⏳ Load Document (0 / 0 pages)

Semantic Search Demonstration

This is a demonstration of semantic search. We have a document with nearly 600 pages, which has been converted into a vector database. The document was embedded into 256-character chunks, with a maximum overlap of 25 characters.

How to Use

1. Enter a search term into the input field at the bottom.
2. The system will find the most similar chunk.
3. The result will display:
   • The page number where the chunk is located.
   • The similarity score in percentage.
   • A color-coded similarity indicator:
     • Green → Higher similarity.
     • Red → Lower similarity.
   • Text chunk found in original document

Important Notes

• The page numbers displayed in the system differ from those in the document’s footer, as the first five pages are unnumbered.
• The background color of the result window reflects the similarity score, making it easier to interpret.

Happy searching! 🚀