Machine learning on search sessions to provide intelligent search results
AssigneeDell Products L.P.
InventorsMing Qian, Jaeyoo Jung
Life of this patent
Estimating the term…
Abstract
Techniques for training and/or using a machine learning (ML) algorithm to generate search results are disclosed. An ML algorithm is configured (i) to identify a search starting event and a search terminating event for a search session, where the search session includes multiple navigations across multiple different webpages, and (ii) to derive a dynamic score for the search session, where the dynamic score reflects whether the search session successfully identified an end result that was initially unknown at a time when the search starting event occurred. The trained ML algorithm can then be used during later search sessions to promote better search results and help users identify end targets or results in a faster and more intuitive manner.
Claims20
A training method for training a machine learning (ML) algorithm (i) to identify a search starting event and a search terminating event for a search session, where the search session includes multiple navigations across multiple different webpages, and (ii) to derive a dynamic score for the search session, where the dynamic score reflects whether the search session successfully identified an end result that was initially unknown at a time when the search starting event occurred, said training method comprising:
- causing an ML algorithm to identify a search starting event for a search session, wherein the search starting event is identified (i) based on a detection of one or more search terms being entered in a browser to facilitate execution of an initial search or (ii) based on the execution of the initial search, and wherein a webpage is displayed in the browser in response to the execution of the initial search;
- after the webpage is displayed, causing the ML algorithm to track intermediary events that include multiple navigations to multiple different webpages, wherein a determination is made that the intermediary events are related to the initial search despite the navigations to the multiple different webpages, such that the search session is configured to include the webpage resulting from the initial search as well as the multiple different webpages resulting from the multiple navigations;
- after the intermediary events are tracked, causing the ML algorithm to identify a search terminating event for the search session;
- causing the ML algorithm to identify a navigation route that was generated between a time when the search starting event occurred and a time when the search terminating event occurred, wherein the navigation route includes the multiple navigations to the multiple different webpages;
- causing the ML algorithm to determine an outcome of the search session;
- and
- based on the determined outcome, causing the ML algorithm to derive a dynamic score for the navigation route of the search session, wherein the dynamic score reflects a quantitative confidence as to whether the search session resulted in successful identification of an end result that was initially unknown at the time when the search starting event occurred.
A method for using a trained machine learning (ML) algorithm that has been trained to identify correlations between a new search session and previously learned search sessions in order to provide refined search results for the new search session, where the refined search results are designed in an attempt to reduce a number of webpage navigations that will potentially be performed during the new search session and where the refined search results are generated based on webpage navigations that were tracked as parts of the previously learned search sessions, said method comprising:
- detecting a search starting event for the new search session, wherein the search starting event is identified (i) based on a detection of one or more search terms being entered in a browser to facilitate execution of an initial search or (ii) based on the execution of the initial search;
- causing the trained ML algorithm to attempt to identify correlations between (i) the one or more search terms and (ii) one or more previous navigations between various webpages, where the one or more previous navigations were tracked as parts of the previously learned search sessions;
- generating a set of search results based on the one or more search terms and further based on a correlation identified by the trained ML algorithm;
- and
- based on new data collected during the new search session, causing the ML algorithm to further generate search results that are generated based on one or more new correlations that are identified by the trained ML algorithm.
A method for using a trained machine learning (ML) algorithm that has been trained to identify correlations between a new search session and previously learned search sessions in order to provide refined search results for the new search session, where the refined search results are designed in an attempt to reduce a number of webpage navigations that will potentially be performed during the new search session and where the refined search results are generated based on webpage navigations that were tracked as parts of the previously learned search sessions, said method comprising:
- detecting a search starting event for the new search session, wherein the search starting event is identified (i) based on a detection of one or more search terms being entered in a browser to facilitate execution of an initial search or (ii) based on the execution of the initial search;
- causing the trained ML algorithm to attempt to identify correlations, wherein the correlations include correlations between (i) the one or more search terms and (ii) one or more previous navigations between various webpages, where the one or more previous navigations were tracked as parts of the previously learned search sessions, and wherein the correlations include correlations between characteristics of a user who is conducting the new search session and characteristics of users who conducted the previously learned search sessions;
- generating a set of search results based on the one or more search terms and further based on a set of correlations identified by the trained ML algorithm;
- and
- based on new data collected during the new search session, causing the ML algorithm to further generate search results that are generated based on one or more new correlations that are identified by the trained ML algorithm.
Dependencies are inferred from the claim text.