How to beginner · 3 min read

How to use structured outputs for sentiment analysis

Q: How to use structured outputs for sentiment analysis

Use the structured_output parameter in AI API calls to define a JSON schema for sentiment analysis, enabling the model to return sentiment labels and scores in a predictable format. This approach improves reliability and parsing ease compared to free-text responses.

Quick answer

Use the structured_output parameter in AI API calls to define a JSON schema for sentiment analysis, enabling the model to return sentiment labels and scores in a predictable format. This approach improves reliability and parsing ease compared to free-text responses.

PREREQUISITES

Python 3.8+
OpenAI API key (free tier works)
pip install openai>=1.0

Setup

Install the openai Python package and set your API key as an environment variable.

Install the SDK: pip install openai>=1.0
Set environment variable: export OPENAI_API_KEY='your_api_key' (Linux/macOS) or setx OPENAI_API_KEY "your_api_key" (Windows)

bash

pip install openai>=1.0

Step by step

This example uses the OpenAI SDK to send a chat completion request with a response_format JSON schema for sentiment analysis. The model returns sentiment and confidence in a structured JSON format.

python

import os
from openai import OpenAI

client = OpenAI(api_key=os.environ["OPENAI_API_KEY"])

schema = {
    "type": "object",
    "properties": {
        "sentiment": {
            "type": "string",
            "enum": ["positive", "neutral", "negative"]
        },
        "confidence": {
            "type": "number",
            "minimum": 0,
            "maximum": 1
        }
    },
    "required": ["sentiment", "confidence"]
}

messages = [
    {"role": "user", "content": "I love this product! It works perfectly."}
]

response = client.chat.completions.create(
    model="gpt-4o",
    messages=messages,
    response_format={"type": "json_schema", "json_schema": schema}
)

result = response.choices[0].message.content
import json
result_json = json.loads(result)
print("Sentiment:", result_json["sentiment"])
print("Confidence:", result_json["confidence"])

output

Sentiment: positive
Confidence: 0.95

Common variations

You can adapt this approach by:

Using different models like gpt-4o-mini for faster, cheaper inference.
Applying the same structured output schema with Anthropic's claude-3-5-sonnet-20241022 model using its system and messages parameters.
Implementing asynchronous calls or streaming if supported by your SDK.

python

import anthropic
import os

client = anthropic.Anthropic(api_key=os.environ["ANTHROPIC_API_KEY"])

schema = {
    "type": "object",
    "properties": {
        "sentiment": {
            "type": "string",
            "enum": ["positive", "neutral", "negative"]
        },
        "confidence": {
            "type": "number",
            "minimum": 0,
            "maximum": 1
        }
    },
    "required": ["sentiment", "confidence"]
}

response = client.messages.create(
    model="claude-3-5-sonnet-20241022",
    system="You are a sentiment analysis assistant. Return output strictly as JSON matching the schema.",
    messages=[{"role": "user", "content": "I love this product! It works perfectly."}],
    max_tokens=200
)

print("Sentiment:", response.content)

output

Sentiment: {"sentiment": "positive", "confidence": 0.95}

Troubleshooting

If the model returns free text instead of structured JSON, verify your response_format parameter is correctly set and the model supports structured outputs.
For JSON parsing errors, ensure the schema is valid JSON Schema and the model response matches it.
If confidence scores seem off, adjust prompt instructions or try a more capable model like gpt-4o or claude-3-5-sonnet-20241022.

✅

Key Takeaways

Use response_format with a JSON schema to get reliable sentiment analysis results.
Structured outputs simplify parsing and downstream processing compared to free-text responses.
Both OpenAI and Anthropic SDKs support structured outputs with similar JSON schema definitions.

Verified 2026-04 · gpt-4o, gpt-4o-mini, claude-3-5-sonnet-20241022

Verify ↗