Filtering information using targeted filtering schemes

• US 8,725,746 B2
• Filed: 06/26/2009
• Issued: 05/13/2014
• Est. Priority Date: 06/26/2008
• Status: Active Grant

• Current Assignee: Alibaba Group Holding Ltd.
• Original Assignee: Alibaba Group Holding Ltd.
• Inventors: Yang, Junjie, Ni, Liang, Zhang, Zhenghua, Zhang, Zhenyu
• Primary Examiner(s): Bhatia, Ajay
• Assistant Examiner(s): PARK, GRACE A
• Application Number:
  US12/667,145
• Publication Number:
  US 20110010374A1
• Time in Patent Office: 1,782 Days
• Field of Search: None
• US Class Current: 707/754
• CPC Class Codes: G06F 16/337 Profile generation, learning or modification
• CPC Class Codes: G06F 16/9535 Search customisation based on user profiles and personalisation
• CPC Class Codes: G06F 21/55 Detecting local intrusion or implementing counter-measures
• CPC Class Codes: G06Q 30/02 Marketing; Price estimation or determination; Fundraising
• CPC Class Codes: H04L 51/04 Real-time or near real-time messaging, e.g. instant messaging [IM]
• CPC Class Codes: H04L 51/212 using filtering or selective blocking
• CPC Class Codes: H04L 63/0245 Filtering by information in the payload

Abstract

A method for filtering user information takes into account not only specific keywords in the user information, but also related user-characteristic data (e.g., user activity data), and allows targeted user characteristics to be determined from multiple aspects of user activities. In one aspect, the disclosed method adopts different filtering schemes for different types of targeted users to improve the recognition accuracy with respect to the target user information. The method determines a suitable filtering scheme using a correspondence relationship between the filtering scheme and keywords and user-characteristic data. The method uses modeling of sample users and multiple candidate filtering schemes to formulate targeted filtering scheme. An apparatus for implementing the method is also disclosed.

Claims

1. An apparatus comprising:

a computing device;

one or more storage media configured with executable instructions that, when executed by the computing device, instruct the computing device to perform acts comprising;

generating a plurality of filtering schemes for a plurality of types of target users, the generating comprising;

setting keywords and user-characteristic data for each type of target users;

eliminating invalid data from the user-characteristic data of each type of target users;

computing derived variables for each type of target users using remaining data in the user-characteristic data of the respective type of target user, the derived variables comprising;

an aggregated variable representing a number of targeted keyword types that appear,

a ratio variable representing a ratio between an amount of information sent and an amount of information received, and

an average variable representing an average frequency of an appearance of a keyword type;

filtering parameters of the user-characteristic data of each type of target users, the parameters comprising at least one of the derived variables; and

modeling each type of target users based on the filtered parameters of the user-characteristic of the respective type of target users, the modeling comprising;

computing multiple models for each type of target users based on different modeling algorithms and different filtered parameters of the user-characteristic data of the respective type of target users,

testing the multiple models of each type of target users using training data which includes information of the target users of the respective type whose behavior is known, and

selecting a respective model having the highest accuracy among the multiple models of each type of target users as a filtering scheme for the respective type of target users.

2. The apparatus as recited in claim 1, the acts further comprising:

obtaining keywords and user-characteristic data of a user;

selecting a filtering scheme from the plurality of filtering schemes according to a correspondence relationship between the filtering scheme and the keywords and user-characteristic data of the user, the selecting comprising;

comparing the keywords and the user-characteristic data of the user with keywords and user-characteristic data corresponding to each filtering scheme of the plurality of filtering schemes; and

selecting a filtering scheme having the highest match with the keywords and user-characteristic data of the user to be the selected filtering scheme; and

filtering information of the user according to the selected filtering scheme.

3. The apparatus as recited in claim 2, the acts further comprising:

giving a score to the user based on the selected filtering scheme;

in response to determining that the score of the user exceeds a preset threshold,

enlisting the user into a monitoring system, and

penalizing the user if the user meets one or more penalty requirements.

4. The apparatus as recited in claim 2, wherein the user-characteristic data of the user comprises user activity data, user information data and network user-characteristic data.

5. The apparatus as recited in claim 4, wherein:

the user activity data includes one or more of;

frequencies of characteristic word groups in information sent by the user within a time frame, the number of times the user sends/receives the information, and the amount of information the user sends/receives;

the user information data includes one or more of;

an initial login time of the user, user activities after login, and number of contacts of the user; and

the network user-characteristic data includes one or more of;

a number of users associated with a same IP of the user, and a number of users associated with a same machine ID of the user.

6. The apparatus as recited in claim 1, wherein filtering the parameters of the user-characteristic data of each type of target users comprises filtering out anomalies in the parameters of the user-characteristic data of each type of target users.

7. The apparatus as recited in claim 6, wherein filtering out the anomalies comprises:

replacing a missing value in a first parameter of user-characteristic data of a first type of target users by a replacement value;

or

replacing an irregular data in a second parameter of user-characteristic data of a second type of target user by a regular value.

8. The apparatus as recited in claim 1, wherein generating the plurality of filtering schemes further comprises:

for a type of target users,

identifying valid data in keywords and the user-characteristic data of the type of target users;

selecting one or more sample users based on the valid data; and

obtaining parameters of user-characteristic data of the type of target users based on keywords and user-characteristic data of the one or more sample users.

9. A method comprising:

under control of a computing device configured with executable instructions;

generating a plurality of filtering schemes for a plurality of types of target users, the generating comprising;

setting keywords and user-characteristic data for each type of target users;

eliminating invalid data from the user-characteristic data of each type of target users;

computing derived variables for each type of target users using remaining data in the user-characteristic data of the respective type of target user, the derived variables comprising;

an aggregated variable representing a number of targeted keyword types that appear,

a ratio variable representing a ratio between an amount of information sent and an amount of information received, and

an average variable representing an average frequency of an appearance of a keyword type;

filtering parameters of the user-characteristic data of each type of target users, the parameters comprising at least one of the derived variables; and

modeling each type of target users based on the filtered parameters of the user-characteristic of the respective type of target users, the modeling comprising;

computing multiple models for each type of target users based on different modeling algorithms and different filtered parameters of the user-characteristic data of the respective type of target users,

testing the multiple models of each type of target users using training data which includes information of the target users of the respective type whose behavior is known, and

selecting a respective model having the highest accuracy among the multiple models of each type of target users as a filtering scheme for the respective type of target users.

10. The method as recited in claim 9, further comprising:

obtaining keywords and user-characteristic data of a user;

selecting a filtering scheme from the plurality of filtering schemes according to a correspondence relationship between the filtering scheme and the keywords and user-characteristic data of the user, the selecting comprising;

comparing the keywords and the user-characteristic data of the user with keywords and user-characteristic data corresponding to each filtering scheme of the plurality of filtering schemes; and

selecting a filtering scheme having the highest match with the keywords and user-characteristic data of the user to be the selected filtering scheme; and

filtering information of the user according to the selected filtering scheme.

11. The method as recited in claim 10, further comprising:

giving a score to the user based on the selected filtering scheme;

in response to determining that the score of the user exceeds a preset threshold,

enlisting the user into a monitoring system, and

penalizing the user if the user meets one or more penalty requirements.

12. The method as recited in claim 10, wherein the user-characteristic data of the user comprises user activity data, user information data and network user-characteristic data.

13. The method as recited in claim 12, wherein:

the user activity data includes one or more of;

frequencies of characteristic word groups in information sent by the user within a time frame, the number of times the user sends/receives the information, and the amount of information the user sends/receives;

the user information data includes one or more of;

an initial login time of the user, user activities after login, and number of contacts of the user; and

the network user-characteristic data includes one or more of;

a number of users associated with a same IP of the user, and a number of users associated with a same machine ID of the user.

14. The method as recited in claim 9, wherein filtering the parameters of the user-characteristic data of each type of target users comprises filtering out anomalies in the parameters of the user-characteristic data of each type of target users.

15. The method as recited in claim 14, wherein filtering out the anomalies comprises:

replacing a missing value in a first parameter of user-characteristic data of a first type of target users by a replacement value;

or

replacing an irregular data in a second parameter of user-characteristic data of a second type of target user by a regular value.

16. The method as recited in claim 9, wherein generating the plurality of filtering schemes further comprises:

for a type of target users,

identifying valid data in keywords and the user-characteristic data of the type of target users;

selecting one or more sample users based on the valid data; and

obtaining parameters of user-characteristic data of the type of target users based on keywords and user-characteristic data of the one or more sample users.

17. One or more memory devices storing executable instructions that, when executed by a computing device, cause the computing device to perform acts comprising:

generating a plurality of filtering schemes for a plurality of types of target users, the generating comprising;

setting keywords and user-characteristic data for each type of target users;

eliminating invalid data from the user-characteristic data of each type of target users;

computing derived variables for each type of target users using remaining data in the user-characteristic data of the respective type of target user, the derived variables comprising;

an aggregated variable representing a number of targeted keyword types that appear,

a ratio variable representing a ratio between an amount of information sent and an amount of information received, and

an average variable representing an average frequency of an appearance of a keyword type;

filtering parameters of the user-characteristic data of each type of target users, the parameters comprising at least one of the derived variables; and

modeling each type of target users based on the filtered parameters of the user-characteristic of the respective type of target users, the modeling comprising;

computing multiple models for each type of target users based on different modeling algorithms and different filtered parameters of the user-characteristic data of the respective type of target users,

testing the multiple models of each type of target users using training data which includes information of the target users of the respective type whose behavior is known, and

selecting a respective model having the highest accuracy among the multiple models of each type of target users as a filtering scheme for the respective type of target users.

18. The one or more memory devices as recited in claim 17, the acts further comprising:

obtaining keywords and user-characteristic data of a user;

selecting a filtering scheme from the plurality of filtering schemes according to a correspondence relationship between the filtering scheme and the keywords and user-characteristic data of the user, the selecting comprising;

comparing the keywords and the user-characteristic data of the user with keywords and user-characteristic data corresponding to each filtering scheme of the plurality of filtering schemes; and

selecting a filtering scheme having the highest match with the keywords and user-characteristic data of the user to be the selected filtering scheme; and

filtering information of the user according to the selected filtering scheme.

19. The one or more memory devices as recited in claim 18, the acts further comprising:

giving a score to the user based on the selected filtering scheme;

in response to determining that the score of the user exceeds a preset threshold,

enlisting the user into a monitoring system, and

penalizing the user if the user meets one or more penalty requirements.

20. The one or more memory devices as recited in claim 17, wherein generating the plurality of filtering schemes further comprises:

for a type of target users,

identifying valid data in keywords and the user-characteristic data of the type of target users;

selecting one or more sample users based on the valid data; and

obtaining parameters of user-characteristic data of the type of target users based on keywords and user-characteristic data of the one or more sample users.