Method and system for determining result for task executed in crowd-sourced environment

1. A method for completing a task executed in a crowd-sourced environment, the task being of a given type of tasks, the method being executed by a server, the method comprising, at a training phase:

transmitting, by the server, a training task to a plurality of users of the crowd-sourced environment, the training task being associated with a predefined training answer;

acquiring, by the server, a plurality of training results responsive to the training task having been submitted by the plurality of users;

acquiring, by the server, a respective user activity history associated with each user of the plurality of users, the respective user activity history including previously collected user interactions and characteristics of the user interactions within the crowd-sourced environment;

for each of the plurality of training results, assigning a label value, the label value being one of a first label value and a second label value, the first label value being assigned to one or more training results corresponding to the predefined training answer, and the second label value being assigned to one or more training results not corresponding to the predefined training answer;

generating, by the server, a set of triples of training data, each triples of training data being associated with a given training result of the plurality of training results, a given triples of training data including:

the given training result;

the user activity history associated with the given training result;

the label value assigned to the given training result;

training a machine learning algorithm (MLA), the training including:

determining for each triples of training data, a set of features representative of a property of the triples of training data; and

generating an inferred function based on the set of features, the inferred function being configured to determine a error parameter of a given result received from a given user to a given task, the error parameter representing a probability of the given result being an incorrect result, the error parameter being determined based on the user activity history of the given user.

2. The method of claim 1 , wherein the MLA is a first MLA, the method further comprising:

at an in-use phase:

acquiring, by the server, a result to the task received by a user, the task being different from the training task;

acquiring, by the server, the user activity history associated with the user;

determining, using the first MLA, the error parameter of the result based on the user activity history associated with the user;

inputting the error parameter into a second MLA trained to generate a checking score, the checking score being indicative of a necessity to submit the result for additional verification; and

in response to the checking score being above a predetermined threshold, transmitting the task and the result to a human assessor for additional verification.

3. The method of claim 1 , wherein the error parameter is further indicative of a necessity to submit the given result for additional verification.

4. The method of claim 3 , the method further comprising:

at an in-use phase:

acquiring, by the server, a result to the task having been submitted by a user, the task being different from the training task;

acquiring, by the server, the user activity history associated with the user;

determining, using the MLA, the error parameter of the result based on the user activity history associated with the user; and

in response to the error parameter being above a predetermined threshold, transmitting the task and the result to a human assessor for assigning the label to the result.

5. The method of claim 4 , the method further comprising:

acquiring, by the server, the assigned label value from the human assessor; and

in response to the assigned label value being indicative of the result being the incorrect result to the task, generating an additional training triple of data, the additional triple of data being different from the triple of training data included within the set of triples of training data, the additional triples of data comprising:

the result;

the user activity history associated with the user;

the assigned label value; and

training the MLA, the training including updating the inferred function based on the set of features representative of a property of the additional triples of data, for a subsequent determination of the error parameter.

6. The method of claim 5 , further comprising:

receiving, by the server, a quality score associated with the user, the quality score being indicative of a reliability of the user; and

in response to the assigned label value being indicative of the result being the incorrect answer to the task, lowering the quality score associated with the user.

7. The method of claim 6 , further comprising:

upon determining that the lowered quality score is below a predetermined threshold, applying a restrictive action to the user, the restrictive action including suspending an account of the user within the crowd-sourced environment.

8. The method of claim 1 , wherein the given type of task is one of a labelling task and a picture uploading task.

9. The method of claim 5 , wherein in response to the assigned label value being indicative of the result being the incorrect answer to the task, the method further comprises re-training the MLA using an output of the comparison of the result and the assigned label.

10. The method of claim 9 , wherein the retraining is further based on features of the result and features associated with the human assessor.

11. The method of claim 10 , features of the result are representative of the features associated with the incorrect answer and features associated with a correct result, respectively.

12. The method of claim 11 , wherein the task is associated with a list of executional instructions for the user to follow, and wherein the correct result is based on the user following the executional instructions and the incorrect result is based on the user not following the executional instructions.

13. The method of claim 12 , wherein the features of the result are further representative: the human assessor following the executional instructions and the human assessor not following the executional instructions.

14. A system for completing a task executed in a crowd-sourced environment, the task being of a given type of tasks, the system comprising a server, the server comprising a processor configured to, in a training phase:

transmit, by the server, a training task to a plurality of users of the crowd-sourced environment, the training task being associated with a predefined training answer;

acquire, by the server, a plurality of training results responsive to the training task having been submitted by the plurality of users;

acquire, by the server, a respective user activity history associated with each user of the plurality of users, the respective user activity history including previously collected user interactions and characteristics of the user interactions within the crowd-sourced environment;

for each of the plurality of training results, assign a label value, the label value being one of a first label value and a second label value, the first label value being assigned to one or more training results corresponding to the predefined training answer, and the second label value being assigned to one or more training results not corresponding to the predefined training answer;

generate, by the server, a set of triples of training data, each triples of training data being associated with a given training result of the plurality of training results, a given triples of training data including:

the given training result;

the user activity history associated with the given training result;

the label value assigned to the given training result;

train a machine learning algorithm (MLA), the training including:

determining for each triples of training data, a set of features representative of a property of the triples of training data; and

generate an inferred function based on the set of features, the inferred function being configured to determine a error parameter of a given result received from a given user to a given task, the error parameter representing a probability of the given result being an incorrect result, the error parameter being determined based on the user activity history of the given user.

15. The system of claim 14 , wherein the MLA is a first MLA, the processor is further configured to:

at an in-use phase:

acquire, by the server, a result to the task received by a user, the task being different from the training task;

acquire, by the server, the user activity history associated with the user;

determine, using the first MLA, the error parameter of the result based on the user activity history associated with the user;

input the error parameter into a second MLA trained to generate a checking score, the checking score being indicative of a necessity to submit the result for additional verification; and

in response to the checking score being above a predetermined threshold, transmit the task and the result to a human assessor for additional verification.

16. The system of claim 14 , wherein the error parameter is further indicative of a necessity to submit the given result for additional verification.

17. The system of claim 16 , the processor being further configured to:

at an in-use phase:

acquire, by the server, a result to the task having been submitted by a user, the task being different from the training task;

acquire, by the server, the user activity history associated with the user;

determine, using the MLA, the error parameter of the result based on the user activity history associated with the user; and

in response to the error parameter being above a predetermined threshold, transmit the task and the result to a human assessor for assigning the label to the result.

18. The system of claim 17 , the processor being further configured to:

acquire, by the server, the assigned label value from the human assessor; and

in response to the assigned label value being indicative of the result being the incorrect result to the task, generate an additional training triple of data, the additional triple of data being different from the triple of training data included within the set of triples of training data, the additional triples of data comprising:

the result;

the user activity history associated with the user;

the assigned label value; and

train the MLA, the training including updating the inferred function based on the set of features representative of a property of the additional triples of data, for a subsequent determination of the error parameter.

19. The system of claim 17 , the processor being further configured to:

receive, by the server, a quality score associated with the user, the quality score being indicative of a reliability of the user; and

in response to the assigned label value being indicative of the result being the incorrect answer to the task, lower the quality score associated with the user.

20. The system of claim 19 , the processor being further configured to:

upon determining that the lowered quality score is below a predetermined threshold, apply a restrictive action to the user, the restrictive action including suspending an account of the user within the crowd-sourced environment.