CatgirlIntelligenceAgency/code/features-convert/summary-extraction
2023-06-28 15:41:10 +02:00
..
src Small optimizations that shave an hour of processing time :D 2023-06-28 15:41:10 +02:00
build.gradle Consider keyword relevance signals when creating the document summary using the DOM walker. 2023-06-19 17:58:19 +02:00
readme.md Clean up summary extractor module. 2023-03-18 10:29:25 +01:00

Summary Extraction

This feature attempts to find a descriptive passage of text that summarizes what a search result "is about". It's the text you see below a search result.

It must solve two problems:

  1. Identify which part of the document that contains "the text". The crux is that the document may be anywhere from 1993 to the present, with era-appropriate formatting. It may be formatted with <center>ed <font>-tags, or semantic HTML5.

  2. Identify which part of "the text" best describes the document.

It uses several naive heuristics to try to find something that makes sense, and there is probably room for improvement.

There are many good techniques for doing this, but they've sadly not proved particularly fast. Whatever solution is used needs to be able to summarize of order of a 100,000,000 documents with a time budget of a couple of hours.

Central Classes