Analysis of topic dynamics of web search

• US 20070005646A1
• Filed: 06/30/2005
• Published: 01/04/2007
• Est. Priority Date: 06/30/2005
• Status: Abandoned Application

• Current Assignee: Microsoft Technology Licensing LLC (Microsoft Corporation)
• Original Assignee: Microsoft Corporation
• Inventors: Horvitz, Eric, Dumais, Susan, Shen, Xuehua
• Application Number:
  US11/171,123
• Publication Number:
  US 20070005646A1
• Time in Patent Office: Days
• US Class Current: 1/1
• CPC Class Codes: G06F 16/9535 Search customisation based on user profiles and personalisation
• CPC Class Codes: G06F 2216/03 Data mining

Abstract

The subject invention relates to probabilistic models that are trained from transitions among various topics of pages visited by a sample population of search users. In one aspect, probabilistic models of topic transitions are learned for individual users and groups of users. Topic transitions for individuals versus larger groups are analyzed, wherein the relative accuracies of personal models of topic dynamics with models constructed from sets of pages drawn from similar groups and from a larger population of users are compared. To exploit temporal dynamics, the accuracy of these models are tested for predicting transitions in topics of visits at increasingly more distant times in the future. The models can be applied to search topic dynamics of tagged pages, and then utilized to predict topics of subsequent pages visited by users.

Claims

1. A topic analysis system, comprising:

at least one learning model that is trained from information access data from a plurality of web sites; and

a search component that employs the learning model to predict potential future web sites or topics of interest.

2. The system of claim 1, the learning model is a Marginal model, a Markov model or a time-specific Markov model.

3. The system of claim 1, further comprising an evaluation data subset derived from a web access or search log.

4. The system of claim 3, the evaluation data subset includes basic data characteristics, topic categories, and sample log data.

5. The system of claim 1, the learning model is trained from topical categories associated with queries and/or universal resource locators (URLs) visited over time.

6. The system of claim 1, the learning model is trained from individuals, groups of individuals, and populations of users as a whole over time.

7. The system of claim 1, the learning model determines a probability that a user will transition from a given topic to another topic or to the same topic.

8. The system of claim 1, further comprising an analysis component to estimate model parameters and to apply smoothing to estimate model distributions.

9. The system of claim 1, the analysis component includes a maximum likelihood estimation process.

10. The system of claim 1, further comprising a component to collect training data, the training data including user queries, lists of search results returned, one or more URLs visited, a client identification, a time stamp, an action, and an action value.

11. The system of claim 10, further comprising a web directory component to facilitate collection of training data.

12. The system of claim 1, a divergence component for determining differences between topic distributions.

13. The system of claim 1, further comprising a scoring component to determine model accuracy based on an overlap between actual topic categories and predicted topic categories.

15. A computer readable medium having computer readable instructions stored thereon for executing the components of claim 1.

14. The system of 13, the scoring component includes a text classification predictor for automatically assigning topic tags.

16. A method for performing automated topic predictions, comprising:

automatically measuring a plurality of past user or group actions from a search log;

training at least one model from the past user or group actions; and

automatically predicting future topic selections based in part on the past user or group actions.

17. The method of claim 16, further comprising analyzing the past user or group actions in terms of topic transitions, topic dynamics, and temporal dynamics.

18. The method of claim 16, further comprising automatically analyzing universal resource locators visited by users or groups of users.

19. The method of claim 16, further comprising analyzing the model over varying degrees of time.

20. A system to facilitate automated topical searches, comprising:

means for collecting past user or group search data;

means for analyzing the past user or group search data; and

means for predicting future topics of interest from past user or group search data.