Automatic, personalized online information and product services

• US 8,825,853 B1
• Filed: 05/31/2013
• Issued: 09/02/2014
• Est. Priority Date: 12/28/1999
• Status: Expired due to Term

• Current Assignee: Personalized User Model LLP
• Original Assignee: Personalized User Model LLP
• Inventors: Konig, Yochai, Twersky, Roy, Berthold, Michael R
• Primary Examiner(s): BAROT, BHARAT
• Application Number:
  US13/907,115
• Time in Patent Office: 459 Days
• Field of Search: 709202-203, 709223-228, 715736-743, 707E17107-E17109
• US Class Current: 709/224
• CPC Class Codes: G06F 16/9535 Search customisation based on user profiles and personalisation
• CPC Class Codes: G06N 20/00 Machine learning
• CPC Class Codes: H04L 67/306 User profiles
• CPC Class Codes: H04L 67/53 using third party service providers
• CPC Class Codes: H04L 67/55 Push-based network services
• CPC Class Codes: H04L 67/75 Indicating network or usage conditions on the user display
• CPC Class Codes: H04N 21/44224 Monitoring of user activity on external systems, e.g. Internet browsing

Abstract

A method for providing automatic, personalized information services to a computer user includes the following steps: transparently monitoring user interactions with data during normal use of the computer; updating user-specific data files including a set of user-related documents; estimating parameters of a learning machine that define a User Model specific to the user, using the user-specific data files; analyzing a document to identify its properties; estimating the probability that the user is interested in the document by applying the document properties to the parameters of the User Model; and providing personalized services based on the estimated probability. Personalized services include personalized searches that return only documents of interest to the user, personalized crawling for maintaining an index of documents of interest to the user; personalized navigation that recommends interesting documents that are hyperlinked to documents currently being viewed; and personalized news, in which a third party server customized its interaction with the user.

Claims

1. A computer-implemented method for providing automatic, personalized information services to a user u, the method comprising:

a) transparently monitoring user interactions with data while the user is engaged in normal use of a computer;

b) updating user-specific data files, wherein the user-specific data files comprise the monitored user interactions with the data and a set of documents associated with the user;

c) estimating parameters of a learning machine, wherein the parameters define a User Model specific to the user and wherein the parameters are estimated in part from the user-specific data files;

d) analyzing a document d to identify properties of the document, the properties including (a) words of the document and (b) at least one additional property, the words of the document having (i) a first word frequency within the document and (ii) a second word frequency within other documents;

e) estimating a probability P(u|d) that an unseen document d is of interest to the user u, wherein the probability P(u|d) is estimated by applying the identified properties of the document to the learning machine having the parameters defined by the User Model; and

f) using the estimated probability to provide automatic, personalized information services to the user.

2. The method of claim 1 wherein the user-specific data files include documents of interest to the user u and documents that are not of interest to the user u, and wherein estimating the parameters comprises distinct treatment of the documents of interest and the documents that are not of interest.

3. The method of claim 1 wherein analyzing the document d provides for the analysis of documents having multiple distinct media types.

4. The method of claim 1 wherein transparently monitoring user interactions with data comprises monitoring multiple distinct modes of user interaction with network data.

5. The method of claim 4 wherein the multiple distinct modes of user interaction comprise a mode selected from the group consisting of a network searching mode, a network navigation mode, a network browsing mode, an email reading mode, an email writing mode, a document writing mode, a viewing “

pushed”

information mode, a finding expert advice mode, and a product purchasing mode.

6. The method of claim 1 wherein the identified properties of the document d comprise a user u-independent property selected from the group consisting of:

a) a probability P(t,d) that the document d is of interest to users interested in a topic t;

b) a topic classifier discrete probability distribution P(t|d);

c) a product model discrete probability distribution P(p|d);

d) product feature values extracted from the document d;

e) an author of the document d;

f) an age of the document d;

g) a list of documents linked to the document d;

h) a language of the document d;

i) a number of users who have accessed the document d;

j) a number of users who have saved the document d in a favorite document list; and

k) a list of users previously interested in the document d.

7. The method of claim 1 wherein the parameters of the learning machine define a user u-dependent function selected from the group consisting of:

a) a user topic probability distribution P(t|u) representing interests of the user u in various topics t;

b) a user product probability distribution P(p|u) representing interests of the user u in various products p;

c) a user product feature probability distribution P(f|u,p) representing interests of the user u in various features f of each of the various products p;

d) a web site probability distribution P(s|u) representing interests of the user u in various web sites s;

e) a cluster probability distribution P(c(u)|u) representing similarity of the user u to users in various clusters c(u);

f) a phrase model probability distribution P(w|u) representing interests of the user u in various phrases w;

g) an information theory based measure I(I_w;

I_u) representing mutual information between various phrases w and the user u;

h) an information theory based measure I(I_t;

I_u) representing mutual information between various topics t and the user u;

i) an information theory based measure I(I_s;

I_u) representing mutual information between various web sites s and the user u;

j) an information theory based measure I(I_p;

I_u) representing mutual information between various products p and the user u; and

k) an information theory based measure I(I_f;

I_u) representing mutual information between various features f of each of the various products p and the user u.

8. The method of claim 1 wherein the parameters of the learning machine define:

a) a user product probability distribution P(p|u) representing interacts of the user u in various products p; and

b) a user product feature probability distribution P(f|u,p) representing interests of the user u in various features f of each of the various products p; and

wherein the method further comprises estimating a probability P(u|d, product described=p) that a document d that describes a product p is of interest to the user u, wherein the probability is estimated in part from the user product probability distribution and the user product feature probability distribution.

9. The method of claim 1 further comprising estimating a posterior probability P(u|d,q) that the document d is of interest to the user u, given a query q submitted by the user.

10. The method of claim 9 wherein estimating the posterior probability comprises estimating a probability P(q|d,u) that the query q is expressed by the user u with an information need in the document d.

11. The method of claim 1 further comprising parsing the document d for hyperlinks, and separately estimating for each of the hyperlinks a probability that the hyperlink is of interest to the user u.

12. The method of claim 1 further comprising sending to a third party web server user interest information derived from the User Model, whereby the third party web server may customize its interaction with the user.

13. The method of claim 1 wherein the monitored user interactions include a sequence of interaction times.

14. The method of claim 1 further comprising initializing the User Model using information selected from the group consisting of a set of documents provided by the user, a web browser history file associated with the user, a web browser bookmarks file associated with the user, ratings by the user of a set of documents, and previous product purchases made by the user.

15. The method of claim 1 further comprising providing to the user a score for a document identified by the user, wherein the score is derived from the estimated probability.

16. The method of claim 1 further comprising temporarily using a User Model that is built from a set of predetermined parameters of a profile selected by the user.

17. The method of claim 1 further comprising initializing the User Model by selecting a set of predetermined parameters of a prototype user selected by the user.

18. The method of claim 1 further comprising identifying a set of users interested in the document d.

19. A program storage device accessible by a central computer, tangibly embodying a program of instructions executable by the central computer to perform method steps for providing automatic, personalized information services to a user u, the method steps comprising:

a) transparently monitoring user interactions with data while the user is engaged in normal use of a client computer in communication with the central computer;

b) updating user-specific data files, wherein the user-specific data files comprise the monitored user interactions with the data and a set of documents associated with the user;

c) estimating parameters of a learning machine, wherein the parameters define a User Model specific to the user and wherein the parameters are estimated in part from the user-specific data files;

d) analyzing a document d to identify properties of the document, the properties including (a) words of the document and (b) at least one additional property, the words of the document having (i) a first word frequency within the document and (ii) a second word frequency within other documents;

e) estimating a probability P(u|d) that an unseen document d is of interest to the user u, wherein the probability P(u|d) is estimated by applying the identified properties of the document to the learning machine having the parameters defined by the User Model; and

f) using the estimated probability to provide automatic, personalized information services to the user.