Utilizing Sentence Function Tagging to Retrieve Related Text Excerpts

摘要

A method is described for utilizing a fast, rule-based thematic tagger on excerpts of texts within a SERP (Search Engine Results Page) as a context model for finding snippets from deeper in the SERP that are relevant to those found higher up. This method enables a "more like this" functionality which is found to have accuracy materially better that that accomplished merely by keyword matching, without introducing too much latency for the user.