Cursor Activity Evaluation For Search Result Enhancement

• US 20130246383A1
• Filed: 03/18/2012
• Published: 09/19/2013
• Est. Priority Date: 03/18/2012
• Status: Abandoned Application

• Current Assignee: Microsoft Technology Licensing LLC (Microsoft Corporation)
• Original Assignee: Microsoft Corporation
• Inventors: White, Ryen William, Dumais, Susan T., Huang, Jeff, Wang, Kuansan, Diriye, Abdigani M., Buscher, Georg LW
• Application Number:
  US13/423,243
• Publication Number:
  US 20130246383A1
• Time in Patent Office: Days
• US Class Current: 707/706
• CPC Class Codes: G06F 16/9535 Search customisation based on user profiles and personalisation

Abstract

The subject disclosure is directed towards using cursor activity with respect to search result pages to enhance search engine operation. Data associated with the cursor activity may be translated into cursor events representing user interactions with a search result page. Based on the cursor events, user behavior indicia may be identified via a user intent prediction model corresponding to various search result page related user actions. The user behavior indicia and/or the user intent prediction model may be used to produce search result pages for current search queries from the user.

Claims

1. In a computing environment, a method performed at least in part on at least one processor, comprising, using cursor data to improve search engine performance, including, processing the cursor data associated with web pages being presented to a user, extracting cursor events from the cursor data, identifying user behavior indicia from the cursor events based upon a user intent prediction model, and when the user submits a search query, producing a search result page using the user behavior indicia.

2. The method of claim 1 further comprising detecting a lack of cursor movement on a set of web pages indicating bot-traffic.

3. The method of claim 1 further comprising reordering relevant document links on the search result page based on the user intent prediction model.

4. The method of claim 1, wherein identifying the user behavior indicia further comprises generating feature data from the cursor data and search logs.

5. The method of claim 4, wherein generating the feature data further comprises identifying at least one of user groups and query groups according to the feature data.

6. The method of claim 4, wherein generating the feature data further comprises comparing the feature data with the user intent prediction model to assign a query group to the search query.

7. The method of claim 4, wherein generating the feature data further comprises comparing the feature data with the user intent prediction model to assign a user group to the user.

8. The method of claim 4, wherein generating the feature data further comprises identifying one or more areas of interest on the search result page based upon the cursor events.

9. The method of claim 1, wherein producing the search result page further comprises combining the user intent prediction model with the information retrieval model and transforming the information retrieval model and the user behavior indicia into the search result page.

10. The method of claim 1 further comprising producing a refined search result page using the user intent prediction model.

16. The method of claim 1 further comprising training the user intent prediction model with the user behavior indicia and deploying the user intent prediction model in an information retrieval model.

11. In a computing environment, a system, comprising, an extraction mechanism configured to generate a user intent prediction model for providing search results, wherein the extraction mechanism is further configured to identify a set of cursor events representing user interactions with search result pages, to generate feature data based on the set of cursor events, to produce a correlation result between user actions and the feature data, to update the user intent prediction model with the correlation result, and to incorporate the user intent prediction model into a search engine information retrieval model.

12. The system of claim 11, wherein the extraction mechanism is further configured to identify one or more abandonments of search result pages using the feature data and to build an abandonment model with the one or more abandonments.

13. The system of claim 11, wherein the extraction mechanism is further configured to process relevance values for one or more search result pages and to build a relevance model with the relevance values and a user action associated with each search result page.

14. The system of claim 11, wherein the extraction mechanism is further configured to process click-prediction rates for relevant electronic documents in each search result page and build a click-prediction model with the click-prediction rates and corresponding cursor activity.

15. The system of claim 11, wherein the extraction mechanism is further configured to access gaze tracking data for each search result page and build a gaze prediction model with the gaze tracking data and corresponding cursor activity.

17. One or more computer-readable media having computer-executable instructions stored thereon, which in response to execution by a computer, cause the computer to perform steps comprising:

monitoring a user computer to capture cursor activity associated with web pages;

examining a user intent prediction model to interpret a user behavior profile comprising feature data corresponding to the cursor activity;

determining a user action prediction for a current search query based on the user behavior profile; and

producing a search result page for ranking electronic documents according to a likelihood of at least one pointer click by the user.

18. The one or more computer-readable media of claim 17 having further computer-executable instructions, which in response to execution by the computer, cause the computer to perform further steps comprising:

accessing the information retrieval model if the user action prediction indicates an abandonment;

classifying the abandonment according to user rationale; and

if the abandonment is predicted to be negative, modifying one or more portions of the search result page in response to the user intent prediction model.

19. The one or more computer-readable media of claim 18 having further computer-executable instructions, which in response to execution by the computer, cause the computer to perform further steps comprising:

enhancing search results with expanded document summaries when producing the search result page.

20. The one or more computer-readable media of claim 17 having further computer-executable instructions, which in response to execution by the computer, cause the computer to perform further steps comprising:

identifying one or more areas of interest on the search result page;

generating another search result page based on the one or more areas of interest and an information retrieval model.