Amazon Product Autocomplete

This notebook provides a solution for providing product recommendations based on the search query from a user. The notebook implements two approachs:

• Keyword based search using TheFuzz
• Semantic search using sentence-transformers and Faiss

The steps implemented in this notebook are:

• Data loading and basic analysis
• Data cleaning
• Data preprocessing
• Search implementation
• Testing

For testing, we have considered three strings,

• "Fire TV" straightforward keyword based search
• "A birthday gift for kids party" for semantic based search
• "Dire tablet" for misspelling text search

Further improvements needed:

• Optimize search
• Semantic search needs improvement. Currently, it is done only on product names (extracted from the data); ideally it should be done on all names
• More comprehensive testing