Building Knowledge Graphs from Unstructured Text with Claude

You have a pile of unstructured documents and need to answer questions that span them — "who works with people who worked on project X", "which vendors are connected to this incident". No single document contains the answer. RAG retrieval won't chain the facts for you. You need a knowledge graph: entities as nodes, typed relations as edges, so that multi-hop reasoning becomes graph traversal.

Building one used to mean training a named-entity recognizer on your domain, training a relation classifier, writing entity-resolution heuristics, and maintaining all three as your data shifted. With Claude, each of those stages becomes a prompt.

What You'll Learn

By the end of this guide you will be able to:

• Use structured outputs to extract typed entities and subject–predicate–object triples from arbitrary text with no training data
• Apply Claude-driven entity resolution to collapse surface-form variants into canonical nodes, replacing brittle string-similarity heuristics
• Assemble and query an in-memory graph, and run multi-hop questions by serializing subgraphs back to Claude
• Measure extraction quality with precision/recall against a gold set and reason about the cost/quality tradeoff between Haiku and Sonnet

Everything runs in memory with no database. The techniques transfer directly to Neo4j, Neptune, or a Postgres adjacency table when you need to scale.

Prerequisites

• Python 3.11+
• Anthropic API key (get one here)
• Basic familiarity with graphs (nodes, edges, traversal)

Setup

We use two models. Haiku handles the high-volume, schema-constrained extraction work where speed and cost matter more than nuance. Sonnet handles entity resolution and summarization, where the model needs to weigh conflicting evidence across documents.

import anthropic
from pydantic import BaseModel, Field
from typing import List, Optional
client = anthropic.Anthropic()
Define your extraction schema
class Entity(BaseModel):
    name: str = Field(description="The canonical name of the entity")
    type: str = Field(description="Entity type: PERSON, ORGANIZATION, LOCATION, EVENT, etc.")
    description: str = Field(description="One-line description for disambiguation")
class Relation(BaseModel):
    subject: str = Field(description="Name of the subject entity")
    predicate: str = Field(description="Relation type (e.g., 'works_at', 'launched', 'commanded')")
    object: str = Field(description="Name of the object entity")
class Extraction(BaseModel):
    entities: List[Entity]
    relations: List[Relation]

Building a Corpus

We need a handful of documents that talk about overlapping entities, so that entity resolution has real work to do. The Apollo program is a good test bed: six short Wikipedia summaries that all mention NASA, the Moon, several astronauts, and a launch vehicle — but each article names them slightly differently.

We fetch summaries from the Wikipedia REST API rather than full articles to keep token costs low. For a production pipeline you would chunk full documents; the extraction logic is identical.

Entity and Relation Extraction

Classical NER tags spans of text with labels (PERSON, ORG, LOC). Classical relation extraction then classifies pairs of spans into relation types. Both traditionally require labeled training data per domain.

We collapse both stages into a single Claude call per document. The key is structured outputs: we define the output shape as a Pydantic model and pass it to client.messages.parse(). Claude's response is guaranteed to validate against that schema and comes back as a typed Python object — no regex parsing, no JSON decode errors, no defensive isinstance checks.

def extract_from_document(text: str) -> Extraction:
    response = client.messages.parse(
        model="claude-3-haiku-20240307",
        max_tokens=2000,
        system="You are an expert information extraction system. Extract all named entities and their relationships from the text. Be thorough but precise.",
        messages=[{"role": "user", "content": f"Extract entities and relations from this text:\n\n{text}"}],
        response_model=Extraction
    )
    return response

Let's look at what was extracted. Notice how the same real-world entity appears under different surface forms across documents — this is the entity resolution problem we solve next.

Entity Resolution

The raw extraction gives us overlapping mentions: "NASA" and "National Aeronautics and Space Administration", "Neil Armstrong" and "Armstrong", possibly "the Moon" and "Moon". If we build a graph directly from this, we get a fractured mess where the same concept is split across disconnected nodes.

Traditional approaches use string similarity (edit distance, Jaccard on tokens) plus blocking rules. That works for typos but fails on "Edwin Aldrin" vs "Buzz Aldrin" — two names with zero character overlap that refer to the same person.

We instead ask Claude to cluster entities of each type, using the one-line descriptions from extraction as disambiguation context. The descriptions matter: "Armstrong — first person to walk on the Moon" and "Armstrong — jazz trumpeter" have the same name but should not merge.

def resolve_entities(entities: List[Entity]) -> dict:
    """Cluster entities by type and return alias->canonical mapping."""
    # Group entities by type
    by_type = {}
    for e in entities:
        by_type.setdefault(e.type, []).append(e)
    
    alias_to_canonical = {}
    
    for entity_type, group in by_type.items():
        # Prepare the prompt with descriptions for disambiguation
        entity_list = "\n".join([f"- {e.name}: {e.description}" for e in group])
        prompt = f"""Cluster these {entity_type} entities into groups that refer to the same real-world entity.
Use the descriptions for disambiguation.
{entity_list}
Return a JSON mapping from each surface form to its canonical name."""
        
        response = client.messages.parse(
            model="claude-3-sonnet-20240229",
            max_tokens=1000,
            system="You are an entity resolution system. Merge variants that refer to the same entity.",
            messages=[{"role": "user", "content": prompt}],
            response_model=dict
        )
        alias_to_canonical.update(response)
    
    return alias_to_canonical

Two failure modes to watch for. First, any raw name Claude leaves out of every cluster silently disappears from the graph — a production resolver should fall back to a single-element cluster for unmatched names so nothing is lost. Second, the resolver can over-merge: a specific mission like "Gemini 12" may get folded into the broader "Project Gemini" because the descriptions overlap. The first loses nodes, the second loses precision. Both are worth spot-checking in the output below.

Assembling the Graph

With a clean alias map, we rewrite every relation endpoint to its canonical form and load the result into NetworkX. We use a MultiDiGraph because two entities can be connected by several distinct predicates ("launched from" and "operated by"), and direction matters ("Armstrong commanded Apollo 11" is not the same edge as "Apollo 11 commanded Armstrong").

Each node carries its type, the source document IDs, and the original surface forms — metadata that becomes critical when you need to explain how you know something.

import networkx as nx
def build_graph(extractions: List[Extraction], alias_map: dict) -> nx.MultiDiGraph:
    G = nx.MultiDiGraph()
    
    for extraction in extractions:
        for entity in extraction.entities:
            canonical = alias_map.get(entity.name, entity.name)
            G.add_node(canonical, type=entity.type, description=entity.description)
        
        for relation in extraction.relations:
            subj = alias_map.get(relation.subject, relation.subject)
            obj = alias_map.get(relation.object, relation.object)
            G.add_edge(subj, obj, predicate=relation.predicate)
    
    return G

Querying the Graph

Now the fun part. With a clean graph, you can answer multi-hop questions by traversing edges. For complex questions, serialize the relevant subgraph back to Claude for reasoning.

def query_graph(G: nx.MultiDiGraph, question: str) -> str:
    # Simple traversal for direct questions
    if "who worked on" in question.lower():
        # Extract entity name and find connected nodes
        # ... traversal logic ...
        pass
    
    # For complex questions, serialize subgraph and ask Claude
    subgraph_text = serialize_subgraph(G, relevant_nodes)
    
    response = client.messages.create(
        model="claude-3-sonnet-20240229",
        max_tokens=1000,
        system="You are a knowledge graph query engine. Answer questions based only on the provided graph data.",
        messages=[{
            "role": "user", 
            "content": f"Graph data:\n{subgraph_text}\n\nQuestion: {question}"
        }]
    )
    return response.content[0].text

Measuring Quality

To trust your graph, you need to measure precision and recall against a gold standard. Create a small annotated set of documents with known entities and relations, then compare Claude's extraction against it.

def evaluate_extraction(gold: Extraction, predicted: Extraction) -> dict:
    gold_entities = set((e.name, e.type) for e in gold.entities)
    pred_entities = set((e.name, e.type) for e in predicted.entities)
    
    true_positives = gold_entities & pred_entities
    false_positives = pred_entities - gold_entities
    false_negatives = gold_entities - pred_entities
    
    precision = len(true_positives) / (len(true_positives) + len(false_positives)) if pred_entities else 0
    recall = len(true_positives) / (len(true_positives) + len(false_negatives)) if gold_entities else 0
    
    return {
        "precision": precision,
        "recall": recall,
        "f1": 2  precision  recall / (precision + recall) if (precision + recall) > 0 else 0
    }

Cost/Quality Tradeoffs

Haiku costs ~$0.25 per million input tokens and handles extraction well for clean text. Sonnet costs ~$3 per million input tokens but catches nuanced relations and handles entity resolution better. A typical pipeline uses Haiku for extraction (the high-volume step) and Sonnet for resolution and querying (the reasoning-heavy steps).

For production, consider:

• Batch extraction with Haiku to keep costs low
• Sonnet for entity resolution where context matters
• Caching extraction results to avoid re-processing
• Incremental updates — only re-extract changed documents

Key Takeaways

• No training data needed: Claude's structured outputs let you define entity and relation schemas with Pydantic models, eliminating the need for labeled training data or custom NLP pipelines.
• AI-driven entity resolution beats heuristics: Claude can resolve "Buzz Aldrin" vs "Edwin Aldrin" using semantic context, not just string similarity — catching cases that traditional approaches miss.
• Multi-hop reasoning becomes graph traversal: Once entities are resolved and relations are typed, complex questions that span multiple documents become simple graph queries or subgraph serializations for Claude to reason over.
• Measure before you trust: Always evaluate precision and recall against a small gold set. The cost/quality tradeoff between Haiku and Sonnet means you should use Haiku for high-volume extraction and Sonnet for resolution and reasoning.
• Start in memory, scale to databases: The techniques shown here with NetworkX transfer directly to Neo4j, Neptune, or Postgres when your graph grows beyond memory limits.