A keyword search for "warm jacket for hiking" misses a product titled "insulated trail parka". A semantic product recommender fixes that by comparing meaning instead of exact words. Each product description is turned into an embedding, a list of numbers that places similar products close together, so you can recommend the nearest ones.
The lessons start with semantic search and how sentence embeddings capture meaning. Next you work with the Sentence Transformers library to compute embeddings and similarity. Then you learn how to store and query them in a vector database such as FAISS, Chroma or Pinecone, and finally see how vector embeddings are used in real e-commerce search.
Watch the lessons in order and code along in Google Colab or a Jupyter notebook. For your own version, grab a product dataset (for example from Kaggle), embed the titles and descriptions, and return the top similar products for any item or search phrase.
A beginner-friendly tutorial on semantic vs keyword search, how sentence embeddings and vector similarity work, and a Python implementation with Sentence Transformers.
Video 19minUse Sentence Transformers to create sentence embeddings, measure similarity between texts, run semantic search and cluster similar items.
Video 33minA guide to storing and searching embeddings at scale with three popular vector databases: FAISS, Chroma and Pinecone.
Video 52minHow vector embeddings are used in search applications, with examples based on e-commerce product search.
Video 15min