SYSTEMS AND METHODS FOR TRAINING A MACHINE LEARNING MODEL FOR A SECOND LANGUAGE BASED ON A MACHINE LEARNING MODEL FOR A FIRST LANGUAGE
Systems, methods, and non-transitory computer readable media can train a machine learning model for a first language to determine a classification for a content item in the first language. Machine translation can be performed to generate respective machine translations of a plurality of content items in a second language into the first language. Respective classifications for the plurality of content items in the second language can be determined based on the machine translations of the plurality of content items in the second language and the machine learning model for the first language. Training data in the second language can be automatically generated, where the training data in the second language includes the plurality of content items in the second language and the respective classifications.
1 . A computer-implemented method comprising:
providing, by a computing system, a first machine learning model to determine classifications for a first plurality of content items in a first language;
performing, by the computing system, machine translation of a second plurality of content items originally in a second language to generate the second plurality of content items in the first language, classifications for the second plurality of content items to be determined;
in response to performance of machine translation of the second plurality of content items to the first language, determining, by the computing system, based on the first machine learning model, respective classifications for the second plurality of content items in the first language;
generating, by the computing system, training data in the second language to train a second machine learning model to determine a classification for a content item in the second language, the training data in the second language including i) the second plurality of content items in the second language and ii) labels constituting the respective classifications of the second plurality of content items determined by the first machine learning model; and
training, by the computing system, the second machine learning model based on the training data in the second language including i) the second plurality of content items in the second language and ii) labels constituting the respective classifications of the second plurality of content items determined by the first machine learning model after the second plurality of content items were translated from the first language to the second language.
2 . The computer-implemented method of claim 1 , wherein the first machine learning model is trained based on training data in the first language that includes content items in the first language and their respective classifications.
3 . The computer-implemented method of claim 1 , wherein the first machine learning model is trained to output scores indicative of a predicted likelihood of the first plurality of content items in the first language being associated with the classifications in the first language.
4 . (canceled)
5 . The computer-implemented method of claim 1 , wherein the second machine learning model is trained to output a score indicative of a predicted likelihood of the content item in the second language being associated with the classification for the content item in the second language.
6 . The computer-implemented method of claim 1 , further comprising:
obtaining, by the computing system, a particular content item in the second language; and
determining, by the computing system, a classification for the particular content item in the second language based on the second machine learning model.
7 . The computer-implemented method of claim 1 , further comprising:
refining, by the computing system, the second machine learning model based at least on a verified portion of the training data in the second language.
8 . (canceled)
9 . The computer-implemented method of claim 1 , wherein performing machine translation to the second plurality of content items in the second language to generate the machine translations of the second plurality of content items in the first language includes translating text associated with the second plurality of content items in the second language from the second language into the first language.
10 . The computer-implemented method of claim 2 , wherein the training data in the first language and the training data in the second language include features relating to one or more of: content attributes, user attributes, comment attributes, or reaction attributes.
11 . A system comprising:
at least one hardware processor; and
a memory storing instructions that, when executed by the at least one processor, cause the system to perform:
providing a first machine learning model to determine classifications for a first plurality of content items in a first language;
performing machine translation of a second plurality of content items originally in a second language to generate the second plurality of content items in the first language, classifications for the second plurality of content items to be determined;
in response to performance of machine translation of the second plurality of content items to the first language, determining, based on the first machine learning model, respective classifications for the second plurality of content items in the first language;
generating training data in the second language to train a second machine learning model to determine a classification for a content item in the second language, the training data in the second language including i) the second plurality of content items in the second language and ii) labels constituting the respective classifications of the second plurality of content items determined by the first machine learning model; and
training the second machine learning model based on the training data in the second language including i) the second plurality of content items in the second language and ii) labels constituting the respective classifications of the second plurality of content items determined by the first machine learning model after the second plurality of content items were translated from the first language to the second language.
12 . The system of claim 11 , wherein the instructions, when executed, further cause the system to perform:
training the second machine learning model based on the training data in the second language to determine a classification for a content item in the second language.
13 . The system of claim 12 , wherein the second machine learning model is trained to output a score indicative of a predicted likelihood of the content item in the second language being associated with the classification for the content item in the second language.
14 . The system of claim 12 , wherein the instructions, when executed, further cause the system to perform:
obtaining a particular content item in the second language; and
determining a classification for the particular content item in the second language based on the second machine learning model.
15 . The system of claim 12 , wherein the instructions, when executed, further cause the system to perform:
refining the second machine learning model based at least on a verified portion of the training data in the second language.
16 . A non-transitory computer readable medium including instructions that, when executed by at least one hardware processor of a computing system, cause the computing system to perform a method comprising:
providing a first machine learning model to determine classifications for a first plurality of content items in a first language;
performing machine translation of a second plurality of content items originally in a second language to generate the second plurality of content items in the first language, classifications for the second plurality of content items to be determined;
in response to performance of machine translation of the second plurality of content items to the first language, determining, based on the first machine learning model, respective classifications for the second plurality of content items in the first language;
generating training data in the second language to train a second machine learning model to determine a classification for a content item in the second language, the training data in the second language including i) the second plurality of content items in the second language and ii) labels constituting the respective classifications of the second plurality of content items determined by the first machine learning model; and
training the second machine learning model based on the training data in the second language including i) the second plurality of content items in the second language and ii) labels constituting the respective classifications of the second plurality of content items determined by the first machine learning model after the second plurality of content items were translated from the first language to the second language.
17 . The non-transitory computer readable medium of claim 16 , wherein the instructions, when executed, further cause the computing system to perform:
training the second machine learning model based on the training data in the second language to determine a classification for a content item in the second language.
18 . The non-transitory computer readable medium of claim 17 , wherein the second machine learning model is trained to output a score indicative of a predicted likelihood of the content item in the second language being associated with the classification for the content item in the second language.
19 . The non-transitory computer readable medium of claim 17 , wherein the instructions, when executed, further cause the computing system to perform:
obtaining a particular content item in the second language; and
determining a classification for the particular content item in the second language based on the second machine learning model.
20 . The non-transitory computer readable medium of claim 17 , wherein the instructions, when executed, further cause the computing system to perform:
refining the second machine learning model based at least on a verified portion of the training data in the second language.
21 . The computer-implemented method of claim 1 , wherein the training data for the first machine learning model includes training examples labeled by humans.
22 . The computer-implemented method of claim 1 , wherein the training data for the second machine learning model is generated without human labeling.