User intention modeling for web navigation

1. A method for modeling a user intention during network navigation, the method comprising:

predicting, based on a statistical multi-step n-gram probability model, an optimal information goal of the user, the optimal information goal being based on a sequence of previously visited network content pieces and a globally optimized navigation path through the sequence, the optimal information goal being predicted as follows:

recording a history of user action, the history comprising information corresponding to user navigation to a plurality of networked content pieces, the information indicating at least the sequence of previously visited network content pieces;

for at least a portion of the sequence data, calculating respective probabilities that a user would visit a particular content piece n in the sequence from a content piece n−1 in the sequence, a prediction of the optimal information goal being based on the respective probabilities, the calculating comprising:

Pr

(

w

i

⁢



w

1

,

⩓

,

w

i

-

1

)

≈

Pr

⁡

(

w

i

|

w

i

-

n

+

1

,

…

,

w

i

-

2

,

w

i

-

1

)

=

Pr

⁡

(

w

i

-

n

+

1

,

…

,

w

i

-

2

,

w

i

-

1

,

w

i

)

Pr

⁡

(

w

i

-

n

+

1

,

…

,

w

i

-

2

,

w

i

-

1

)

=

C

⁡

(

w

i

-

n

+

1

,

…

,

w

i

-

2

,

w

i

-

1

,

w

i

)

/

C

n

C

⁡

(

w

i

-

n

+

1

,

…

,

w

i

-

2

,

w

i

-

1

)

/

C

n

-

1

=

C

⁡

(

w

i

-

n

+

1

,

…

,

w

i

-

2

,

w

i

-

1

,

w

i

)

C

⁡

(

w

i

-

n

+

1

,

…

,

w

i

-

2

,

w

i

-

1

)

*

C

;

wherein Pr represents the probability;

wherein user navigation to the plurality of networked content pieces is represented as w 1 , w 2 , Λ, w i , Λ, w L , where w i is the ith visited content piece in the sequence; and

wherein C(w i−n+1 , . . . , w i−2 , w i−1 w i ) denotes the count of an n-Gram (w i−n+1 , . . . , w i−2 , w i−1 , w i ) appearing in training data, C n is a total number of the n-grams, C n−1 is a total number of the (n−1)-grams, C equals to C n /C n−1 , C n , C n− , and C are constants.

2. The method of claim 1 , wherein a content piece of the sequence comprises a Web page.

3. The method of claim 1 , wherein predicting the globally optimized navigation path is dynamically performed responsive to user Web navigation.

4. The method of claim 1 , wherein the portion comprises a session of user access to particular ones of the networked content pieces.

5. The method of claim 1 , wherein the history further comprises a URL, a hyperlink, a user name, a start time, text corresponding to a URL, a query, elapsed time, or an object name.

6. The method of claim 1 , wherein the globally optimized navigation path is based on the following:

argmax

w

i

⁢

∏

i

=

k

+

1

∞

⁢

⁢

Pr

⁡

(

w

i

|

w

i

-

2

⁢

w

i

-

1

)

;

and

,

wherein k represents a number of networked content pieces that the user has already visited.

7. The method of claim 1 , wherein the globally optimized navigation path is based on the following:

1

t

⁢

∑

i

=

k

+

1

i

=

k

+

t

⁢

⁢

log

⁡

(

Pr

⁡

(

w

i

|

w

i

-

2

⁢

w

i

-

1

)

)

;

wherein k represents a number of networked content pieces that the user has already visited; and

wherein t indicates how many steps are predicted.

8. The method of claim 7 further comprising dynamically determining how many steps to the optimal information goal are to be predicted by employing a perplexity that reflects the entropy of the globally optimized navigation path.

9. The method of claim 7 , wherein the optimal information goal is determined according to the following:

argmax

t

⁡

(

1

t

⁢

∑

i

=

k

+

1

i

=

k

+

t

⁢

⁢

log

⁢

⁢

Pr

⁡

(

w

i

|

w

i

-

2

⁢

w

i

-

1

)

)

.

10. A computer-readable medium for modeling a user intention during network navigation, the computer-readable medium comprising computer-executable instructions for:

predicting, based on a statistical multi-step n-gram probability model, an optimal information goal of the user, the optimal information goal being based on a sequence of previously visited network content pieces and a globally optimized navigation path through the sequence, the optimal information goal being predicted as follows:

recording a history of user action, the history comprising information corresponding to user navigation to a plurality of networked content pieces, the information indicating at least the sequence of previously visited network content pieces;

for at least a portion of the sequence data, calculating respective probabilities that a user would visit a particular content piece n in the sequence from a content piece n−1 in the sequence, a prediction of the optimal information goal being based on the respective probabilities, the calculating comprising:

Pr

(

w

i

⁢



w

1

,

⩓

⁢

,

w

i

-

1

)

≈

⁢

Pr

⁡

(

w

i

|

w

i

-

n

+

1

,

…

⁢

,

w

i

-

2

,

w

i

-

1

)

=

⁢

Pr

⁡

(

w

i

-

n

+

1

,

…

⁢

,

w

i

-

2

,

w

i

-

1

,

w

i

)

Pr

⁡

(

w

i

-

n

+

1

,

…

⁢

,

w

i

-

2

,

w

i

-

1

)

=

⁢

C

⁡

(

w

i

-

n

+

1

,

…

⁢

,

w

i

-

2

,

w

i

-

1

,

w

i

)

/

C

n

C

⁡

(

w

i

-

n

+

1

,

…

⁢

,

w

i

-

2

,

w

i

-

1

)

/

C

n

-

1

=

⁢

C

⁡

(

w

i

-

n

+

1

,

…

⁢

,

w

i

-

2

,

w

i

-

1

,

w

i

)

C

⁡

(

w

i

-

n

+

1

,

…

⁢

,

w

i

-

2

,

w

i

-

1

)

*

C

;

wherein Pr represents the probability;

wherein user navigation to the plurality of networked content pieces is represented as w 1 , w 2 , Λ, w i , Λ, w L , where w i is the ith visited content piece in the sequence; and

wherein C(w i−n+1 , . . . , w i−2 , w i−1 w i ) denotes the count of an n-Gram (w i−n+1 , . . . , w i−2 , w i−1 , w i ) appearing in training data, C n is a total number of the n-grams, C n−1 is a total number of the (n−1)-grams, C equals to C n /C n−1 , C n , C n− , and C are constants.

11. The computer-readable medium of claim 10 , wherein a content piece of the sequence comprises a Web page.

12. The computer-readable medium of claim 10 , wherein predicting the globally optimized navigation path is dynamically performed responsive to user Web navigation.

13. The computer-readable medium of claim 10 , wherein the portion comprises a session of user access to particular ones of the networked content pieces.

14. The computer-readable medium of claim 10 , wherein the history further comprises a, URL, a hyperlink, a user name, a start time, text corresponding to a URL, a query, elapsed time, or an object name.

15. The computer-readable medium of claim 10 , wherein the globally optimized navigation path is based on the following:

argmax

w

i

⁢

∏

i

=

k

+

1

∞

⁢

⁢

Pr

⁡

(

w

i

|

w

i

-

2

⁢

w

i

-

1

)

;

and

,

wherein k represents a number of networked content pieces that the user has already visited.

16. The computer-readable medium of claim 10 , wherein the globally optimized navigation path is based on the following:

1

t

⁢

∑

i

=

k

+

1

i

=

k

+

t

⁢

⁢

log

⁡

(

Pr

⁡

(

w

i

|

w

i

-

2

⁢

w

i

-

1

)

)

;

wherein k represents a number of networked content pieces that the user has already visited; and

wherein t indicates how many steps are predicted.

17. The computer-readable medium of claim 16 further comprising instructions for dynamically determining how many steps to the optimal information goal are to be predicted by employing a perplexity that reflects the entropy of the globally optimized navigation path.

18. The computer-readable medium of claim 16 , wherein the optimal information goal is determined according to the following:

argmax

t

(

1

t

⁢

∑

i

=

k

+

1

i

=

k

+

t

⁢

log

⁡

(

Pr

⁡

(

w

i

|

w

i

-

2

⁢

w

i

-

1

)

)

.

19. A computing device for modeling a user intention during network navigation, the computing device comprising:

a processor; and

a memory coupled to the processor, the memory comprising computer-executable instructions that are fetched and executed by the processor for:

predicting, based on a statistical multi-step n-gram probability model, an optimal information goal of the user, the optimal information goal being based on a sequence of previously visited network content pieces and a globally optimized navigation path trough the sequence, the optimal information goal being predicted as follows:

recording a history of user action, the history comprising information corresponding to user navigation to a plurality of networked content pieces, the information indicating at least the sequence of previously visited network content pieces;

for at least a portion of the sequence data, calculating respective probabilities that a user would visit a particular content piece n in the sequence from a content piece n−1 in the sequence, a prediction of the optimal information goal being based on the respective probabilities, the calculating comprising:

Pr

(

w

i

⁢



w

1

,

⩓

,

w

i

-

1

)

≈

Pr

⁡

(

w

i

|

w

i

-

n

+

1

,

…

,

w

i

-

2

,

w

i

-

1

)

=

Pr

⁡

(

w

i

-

n

+

1

,

…

,

w

i

-

2

,

w

i

-

1

,

w

i

)

Pr

⁡

(

w

i

-

n

+

1

,

…

,

w

i

-

2

,

w

i

-

1

)

=

C

⁡

(

w

i

-

n

+

1

,

…

,

w

i

-

2

,

w

i

-

1

,

w

i

)

/

C

n

C

⁡

(

w

i

-

n

+

1

,

…

,

w

i

-

2

,

w

i

-

1

)

/

C

n

-

1

=

C

⁡

(

w

i

-

n

+

1

,

…

,

w

i

-

2

,

w

i

-

1

,

w

i

)

C

⁡

(

w

i

-

n

+

1

,

…

,

w

i

-

2

,

w

i

-

1

)

*

C

;

wherein Pr represents the probability;

wherein user navigation to the plurality of networked content pieces is represented as w 1 , w 2 , Λ, w i , Λ, w L , where w i is the ith visited content piece in the sequence; and

wherein C(w i−n+1 , . . . , w i−2 , w i−1 , w i ) denotes the count of an n-Gram (w i−n+1 , . . . , w i−2 , w i−1 , w i ) appearing in training data, C n is a total number of the n-grams, C n−1 is a total number of the (n−1)-grams, C equals to C n /C n−1 , C n , C n− , and C are constants.

20. The computing device of claim 19 , wherein a content piece of the sequence comprises a Web page.

21. The computing device of claim 19 , wherein predicting the globally optimized navigation path is dynamically performed responsive to user Web navigation.

22. The computing device of claim 19 , wherein the portion comprises a session of user access to particular ones of the networked content pieces.

23. The computing device of claim 19 , wherein the history further comprises a URL, a hyperlink, a user name, a start time, text corresponding to a URL, a query, elapsed time, or an object name.

24. The computing device of claim 19 , wherein the globally optimized navigation path is based on the following:

argmax

w

i

⁢

∏

i

=

k

+

1

∞

⁢

⁢

Pr

⁡

(

w

i

|

w

i

-

2

⁢

w

i

-

1

)

;

and

,

wherein k represents a number of networked content pieces that the user has already visited.

25. The computing device of claim 19 , wherein the globally optimized navigation path is based on the following:

1

t

⁢

∑

i

=

k

+

1

i

=

k

+

t

⁢

log

⁡

(

Pr

⁡

(

w

i

|

w

i

-

2

⁢

w

i

-

1

)

)

;

wherein k represents a number of networked content pieces that the user has already visited; and

wherein t indicates how many steps are predicted.

26. computing device of claim 25 , further comprising instructions for dynamically determining how many steps to the optimal information goal are to be predicted by employing a perplexity that reflects the entropy of the globally optimized navigation path.

27. computing device of claim 25 , wherein the optimal information goal is determined according to the following:

argmax

t

(

1

t

⁢

∑

i

=

k

+

1

i

=

k

+

t

⁢

log

⁡

(

Pr

⁡

(

w

i

|

w

i

-

2

⁢

w

i

-

1

)

)

.

28. A computer device for modeling a user intention during network navigation, the computing device comprising processing means for:

predicting, based on a statistical multi-step n-gram probability model, an optimal information goal of the user, the optimal information goal being based on a sequence of previously visited network content pieces and a globally optimized navigation path through the sequence, the optimal information goal being predicted as follows:

recording a history of user action, the history comprising information corresponding to user navigation to a plurality of networked content pieces, the information indicating at least the sequence of previously visited network content pieces;

for at least a portion of the sequence data, calculating respective probabilities that a user would visit a particular content piece n in the sequence from a content piece n−1 in the sequence, a prediction of the optimal information goal being based on the respective probabilities, the calculating comprising:

Pr

(

w

i

⁢



w

1

,

⩓

,

w

i

-

1

)

≈

Pr

⁡

(

w

i

|

w

i

-

n

+

1

,

…

,

w

i

-

2

,

w

i

-

1

)

=

Pr

⁡

(

w

i

-

n

+

1

,

…

,

w

i

-

2

,

w

i

-

1

,

w

i

)

Pr

⁡

(

w

i

-

n

+

1

,

…

,

w

i

-

2

,

w

i

-

1

)

=

C

⁡

(

w

i

-

n

+

1

,

…

,

w

i

-

2

,

w

i

-

1

,

w

i

)

/

C

n

C

⁡

(

w

i

-

n

+

1

,

…

,

w

i

-

2

,

w

i

-

1

)

/

C

n

-

1

=

C

⁡

(

w

i

-

n

+

1

,

…

,

w

i

-

2

,

w

i

-

1

,

w

i

)

C

⁡

(

w

i

-

n

+

1

,

…

,

w

i

-

2

,

w

i

-

1

)

*

C

;

wherein Pr represents the probability;

wherein user navigation to the plurality of networked content pieces is represented as w 1 , w 2 , Λ, w i , Λ, w L , where w i is the ith visited content piece in the sequence; and

wherein C(w i−n+1 , . . . , w i−2 , w i−1 , w i ) denotes the count of an n-Gram (w i−n+1 , . . . , w i−2 , w i−1 , w i ) appearing in training data, C n is a total number of the n-grams, C n−1 is a total number of the (n−1)-grams, C equals to C n /C n−1 , C n , C n− , and C are constants.

29. The computing device of claim 28 , wherein a content piece of the sequence comprises a Web page.

30. The computing device of claim 28 , wherein predicting the globally optimized navigation path is dynamically performed responsive to user Web navigation.

31. The computing device of claim 28 , wherein the portion comprises a session of user access to particular ones of the networked content pieces.

32. The computing device of claim 28 , wherein the history further comprises a URL, a hyperlink, a user name, a start time, text corresponding to a URL, a query, elapsed time, or an object name.

33. The computing device of claim 28 , wherein the globally optimized navigation path is based on the following:

argmax

w

i

⁢

∏

i

=

k

+

1

∞

⁢

⁢

Pr

⁡

(

w

i

|

w

i

-

2

⁢

w

i

-

1

)

;

and

,

wherein k represents a number of networked content pieces that the user has already visited.

34. The computing device of claim 28 , wherein the globally optimized navigation path is based on the following:

1

t

⁢

∑

i

=

k

+

1

i

=

k

+

t

⁢

log

⁡

(

Pr

⁡

(

w

i

|

w

i

-

2

⁢

w

i

-

1

)

)

;

wherein k represents a number of networked content pieces that the user has already visited; and

wherein t indicates how many steps are predicted.

35. The computing device of claim 34 further comprising means for dynamically determining how many steps to the optimal information goal are to be predicted by employing a perplexity that reflects the entropy of the globally optimized navigation path.

36. The computing device of claim 34 , wherein the optimal information goal is determined according to the following:

arg

⁢

⁢

max

t

⁢

(

1

t

⁢

∑

i

=

k

+

1

i

=

k

+

t

⁢

log

⁢

⁢

Pr

⁡

(

w

i

|

w

i

-

2

⁢

w

i

-

1

)

)

.