Iterative classifier training on online social networks
First Claim
1. A method comprising, by one or more computing devices associated with an online social network:
- accessing, by one or more of the computing devices, a first set of objects posted to the online social network by a plurality of client systems associated with a plurality of users of the online social network, the objects being stored in one or more data stores associated with the online social network, each object being associated with one or more comments, wherein each comment is posted to the online social network with respect to the corresponding object by a client system associated with a user of the online social network;
generating, by one or more of the computing devices, a second set of objects by adding one or more objects in the first set of objects meeting a first filtering criteria to the second set;
scoring, by one or more of the computing devices, each object in the second set of objects by comparing the comments associated with each object to a pre-determined set of terms associated with a first object-classification;
generating, by one or more of the computing devices, a training set of objects from the second set of objects by selecting each object from the second set of objects having a score greater than a first threshold score, each object in the training set being pre-classified with the first object-classification;
generating, by one or more of the computing devices, an object-classifier algorithm for the first object-classification, the object-classifier algorithm comprising one or more features each comprising a text expression and stored in association with a feature score indicating a level of correlation between the corresponding feature and the first object-classification and being determined through an iterative training process performed one or more times, each iteration of the iterative training process comprising;
training an initial object-classifier algorithm based on the comments associated with the objects in the training set of objects;
accessing a third set of objects posted to the online social network by a plurality of client systems associated with a plurality of users of the online social network, the objects being stored in one or more data stores associated with the online social network, each object being associated with one or more comments, wherein each comment is posted to the online social network with respect to the corresponding object by a client system associated with a user of the online social network;
classifying, using the initial object-classifier algorithm, each object in the third set of objects based on an analysis of the comments associated with each object, one or more of the objects in the third set of objects being classified with the first object-classification, wherein the analysis comprises comparing the comments with the features;
training a revised object-classifier algorithm based on the comments associated with the objects in the third set of objects having the first object-classification, wherein the training comprises revising one or more features of the initial object-classifier algorithm;
accessing a fourth set of objects stored in one or more data stores associated with the online social network, the fourth set of objects being generated by applying a second filtering criteria to a fifth set of objects posted to the online social network by a plurality of client systems associated with a plurality of users, the objects being stored in one or more data stores associated with the online social network, each object being associated with one or more comments, wherein each comment is posted to the online social network with respect to the corresponding object by a client system associated with a user of the online social network;
classifying, using the revised object-classifier algorithm, each object in the fourth set of objects based on an analysis of the comments associated with each object, one or more objects in the fourth set of objects being classified with the first object-classification, wherein the analysis comprises comparing the comments with the modified features;
determining a level of precision associated with the object-classifier algorithm; and
if the determined level of precision does not satisfy a specified threshold level of precision;
then, generating a sixth set of objects from the fourth set of objects by selecting each object from the fourth set of objects having a score greater than a second threshold score, each object in the sixth set of objects being classified with the first object-classification, wherein the sixth set of objects is to be used as the training set in a next iteration of the iterative training process;
else, terminating the iterative training process and storing the revised object-classifier algorithm; and
classifying, using the generated object-classifier algorithm, a particular object posted by a client system associated with a particular user, wherein the particular object is classified with the first object-classification; and
determining a status of the particular user posting the particular object based at least in part on the classification of the particular object.
2 Assignments
0 Petitions
Accused Products
Abstract
In one embodiment, a method includes accessing a first set of objects associated with an online social network, each object being associated with one or more comments. The method also includes generating a second set of objects from the first set of objects by applying a first filtering criteria to the first set of objects and scoring each object in the second set of objects based on the comments associated with each object. The method further includes generating a training set of objects from the second set of objects by selecting each object from the second set of objects having a score greater than a first threshold score, each object in the training set being associated with a first object-classification. The method further includes determining an object-classifier algorithm for the first object-classification, the object-classifier algorithm being determined through an iterative training process performed one or more times.
-
Citations
20 Claims
-
1. A method comprising, by one or more computing devices associated with an online social network:
-
accessing, by one or more of the computing devices, a first set of objects posted to the online social network by a plurality of client systems associated with a plurality of users of the online social network, the objects being stored in one or more data stores associated with the online social network, each object being associated with one or more comments, wherein each comment is posted to the online social network with respect to the corresponding object by a client system associated with a user of the online social network; generating, by one or more of the computing devices, a second set of objects by adding one or more objects in the first set of objects meeting a first filtering criteria to the second set; scoring, by one or more of the computing devices, each object in the second set of objects by comparing the comments associated with each object to a pre-determined set of terms associated with a first object-classification; generating, by one or more of the computing devices, a training set of objects from the second set of objects by selecting each object from the second set of objects having a score greater than a first threshold score, each object in the training set being pre-classified with the first object-classification; generating, by one or more of the computing devices, an object-classifier algorithm for the first object-classification, the object-classifier algorithm comprising one or more features each comprising a text expression and stored in association with a feature score indicating a level of correlation between the corresponding feature and the first object-classification and being determined through an iterative training process performed one or more times, each iteration of the iterative training process comprising; training an initial object-classifier algorithm based on the comments associated with the objects in the training set of objects; accessing a third set of objects posted to the online social network by a plurality of client systems associated with a plurality of users of the online social network, the objects being stored in one or more data stores associated with the online social network, each object being associated with one or more comments, wherein each comment is posted to the online social network with respect to the corresponding object by a client system associated with a user of the online social network; classifying, using the initial object-classifier algorithm, each object in the third set of objects based on an analysis of the comments associated with each object, one or more of the objects in the third set of objects being classified with the first object-classification, wherein the analysis comprises comparing the comments with the features; training a revised object-classifier algorithm based on the comments associated with the objects in the third set of objects having the first object-classification, wherein the training comprises revising one or more features of the initial object-classifier algorithm; accessing a fourth set of objects stored in one or more data stores associated with the online social network, the fourth set of objects being generated by applying a second filtering criteria to a fifth set of objects posted to the online social network by a plurality of client systems associated with a plurality of users, the objects being stored in one or more data stores associated with the online social network, each object being associated with one or more comments, wherein each comment is posted to the online social network with respect to the corresponding object by a client system associated with a user of the online social network; classifying, using the revised object-classifier algorithm, each object in the fourth set of objects based on an analysis of the comments associated with each object, one or more objects in the fourth set of objects being classified with the first object-classification, wherein the analysis comprises comparing the comments with the modified features; determining a level of precision associated with the object-classifier algorithm; and if the determined level of precision does not satisfy a specified threshold level of precision; then, generating a sixth set of objects from the fourth set of objects by selecting each object from the fourth set of objects having a score greater than a second threshold score, each object in the sixth set of objects being classified with the first object-classification, wherein the sixth set of objects is to be used as the training set in a next iteration of the iterative training process; else, terminating the iterative training process and storing the revised object-classifier algorithm; and classifying, using the generated object-classifier algorithm, a particular object posted by a client system associated with a particular user, wherein the particular object is classified with the first object-classification; and determining a status of the particular user posting the particular object based at least in part on the classification of the particular object. - View Dependent Claims (2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18)
-
-
19. One or more computer-readable non-transitory storage media embodying software that is operable when executed to:
-
access, by one or more of the computing devices, a first set of objects posted to the online social network by a plurality of client systems associated with a plurality of users of the online social network, the objects being stored in one or more data stores associated with an online social network, each object being associated with one or more comments, wherein each comment is posted to the online social network with respect to the corresponding object by a client system associated with a user of the online social network; generate, by one or more of the computing devices, a second set of objects by adding one or more objects in the first set of objects meeting a first filtering criteria to the second set; score, by one or more of the computing devices, each object in the second set of objects by comparing the comments associated with each object to a pre-determined set of terms associated with a first object-classification; generate, by one or more of the computing devices, a training set of objects from the second set of objects by selecting each object from the second set of objects having a score greater than a first threshold score, each object in the training set being pre-classified with the first object-classification; generate, by one or more of the computing devices, an object-classifier algorithm for the first object-classification, the object-classifier algorithm comprising one or more features each comprising a text expression and stored in association with a feature score indicating a level of correlation between the corresponding feature and the first object-classification and being determined through an iterative training process performed one or more times, each iteration of the iterative training process comprising; train an initial object-classifier algorithm based on the comments associated with the objects in the training set of objects; access a third set of objects posted to the online social network by a plurality of client systems associated with a plurality of users of the online social network, the objects being stored in one or more data stores associated with the online social network, each object being associated with one or more comments, wherein each comment is posted to the online social network with respect to the corresponding object by a client system associated with a user of the online social network; classify, using the initial object-classifier algorithm, each object in the third set of objects based on an analysis of the comments associated with each object, one or more of the objects in the third set of objects being classified with the first object-classification, wherein the analysis comprises comparing the comments with the features; train a revised object-classifier algorithm based on the comments associated with the objects in the third set of objects having the first object-classification, wherein the training comprises revising one or more features of the initial object-classifier algorithm; access a fourth set of objects stored in one or more data stores associated with the online social network, the fourth set of objects being generated by applying a second filtering criteria to a fifth set of objects posted to the online social network by a plurality of client systems associated with a plurality of users, the objects being stored in one or more data stores associated with the online social network, each object being associated with one or more comments, wherein each comment is posted to the online social network with respect to the corresponding object by a client system associated with a user of the online social network; classify, using the revised object-classifier algorithm, each object in the fourth set of objects based on an analysis of the comments associated with each object, one or more objects in the fourth set of objects being classified with the first object-classification, wherein the analysis comprises comparing the comments with the modified features; determine a level of precision associated with the object-classifier algorithm; and if the determined level of precision does not satisfy a specified threshold level of precision; then, generate a sixth set of objects from the fourth set of objects by selecting each object from the fourth set of objects having a score greater than a second threshold score, each object in the sixth set of objects being classified with the first object-classification, wherein the sixth set of objects is to be used as the training set in a next iteration of the iterative training process; else, terminate the iterative training process and storing the revised object-classifier algorithm; and classify, using the generated object-classifier algorithm, a particular object posted by a client system associated with a particular user, wherein the particular object is classified with the first object-classification; and determine a status of the particular user posting the particular object based at least in part on the classification of the particular object.
-
-
20. A system comprising:
- one or more processors; and
a memory coupled to the processors comprising instructions executable by the processors, the processors operable when executing the instructions to;access, by one or more of the computing devices, a first set of objects posted to the online social network by a plurality of client systems associated with a plurality of users of the online social network, the objects being stored in one or more data stores associated with an online social network, each object being associated with one or more comments, wherein each comment is posted to the online social network with respect to the corresponding object by a client system associated with a user of the online social network; generate, by one or more of the computing devices, a second set of objects by adding one or more objects in the first set of objects meeting a first filtering criteria to the second set; score, by one or more of the computing devices, each object in the second set of objects by comparing the comments associated with each object to a pre-determined set of terms associated with a first object-classification; generate, by one or more of the computing devices, a training set of objects from the second set of objects by selecting each object from the second set of objects having a score greater than a first threshold score, each object in the training set being pre-classified with the first object-classification; generate, by one or more of the computing devices, an object-classifier algorithm for the first object-classification, the object-classifier algorithm comprising one or more features each comprising a text expression and stored in association with a feature score indicating a level of correlation between the corresponding feature and the first object-classification and being determined through an iterative training process performed one or more times, each iteration of the iterative training process comprising; train an initial object-classifier algorithm based on the comments associated with the objects in the training set of objects; access a third set of objects posted to the online social network by a plurality of client systems associated with a plurality of users of the online social network, the objects being stored in one or more data stores associated with the online social network, each object being associated with one or more comments, wherein each comment is posted to the online social network with respect to the corresponding object by a client system associated with a user of the online social network; classify, using the initial object-classifier algorithm, each object in the third set of objects based on an analysis of the comments associated with each object, one or more of the objects in the third set of objects being classified with the first object-classification, wherein the analysis comprises comparing the comments with the features; train a revised object-classifier algorithm based on the comments associated with the objects in the third set of objects having the first object-classification, wherein the training comprises revising one or more features of the initial object-classifier algorithm; access a fourth set of objects stored in one or more data stores associated with the online social network, the fourth set of objects being generated by applying a second filtering criteria to a fifth set of objects posted to the online social network by a plurality of client systems associated with a plurality of users, the objects being stored in one or more data stores associated with the online social network, each object being associated with one or more comments, wherein each comment is posted to the online social network with respect to the corresponding object by a client system associated with a user of the online social network; classify, using the revised object-classifier algorithm, each object in the fourth set of objects based on an analysis of the comments associated with each object, one or more objects in the fourth set of objects being classified with the first object-classification, wherein the analysis comprises comparing the comments with the modified features; determine a level of precision associated with the object-classifier algorithm; and if the determined level of precision does not satisfy a specified threshold level of precision; then, generate a sixth set of objects from the fourth set of objects by selecting each object from the fourth set of objects having a score greater than a second threshold score, each object in the sixth set of objects being classified with the first object-classification, wherein the sixth set of objects is to be used as the training set in a next iteration of the iterative training process; else, terminate the iterative training process and storing the revised object-classifier algorithm; and classify, using the generated object-classifier algorithm, a particular object posted by a client system associated with a particular user, wherein the particular object is classified with the first object-classification; and determine a status of the particular user posting the particular object based at least in part on the classification of the particular object.
- one or more processors; and
Specification