03 Mar Question Answering using Hugging Face

In this lesson, we will learn how to use the Hugging Face Transformers library for Question-Answering. Learn how to use the Hugging Face Transformers library for Question-Answering. We will see the steps with examples. Using a Hugging Face model for question answering (QA) is quite easy with the Transformers library.

Before moving further, we’ve prepared a video tutorial to implement Question-Answering with Hugging Face:

Note: We will run the codes on Google Colab

Questions Answering – Coding Example

Below is a step-by-step guide to using a pre-trained QA model from Hugging Face.

Step 1: Install the Required Libraries

Make sure you have the transformers and torch (or tensorflow) libraries installed. On Google Colab, use the following command to install:

!pip install transformers torch

Step 2: Load a Pre-trained QA Model

Hugging Face provides several pre-trained models for question answering, such as bert-large-uncased-whole-word-masking-finetuned-squad or distilbert-base-cased-distilled-squad. You can load a model and its tokenizer as follows:

from transformers import pipeline

# Load a pre-trained QA pipeline
qa_pipeline = pipeline("question-answering")

Alternatively, you can load a specific model and tokenizer:

from transformers import AutoModelForQuestionAnswering, AutoTokenizer

model_name = "bert-large-uncased-whole-word-masking-finetuned-squad"
tokenizer = AutoTokenizer.from_pretrained(model_name)
model = AutoModelForQuestionAnswering.from_pretrained(model_name)

Step 3: Prepare the Input

For QA tasks, you need:

1. A context (a paragraph or text where the answer might be found).
2. A question (the question you want to answer).

Example:

context = """
I am Amit Diwan from Delhi. My interests include programming and cricket. I have created courses
not only in programming, but web dev, and database technologies as well. Yes, AI is trending,
and we all are a part of it. Artificial Intelligence (AI) is transforming the world by enabling machines 
to learn, adapt, and perform tasks that typically require human intelligence.
"""

question = "Where Amit Diwan is based?"

Step 4: Tokenize the Input

Tokenize the context and question using the tokenizer:

inputs = tokenizer(question, context, return_tensors="pt")

Step 5: Get the Model’s Prediction

Pass the tokenized input to the model to get the answer:

outputs = model(**inputs)

# Extract the start and end scores
start_scores = outputs.start_logits
end_scores = outputs.end_logits

# Get the most likely start and end positions
start_index = start_scores.argmax()
end_index = end_scores.argmax()

# Convert token IDs back to words
answer_tokens = inputs["input_ids"][0][start_index : end_index + 1]
answer = tokenizer.decode(answer_tokens, skip_special_tokens=True)

print(f"Answer: {answer}")

For the example above, the output might look like:

Answer: delhi

or

Alternative code: Use the Pipeline (Simpler Alternative)

If you don’t want to handle tokenization and model outputs manually, you can use the pipeline API, which simplifies the process:

from transformers import pipeline

qa_pipeline = pipeline("question-answering")

result = qa_pipeline(question=question, context=context)
print(f"Answer: {result['answer']}")
print(f"Score: {result['score']}")

For the alternative example above, the output might look like:

The above screenshot displayed the following output:

Answer: Delhi
Score: 0.9925423860549927

If you liked the tutorial, spread the word and share the link and our website Studyopedia with others.

For Videos, Join Our YouTube Channel: Join Now

Read More:

• RAG Tutorial
• Generative AI Tutorial
• Machine Learning Tutorial
• Deep Learning Tutorial
• Ollama Tutorial
• Retrieval Augmented Generation (RAG) Tutorial
• Copilot Tutorial
• ChatGPT Tutorial