Multi-Stage Query Processing at Google
Deciding how a term or phrase is used in the context of a page can help decide how relevant that page is for a query from a searcher.
A patent application from Google this week looks at ways to consider the context of those words. It also describes a multi-stage query processing approach to determine relevancy and find results in a search.
The document is complex, but possible actions that can be taken during the different stages are:
Multi-Stage Query Processing Stage 1:
a) Deletion of stop words.
b) Term Stemming
c) Expanding queries to use synonyms and related terms that often co-occur with them.
d) Relevancy scores between a query and each document using one or more scoring algorithms, such as:
….the presence or absence of query term(s), term frequency, Boolean logic fulfillment, query term weights, the popularity of the documents (e.g., a query independent score of the document’s importance or popularity or interconnectedness), the proximity of the query terms to each other, context, attributes, etc.
Multi-Stage Query Processing Stage 2:
Adjacency and Proximity of terms are used to rank documents.
Multi-Stage Query Processing Stage 3:
Term attributes, such as whether terms are in titles, headings, metadata, and have certain font characteristics, are reviewed.
Multi-Stage Query Processing Stage 4:
Generation of snippets to return with results
Other relevance feedback algorithms might be used, such as:
….pseudo-relevance feedback algorithms based on a full document approach (pseudo relevance feedback based on a whole web page), Document Object Model (DOM) segmentation, Vision-based Page Segmentation (VIPS), conceptual relevance feedback using concept lattices, etc.
Multi-stage query processing system and method for use with tokenspace repository
The patent application
Inventors: Jeffrey Adgate Dean, Paul G. Haahr, Olcan Sercinoglu, and Amitabh K. Singhal
US Patent Application 20060036593
Filed: August 13, 2004
Published February 16, 2006
Abstract:
A multi-stage query processing system and method enables multi-stage query scoring, including “snippet” generation, through incremental document reconstruction facilitated by a multi-tiered mapping scheme. At one or more stages of a multi-stage query processing system, a set of relevancy scores is used to select a subset of documents for presentation as an ordered list to a user. The set of relevancy scores can be derived in part from one or more sets of relevancy scores determined in prior stages of the multi-stage query processing system. In some embodiments, the multi-stage query processing system is capable of executing one or more passes on a user query and using information from each pass to expand the user query for use in a subsequent pass to improve the relevancy of documents in the ordered list.
Last Updated July 4, 2019.
The approach is quite interesting. But, I guess, Google must have modified and updated its algorithms to determine relevancy quite a bit since they filed this patent back in 2006.
Hi Robert,
Thanks.
Google has changed around a number of the things they do, and their approaches since this patent application was published, but it’s quite possible that Google is using a multi-staged approach to processing queries that’s probably still similar to what is described in that document. I know they do treat stop words differently, and there may be some other stages as well know, but the basic idea, that queries can be expanded in a number of ways, such as finding appropriate synonyms, and so on, is probably still on point.