TensorLearn
Back to Course
NLP Specialist: BERT & Beyond
Module 10 of 11

10. Semantic Search

1. Keyword vs Semantic

  • Keyword: "Python" matches "Python Scope".
  • Semantic: "Snake programming" matches "Python".

2. Bi-Encoders vs Cross-Encoders

  • Bi-Encoder: Fast. Pre-compute vector for every doc. Compare cos(A, B).
  • Cross-Encoder: Slow but accurate. Feed (Doc, Query) into BERT together to get a score. Use Bi-Encoder to retrieve Top 100, then Cross-Encoder to re-rank Top 10.

Mark as Completed

TensorLearn - AI Engineering for Professionals