Browse Prior Art Database

Method and System for Organizing Potential Search Suggestions in a Database Using a Novel Multi-Token Indexing Mechanism

IP.com Disclosure Number: IPCOM000242948D
Publication Date: 2015-Sep-02
Document File: 3 page(s) / 24K

Publishing Venue

The IP.com Prior Art Database

Related People

Zhongqiang Chen: INVENTOR [+7]

Abstract

A method and system for organizing potential search suggestions in a database using a novel multi-token indexing mechanism is disclosed. The multi-token indexing mechanism helps to quickly retrieve suggestion candidates for user submitted search query. The suggestion candidates are ranked according to a variety of features including number of tokens shared with user's query as well as length and occurring frequency of candidate suggestions.

This text was extracted from a Microsoft Word document.
This is the abbreviated version, containing approximately 36% of the total text.

Method and System for Organizing Potential Search Suggestions in a Database Using a Novel Multi-Token Indexing Mechanism

Abstract

A method and system for organizing potential search suggestions in a database using a novel multi-token indexing mechanism is disclosed.  The multi-token indexing mechanism helps to quickly retrieve suggestion candidates for user submitted search query.  The suggestion candidates are ranked according to a variety of features including number of tokens shared with user's query as well as length and occurring frequency of candidate suggestions.

Description

In order to provide search assistance to end users search engine typically maintains a huge database of potential suggestion candidates that are usually derived from search query logs or fed from other sources.  Whenever user-submitted queries are received, suggestions are fetched from the database.  Meanwhile, it is desired that a search engine strive to achieve excellent coverage and relevance in search suggestions to enhance search experience for a user.  However, it is challenging to identify relevant suggestions for a given query, from a large database at an extremely fast speed.  The stringent requirement on timely response to a user query renders it impractical to naively search an entire database from beginning to end in order to discover suggestion candidates that are of good quality.

Disclosed is a method and system for organizing potential search suggestions in a database using a novel multi-token indexing mechanism.  The database is partitioned into multiple segments that are prioritized or ranked according to importance.  Therafter, every suggestion within each partition is treated as a document and indices are built based on multiple tokens.  When suggestions are needed for a given query, a set of multi-token keys for the query are formed and are used to search for suggestion candidates from partitions in due order based on prioritized order of the suggestion candidates.  The multi-token indexing mechanism helps to quickly retrieve suggestion candidates for user submitted search query.

In accordance with the method and system, a multi-token indexing mechanism consists of two subsystems namely a offline subsystem and a online subsystem.  In a offline subsystem a database of potential suggestions is partitioned according to importance, which is a function of multiple features including number of tokens shared with query, occurring frequency, length of suggestions, orders of matched tokens in query and suggestion and distribution of gaps between matched tokens in a suggestion.  Each suggestion within a partition is tokenized and the suggestion is indexed with a set of keys, which are formed by all possible combination of n tokens among all the tokens for the suggestion.  The method and system facilitates merging of indices for all suggestions with a partition to form the indexing structure.  The resulting indexing structure may be  pruned...