PROCESS OF TIME-SPACE COLLABORATIVE FILTERING OF INFORMATION

• US 20080319986A1
• Filed: 06/03/2008
• Published: 12/25/2008
• Est. Priority Date: 06/19/2007
• Status: Abandoned Application

• Current Assignee: Deutsche Telekom AG
• Original Assignee: Deutsche Telekom AG
• Inventors: Shoval, Peretz, Shapira, Bracha Leah, Steinbock, Nimrod
• Application Number:
  US12/132,043
• Publication Number:
  US 20080319986A1
• Time in Patent Office: Days
• US Class Current: 707/5
• CPC Class Codes: G06F 16/9535 Search customisation based on user profiles and personalisation

Abstract

The invention is a process for collaborative filtering of information called Time-Space Filtering (TSIF). The invention is used in the fields of information filtering and publishing and is particularly useful in the field of providing web-based information, e.g. electronic newspapers. TSIF is a process of filtering and ranking the relevance of an article'"'"'s content to specific readers, taking into account the time dimension of information as well as the factors traditionally considered by content-based or collaborative filtering.

Claims

1. A process for collaborative time-space filtering of information comprising the steps:

Step 1;

receiving a request to provide a relevancy-ranked list of items for user X;

Step 2;

using a content-based filter to rank the items in the repository for user X;

Step 3;

using a time-dependant collaborative filter to rank the items in the repository for user X;

Step 4. computing time-dependant weights for combining the rankings from steps 2 and 3; and

Step 5;

computing the overall ranking of the items by using the time-dependant weights to combine the two rankings.

2. A process according to claim 1, wherein the information is web based information.

3. A process according to claim 2, wherein the web based information is used to publish an electronic newspaper.

4. A process according to claim 1, wherein the content-based filter ranks the items in the repository for user X based on the degree of similarity/match between user X'"'"'s profile and the profile of each item.

5. A process according to claim 1, wherein time-dependant collaborative filter ranks the items in the repository for user X by taking into consideration the following assumptions:

a. that recent user opinions have a bigger influence on the ranking than older opinions; and

b. that opinions from users similar to user X have a bigger influence on the ranking than opinions from less-similar users

6. A process according to claim 1, wherein the time-dependant weights for combining the two rankings are computed assuming that the weight of the collaborative filter is higher for older items.

7. A process according to claim 1, wherein the collaborative filter compiles a User'"'"'s Collaborative Profile comprising for each item:

a. a unique identifier of each item recommended to the user in the past;

b. a time-stamp of the time the item was recommended to the user; and

c. an indicator for the user'"'"'s opinion of the item.

8. A process according to claim 7, wherein the collaborative filter uses the User'"'"'s Collaborative Profile to compute a User Similarity Score for user X with each user J in the user repository.

9. A process according to claim 8, wherein instead of comparing the entire list of collaborative profiles of the users, only the profiles accumulated over a selected period of time are compared,

10. A process according to claim 1, wherein the collaborative filter ranks all the items by computing for each item k a value, which is the relative score given to said item, based on other users'"'"' recommendations.

11. A process according to claim 10, wherein the weight of a recommendation from user J to item k is discounted according to the time that has passed since the recommendation.

12. A process according to claim 11, wherein different information types receive different discounting factors.

13. A process according to claim 1, wherein the time-dependant weight is used to combine the two rankings by giving a bigger weight to the collaborative filter'"'"'s ranking for older items.

14. A process according to claim 1, wherein said process removes obsolete items from the items repository and archives them.

15. A filtering engine comprising conventional hardware including input/output devices, memory modules, and servers, configured to collect and process the required information according to the steps of claim 1;

wherein said components are configured to form the following main modules;

a. the collaborative filter module;

b. the rankings combination module;

c. the update the collaborative filter module; and

d. the remove decayed items module.

16. A filtering engine according to claim 15, wherein the purpose of the collaborative filter module is to arrange a list items into a ranked list.

17. A filtering engine according to claim 15, wherein the purpose of the rankings combination module is to combine the rankings for the content-based filter and the collaborative filter.

18. A filtering engine according to claim 15, wherein the purpose of the update the collaborative filter module is to update the profile of a user after the user was presented with an item.

19. A filtering engine according to claim 15, wherein the purpose of the remove decayed items module is to determine when an item became obsolete and define it as Archived.