Task-level answer confidence estimation for worker assessment
Particular embodiments may receive results of a task performed by a plurality of workers, wherein fraud analysis information is provided for each of the results. An accuracy confusion matrix may be determined based on the results. For each of the results, an accuracy confidence score may be determined based on the fraud analysis information. Based on the accuracy confusion matrix and the accuracy confidence score, accuracy of workers performing the task may be predicted.
1. A method comprising:
receiving, by one or more processors of an information processing system, results of a task performed by a plurality of workers, wherein fraud analysis information is provided for each of the results;
determining, by the one or more processors, an accuracy confusion matrix for each worker of the plurality of workers based on the results;
determining, by the one or more processors, for each result of the plurality of results, an accuracy confidence score based at least on the fraud analysis information provided for each of the results;
receiving, by the one or more processors, an additional result of the task;
determining, by the one or more processors, that the additional result causes the plurality of results to fail to satisfy an accuracy quality threshold; and
predicting, by the one or more processors, a number of additional workers who will cause the plurality of results to satisfy the accuracy quality threshold, wherein the predicting is based on predicting an accuracy of each of the additional workers in performing the task based on the accuracy confusion matrices and the accuracy confidence scores.
2. The method of claim 1 , further comprising:
predicting a number of additional results required to satisfy the accuracy quality threshold for the task, wherein the predicting the number of additional workers who will cause the plurality of results to satisfy the accuracy quality threshold is based on the predicted number of additional results.
3. The method of claim 1 , further comprising:
requesting additional results of the task from one or more additional workers, wherein the requesting involves at least the predicted number of additional workers.
4. The method of claim 1 , wherein the accuracy confidence scores are determined based on attributes of one or more confirmed instances of fraud assessment.
5. The method of claim 1 , further comprising:
identifying a vector of shared features based on skills needed to accurately perform the task; and
wherein the accuracy confidence scores are determined based on the vector of shared features.
6. The method of claim 5 , further comprising:
determining, using a Transfer Learning Independent Bayesian Classifier Combination (TLIBCC) algorithm, optimal values for each of the shared features, wherein the predicted accuracy of each of the additional workers is further based on the optimal values.
7. The method of claim 1 , wherein tasks are classified under one or more task types, wherein the task types are based on one or more actions involved in performing the task, a level of difficulty of the task, pre-requisites required for the task, a level of compensation provided for the task, a duration of the task, a minimum number of iterations required for the task, or a maximum number of iterations required for the task.
8. One or more computer-readable non-transitory storage media embodying software comprising instructions operable when executed to:
receive, by one or more processors of an information processing system, results of a task performed by a plurality of workers, wherein fraud analysis information is provided for each of the results;
determine, by the one or more processors, an accuracy confusion matrix for each worker of the plurality of workers based on the results;
determine, by the one or more processors, for each result of the plurality of results, an accuracy confidence score based at least on the fraud analysis information provided for each of the results;
receive, by the one or more processors, an additional result of the task;
determine, by the one or more processors, that the additional result causes the plurality of results to fail to satisfy an accuracy quality threshold; and
predict, by the one or more processors, a number of additional workers who will cause the plurality of results to satisfy the accuracy quality threshold, wherein the predicting is based on predicting an accuracy of each of the additional workers in performing the task based on the accuracy confusion matrices and the accuracy confidence scores.
9. The media of claim 8 , the software further comprising instructions operable when executed to: predict a number of additional results required to satisfy the accuracy quality threshold for the task, wherein the predicting the number of additional workers who will cause the plurality of results to satisfy the accuracy quality threshold is based on the predicted number of additional results.
10. The media of claim 8 , the software further comprising instructions operable when executed to:
request additional results of the task from one or more additional workers, wherein the requesting involves at least the predicted number of additional workers.
11. The media of claim 8 , wherein the accuracy confidence scores are determined based on attributes of one or more confirmed instances of fraud assessment.
12. The media of claim 8 , the software further comprising instructions operable when executed to:
identify a vector of shared features based on skills needed to accurately perform the task; and
wherein the accuracy confidence scores are determined based on the vector of shared features.
13. The media of claim 12 , the software further comprising instructions operable when executed to:
determine, using a Transfer Learning Independent Bayesian Classifier Combination (TLIBCC) algorithm, optimal values for each of the shared features, wherein the predicted accuracy of each of the additional workers is further based on the optimal values.
14. The media of claim 8 , wherein tasks are classified under one or more task types, wherein the task types are based on one or more actions involved in performing the task, a level of difficulty of the task, pre-requisites required for the task, a level of compensation provided for the task, a duration of the task, a minimum number of iterations required for the task, or a maximum number of iterations required for the task.
15. A system comprising one or more processors and a memory coupled to the processors comprising instructions executable by the processors, the processors being operable when executing the instructions to:
receive results of a task performed by a plurality of workers, wherein fraud analysis information is provided for each of the results;
determine an accuracy confusion matrix for each worker of the plurality of workers based on the results;
determine, for each result of the plurality of results, an accuracy confidence score based at least on the fraud analysis information provided for each of the results;
receive an additional result of the task;
determine that the additional result causes the plurality of results to fail to satisfy an accuracy quality threshold; and
predict a number of additional workers who will cause the plurality of results to satisfy the accuracy quality threshold, wherein the predicting is based on predicting an accuracy of each of the additional workers in performing the task based on the accuracy confusion matrices and the accuracy confidence scores.
16. The system of claim 15 , the processors being further operable when executing the instructions to: predict a number of additional results required to satisfy the accuracy quality threshold for the task, wherein the predicting the number of additional workers who will cause the plurality of results to satisfy the accuracy quality threshold k based on the predicted number of additional results.
17. The system of claim 15 , the processors being further operable when executing the instructions to:
request additional results of the task from one or more additional workers, wherein the requesting involves at least the predicted number of additional workers.
18. The method of claim 1 , wherein:
a task is one or more actions performed by a worker;
a result for a task is an output generated by a worker performing that task; and
wherein the task is part of a workflow configured to determine at least one answer to at least one question.
19. The method of claim 1 , wherein:
the plurality of results satisfy the accuracy quality threshold when the plurality of results include a number of errors below a set amount.