Dismiss
InnovationQ will be updated on Sunday, Oct. 22, from 10am ET - noon. You may experience brief service interruptions during that time.
Browse Prior Art Database

Method and System for Providing Personalized Term Suggestions Using Bloom Filters

IP.com Disclosure Number: IPCOM000241870D
Publication Date: 2015-Jun-05
Document File: 2 page(s) / 35K

Publishing Venue

The IP.com Prior Art Database

Related People

Rodrigo Setti: INVENTOR

Abstract

A method and system is disclosed for providing personalized term suggestions using bloom filters. The method and system suggests relevant, personalized terms for a user, as the user enter text in an application. The personalization is suggested based on a user data corpus located in cloud and not locally in a user device.

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 Providing Personalized Term Suggestions Using Bloom Filters

Abstract

A method and system is disclosed for providing personalized term suggestions using bloom filters.  The method and system suggests relevant, personalized terms for a user, as the user enter text in an application.  The personalization is suggested based on a user data corpus located in cloud and not locally in a user device.

Description

Currently, applications perform personalized term or spell suggestions by calling a remote server, via Hypertext Transfer Protocol (HTTP) calls for an entered term.  As a user types, a server application matches a prefix against a user data corpus and returns suggestions based on frequent terms.  However, the latency may be too high in cellular/wireless networks, such as mobile applications connected to 2G or low signal wireless network.  Thus, the user experience is impacted.  The performance is required to be real time to account for the fact that as the user types and edits, suggestions get irrelevant quickly.  Another approach which is used to suggest terms is by downloading an entire corpus index and keep in the application for offline query of term suggestions.  Although the latency problem is solved, the entire corpus index takes up the memory and sometimes can become prohibitory if dealing with a huge corpus.  On mobile devices, term matching is also costly in terms of Central Processing Unit (CPU), which would introduce latency regardless.  Reducing the size of the corpus is a feasible solution to mitigate both memory and CPU costs, but that penalizes quality.

Disclosed is a method and system for providing personalized term suggestions using bloom filters.  The method and system suggests relevant, personalized terms for a user, as the user enter text in an application.  The personalized suggestions are based on a user data corpus located in cloud and not locally in a user device.  The bloom filter is a compact, hash representation of a set, where pertinence tests can be made quickly at the cost of a (low) probability of false positives.

The method and system performs zero-latency personalized term suggestions for the application with a large corpus of the user data.  The application downloads a compact representation of a user index term corpus encoded as a bloom filter.  Thereafter, for each entered term, the system decides if there are suggestions available for the term by checking against the bloom filter.  If the term is absent (i.e.  needs suggestion), the system starts exploring an edit distance neighborhood for terms that are present in the bloom filter until there is a significant number that matches, or the distance from the original entered term is high.  The method and system does not sacrifice quality to...