User-trained searching application system and method
System, apparatus, user equipment, and associated computer program and computing methods are provided for suggesting websites that are relevant based on the user's browsing history and past search results. In one aspect, a hosted computer application stores the user's browsing history and search results using a cloud-based storage facility, and computing methods, using machine learning techniques, are operative to predict websites the user may want to visit next. Example machine learning techniques may be configured to use non-parsed and unstructured data to identify patterns and map hundreds of thousands of data elements, to predict which website(s) the user might like to visit in a search/browsing session. Example machine learning techniques may be further operative to recognize patterns and analyze data at each interaction with the user. The training of example machine learning techniques is driven by user interaction, allowing the removal of non-relevant or less relevant websites from the suggested websites via a suitable user interface.
1. An Internet searching application method, comprising:
receiving a user query with respect to an Internet searching session established at a user equipment (UE) device;
determining search results responsive to the user query based on a user-trainable predictive analytical model, the search results including websites relevant to the user query, wherein relevance is predicted responsive to pair-wise comparisons between websites classified into multiple classes;
providing the search results for presentation via a user interface displayed at the UE device;
receiving a user interaction response comprising at least one of deleting a particular search result based on the user's perception of whether the particular search result is relevant to the user query; and
utilizing the user interaction response as an input to a training machine used in association with the predictive analytical model, wherein the user interaction response is used at least in part in selecting a classifier variable for classifying websites into the multiple classes, thereby modulating predictive behavior of the predictive analytical model with respect to future searching sessions of the user.
2. The Internet searching application method as recited in claim 1 , wherein the user query comprises an alphanumerical search term entered in a command line of at least one of a search engine and a browser application launched at the UE device by the user.
3. The Internet searching application method as recited in claim 1 , wherein the user query comprises a website's Uniform Resource Locator (URL) entered in a command line of a browser application launched at the UE device by the user.
4. The Internet searching application method as recited in claim 1 , wherein the user query comprises selecting a website's URL in a search results page generated in the Internet searching session.
5. The Internet searching application method as recited in claim 1 , wherein the user-trainable predictive analytical model comprises at least one of a support vector network, a support vector machine, a machine learning mechanism and a neural network, the user-trainable predictive analytical model further configured to suggest a plurality of websites responsive to the user query based on the user's browsing history data, search history data, and historical data relating to previous deletions of suggested websites by the user.
6. The Internet searching application method as recited in claim 5 , wherein the search results are presented via the user interface as a visualization format comprising a first portion having a constellation of previously visited websites displayed on one side of a display area of the user interface and a second portion having a constellation of suggested websites displayed on another side of the display area of the user interface, with most recently visited website being disposed therebetween.
7. The Internet searching application method as recited in claim 6 , wherein the constellation of previously visited websites are displayed in an interconnected graphical tree form based on a traversal history of the user regarding how the user navigated from one website to another in one or more Internet searching sessions.
8. The Internet searching application method as recited in claim 7 , wherein the traversal history includes the user's navigation data gathered from a plurality of UE devices associated with the user.
9. The Internet searching application method as recited in claim 6 , wherein the constellation of suggested websites are displayed in an interconnected graphical tree form based on their relationship predicted by the user-trainable predictive analytical model.
10. One or more network nodes configured to effectuate Internet searching based on user input, the one or more network nodes comprising:
one or more processors distributed among the one or more network nodes; and
one or more persistent memory modules distributed among the one or more network nodes and coupled to the one or more processors, the one or more persistent memory modules having program instructions stored thereon which, when executed by the one or more processors, are configured to:
responsive to receiving a user query with respect to an Internet searching session established at a user equipment (UE) device, determine a plurality of search results based on a user-trainable predictive analytical model, the search results including websites relevant to the user query, wherein relevance is predicted responsive to pair-wise comparisons between websites classified into multiple classes;
provide the search results to be presented via a user interface displayed at the UE device;
receive a user interaction response comprising at least one of deleting a particular search result based on the user's perception of whether the particular search result is relevant to the user query; and
process the user interaction response to utilize the user interaction response as an input to a training machine used in association with the predictive analytical model, wherein the user interaction response is used at least in part in selecting a classifier variable for classifying websites into the multiple classes, thereby modulate predictive behavior of the predictive analytical model with respect to future searching sessions of the user.
11. The one or more network nodes as recited in claim 10 , wherein the user query comprises an alphanumerical search term entered in a command line of at least one of a search engine and a browser application launched at the UE device by the user.
12. The one or more network nodes as recited in claim 10 , wherein the user query comprises a website's Uniform Resource Locator (URL) entered in a command line of a browser application launched at the UE device by the user.
13. The one or more network nodes as recited in claim 10 , wherein the user query comprises selecting a website's URL in a search results page generated in the Internet searching session.
14. The one or more network nodes as recited in claim 10 , wherein the user-trainable predictive analytical model comprises at least one of a support vector network, a support vector machine, a machine learning mechanism and a neural network, the predictive analytical model configured to suggest a plurality of websites responsive to the user query based on the user's browsing history data, search history data, and historical data relating to previous deletions of suggested websites by the user.
15. The one or more network nodes as recited in claim 14 , wherein the program instructions are further configured to provide the search results via the user interface in a presentation view comprising a first portion having a constellation of previously visited websites displayed on one side of a display area of the user interface and a second portion having a constellation of suggested websites displayed on another side of the display area of the user interface, with most recently visited website being disposed therebetween.
16. The one or more network nodes as recited in claim 15 , wherein the constellation of previously visited websites are displayed in an interconnected graphical tree form based on a traversal history of the user regarding how the user navigated from one website to another in one or more Internet searching sessions.
17. The one or more network nodes as recited in claim 16 , wherein the traversal history includes the user's navigation data gathered from a plurality of UE devices associated with the user.
18. The one or more network nodes as recited in claim 15 , wherein the constellation of suggested websites are displayed in an interconnected graphical tree form based on their relationship predicted by the user-trainable predictive analytical model.
19. One or more non-transitory computer-readable media containing instructions stored thereon which, when executed by one or more processors of a distributed server environment, effectuate Internet searching based on user input, the one or more non-transitory computer-readable media comprising:
a code portion, responsive to receiving a user query with respect to an Internet searching session established at a user equipment (UE) device, configured to determine a plurality of search results based on a user-trainable predictive analytical model, the search results including websites relevant to the user query, wherein relevance is predicted responsive to pair-wise comparisons between websites classified into multiple classes;
a code portion for facilitating presentation of the search results via a user interface displayed at the UE device;
a code portion for processing a user interaction response received from the user, the user interaction response comprising at least one of deleting a particular search result based on the user's perception of whether the particular search result is relevant to the user query; and
a code portion for utilizing the user interaction response as an input to a training machine used in association with the predictive analytical model, wherein the user interaction response is used at least in part in selecting a classifier variable for classifying websites into the multiple classes, thereby modulating predictive behavior of the predictive analytical model with respect to future searching sessions of the user.