Code beginner · 3 min read

How to do sentiment analysis with Python

Direct answer

Use the OpenAI Python SDK to send text to a chat model like gpt-4o with a prompt instructing it to classify sentiment, then parse the response from response.choices[0].message.content.

Setup

Install

bash

pip install openai

Env vars

OPENAI_API_KEY

Imports

python

import os
from openai import OpenAI

Examples

inI love this product! It works perfectly.

outSentiment: Positive

inThe service was okay, nothing special.

outSentiment: Neutral

inI'm very disappointed with the quality.

outSentiment: Negative

Integration steps

Import the OpenAI SDK and initialize the client with the API key from os.environ
Create a chat completion request with model 'gpt-4o' and a prompt asking for sentiment classification
Send the user text as a message in the chat completion request
Receive the response and extract the sentiment label from response.choices[0].message.content
Print or use the sentiment result in your application

Full code

python

import os
from openai import OpenAI

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

text_to_analyze = "I love this product! It works perfectly."

messages = [
    {"role": "user", "content": f"Classify the sentiment of this text as Positive, Neutral, or Negative:\n\n{text_to_analyze}"}
]

response = client.chat.completions.create(
    model="gpt-4o",
    messages=messages
)

sentiment = response.choices[0].message.content.strip()
print(f"Sentiment: {sentiment}")

output

Sentiment: Positive

API trace

Request

json

{"model": "gpt-4o", "messages": [{"role": "user", "content": "Classify the sentiment of this text as Positive, Neutral, or Negative:\n\nI love this product! It works perfectly."}]}

Response

json

{"choices": [{"message": {"content": "Positive"}}], "usage": {"prompt_tokens": 20, "completion_tokens": 1, "total_tokens": 21}}

Extractresponse.choices[0].message.content

Variants

Streaming sentiment analysis ›

Use streaming to get partial sentiment results faster or for UI with live updates.

python

import os
from openai import OpenAI

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

text_to_analyze = "I love this product! It works perfectly."

messages = [
    {"role": "user", "content": f"Classify the sentiment of this text as Positive, Neutral, or Negative:\n\n{text_to_analyze}"}
]

stream = client.chat.completions.create(
    model="gpt-4o",
    messages=messages,
    stream=True
)

sentiment = ""
for chunk in stream:
    delta = chunk.choices[0].delta.content or ""
    sentiment += delta
print(f"Sentiment: {sentiment.strip()}")

Async sentiment analysis ›

Use async when integrating sentiment analysis into asynchronous Python applications or web servers.

python

import os
import asyncio
from openai import OpenAI

async def analyze_sentiment():
    client = OpenAI(api_key=os.environ["OPENAI_API_KEY"])
    text_to_analyze = "I love this product! It works perfectly."
    messages = [
        {"role": "user", "content": f"Classify the sentiment of this text as Positive, Neutral, or Negative:\n\n{text_to_analyze}"}
    ]
    response = await client.chat.completions.acreate(
        model="gpt-4o",
        messages=messages
    )
    sentiment = response.choices[0].message.content.strip()
    print(f"Sentiment: {sentiment}")

asyncio.run(analyze_sentiment())

Using Anthropic Claude for sentiment analysis ›

Use Anthropic Claude models if you prefer Claude's style or have an Anthropic API key.

python

import os
from anthropic import Anthropic

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

text_to_analyze = "I love this product! It works perfectly."

response = client.messages.create(
    model="claude-3-5-sonnet-20241022",
    system="You are a helpful assistant that classifies sentiment.",
    messages=[{"role": "user", "content": f"Classify the sentiment of this text as Positive, Neutral, or Negative:\n\n{text_to_analyze}"}]
)

sentiment = response.content.strip()
print(f"Sentiment: {sentiment}")

Performance

Latency~800ms for gpt-4o non-streaming

Cost~$0.002 per 500 tokens for gpt-4o

Rate limitsTier 1: 500 RPM / 30K TPM

Keep prompts concise to reduce token usage.
Avoid sending unnecessary system messages for simple classification.
Cache repeated sentiment results for identical inputs.

Approach	Latency	Cost/call	Best for
Standard chat completion	~800ms	~$0.002	Simple sentiment classification
Streaming chat completion	~500ms initial	~$0.002	UI with live updates
Async chat completion	~800ms	~$0.002	Concurrent or web apps
Anthropic Claude	~900ms	Check Anthropic pricing	Alternative model style

✓

Quick tip

Frame your prompt clearly to classify sentiment as 'Positive', 'Neutral', or 'Negative' for consistent results.

⚠

Common mistake

Beginners often forget to strip whitespace from the model's response, leading to parsing errors.

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

Verify ↗

Community Notes

No notes yetBe the first to share a version-specific fix or tip.