Feedback Categorization using AI (OpenAI)

Leia em Português

Objective

This POC aims to categorize customer comments using OpenAI.

A analysis will be done based on the text of the comment to assign a specific category that best describes the feedback.

The categories considered are:

• ATTENDANCE
• PAYMENT
• COMFORT
• WAITING_TIME
• PRODUCT
• PRODUCT_AVAILABILITY
• PRICE
• PRODUCT_QUALITY
• COMPLIMENT
• NONE
• INTERESTS
• FINANCIAL_PRODUCTS

The category property in the object will be filled by the IA after the analysis of the comment.

Input (comment)	Output (category)
I was very well attended by the sellers, all friendly and helpful.	ATTENDANCE
The store is well organized and comfortable, I liked the spacious area.	COMFORT
I bought a shoe that had a defect, but it was replaced quickly.	PRODUCT_QUALITY

A log will be generated with the number of tokens used and the cost in dollars, as below:

General Summary

• This code creates a Flask API with a route /categorization.
• The categorization logic is separated in the controller module.

Environment Configuration and Application Initialization

Prerequisites:

• Python 3.3+
• Get an api-key from the OpenAI site - https://platform.openai.com/api-keys

1. Create and activate the virtual environment

  python -m venv venv
  source venv/bin/activate  # Windows: `venv\Scripts\activate`

2. Install the dependencies

  pip install -r requirements.txt

3. Start the application

cd app
python app.py

4. Access the route

Access the route /categorization to perform categorization of comments:

Request (with cURL):

In the command line, replace <api-key> with your OpenAI api-key.

curl --location 'http://127.0.0.1:5000/categorization' \
--header 'api-key: <api-key>' \
--header 'model: gpt-4o' \
--header 'Content-Type: application/json' \
--data '{
    "categories": ["ATTENDANCE", "PAYMENT", "COMFORT", "WAITING_TIME", "PRODUCT", "PRODUCT_AVAILABILITY", "PRICE", "PRODUCT_QUALITY", "COMPLIMENT", "NONE", "INTERESTS", "FINANCIAL_PRODUCTS"],
    "comments": [
        {
            "id": 1,
            "comment": "I was very well attended by the sellers, all friendly and helpful."
        },
        {
            "id": 2,
            "comment": "The store is well organized and comfortable, I liked the spacious area."
        },
        {
            "id": 3,
            "comment": "I bought a shoe that had a defect, but it was replaced quickly."
        }
    ],
    "instructions": "You are an expert in categorizing comments. Categorize each comment with its respective category."
}'

Input Data

request body (JSON):

{
    "categories": ["ATTENDANCE", "PAYMENT", "COMFORT", "WAITING_TIME", "PRODUCT", "PRODUCT_AVAILABILITY", "PRICE", "PRODUCT_QUALITY", "COMPLIMENT", "NONE", "INTERESTS", "FINANCIAL_PRODUCTS"],
    "comments": [
        {
            "id": 1,
            "comment": "I was very well attended by the sellers, all friendly and helpful."
        },
        {
            "id": 2,
            "comment": "The store is well organized and comfortable, I liked the spacious area."
        },
        {
            "id": 3,
            "comment": "I bought a shoe that had a defect, but it was replaced quickly."
        }
    ],
    "instructions": "You are an expert in categorizing comments. Categorize each comment with its respective category."
}

Response:

{
    "data": [
        {
            "category": "ATTENDANCE",
            "comment": "I was very well attended by the sellers, all friendly and helpful.",
            "id": 1
        },
        {
            "category": "COMFORT",
            "comment": "The store is well organized and comfortable, I liked the spacious area.",
            "id": 2
        },
        {
            "category": "PRODUCT_QUALITY",
            "comment": "I bought a shoe that had a defect, but it was replaced quickly.",
            "id": 3
        }
    ],
    "ok": true
}

Observations

Mock Data File: More examples of comments without categorization can be found in the mockdata.json file.

Required Headers: Ensure that the headers api-key and model are included with appropriate values.

Customization: The list of categories and comments can be adapted according to the project's needs.