NLP-Based Sentiment Analysis of E-Commerce Customer Feedback

Problem Statement

E-commerce websites receive large amounts of customer feedback about products and services. Manually analyzing this feedback is time-consuming. The objective of this practical is to apply Natural Language Processing (NLP) techniques to preprocess customer feedback and perform sentiment analysis to identify customer opinions.

Dataset Description

The dataset contains 300 synthetic customer feedback statements similar to reviews collected from an e-commerce website. The feedback covers areas such as product quality, delivery, packaging, payment, refund, website experience, and customer service. The data is stored as a raw text (.txt) file, with one customer feedback statement per line. 

Download Dataset ecommerce_customer_feedback   

VADER Technique

ADER → Valence Aware Dictionary and sEntiment Reasoner.

VADER is a lexicon-based sentiment analysis technique that identifies the emotional polarity of text using a predefined dictionary of words with sentiment values.

It does not require model training because it uses predefined sentiment scores associated with words and expressions.

For example:

excellent → Positive sentiment
good → Positive sentiment
poor → Negative sentiment
disappointed → Negative sentiment

VADER provides four scores:

Positive (pos) → Amount of positive sentiment.

Negative (neg) → Amount of negative sentiment.

Neutral (neu) → Amount of neutral sentiment.

Compound (compound) → Overall combined sentiment score ranging from -1 to +1.

VADER is a Lexicon-based means sentiment is identified using a predefined dictionary (lexicon) of words and their sentiment values, rather than training a machine-learning model.

For example:

Word	Sentiment
excellent	Positive
happy	Positive
good	Positive
poor	Negative
disappointed	Negative
terrible	Negative

If the feedback is:

“The product is excellent but delivery was terrible.”

VADER looks at sentiment-bearing words such as excellent and terrible, considers their predefined sentiment strengths and context, and calculates an overall sentiment score.

Ecommerce Customer Feedback Sentimental Analysis