Explore the inner workings of Spotify's podcast search functionality and learn how to implement a similar system in this 50-minute video. Dive into the world of natural language processing in semantic search, understanding the rise of transformer models and sentence transformers. Discover how Spotify built their podcast search, including data sourcing, fine-tuning, and evaluation processes. Follow along with code implementation, dataset preparation, query generation, and model fine-tuning specific to podcasts. Evaluate the system's performance and consider scalability issues. Gain insights into sharing your work and implementing advanced search capabilities in the growing podcast market.
Overview
Syllabus
Intro
NLP in Semantic Search
Why Now?
Transformer Models
Sentence Transformers
Vector Search
How Spotify Built Podcast Search
Data Source, Fine-tuning, and Eval
Code Implementation, Dataset
Data Preparation
Query Generation
Fine-tuning a Podcast Model
Evaluation
Does it Scale?
Sharing Your Work
Taught by
James Briggs
