Feedback Sentiment Analysis

• Topic: Customer Feedback
• This experiments Prompt Engineering from API OPEN AI to get inferencing (sentiment) and summary for User Feedback on Products

Dataset and Model

• Women Clothing Review from Kaggle: https://www.kaggle.com/datasets/nicapotato/womens-ecommerce-clothing-reviews
• Amount of dataset: 1200 rows
• Model applied for Prompt Engineering: gpt-turbo-3.5
• For visualizing, timeframe is added with purpose to explore more metrics on PowerBI related to the Customer Feedback

Expected Result from Prompt

This include 2 seperate process: Prompt Engineering and Data Prep

Prompt Engineering part

Based on review_text column, create a dataframe with 4 columns:

• Department: define 5 departments Marketing, Sales, Product, Logistics, and Inventory to categorize the reviews of customers
• sentiment: positive or negative
• emotions: extract 5 most relevant emotions
• summary: summarize the text within 10 words

Data Prep

After prompt engineering is done, we will merge the result into dataset. Lately, adding timeframe to dataset

Visualization

Using PowerBI for visualizing the insight and metrics