Worker group identification
AssigneeMicrosoft Technology Licensing, LLC
InventorsBo Zhao, Ariel Fuxman, Hongwei Li
Life of this patent
Estimating the term…
Abstract
Various techniques for identifying a target worker group are described herein. In one example, a method includes detecting a response to a task from each worker in a group of workers and detecting a set of characteristics that correspond to each worker, wherein each characteristic comprises at least one attribute. The method can also include detecting a first attribute that corresponds to workers that provide responses with an accuracy above a threshold value. Furthermore, the method can include identifying the target worker group, the target worker group comprising the workers corresponding to the detected first attribute. The method may also include sending an additional task to the target worker group.
Claims17
A method for identifying a target worker group comprising:
- detecting, via a processor, a response to a task from each worker in a group of workers;
- detecting, via the processor, a set of characteristics that correspond to each worker, wherein at least one characteristic comprises a plurality of attributes;
- generating, via the processor, the target worker group based on a top-down technique comprising:
- detecting that the at least one characteristic has a higher effect value than another characteristic of the set of characteristics, wherein the effect value indicates a likelihood that workers having the characteristic will provide an accurate response to a particular task;
- calculating an effect value for each attribute of the plurality of attributes of the at least one characteristic, wherein the effect value indicates a likelihood that workers having the attribute for the characteristic will provide an accurate response to the particular task;
- ranking the attributes by the effect values to generate a subgroup based on a higher ranked attribute;
- modifying the target worker group to include workers from the subgroup, the target worker group comprising the workers corresponding to a first attribute;
- and
- providing the particular task to the target work group;
- and
- modifying, via the processor, a machine learning model to perform a classification task via training on a label identified in a response of each worker in the target worker group to the particular task.
One or more computer-readable storage devices for identifying a target worker group comprising a plurality of instructions that, when executed by a processor, cause the processor to:
- detect a response to a task from each worker in a group of workers;
- detect a set of characteristics that correspond to each worker, wherein at least one characteristic comprises at least one a plurality of attributes;
- generate the target worker group based on a top-down technique comprising:
- detecting that the at least one characteristic has a higher effect value than another characteristic of the set of characteristics, wherein the effect value indicates a likelihood that workers having the characteristic will provide an accurate response to a particular task;
- calculating an effect value for each attribute of the plurality of attributes of the at least one characteristic, wherein the effect value indicates a likelihood that workers having the attribute for the characteristic will provide an accurate response to the particular task;
- ranking the attributes by the effect values to generate a subgroup based on a higher ranked attribute;
- modifying the target worker group to include workers from the subgroup, the target worker group comprising the workers corresponding to a first attribute;
- and
- providing the particular task to the target work group;
- and
- modify a machine learning model to perform a classification task via training on a label identified in a response of each worker in the target worker group to the particular task.
A computing device for identifying a target worker group comprising:
- a processor to:
- detect a response to a task from each worker in a group of workers;
- detect a set of characteristics that correspond to each worker, wherein at least one characteristic comprises a plurality of attributes;
- detect the target worker group based on a top-down technique comprising:
- detecting that the at least one characteristic has a higher effect value than another characteristic of the set of characteristics, wherein the effect value indicates a likelihood that workers having the characteristic will provide an accurate response to a particular task;
- calculating an effect value for each attribute of the plurality of attributes of the at least one characteristic, wherein the effect value indicates a likelihood that workers having the attribute for the characteristic will provide an accurate response to the particular task;
- ranking the attributes by the effect values to generate a subgroup based on a higher ranked attribute;
- modifying the target worker group to include workers from the subgroup, the target worker group comprising the workers corresponding to a first attribute;
- and
- providing the particular task to the target work group;
- and
- modify a machine learning model to perform a classification task via training on a label identified in a response of each worker in the target worker group to the particular task.
Dependencies are inferred from the claim text.