Claude Sentiment Analysis with Python

Working Python code to run sentiment analysis with the Claude API in 2026. Zero-shot classifier, custom sentiment scales, aspect-based analysis, and batch processing patterns.

Claude's language understanding makes it an accurate zero-shot sentiment classifier — no training data or fine-tuning required. This guide covers every sentiment analysis pattern from simple positive/negative to aspect-level JSON extraction.

Installation

pip install anthropic

Simple positive / negative / neutral classifier

import anthropic
import json

client = anthropic.Anthropic()

def analyze_sentiment(text: str) -> dict:
    response = client.messages.create(
        model="claude-haiku-4-5-20251001",  # cheapest model — sufficient for sentiment
        max_tokens=128,
        temperature=0,  # deterministic output
        system=(
            "You are a sentiment classifier. "
            "Return ONLY valid JSON with this schema: "
            '{"sentiment": "positive"|"negative"|"neutral", "confidence": 0.0-1.0}'
        ),
        messages=[{"role": "user", "content": f"Classify the sentiment of: {text}"}],
    )
    return json.loads(response.content[0].text)

# Example
result = analyze_sentiment("The delivery was fast but the product broke after two days.")
print(result)
# {"sentiment": "negative", "confidence": 0.82}

Five-point sentiment scale with rationale

def analyze_sentiment_detailed(text: str) -> dict:
    response = client.messages.create(
        model="claude-haiku-4-5-20251001",
        max_tokens=256,
        temperature=0,
        system=(
            "You are a sentiment analyst. Return ONLY valid JSON: "
            '{"score": 1-5, "label": "very negative"|"negative"|"neutral"|"positive"|"very positive", '
            '"confidence": 0.0-1.0, "rationale": "one sentence"}'
        ),
        messages=[{"role": "user", "content": text}],
    )
    return json.loads(response.content[0].text)

result = analyze_sentiment_detailed(
    "I've been using this app for months. It's mostly fine but crashes weekly and support ignores tickets."
)
print(result)
# {"score": 2, "label": "negative", "confidence": 0.88,
#  "rationale": "Crashes and unresponsive support outweigh functional adequacy."}

Aspect-based sentiment analysis (ABSA)

def aspect_sentiment(text: str, aspects: list[str]) -> dict:
    aspect_list = ", ".join(aspects)
    response = client.messages.create(
        model="claude-sonnet-4-6",  # better reasoning for multi-aspect
        max_tokens=512,
        temperature=0,
        system=(
            "You are an aspect-based sentiment analyst. "
            "Return ONLY valid JSON: a dict mapping each aspect to "
            '{"sentiment": "positive"|"negative"|"neutral"|"not_mentioned", "evidence": "quote or null"}'
        ),
        messages=[{
            "role": "user",
            "content": f"Aspects to evaluate: {aspect_list}

Review: {text}"
        }],
    )
    return json.loads(response.content[0].text)

review = (
    "The camera takes stunning photos in daylight but struggles at night. "
    "Battery lasts two days easily. The price is steep for what you get."
)
result = aspect_sentiment(review, ["camera", "battery", "price", "design"])
# {
#   "camera": {"sentiment": "positive", "evidence": "takes stunning photos in daylight"},
#   "battery": {"sentiment": "positive", "evidence": "lasts two days easily"},
#   "price": {"sentiment": "negative", "evidence": "steep for what you get"},
#   "design": {"sentiment": "not_mentioned", "evidence": null}
# }

Batch sentiment analysis (50% cost with Batch API)

import anthropic

client = anthropic.Anthropic()

reviews = [
    "Great product, fast shipping!",
    "Broke after a week. Very disappointed.",
    "Average quality, nothing special.",
    # ... thousands more
]

# Build batch requests
requests = [
    {
        "custom_id": f"review-{i}",
        "params": {
            "model": "claude-haiku-4-5-20251001",
            "max_tokens": 64,
            "temperature": 0,
            "system": 'Return ONLY JSON: {"sentiment": "positive"|"negative"|"neutral"}',
            "messages": [{"role": "user", "content": text}],
        },
    }
    for i, text in enumerate(reviews)
]

# Submit batch (50% cheaper than real-time, processes within 24h)
batch = client.messages.batches.create(requests=requests)
print(f"Batch ID: {batch.id}")  # poll this later

# Poll for results
import time
while True:
    status = client.messages.batches.retrieve(batch.id)
    if status.processing_status == "ended":
        break
    time.sleep(60)

results = {}
for result in client.messages.batches.results(batch.id):
    if result.result.type == "succeeded":
        data = json.loads(result.result.message.content[0].text)
        results[result.custom_id] = data["sentiment"]

print(results)

Domain-specific sentiment (financial / medical)

def financial_sentiment(headline: str) -> dict:
    """Classify financial news headlines as bullish / bearish / neutral."""
    response = client.messages.create(
        model="claude-haiku-4-5-20251001",
        max_tokens=64,
        temperature=0,
        system=(
            "You are a financial news sentiment analyst. "
            'Return ONLY JSON: {"sentiment": "bullish"|"bearish"|"neutral", "confidence": 0.0-1.0}'
        ),
        messages=[{"role": "user", "content": headline}],
    )
    return json.loads(response.content[0].text)

print(financial_sentiment("Fed signals rate pause amid cooling inflation"))
# {"sentiment": "bullish", "confidence": 0.79}

Claude vs VADER vs TextBlob comparison

Use the Claude API Cost Calculator to estimate cost at your review volume. For a general-purpose classifier (multi-class, not just sentiment), see the text classification guide. For bulk document summarization, see the summarization guide.

Frequently asked questions

Free tools

More examples