Language-understanding training database action pair augmentation using bidirectional translation
View Patent ↗Embodiments of the present invention relate to a method and system for augmenting a training database of an automated language-understanding system. In one embodiment, a training example in a first language is received from the training database. The first language-training example is translated to a second language output. The second language output is translated to a first variant of the first language-training example. An action pair including the first variant of the first language-training example and an action command associated with the first language-training example is stored in an augmented training database.
1. A method for augmenting an initial training database, comprising:
receiving a training example from the initial training database, wherein the training example is from an action pair comprising the training example and an action command associated with the training example;
translating the training example to an output, wherein the output comprises a variation of the training sample;
translating the output to a first variant of the training example; and
storing an augmented action pair including the first variant of the training example and the action command associated with the training example in an augmented training database.
2. The method of claim 1 , further comprising:
translating the output to a second output;
translating the second output to a second variant of the training example; and
outputting the second variant of the training example to an output manager.
3. The method of claim 2 , further comprising:
storing an augmented action pair including the second variant of the training example and the action command associated with the training example in the augmented training database.
4. The method of claim 1 , wherein the action command is a computer executable command.
5. The method of claim 1 , wherein the training example comprises a first phrase.
6. The method of claim 5 , wherein the output comprises a different phrase from the first phrase.
7. The method of claim 1 , wherein the first variant is in a natural language.
8. A system for augmenting an initial training database, comprising:
a first translation unit configured to translate a training example to an output, wherein the training example is from an action pair comprising the training example and an action command associated with the training example;
a second translation unit configured to translate the output to a first variant of the training example, wherein the output comprises a variation of the training sample; and
a database configured to store an augmented action pair including the first variant of the training example and the action command associated with the training example.
9. The system of claim 8 , further comprising:
a translation unit configured to translate the output to a second output; and
a translation unit configured to translate the second output to a second variant of the training example.
10. The system of claim 9 , further comprising:
an output manager configured to output the second variant of the training example.
11. The system of claim 9 , wherein the database is configured to store the second variant of the training example and the action command associated with the training example.
12. A machine-readable medium having stored thereon a plurality of executable instructions, the plurality of instructions comprising instructions to:
receive a training example from an initial training database, wherein the training example is from an action pair comprising the training example and an action command associated with the training example;
translate the training example to an output, wherein the output comprises a variation of the training sample;
translate the output to a first variant of the training example; and
store an augmented action pair including the first variant of the training example and the action command associated with the training example in an augmented training database.
13. The machine-readable medium of claim 12 having stored thereon additional executable instructions, the additional instructions comprising instructions to:
translate the output to a second output;
translate the second output to a second variant of the training example; and
output the second variant of the training example to an output manager.
14. The machine-readable medium of claim 13 having stored thereon additional executable instructions, the additional instructions comprising instructions to:
store an augmented action pair including the second variant of the training example and the action command associated with the training example in the augmented training database.
15. A system for augmenting an initial training database, comprising:
a translation unit configured to translate a training example to an output and configured to translate the output to a first variant of the training example, wherein the training example is from an action pair comprising the training example and an action command associated with the training example, wherein the output comprises a variation of the training sample; and
a database for storing an augmented action pair including the first variant of the training example and the action command associated with the training example.
16. The system of claim 15 , wherein the translation unit translates the output to a second output and translates the second output to a second variant of the training example.
17. The system of claim 16 , further comprising:
an output manager for outputting the second variant of the training example.
18. The system of claim 16 , wherein the database is for storing the second variant of the training example and the action command associated with the training example.