Open

I am K R S Alpana, a second-year engineering student at R V College of Engineering, Bengaluru, India. I am pursuing a Bachelor of Engineering in Computer Science with specialization in Data Science. My strong academic record demonstrates my consistency and work ethic. I have developed proficiency in C, C++, and Python programming languages. My interests lie in machine learning, and I have practical knowledge of various models and their applications, which I've implemented in my projects. Beyond academics, I am an active member of clubs like TEDx RVCE, which has enhanced my teamwork and collaboration skills. This combination of technical knowledge and soft skills positions me well for future opportunities in the field of Computer Science and Data Science.

Poster Abstract

In the ever-evolving e-commerce sector, understanding customer sentiment is vital for clothing brands to maintain competitiveness and enhance customer satisfaction. E-commerce brands, specifically, can employ sentiment analysis understand overall consumer preferences and trends across their vast collections, using which they can tailor their marketing strategies to enhance the shopping experience and drive purchases. Through our research, we have found very little deployment of modern NLP models for sentiment analysis for Indian e-commerce reviews. This project explores the application of smart sentiment analysis techniques to reviews from leading Indian clothing brands like Myntra, Nykaa and Ajio, utilizing the HPCC Systems platform for large-scale data analytics.

The model makes use of BERT (Bidirectional Encoder Representations from Transformers) to accurately classify and interpret sentiments from customer reviews.

The methodology begins with creating a comprehensive training dataset of e-commerce clothing reviews tailored to the Indian market. This dataset captures the nuances of Indian linguistic patterns and context-specific sentiments. HPCC Systems' text analytics capabilities can be leveraged to clean and preprocess this unstructured data, addressing challenges such as regional language variations, slang, and user-generated content.

BERT model shall be trained across multiple nodes, significantly reducing training time and allowing for efficient hyperparameter tuning. This approach will help in handling large volumes of data and iterate quickly on improvements in the model.

For validation, trained model will be applied to a test set of Myntra product reviews. The results can be analyzed to extract key findings, including identifying brands associated with high and low ratings, determining main factors contributing to negative ratings, and distinguishing between issues related to the e-commerce platform and those specific to the brands, like fitting or quality related issues.

HPCC Systems can be used to deploy the model for real-time scoring of new reviews. This allows an immediate integration of insights into the platforms, enabling a rapid response to customer feedback and the ever-changing market trends.

These findings provide valuable insights for competitive analysis and strategic marketing, enabling e-commerce brands to refine their product offerings and promotional strategies based on customer sentiment. The real-time nature of the analysis allows for swift identification of emerging trends and potential issues, allowing brands to proactively address customer concerns.

A multilingual analysis to incorporate Indian regional languages, integration with complementary models, and exploration of more advanced models for potentially higher accuracy in sentiment analysis.

By combining the massive processing power of HPCC systems platform with state-of-the-art NLP models, the model provides Indian e-commerce brands with a sophisticated tool for understanding and responding to customer sentiment, ultimately driving improvements in product quality, customer satisfaction, and market positioning.

Presentation

In this Video Recording, Kuppilli provides a tour and explanation of her poster content.

Sentiment Analysis of Indian E-commerce Clothing Brand Reviews using BERT:

Click on the poster for a larger image.

Open