AUTOMATED ACCURACY ASSESSMENT IN TASKING SYSTEM
Disclosed is a system (and process) for determining the accuracy of computerized tasks in a task batch. The system calculates a number of reviews to assess the accuracy of a task based on a source accuracy and a reviewer accuracy. The source accuracy is based factors calculated by a predictive model, the factors including a historical accuracy of an authoring user. The reviewer accuracy is based on a true positive rate and a true negative rate of one or more reviewers of the task batch. The system transmits sourced tasks to a same number of reviewers. The system collects reviews and assesses if the task passes review based on the collected number of reviews.
1 . A method of determining an accuracy of computerized tasks in a task batch, the method comprising:
receiving, at an online system, a sourced task from an authoring user;
calculating, by the online system, a probability that the sourced task is completed correctly based on a source accuracy of the authoring user from one or more factors calculated by a predictive model;
comparing the probability to a quality threshold of a customer; and
responsive to the probability not meeting or exceeding the quality threshold of the customer:
calculating a number of reviews to assess an accuracy of the sourced task based on the source accuracy and a reviewer accuracy, the reviewer accuracy based on a true positive rate and a true negative rate of one or more reviewers of the task batch;
transmitting, by the online system, the sourced task to a same number of reviewers;
receiving, by the online system, the number of reviews from the reviewers, the number of received reviews indicating whether the sourced task passes review or does not pass review;
and transmitting the sourced task to the customer responsive to the number of received reviews indicating the sourced task passes review.
2 . The method of claim 1 , wherein the one or more factors include a historical accuracy of the authoring user based on a proportion of previously completed tasks by the authoring user that have passed review.
3 . The method of claim 1 , wherein the one or more factors includes a behavior pattern of the authoring user, the behavior pattern of the authoring user including at least one of an amount of time for the authoring user to complete the sourced task, a number of points drawn in the sourced task, and a number of words submitted in the sourced task.
4 . The method of claim 1 , wherein the true positive rate is a proportion of tasks of the task batch that the one or more reviewers correctly mark as passing review.
5 . The method of claim 1 , wherein the true negative rate is a proportion of tasks of the task batch that the one or more reviewers correctly mark as not passing review.
6 . The method of claim 1 , wherein the calculating the number of reviews comprises:
calculating an updated probability the sourced task is correctly completed given the sourced task passes review for the number of reviews, the updated probability is a first probability the sourced task passes review and is correctly completed divided by a sum of the first probability and a second probability the sourced task passes review and is incorrectly completed; and
determining if the updated probability meets or exceeds the quality threshold of the customer, wherein the calculated number of reviews corresponds to a smallest number of reviews such that the updated probability meets or exceeds the quality threshold of the customer.
7 . A computer program product stored on a non-transitory computer-readable storage medium comprising stored executable computer program instructions for determining an accuracy of computerized tasks in a task batch, the computer program instructions when executed by a computer processor cause the computer processor to:
receive, at an online system, a sourced task from an authoring user;
calculate, by the online system, a probability that the sourced task is completed correctly based on a source accuracy of the authoring user from one or more factors calculated by a predictive model;
compare the probability to a quality threshold of a customer; and
execute, responsive to the probability not meeting or exceeding the quality threshold of the customer, instructions that further cause the processor to:
calculate a number of reviews to assess an accuracy of the sourced task based on the source accuracy and a reviewer accuracy, the reviewer accuracy based on a true positive rate and a true negative rate of one or more reviewers of the task batch;
transmit, by the online system, the sourced task to a same number of reviewers;
receive, by the online system, the number of reviews from the reviewers, the number of received reviews indicating whether the sourced task passes review or does not pass review; and
transmit the sourced task to the customer responsive to the number of received reviews indicating the sourced task passes review.
8 . The computer readable medium of claim 7 , wherein the one or more factors include a historical source accuracy of the authoring user based on a proportion of previously completed tasks by the authoring user that have passed review.
9 . The computer readable medium of claim 7 , wherein the one or more factors include a behavior pattern of the authoring user, the behavior pattern of the authoring user including at least one of an amount of time for the authoring user to complete the sourced task, a number of points drawn in the sourced task, and a number of words submitted in the sourced task.
10 . The computer readable medium of claim 7 , wherein the true positive rate is a proportion of tasks of the task batch that the one or more reviewers correctly mark as passing review.
11 . The computer readable medium of claim 7 , wherein the true negative rate is a proportion of tasks of the task batch that the one or more reviewers correctly mark as not passing review.
12 . The computer readable medium of claim 7 , wherein the instructions to calculate the number of reviews further comprises instructions that when executed causes the processor to:
calculate an updated probability the sourced task is correctly completed given the sourced task passes review for the number of reviews, the updated probability is a first probability the sourced task passes review and is correctly completed divided by a sum of the first probability and a second probability the sourced task passes review and is incorrectly completed; and
determine if the updated probability meets or exceeds the quality threshold of the customer, wherein the calculated number of reviews corresponds to a smallest number of reviews such that the updated probability meets or exceeds the quality threshold of the customer.
13 . An online system configured for determining an accuracy of computerized tasks in a task batch, the online system configured to:
receive a sourced task from an authoring user;
calculate a probability that the sourced task is completed correctly based on a source accuracy of the authoring user from one or more factors calculated by a predictive model;
compare the probability to a quality threshold of a customer; and
responsive to the probability not meeting or exceeding the quality threshold of the customer:
calculate a number of reviews to assess an accuracy of the sourced task based on the source accuracy and a reviewer accuracy, the reviewer accuracy based on a true positive rate and a true negative rate of one or more reviewers of the task batch;
transmit the sourced task to a same number of reviewers;
receive the number of reviews from the reviewers, the number of received reviews indicating whether the sourced task passes review or does not pass review; and
transmit the sourced task to the customer responsive to the number of received reviews indicating the sourced task passes review.
14 . The online system of claim 13 , wherein the one or more factors include a historical source accuracy of the authoring user based on a proportion of previously completed tasks by the authoring user that have passed review.
15 . The online system of claim 13 , wherein the one or more factors include a behavior pattern of the authoring user, the behavior pattern of the authoring user includes at least one of an amount of time for the authoring user to complete the sourced task, a number of points drawn in the sourced task, and a number of words submitted in the sourced task.
16 . The online system of claim 13 , wherein the true positive rate is a proportion of tasks of the task batch that the one or more reviewers correctly mark as passing review.
17 . The online system of claim 13 , wherein the true negative rate is a proportion of tasks of the task batch that the one or more reviewers correctly mark as not passing review.
18 . The online system of claim 13 , wherein the system configured to calculate the number of reviews further comprises the system configured to:
calculate an updated probability the sourced task is correctly completed given the sourced task passes review for the number of reviews, the updated probability is a first probability the sourced task passes review and is correctly completed divided by a sum of the first probability and a second probability the sourced task passes review and is incorrectly completed; and
determine if the updated probability meets or exceeds the quality threshold of the customer, wherein the calculated number of reviews corresponds to a smallest number of reviews such that the updated probability meets or exceeds the quality threshold of the customer.