Method and system for multi-phase ranking for content personalization
Embodiments of the present teachings disclose method, system, and programs for a multi-phase ranking system for implementation with a personalized content system. The disclosed method, system, and programs utilize a weighted AND system to compute a dot product of the user profile and a content profile in a first phase, a content quality indicator in the second phase and a rules filter in a third phase.
1. A method implemented on a machine having at least one processor, storage, and a communication interface connected to a network for selecting content, the method comprising:
obtaining a user profile of a user characterizing the user's interests;
obtaining a content profile for each piece of content in a first set of content and a content quality indicator characterizing quality of the piece of content in the first set of content;
estimating an affinity measure between each piece of content in the first set of content and the user based on the user profile and the content profile for the piece of content;
selecting a second set of content from the first set of content based on the affinity measures for the pieces of content in the first set of content, wherein the second set is a subset of the first set;
selecting a third set of content from the second set of content based on the content quality indicators and a plurality of contextual parameters including a type of content associated with each piece of content in the second set of content, wherein the content quality indicator of each piece of content is obtained based on engagement of other users with the piece of content and vitality of the piece of content, and wherein the third set is a subset of the second set;
filtering the third set of content based on the plurality of contextual parameters to generate a fourth set of content; and
outputting the fourth set of content on a device operated by the user.
2. The method of claim 1 , wherein
the selecting the second set of content from the first set of content is based on a first model; and
the selecting the third set of content from the second set of content is based on a second model.
3. The method of claim 2 , wherein
the user profile and the content profile are represented by feature vectors, respectively; and
the first model is based on a product of the respective feature vectors of the user and of a piece of content.
4. The method of claim 2 , wherein the second model includes at least one of:
a statistic distribution;
a machine learning model; and
a decision tree.
5. The method of claim 1 , wherein the user profile, the content profile, and the content quality indicator are dynamically updated.
6. The method of claim 1 , wherein the user profile is obtained based on at least one of:
the user's declared interests;
the user's content consumption;
the user's online behaviors; and
the user's demographic information.
7. The method of claim 1 , wherein the content quality indicator is obtained based on at least one of:
timeliness of the piece of content; and
diversity of the piece of content.
8. The method of claim 1 , wherein the plurality of contextual parameters further includes a source of the content, a type of device used by the user to access the content, and a bandwidth of a network over which the content is delivered to the user.
9. A system having at least one processor, a storage, and a communication platform for selecting content, the system comprising:
a user profiler configured to obtain a user profiler of a user that characterizes the user's interests;
a content profiler configured to obtain a content profile for each piece of content in a first set of content and a content quality indicator characterizing quality of the piece of content in the first set of content;
an estimator configured to estimate an affinity measure between each piece of content in the first set of content and the user based on the user profile and the content profile for the piece of content;
a first selector configured to select a second set of content from the first set of content based on the affinity measures for the pieces of content in the first set of content, wherein the second set is a subset of the first set;
a second selector configured to select a third set of content from the second set of content based on the content quality indicators and a plurality of contextual parameters including a type of content associated with each piece of content in the second set of content, wherein the content quality indicator of each piece of content is obtained based on engagement of other users with the piece of content and vitality of the piece of content, and wherein the third set is a subset of the second set;
a filter configured to filter the third set of content based on the plurality of contextual parameters to generate a fourth set of content; and
an output configured to output the fourth set of content on a device operated by the user.
10. The system of claim 9 , wherein
the selecting the second set of content from the first set of content is based on a first model; and
the selecting the third set of content from the second set of content is based on a second model.
11. The system of claim 10 , wherein
the user profile and the content profile are represented by feature vectors, respectively; and
the first model is based on a product of the respective feature vectors of the user and of a piece of content.
12. The system of claim 10 , wherein the second model includes at least one of:
a statistic distribution;
a machine learning model; and
a decision tree.
13. The system of claim 9 , wherein the user profile, the content profile, and the content quality indicator are dynamically updated.
14. The system of claim 9 , wherein the user profile is obtained based on at least one of:
the user's declared interests;
the user's content consumption;
the user's online behaviors; and
the user's demographic information.
15. The system of claim 9 , wherein the content quality indicator is obtained based on at least one of:
timeliness of the piece of content; and
diversity of the piece of content.
16. A non-transitory machine-readable medium having information recorded thereon for selecting content, wherein the information, when read by the machine, causes the machine to perform the following:
obtaining a user profile of a user characterizing the user's interests;
obtaining a content profile for each piece of content in a first set of content and a content quality indicator characterizing quality of the piece of content in the first set of content;
estimating an affinity measure between each piece of content in the first set of content and the user based on the user profile and the content profile for the piece of content;
selecting a second set of content from the first set of content based on the affinity measures for the pieces of content in the first set of content, wherein the second set is a subset of the first set;
selecting a third set of content from the second set of content based on the content quality indicators and a plurality of contextual parameters including a type of content associated with each piece of content in the second set of content, wherein the content quality indicator of each piece of content is obtained based on engagement of other users with the piece of content and vitality of the piece of content, and wherein the third set is a subset of the second set;
filtering the third set of content based on the plurality of contextual parameters to generate a fourth set of content; and
outputting the fourth set of content on a device operated by the user.
17. The non-transitory machine-readable medium of claim 16 , wherein
the selecting the second set of content from the first set of content is based on a first model; and
the selecting the third set of content from the second set of content is based on a second model.
18. The non-transitory machine-readable medium of claim 16 , wherein the user profile, the content profile, and the content quality indicator are dynamically updated.