Browse Prior Art Database

User Activity Prediction To Enable Time-sensitive Computations

IP.com Disclosure Number: IPCOM000238354D
Publication Date: 2014-Aug-19
Document File: 6 page(s) / 30K

Publishing Venue

The IP.com Prior Art Database

Abstract

A recommendation engine, such as a website that recommends websites to a user or a personal assistant included in a smartphone, may anticipate and/or predict recommendation requests from a user and pre-compute recommendations before the anticipated and/or predicted requests so that the requests are already available to the user.

This text was extracted from a PDF file.
This is the abbreviated version, containing approximately 28% of the total text.

Page 01 of 6

User Activity Prediction To Enable Time-sensitive Computations

ABSTRACT

A recommendation engine, such as a website that recommends websites to a user or a personal assistant included in a smartphone, may anticipate and/or predict recommendation requests from a user and pre-compute recommendations before the anticipated and/or predicted requests so that the requests are already available to the user.

Systems that react to a user's state present a challenge when the user's state can change faster than the system can react. In such a case, the system's response may no longer be relevant once it is finally produced. For instance, if a user enters a state for 10 seconds, then a system that takes 20 seconds to respond to the user's change in state is not helpful.

    Therefore, a recommendation engine is proposed which can provide recommendations to a user based on a state of the user on a timescale that is faster than the user usually changes state. The recommendation engine may recommend websites to a user from within a browser, or may provide recommendations, such as restaurants or movies, in response to a voice query into a smartphone. The state of the user may include the user's personal browsing history, such as keyboard inputs including search terms, mouse clicks including hyperlinks clicked on, and time spent at websites, websites that other users with similar interests to the user have recently visited, the present location of the user, and a type of or specific device that the user is using.

    In order to provide fresh and/or relevant recommendations, the system may generate recommendations before the user requests the recommendations, such as before the user visits the website which provides the recommendations or before the user asks a question into his or her smartphone.

    The system may generate recommendations periodically in anticipation of possible requests for recommendations. Or, the system may predict when the user will request a


Page 02 of 6

recommendation based on the user's past behavior. For example, if the user typically requests recommendations and/or engages in activities that require a system response at a specific time of day, then the system may schedule a pre-computation of the recommendation for a time just before the user is expected to request the recommendations. The system may schedule the pre- computation early enough in advance of the predicted request so that the recommendations will be available immediately after the request, but late enough and/or close enough to the request so that the recommendations will still be fresh and/or relevant. In scheduling pre-computations, the system may consider time of day, day of the week, location of the user, and current events that match the user's interests (such as sporting events in a sport the user has previously expressed interest in).

    FIG. 1 is a diagram of a browser 100 according to an example embodiment. The browser 100 may present a page that recommends webs...