Method and apparatus for suggesting/disambiguation query terms based upon usage patterns observed

• US 7,702,690 B2
• Filed: 02/08/2006
• Issued: 04/20/2010
• Est. Priority Date: 12/29/2004
• Status: Expired due to Fees

• Current Assignee: Monetate, Inc. (StartApp, Inc.)
• Original Assignee: Baynote Incorporated (StartApp, Inc.)
• Inventors: Brave, Scott, Bradshaw, Robert, Minson, Christopher, Jia, Jack
• Primary Examiner(s): Alam; Shahid A
• Assistant Examiner(s): Bromell; Alexandria Y
• Application Number:
  US11/351,143
• Publication Number:
  US 20070150466A1
• Time in Patent Office: 1,532 Days
• Field of Search: 707/3, 707/6, 709/224, 715/205
• US Class Current: 707/776
• CPC Class Codes: G06F 16/901 Indexing; Data structures therefor; Storage structures for retrieval from the web G06F16/951
• CPC Class Codes: G06F 16/9535 Search customisation based on user profiles and personalisation

Abstract

The invention comprises a set of complementary techniques that dramatically improve enterprise search and navigation results. The core of the invention is an expertise or knowledge index, called UseRank that tracks the behavior of website visitors. The expertise-index is designed to focus on the four key discoveries of enterprise attributes: Subject Authority, Work Patterns, Content Freshness, and Group Know-how. The invention produces useful, timely, cross-application, expertise-based search and navigation results. In contrast, traditional Information Retrieval technologies such as inverted index, NLP, or taxonomy tackle the same problem with an opposite set of attributes than what the enterprise needs: Content Population, Word Patterns, Content Existence, and Statistical Trends. Overall, the invention emcompasses Baynote Search—a enhancement over existing IR searches, Baynote Guide —a set of community-driven navigations, and Baynote Insights—aggregated views of visitor interests and trends and content gaps.

Claims

1. A computer implemented method for automatically determining any of importance of an on-line asset, topic of said asset, a user'"'"'s current context/topic of interest, and relationships among assets and queries, without asking members of an on-line community directly comprising a processor executing the steps of:

observing usage patterns of said on-line community;

employing automatic techniques to extract patterns from said usage;

learning affinities between and among users, assets, and topics/terms;

identifying usefulness of an online asset by observing user implicit behaviors in connection with said usage patterns of said online asset and by extracting behavioral patterns from said observations;

refining said identified online asset usefulness by context, wherein the context of each online asset is automatically detected based on observed terms/topics from individual and group user behaviors when said online asset is determined to be useful based upon said individual and group user behaviors;

and, given a user query, any of;

determining a user'"'"'s current context/topic based on the user'"'"'s query and refining said identified user context based on previous searches performed and assets used during the user'"'"'s navigation;

comparing the identified context of a current user with the identified context of all assets within the system and assigning to each asset a score based on its similarity to the current user'"'"'s context; and

using said similarity score to present a user with recommended assets;

said observed usage patterns comprising user online search, navigation, and interaction behavior, said behavior including any of searches performed and position in user trail;

assets viewed and position in user trail;

dwell, range, scrolling, think time, and mouse movement on an asset;

anchors and lines used in asset text;

virtual bookmarks and virtual printing; and

explicit downloading, emailing, printing, saving, and removing.

2. The method of claim 1, further comprising the step of:

using said observed usage patterns for any of disambiguating search Terms, suggesting new terms, and automatically inserting search terms into user queries to expand a search based on said terms co-occurrence or lack thereof in identified asset context/topics.

3. The method of claim 1, further comprising the step of:

representing asset topics using a vector-space model and UF-IDF (usage frequency, inverse document frequency) weighting.

4. The method of claim 1, further comprising the step of:

determining co-occurrence of topics in assets by representing said terms using a document vector-space model and UF-IDF (usage frequency, inverse document frequency) weighting.

5. The method of claim 1, further comprising the step of:

classifying usage of a given asset by a given user as successful or unsuccessful use based on a combination of factors comprising any of time spent on an asset, explicit user actions taken on said asset, said actions comprising any of print, save, email, and bookmark, mouse and keyboard behavior, length and distribution of stable pauses between activity, navigation to and from said asset, and browser window state, said state comprising any of foreground, background, minimized, maximized, and closed.

6. The method of claim 1, further comprising the step of:

providing a significantly greater weighting to the importance of implicit observations over the weighting provided to explicit observations.

7. The method of claim 1, further comprising the steps of:

analyzing all observations; and

via said analysis, generating a set of recommendations comprising distilled experiences from on-line community;

wherein said recommendations age over time and are discarded if they have relatively little value; and

wherein recommendations which are most valuable based on repeated usage are stored into a long term memory.

8. The method of claim 1, further comprising the step of:

for a given user who may be anonymous, said user visiting a particular site, and a given context comprising any of what page the user is on and how the user got there, providing recommendations to said user that allow said user to navigate said site more efficiently.

9. The method of claim 1, further comprising the steps of:

generating a set a recommendations that may be applied to a search; and

for a given user who may be anonymous, and a given search query, using said recommendations to refine and augment a resulting search.

10. The method of claim 9, further comprising the step of:

driving said recommendations not just by individual uses, but by the use of said on-line community, leveraging the wisdom of crowds and community emergent behavior.

11. The method of claim 1, further comprising the step of:

identifying said on-line community as comprising any of one or more peer groups and/or expert groups based on an information context;

wherein said one or more peer groups and/or export groups are nested and defined by different levels of contexts.

12. An apparatus for automatically determining any of importance of an on-line asset, topic of said asset, a user'"'"'s current context/topic of interests, and relationships among assets and queries, without asking members of an on-line community directly, comprising:

means for observing usage patterns of said on-line community;

a processor for employing automatic techniques to extract patterns from said usage;

said processor learning affinities between and among users, assets, and topics/terms;

identifying usefulness of an online asset by observing user implicit behaviors in connection with said usage patterns of said online asset and by extracting behavioral patterns from said observations;

refining said identified online asset usefulness by context, wherein the context of each online asset is automatically detected based on observed terms/topics from individual and group user behaviors when said online asset is determined to be useful based upon said individual and group user behaviors;

and, given a user query, any of;

determining a user'"'"'s current context/topic based on the user'"'"'s query and refining said identified user context based on previous searches performed and assets used during the user'"'"'s navigation;

comparing the identified context of a current user with the identified context of all assets within the system and assigning to each asset a score based on its similarity to the current user'"'"'s context; and

using said similarity score to present a user with recommended assets;

said observed usage patterns comprising user online search, navigation, and interaction behavior, said behavior including any of searches performed and position in user trail;

assets viewed and position in user trail;

dwell, range, scrolling, think time, and mouse movement on an asset;

anchors and lines used in asset text;

virtual bookmarks and virtual printing; and

explicit downloading, emailing, printing, saving, and removing.

13. The apparatus of claim 12, further comprising:

means for using said observed usage patterns for any of disambiguating search terms, suggesting new terms, and automatically inserting search terms into user queries to expand a search, based on said terms co-occurrence or lack thereof in identified asset context/topics.

14. The apparatus of claim 12, further comprising:

means for representing asset topics using a vector-space model and UF-IDF (usage frequency, inverse document frequency) weighting.

15. The apparatus of claim 12, further comprising:

means for determining co-occurrence of topics in assets by representing said terms using a document vector-space model and UF-IDF (usage frequency, inverse document frequency) weighting.

16. The apparatus of claim 12, further comprising:

means for classifying usage of a given asset by a given user as successful or unsuccessful use based on a combination of factors comprising any of time spent on an asset, explicit user actions taken on said asset, said actions comprising any of print, save, email, and bookmark, mouse and keyboard behavior, length and distribution of stable pauses between activity, navigation to and from said asset, and browser window state, said state comprising any of foreground, background, minimized, maximized, and closed.

17. The apparatus of claim 12, further comprising:

means for providing a significantly greater weighting to the importance of implicit observations over the weighting provided to explicit observations.

18. The apparatus of claim 12, further comprising:

means for analyzing all observations; and

via said analysis, means for generating a set of recommendations comprising distilled experiences from said on-line community;

wherein said recommendations age over time and are discarded if they have relatively little value; and

wherein recommendations which are most valuable based on repeated usage are stored into a long term memory.

19. The apparatus of claim 12, further comprising:

for a given user who may be anonymous, said user visiting a particular site, and a given context comprising any of what page the user is on and how the user got there, means for providing recommendations to said user that allow said user to navigate said site more efficiently.

20. The apparatus of claim 12, further comprising:

means for generating a set a recommendations that may be applied to a search; and

for a given user who may be anonymous, and a given search query, means for using said recommendations to refine and augment a resulting search.

21. The apparatus of claim 20 further comprising:

means for driving said recommendations not just by individual uses, but by the use of said on-line community, leveraging the wisdom of crowds and community emergent behavior.

22. The apparatus of claim 12, further comprising:

means for identifying said on-line community as comprising any of one or more peer groups and/or expert groups based on an information context;

wherein said one or more peer groups, and/or expert groups are nested and defined by different levels of contexts.