Browse Prior Art Database

Method and System for Automatically Identifying Query Triage using Vector Space Grouping of Multiple Queries

IP.com Disclosure Number: IPCOM000243993D
Publication Date: 2015-Nov-04
Document File: 2 page(s) / 19K

Publishing Venue

The IP.com Prior Art Database

Related People

Max Vladymyrov: INVENTOR [+3]

Abstract

Disclosed is a method and system for automatically identifying query triage using vector representation of multiple queries The method and system groups multiple queries using word embedding model to automatically identify potential error instances of queries to operators that would potentially benefit from query triage

This text was extracted from a Microsoft Word document.
This is the abbreviated version, containing approximately 52% of the total text.

Method and System for Automatically Identifying Query Triage using Vector Space Grouping of Multiple Queries

Abstract

Disclosed is a method and system for automatically identifying query triage using vector representation of multiple queries.  The method and system groups multiple queries using word embedding model to automatically identify potential error instances of queries to operators that would potentially benefit from query triage.

Description

Search engines serve billions of requests on a daily basis to satisfy the information needs of customers.  Such precise, high volume, search engine systems are best built using statistical models and machine learning algorithms.  However, for certain types/classes of queries, the underlying systems that power search engines do not return most relevant results.  During this time, operator intervention process (query triage) is needed to examine the system malfunctions for such queries and implement rectifications on the malfunctions.  Query triage is a user-labor-intensive process and is required every time a query is reported for triage. There is a need to automatically identify and present to operators queries that would potentially benefit from query triage that are related to the original malfunctioning query. 

 

In accordance with the method and system, a mapping of queries to a numerical feature space is built. This mapping can be represented as bag-of-words, part of speech tagging, n-grams, neural network or any other suited features.  The numeric features of the mapped queries represent different similarity information such as, for example, if query A is similar to B semantically, both the queries are placed close enough in the feature space. Conversely, if both the queries are located nearby in the feature space, partial semantic information is shared among both the queries.  Thereafter, class info...