System and method for labeling a content item based on a posterior probability distribution
The present invention is directed towards systems and methods for the classification or scoring of content items. The method according to one embodiment comprises providing at least one labeled content item as input to an initial classification model, a given labeled item having at least one feature, a given feature being associated with a weight, computing a posterior probability of the initial classification model for the given labeled content item and generating an updated classification model using the initial classification model and the weight associated with the given feature in the given labeled content item. The updated classification model is applied to an unlabeled content item to determine a score for the unlabeled content item.
1. A method, implemented on a machine having at least one processor, storage, and a communication platform connected to a network for determining a label for a content item, comprising:
obtaining one or more labeled content items each associated with a label;
determining a model, by the at least one processor, based on the one or more labeled content items and a posterior probability distribution, wherein:
the posterior probability distribution is determined based on a product of a likelihood function and a prior probability distribution that is a first normal distribution determined based on the one or more labeled content items,
the posterior probability distribution is transformed to a second normal distribution based on a transformation function,
a first axis of symmetry for the likelihood function is transformed to a second axis of symmetry based on the transformation function, and
the posterior probability distribution is determined based on the second axis of symmetry; and
determining a label for an unlabeled content item, based on the model.
2. The method according to claim 1 , wherein the one or more labeled content items comprise a training set of one or more labeled content items having a label with either 1 or 0 based on manually labeling using histograms on one or more previous searches.
3. The method according to claim 1 , further comprising:
computing a peak and a second derivative of the product; and
applying a Laplace approximation to the peak and the second derivative to obtain the second normal distribution.
4. The method according to claim 1 , further comprising:
representing the likelihood function as the first axis of symmetry in a solution space with the first normal distribution;
generating the second normal distribution by rescaling, using the transformation function, the first normal distribution to have an equal standard deviation in all directions about a peak thereof;
generating the second axis of symmetry using the transformation function; and
determining a maximum of the product on a solution line intersecting the rescaled second normal distribution and the second axis of symmetry.
5. The method according to claim 4 , wherein the solution line is electronically formed along a diameter of the rescaled second normal distribution and perpendicular to the second axis of symmetry.
6. The method according to claim 1 , comprising annotating the unlabeled content item based on a score of the unlabeled content item by highlighting the content item that is scored above a threshold.
7. The method according to claim 1 , wherein the one or more labeled content items comprise one or more features, wherein the one or more features comprise at least two of a source, an author, and an age of a news story and the one or more features comprise one or more numerical feature vectors.
8. The method according to claim 1 , comprising modifying the likelihood function using a baseline feature.
9. The method according to claim 7 , summing the one or more features having non-zero values associated therewith.
10. The method according to claim 4 , comprising computing a mean and standard deviation; and updating the prior probability distribution with the computed mean and standard deviation.
11. A system, comprising:
at least one processor-based computing device, the at least one processor-based computing device operative at least to:
obtain one or more labeled content items each associated with a label;
determine a model, based on the one or more labeled content items and a posterior probability distribution, wherein:
the posterior probability distribution is determined based on a product of a likelihood function and a prior probability distribution that is a first normal distribution determined based on the one or more labeled content items,
the posterior probability distribution is transformed to a second normal distribution based on a transformation function,
a first axis of symmetry for the likelihood function is transformed to a second axis of symmetry based on the transformation function, and
the posterior probability distribution is determined based on the second axis of symmetry; and
determine a label for an unlabeled content item, based on the model.
12. The system according to claim 11 , wherein the one or more labeled content items comprise a training set of one or more labeled content items having a label with either 1 or 0 based on manually labeling using histograms on one or more previous searches.
13. The system according to claim 11 , the at least one processor-based computing device operative to compute a peak and a second derivative of the product and apply a Laplace approximation to the peak and the second derivative to obtain the second normal distribution.
14. The system according to claim 11 , the at least one processor-based computing device operative to represent the prior probability distribution as the first normal distribution, represent the likelihood function as the first axis of symmetry in a solution space with the first normal distribution, generate the second normal distribution by rescaling the first normal distribution to have an equal standard deviation in all directions about a peak thereof through the use of a transformation function, generate the second axis of symmetry using the transformation function and determine a maximum of the product on a solution line intersecting the rescaled second normal distribution and the second axis of symmetry.
15. The system according to claim 11 , the at least one processor-based computing device operative to annotate the unlabeled content item based on a score of the unlabeled content item by highlighting the content item that is scored above a threshold.
16. The system according to claim 11 , wherein the one or more labeled content items comprise one or more features, wherein the one or more features comprise at least two of source, author, and age of a news story and the one or more features comprise one or more numerical feature vectors.
17. The system according to claim 16 , the at least one processor-based computing device operative to sum the one or more features having non-zero values associated therewith.
18. The system according to claim 11 , the at least one processor-based computing device operative to modify the likelihood function using a baseline feature.
19. The system according to claim 14 , the at least one processor-based computing device operative to form the solution line along a diameter of the rescaled second normal distribution and perpendicular to the second axis of symmetry.
20. The system according to claim 14 , the at least one processor-based computing device operative to compute a mean and standard deviation; and update the prior probability distribution with the computed mean and standard deviation.