Fashion Product Retrieval Using Semantic Search and Natural Language Generation

Title

Team Name

InfoSphere

Email

202318007@daiict.ac.in

Team Member 1 Name

Kavisha Madani

Team Member 1 Id

202318007

Team Member 2 Name

Vishaka Nair

Team Member 2 Id

202318041

Team Member 3 Name

Srushti Bhagchandani

Team Member 3 Id

202318047

Team Member 4 Name

Shubham Gupta

Team Member 4 Id

202318052

Problem Statement

This project aims to build a next-generation fashion product information retrieval and recommendation system by combining Natural Language Understanding (NLU) and semantic search techniques using advanced AI models like BERT and Sentence-BERT. The system will intelligently process user queries, detect their intent (e.g., event-based clothing recommendations like "traditional day" or "casual wear"), and identify relevant entities using NLU techniques. By leveraging neural search models and integrating generative AI (GenAI), the system will generate human-friendly, personalized responses while retrieving semantically aligned products from the H&M fashion dataset.

The project will further explore the integration of Large Language Models (LLMs) to enhance product recommendations, improving the user experience with personalized suggestions. This innovative approach has the potential to advance research in the AI/ML domain, making it suitable for publication in fashion technology or AI research journals.

Evaluation Strategy

Retrieval Accuracy: Measure the accuracy of retrieved products in terms of relevance to the user query (using human evaluation or ground truth labels).
Response Quality: Evaluate the quality of the generated responses using metrics like BLEU score and user satisfaction surveys.
Latency: Measure the time taken for query processing, product retrieval, and response generation.

Dataset

H&M Fashion Dataset

Resources

https://medium.com/@kumawatayushi31/heuristic-evaluation-for-fashion-gpt-on-myntra-a-ux-case-study-a5b4b70006a

parth126 / IT550