Build AI Apps with Python: Product Review Analyzer with JSON | Episode 5

Build AI Apps with Python: Product Review Analyzer with JSON

In this tutorial, we create a product review analyzer that processes any review text and returns a structured JSON response containing sentiment, rating, key points, and a summary. By crafting a precise system prompt and setting the AI temperature to 0.0, we ensure consistent, parseable outputs that can be easily handled in Python.

Code

import openai
import json

# Define a detailed system prompt to enforce JSON output format
system_prompt = """
You are a product review analyzer. Given a product review, respond ONLY with a JSON object in the following format:
{
  "sentiment": "positive" | "neutral" | "negative",
  "rating": 1-5,
  "key_points": ["point1", "point2", "..."],
  "summary": "A concise summary of the review."
}
Do not add any other text or explanation.
"""

def analyze_review(review_text):
  # Compose messages for the chat completion
  messages = [
    {"role": "system", "content": system_prompt},
    {"role": "user", "content": review_text}
  ]

  # Call OpenAI API with temperature=0.0 for deterministic output
  response = openai.ChatCompletion.create(
    model="gpt-4o-mini",
    messages=messages,
    temperature=0.0
  )

  # Extract the JSON string from the AI response
  json_str = response.choices[0].message.content

  # Parse the JSON string into a Python dictionary
  result = json.loads(json_str)

  return result

# Example usage with a sample review
sample_review = "This product exceeded my expectations! The build quality is excellent and it works flawlessly."
analysis = analyze_review(sample_review)

# Pretty-print the structured analysis
print(json.dumps(analysis, indent=2))

Key Points

• Use a detailed system prompt to strictly define the JSON output format for AI responses.
• Set temperature=0.0 in the API call to ensure consistent and reliable structured outputs.
• Parse the AI's JSON response using Python's json.loads() for easy data manipulation.
• Pretty-print parsed JSON with json.dumps(indent=2) for readable output.
• This approach enables predictable, structured data extraction from freeform product reviews.