Methods, systems, and media for identifying errors in predictive models using annotators
Methods, systems, and media for identifying errors in predictive models using annotators are provided. In some embodiments, a method for evaluating predictive models in classification systems is provided, the method comprising: causing an input region to be presented to a user, where the input region receives an instance from the user that corresponds to a predictive model; retrieving a classification conducted by the predictive model for the received instance and a confidence value associated with the classification; determining whether the received instance has been incorrectly classified by the predictive model; determining a reward associated with the incorrect classification made by the predictive model in response to determining that the received instance has been incorrectly classified by the predictive model, where the reward is based on the confidence value associated with the classification of the received instance; and providing the reward to the user.
1. A method for evaluating predictive models in classification systems, the method comprising:
causing, using a hardware processor, an input region to be presented to a user, wherein the input region receives an instance from the user;
in response to receiving the instance from the user, retrieving, using the hardware processor, a classification assigned by a predictive model for the received instance and a confidence value associated with the classification, wherein the classification has been stored in association with the instance from the user in response to determining that the classification exceeds a threshold value, and wherein the threshold value is set based at least in part on a penalty associated with misclassification of the instance;
determining, using the hardware processor, whether the received instance has been incorrectly classified by the predictive model;
in response to determining that the received instance has been incorrectly classified by the predictive model, (i) causing, using the hardware processor, the received instance and a corrected classification provided by the user to be used as training data to update the predictive model, and (ii) determining, using the hardware processor, a reward associated with the incorrect classification made by the predictive model, wherein the reward is directly proportional to the confidence value associated with the classification of the received instance;
updating, using the hardware processor, the predictive model using the received instance as training data; and
causing, using the hardware processor, an indication of the reward to be provided to the user.
2. The method of claim 1 , wherein the instance includes at least one of:
a page and a document.
3. The method of claim 1 , wherein the input region includes task information relating to the predictive model, the method further comprising:
receiving, from an administrator user, a selected predictive model from a plurality of predictive models; and
determining the task information associated with the selected predictive model to display in the input region.
4. The method of claim 1 , further comprising:
receiving a user selection of the predictive model from a plurality of predictive models; and
determining task information associated with the predictive model corresponding to the received user selection to display in the input region.
5. The method of claim 1 , further comprising, in response to determining that the received instance has been incorrectly classified by the predictive model, transmitting the received instance to an administrator user for validation of the incorrect classification.
6. The method of claim 1 , wherein determining whether the received instance has been incorrectly classified by the predictive model further comprises:
receiving an indication from the user that the received instance belongs in a minority class;
determining whether the received instance is classified by the predictive model as belonging in a majority class; and
transmitting the received instance to an administrator user for validation.
7. The method of claim 1 , wherein determining whether the received instance has been incorrectly classified by the predictive model further comprises:
receiving an indication from the user that the received instance belongs in a majority class;
determining whether the instance is classified by the predictive model as belonging in a minority class; and
transmitting the instance to an administrator user for validation.
8. The method of claim 1 , further comprising:
extracting a category identifier and a user classification from the received instance;
comparing the user classification with the classification made by the predictive model; and
inhibiting the received instances from being transmitted to an administrator user for validation based at least in part on the comparison and the confidence value.
9. The method of claim 8 , wherein the received instance is rejected in response to determining that the user classification and the classification made by the predictive model are the same and the confidence value is greater than a second threshold value.
10. The method of claim 9 , further comprising determining the reward in response to determining that the confidence value is below the second threshold value.
11. The method of claim 1 , further comprising determining an error value associated with the classification based at least in part on the confidence value.
12. The method of claim 11 , further comprising determining the reward based at least in part on the error value and the confidence value associated with the classification.
13. The method of claim 11 , further comprising determining a frequency value associated with the classification based at least in part on keywords associated with the classification made by the predictive model.
14. The method of claim 13 , further comprising:
determining the keywords associated with the classification made by the predictive model; and
transmitting a search query for instances matching the keywords associated with the classification made by the predictive model; and
calculating the frequency value based on the number of instances matching the keywords associated with the classification made by the predictive model.
15. The method of claim 13 , further comprising determining the reward based at least in part on the frequency value and the confidence value associated with the classification.
16. The method of claim 13 , further comprising determining the reward based least in part on the confidence value associated with the classification, the error value, and the frequency value.
17. The method of claim 1 , further comprising retrieving a payment distribution associated with the predictive model that includes a first payment associated with a first confidence value and a second payment associated with a second confidence value, wherein the first confidence value is greater than the second confidence value and wherein the first payment is greater than the second payment based on the confidence value.
18. The method of claim 1 , further comprising:
receiving a plurality of gaming parameters from an administrator user, wherein the input region that allows the user to submit one or more instances is presented based on the plurality of gaming parameters.
19. A system for evaluating predictive models in classification systems, the system comprising:
a processor that:
causes an input region to be presented to a user, wherein the input region receives an instance from the user;
in response to receiving the instance from the user, retrieves a classification assigned by a predictive model for the received instance and a confidence value associated with the classification, wherein the classification has been stored in association with the instance from the user in response to determining that the classification exceeds a threshold value, and wherein the threshold value is set based at least in part on a penalty associated with misclassification of the instance;
determines whether the received instance has been incorrectly classified by the predictive model;
in response to determining that the received instance has been incorrectly classified by the predictive model, (i) causing, using the hardware processor, the received instance and a corrected classification provided by the user to be used as training data to update the predictive model, and (ii) determines a reward associated with the incorrect classification made by the predictive model, wherein the reward is directly proportional to the confidence value associated with the classification of the received instance;
updating, using the hardware processor, the predictive model using the received instance as training data; and
causes an indication of the reward to be provided to the user.
20. A non-transitory computer-readable medium containing computer-executable instructions that, when executed by a processor, cause the process to perform a method for evaluating predictive models in classification systems, the method comprising:
causing an input region to be presented to a user, wherein the input region receives an instance from the user;
in response to receiving the instance from the user, retrieving a classification assigned by a predictive model for the received instance and a confidence value associated with the classification, wherein the classification has been stored in association with the instance from the user in response to determining that the classification exceeds a threshold value, and wherein the threshold value is set based at least in part on a penalty associated with misclassification of the instance;
determining whether the received instance has been incorrectly classified by the predictive model;
in response to determining that the received instance has been incorrectly classified by the predictive model, (i) causing, using the hardware processor, the received instance and a corrected classification provided by the user to be used as training data to update the predictive model, and (ii) determining a reward associated with the incorrect classification made by the predictive model, wherein the reward is directly proportional to the confidence value associated with the classification of the received instance;
updating, using the hardware processor, the predictive model using the received instance as training data; and
causing an indication of the reward to be provided to the user.