Target phrase classifier
Exemplary embodiments relate to detecting, removing, and/or replacing objectionable words and phrases in a machine-generated translation. A classifier identifies translations containing target words or phrases. The classifier may be applied to the output translation to remove target words and phrases from the translation, or to prevent target words and phrases from being automatically presented. Further, the classifier may be applied to a translation model to prevent the target words and phrases from appearing in the output translation. Still further, the classifier may be applied to training data so that the translation model is not trained using the target words of phrases. The classifier may remove target words or phrases only when the target words or phrases appear in the output translation but not the source language input data. The classifier may be provided as a standalone service, or may be employed in the context of a machine translation system.
1. A method comprising:
identifying, using a first classifier, target words or phrases in an output of a translation performed by a machine translation system;
identifying words or phrases in an input to the machine translation system that correspond to the identified target words or phrases in the output;
determining, using a second classifier, whether the identified words or phrases in the input are target words or phrases; and
outputting an indication when the identified words or phrases in the input are not target words or phrases;
wherein target words or phrases are words or phrases of a specific type; and
wherein the first and second classifiers comprise support vector machines.
2. The method of claim 1 , wherein the first and second classifiers represent words or phrases as vectors.
3. The method of claim 2 , wherein the first and second classifiers define a maximum-margin hyperplane that separates target vectors from non-target vectors.
4. The method of claim 1 , wherein the first and second classifiers identify the target words or phrases without reference to a dictionary of target words or phrases.
5. The method of claim 1 , further comprising training the first and second classifiers using training data that is labeled as either target training data or non-target training data.
6. The method of claim 1 , wherein the first classifier is trained for a destination language of the output of the machine translation system and wherein the second classifier trained for a source language of the input to the machine translation system.
7. A non-transitory computer-readable medium storing instructions that, when executed by one or more processors, cause the one or more processors to:
identify, using a first classifier, target words or phrases in an output of a translation performed by a machine translation system;
identify words or phrases in an input to the machine translation system corresponding to the identified target words or phrases in the output;
determine, using a second classifier, whether the identified words or phrases in the input are target words or phrases; and
output an indication when the identified words in the input are not target words or phrases;
wherein target words or phrases are words or phrases of a specific type; and
wherein the first and second classifiers comprise support vector machines.
8. The medium of claim 7 , wherein the first and second classifiers represent words or phrases as vectors.
9. The medium of claim 8 , wherein the first and second classifiers define a maximum-margin hyperplane that separates target vectors from non-target vectors.
10. The medium of claim 7 , wherein the first and second classifiers identify the target words or phrases without reference to a dictionary of target words or phrases.
11. The medium of claim 7 , further storing instructions for training the first and second classifiers using training data that is labeled as either target training data or non-target training data.
12. The medium of claim 7 , wherein the first classifier is trained for a destination language of the output of the machine translation system and wherein the second classifier trained for a source language of the input to the machine translation system.
13. An apparatus comprising:
a first classifier configured to identify target words or phrases in an output of a translation performed by a machine translation system;
a processor configured to identify words or phrases in the input to the machine translation system that correspond to the target words or phrases identified in the output;
a second classifier configured to determine whether the identified words or phrases in the input are target words or phrases; and
a processor configured to indicate when the identified words or phrases in the input are not target words or phrases;
wherein target words or phrases are words or phrases of a specific type; and
wherein the first and second classifiers comprise support vector machines.
14. The apparatus of claim 13 , wherein the first and second classifiers represent words or phrases as vectors.
15. The apparatus of claim 13 , wherein the first and second classifiers define a maximum-margin hyperplane that separates target vectors from non-target vectors.
16. The apparatus of claim 13 , wherein the first and second classifiers identify the target words or phrases without reference to a dictionary of target words or phrases.
17. The apparatus of claim 13 , further comprising training logic configured to train the first and second classifiers using training data that is labeled as either target training data or non-target training data.