model and tokenizer

model_name = 'bert-large-uncased-whole-word-masking-finetuned-squad'
model = BertForQuestionAnswering.from_pretrained(model_name)
tokenizer = BertTokenizer.from_pretrained(model_name)

Main Funciton

def robot(context, question):

    # Tokenize the input text and question
    inputs = tokenizer(question, context, return_tensors="pt")

    # Get the answer
    output = model(**inputs)

    # Get start and end scores for answer
    answer_start_scores = output.start_logits
    answer_end_scores = output.end_logits

    # Find start and end of answer
    answer_start = torch.argmax(answer_start_scores)
    answer_end = torch.argmax(answer_end_scores) + 1

    # Convert tokens to string
    answer = tokenizer.convert_tokens_to_string(tokenizer.convert_ids_to_tokens(inputs["input_ids"][0][answer_start:answer_end]))

    return answer

Context of the story

story = """
... Your story here ...
"""

Question to ask about the story

question = "Your question here"
answer = robot(story, question)
print("Answer:", answer)

Replace "Your story here" and "Your question here" with your own context and question. The robot() function will return an answer based on the context.

Requirements

• Python 3.6 or later.
• PyTorch 1.0.0 or later.
• Transformers library from Hugging Face.

Installation

You can install the required packages with pip:

pip install torch transformers