Intermediate

Document Summarization

Generate accurate summaries and extract key information from long documents using extractive, abstractive, and LLM-based approaches.

Summarization Approaches

Approach	Method	Pros	Cons
Extractive	Select key sentences	Faithful, no hallucination	May lack coherence
Abstractive	Generate new text	Natural, concise	Risk of hallucination
Hybrid	Extract then refine	Balanced accuracy/fluency	More complex pipeline
Map-Reduce	Chunk, summarize, merge	Handles long documents	May lose cross-section context

LLM-Based Summarization

Python — Document Summarization Pipeline

from openai import OpenAI

client = OpenAI()

def summarize_document(text, summary_type="executive"):
    """Generate a document summary using an LLM."""
    prompts = {
        "executive": "Write a 3-5 sentence executive summary focusing on key decisions, findings, and action items.",
        "technical": "Write a technical summary covering methodology, data, results, and limitations.",
        "bullet": "Extract the 5-10 most important points as bullet points.",
        "legal": "Summarize key obligations, rights, dates, and risk areas."
    }

    response = client.chat.completions.create(
        model="gpt-4o",
        messages=[{
            "role": "system",
            "content": f"You are a document summarization expert. {prompts[summary_type]} Only include information explicitly stated in the document."
        }, {
            "role": "user",
            "content": f"Summarize this document:\n\n{text}"
        }],
        temperature=0.3  # Lower temperature for factual accuracy
    )
    return response.choices[0].message.content

def summarize_long_document(text, chunk_size=3000):
    """Map-reduce summarization for long documents."""
    chunks = [text[i:i+chunk_size] for i in range(0, len(text), chunk_size)]

    # Map: summarize each chunk
    chunk_summaries = [summarize_document(chunk, "bullet") for chunk in chunks]

    # Reduce: synthesize chunk summaries into final summary
    combined = "\n\n".join(chunk_summaries)
    return summarize_document(combined, "executive")

Key Information Extraction

Named entities: People, organizations, dates, monetary amounts, and locations mentioned in the document.
Key dates: Deadlines, effective dates, expiration dates, and milestones.
Action items: Tasks, responsibilities, and commitments extracted from meeting notes and reports.
Financial figures: Revenue, costs, projections, and financial metrics from business documents.
Risk factors: Identified risks, warnings, and compliance concerns from legal and regulatory documents.

⚠

Verify critical summaries: LLM summaries can hallucinate details not in the source document. For legal, financial, or medical documents, always implement a verification step comparing summary claims against the source text.

← PreviousClassification Next →Workflows