Question Answering & Summarization

# ─── 1. Extractive QA with BERT (HuggingFace) ────────────────
from transformers import pipeline

qa_pipeline = pipeline("question-answering",
                       model="deepset/roberta-base-squad2")

context = """
The Large Hadron Collider (LHC) is the world's largest and most powerful
particle accelerator. It lies in a tunnel 27 kilometres in circumference
and as deep as 175 metres beneath the France-Switzerland border near Geneva.
The LHC was built by CERN between 1998 and 2008 with the aim of allowing
physicists to test the predictions of different theories of particle physics.
It was first started up on 10 September 2008.
"""

questions = [
    "Where is the LHC located?",
    "How long is the tunnel?",
    "When was the LHC first started?",
    "Who built the LHC?",
]

print("Extractive QA (RoBERTa-SQuAD2):")
for q in questions:
    result = qa_pipeline(question=q, context=context)
    print(f"  Q: {q}")
    print(f"  A: {result['answer']} (score={result['score']:.3f})\n")

# ─── 2. RAG with LangChain + FAISS ───────────────────────────
# pip install langchain langchain-community faiss-cpu sentence-transformers
from langchain_community.vectorstores import FAISS
from langchain_community.embeddings import HuggingFaceEmbeddings
from langchain.text_splitter import RecursiveCharacterTextSplitter
from langchain_community.document_loaders import TextLoader
from langchain.chains import RetrievalQA

# Prepare documents
documents = [
    "The Python programming language was created by Guido van Rossum and first released in 1991.",
    "Python emphasizes code readability and uses significant indentation.",
    "Python supports multiple programming paradigms: procedural, object-oriented, and functional.",
    "The name Python comes from the BBC show Monty Python's Flying Circus.",
    "Python 3.0 was released in 2008 and is not fully backward compatible with Python 2.",
    "NumPy and Pandas are popular Python libraries for data science.",
    "TensorFlow and PyTorch are the leading Python frameworks for deep learning.",
]

# Create embeddings and vector store
embeddings = HuggingFaceEmbeddings(model_name="all-MiniLM-L6-v2")
vectorstore = FAISS.from_texts(documents, embeddings)
retriever = vectorstore.as_retriever(search_kwargs={"k": 3})

# Test retrieval
query = "Who created Python and when?"
retrieved = retriever.get_relevant_documents(query)
print("Retrieved documents:")
for i, doc in enumerate(retrieved, 1):
    print(f"  [{i}] {doc.page_content[:80]}")

# ─── 3. Simple RAG Reader ─────────────────────────────────────
def rag_answer(question, retriever, reader_pipeline):
    """Minimal RAG: retrieve then read."""
    docs = retriever.get_relevant_documents(question)
    context = " ".join([d.page_content for d in docs])
    result = reader_pipeline(question=question, context=context)
    return result['answer'], result['score'], docs

reader = pipeline("question-answering", model="deepset/roberta-base-squad2")
answer, score, sources = rag_answer("Who created Python?", retriever, reader)
print(f"\nRAG Answer: {answer} (confidence: {score:.3f})")
print(f"Source: {sources[0].page_content[:100]}")

# ─── 4. Abstractive Summarization ────────────────────────────
from transformers import pipeline as hf_pipeline

summarizer = hf_pipeline("summarization", model="facebook/bart-large-cnn")

long_text = """
Artificial intelligence (AI) has transformed numerous industries over the past decade.
In healthcare, AI algorithms now assist doctors in diagnosing diseases from medical images
with accuracy that rivals experienced specialists. In finance, machine learning models
detect fraudulent transactions in milliseconds, saving billions of dollars annually.
The transportation sector is being revolutionized by self-driving vehicles that use
computer vision and deep learning to navigate complex environments.

Natural language processing, a branch of AI, has enabled voice assistants like Siri,
Alexa, and Google Assistant to understand and respond to human speech. Large language
models such as GPT-4 and Claude can write essays, answer complex questions, and even
generate code. These advances have sparked debates about the future of work, with some
economists predicting significant job displacement while others argue AI will create
more jobs than it eliminates.

Despite these achievements, significant challenges remain. AI systems can perpetuate
and amplify biases present in training data. The environmental cost of training large
models is substantial. Ensuring the safety and alignment of increasingly powerful AI
systems is a major research priority for the field.
"""

summary = summarizer(long_text, max_length=120, min_length=40, do_sample=False)
print("\nAbstractive Summary (BART-CNN):")
print(summary[0]['summary_text'])

# ─── 5. ROUGE Evaluation ─────────────────────────────────────
# pip install rouge-score
from rouge_score import rouge_scorer

scorer = rouge_scorer.RougeScorer(['rouge1', 'rouge2', 'rougeL'], use_stemmer=True)

reference = "Artificial intelligence has transformed healthcare, finance, and transportation. Despite achievements, challenges remain including bias and environmental costs."
hypothesis = summary[0]['summary_text']

scores = scorer.score(reference, hypothesis)
print("\nROUGE Scores:")
for metric, score in scores.items():
    print(f"  {metric}: P={score.precision:.3f} R={score.recall:.3f} F={score.fmeasure:.3f}")