This lesson's interactive features are locked, please to keep using them

Semantic Search

Click to play video

Keyword and metadata search are useful, but they have some serious limits. Have you ever searched "baggins" on Netflix and gotten "The Hobbit"? Keyword search can't do that unless the word "baggins" appears in the title, description, cast, or other searchable metadata.

Semantic search can find results where words have similar meaning, not just similar spelling. It can find relevant documents even when they don't contain your exact search terms.

Bear Movie Search Example:

Query: "exciting adventure"

Keyword Search
- Searches for: "exciting" AND "adventure"
- Misses: Movies described as a "thrilling journey"
Semantic Search
- Understands: You want action-packed films
- Finds: Movies with "gripping," "suspenseful," "journey"

When Semantic Search Is Better

Synonym matching: "Happy movies" → finds "joyful" or "uplifting" films
Conceptual queries: "Movies about friendship" → finds films exploring character bonds
Natural language: "What bear movies are good for kids?" → understands family-friendly content

When Keyword Search Is Better

Exact terms: Medical codes or phrases like "COVID-19"
Proper nouns: Specific titles like "The Matrix"
Technical jargon: Programming terms like "dependency injection"

Assignment

Let's build semantic search. We'll use a new CLI script for semantic search operations.

In the cli directory, create a new file called semantic_search_cli.py and paste in the following code:

import argparse


def main() -> None:
    parser = argparse.ArgumentParser(description="Semantic Search CLI")
    args = parser.parse_args()

    match args.command:
        case _:
            parser.print_help()


if __name__ == "__main__":
    main()

Run and submit the CLI tests.