BanglaEcomReviewCorpus: A dataset for e-commerce product review sentiment analysis

Area of Science:

• Computational Linguistics
• Data Science
• E-commerce Analytics

Background:

• Customer feedback is vital for e-commerce success, influencing business strategies, service quality, and product innovation.
• Understanding consumer behavior and preferences is essential for online retailers.
• Analyzing diverse customer feedback requires comprehensive and well-structured datasets.

Purpose of the Study:

• To introduce a novel, publicly available dataset of 8685 labeled customer feedback items from various Bangladeshi e-commerce platforms.
• To facilitate advanced sentiment analysis and the exploration of consumer behavior in the e-commerce domain.
• To support interdisciplinary research in areas such as sociology, linguistics, and psychology.

Main Methods:

• Data collection from popular e-commerce websites (Daraz, Bikroy.com, Picabbo, Shajgoj, etc.).
• Labeling of 8685 feedback items into positive, negative, and neutral sentiment categories.
• Application of statistical analyses (summary statistics, histograms) and linguistic pattern extraction (unigrams, bigrams, trigrams).
• Utilizing visualizations like word clouds for structural and linguistic diversity insights.
• Ensuring data integrity through rigorous collection, anonymization, and preprocessing techniques.

Main Results:

• A balanced dataset with 3012 positive, 2881 negative, and 2792 neutral feedback entries.
• Identification of linguistic patterns and trends within customer reviews.
• Demonstration of the dataset's utility for natural language processing (NLP) tasks.
• Provision of insights into consumer behavior and preferences across different e-commerce platforms.

Conclusions:

• The dataset is a valuable resource for advancing sentiment analysis and improving business strategies in e-commerce.
• It enables deeper insights into customer behavior, aiding product and service development.
• The dataset supports academic teaching, artificial intelligence (AI) training, and collaborative research across multiple disciplines.