Machine learning models with accurate data labeling
An example method, executed by a processing system, includes receiving data from a data source, routing the data to a plurality of different data labelers to label the data, determining that an inflection point associated with an accuracy of the plurality of different data labelers is reached, re-routing additional data from the data source to a machine learning model, and activating the machine learning model to label the additional data.
1. A method comprising:
receiving, by a processing system comprising at least one processor, data from a data source;
routing, by the processing system, the data to a plurality of different data labelers to label the data;
determining, by the processing system, that an inflection point associated with an accuracy of the plurality of different data labelers is reached, wherein the inflection point comprises when the accuracy of the plurality of different data labelers relative to a consensus reaches an accuracy threshold for the data;
re-routing, by the processing system, additional data from the data source to a machine learning model when the inflection point is reached; and
activating, by the processing system, the machine learning model to label the additional data.
2. The method of claim 1 , further comprising:
controlling, by the processing system, requested data to an endpoint device based on the data and the additional data that are labeled.
3. The method of claim 1 , further comprising:
controlling, by the processing system, a network element within a communication network to connect an endpoint device to a particular data source based on the data and the additional data that are labeled.
4. The method of claim 3 , wherein a label relates to a sentiment.
5. The method of claim 4 , wherein the particular data source comprises a media server that includes media associated with the sentiment.
6. The method of claim 1 , wherein the routing is performed via a gamification interface.
7. The method of claim 6 , wherein the gamification interface comprises a scoreboard that ranks the plurality of different data labelers based on a respective score.
8. The method of claim 7 , wherein the respective score is based on a respective accuracy of the plurality of different data labelers.
9. The method of claim 8 , wherein the respective accuracy is measured relative to the consensus.
10. The method of claim 9 , wherein the consensus comprises a label applied by a highest percentage of the plurality of different data labelers for the data.
11. The method of claim 7 , wherein the respective score is based on a frequency of labeling, a number of labels provided, and an accuracy across different data sets from different data sources.
12. The method of claim 7 , wherein a top N number of data labelers of the plurality of different data labelers is identified to receive a reward.
13. The method of claim 12 , wherein the reward comprises a gift item or a monetary reward.
14. The method of claim 12 , wherein the reward comprises an increase in volume of data for labeling.
15. The method of claim 12 , wherein the reward comprises receiving higher priority data to be labelled.
16. The method of claim 12 , wherein the reward comprises a greater weight for labels applied by the top N number of data labelers.
17. The method of claim 7 , wherein a bottom M number of data labelers of the plurality of different data labelers is identified to receive a penalty.
18. The method of claim 17 , wherein the penalty comprises at least one of: a lower ranking, receiving a lower volume of data for labeling, or being disqualified from participating as a data labeler.
19. A non-transitory computer-readable medium storing instructions which, when executed by a processing system including at least one processor, cause the processing system to perform operations, the operations comprising:
receiving data from a data source;
routing the data to a plurality of different data labelers to label the data;
determining that an inflection point associated with an accuracy of the plurality of different data labelers is reached, wherein the inflection point comprises when the accuracy of the plurality of different data labelers relative to a consensus reaches an accuracy threshold for the data;
re-routing additional data from the data source to a machine learning model when the inflection point is reached; and
activating the machine learning model to label the additional data.
20. An apparatus comprising:
a processing system including at least one processor; and
a computer-readable medium storing instructions which, when executed by the processing system, cause the processing system to perform operations, the operations comprising:
receiving data from a data source;
routing the data to a plurality of different data labelers to label the data;
determining that an inflection point associated with an accuracy of the plurality of different data labelers is reached, wherein the inflection point comprises when the accuracy of the plurality of different data labelers relative to a consensus reaches an accuracy threshold for the data;
re-routing additional data from the data source to a machine learning model when the inflection point is reached; and
activating the machine learning model to label the additional data.