Learning edit machines for robust multimodal understanding

1. A method of processing received data in a grammar, the method causing a computing device to perform steps comprising:

receiving, by a processor of the computing device, input data having a characteristic that the input data cannot be assigned an interpretation by a grammar;

translating, by a processor of the computing device, the input data into translated input data using a 4-edit machine; and

submitting the translated input data into the grammar.

2. The method of claim 1 , further causing the computing device to perform steps comprising:

determining whether the received input data can be assigned an interpretation by the grammar.

3. The method of claim 1 , wherein the translated input data can be assigned an interpretation by the grammar.

4. The method of claim 1 , wherein the received input data is an input sentence that is not in the grammar.

5. The method of claim 4 , wherein the input sentence and the translated input sentence are in the same language.

6. The method of claim 1 , wherein the translation is performed by a weighted finite state edit transducer.

7. The method of claim 4 , wherein the grammar is a multimodal grammar.

8. The method of claim 1 , wherein the translation occurs in a framework which models a noisy channel.

9. The method of claim 1 , wherein the input data is an utterance comprising at least one of: speech data, graffiti and motion.

10. The method of claim 1 , wherein translating the input data further causes the computing device to perform steps comprising coercing a set of strings encoded in a lattice resulting from input recognition into a closest string in the grammar that can be assigned an interpretation.

11. The method of claim 10 , wherein the input data relates to a speech utterance processed by an automatic speech recognition module.

12. The method of claim 1 , wherein translating the input data further causes the computing device to perform steps comprising identifying at least one string having a least costly number of edits that can be assigned an interpretation by the grammar.

13. A method of interpreting input to a computing device, the method causing the computing device to perform steps comprising:

receiving sentences as an input;

generating, by a processor of the computing device, a translation model comprising pairings, each pairing comprising a sentence subset and a sentence, wherein sentence subsets represents the input that is not in a grammar;

when input is received in the form of subsets of sentences, translating the received input using the translation model into meanings using a 4-edit machine; and

inputting the translated input into the grammar.

14. A non-transitory computer-readable storage medium storing instructions for controlling a computing device to perform the steps:

receiving input data having a characteristic that the input data cannot be assigned an interpretation by a grammar;

translating the input data into translated input data using a 4-edit machine; and

submitting the translated input data into the grammar.

15. The non-transitory computer-readable storage medium of claim 14 , the instructions further comprise determining whether the received input data can be assigned an interpretation by the grammar.

16. The non-transitory computer-readable storage medium of claim 14 , wherein the translated input data can be assigned an interpretation by the grammar.

17. The non-transitory computer-readable storage medium of claim 14 , wherein the received input data is an input sentence that is not in the grammar.

18. The non-transitory computer-readable storage medium of claim 17 , wherein the input sentence and the translated input sentence are in the same language.

19. The non-transitory computer-readable storage medium of claim 14 , wherein the translation is performed by a weighted finite state edit transducer.

20. The non-transitory computer-readable storage medium of claim 14 , wherein the grammar is a multimodal grammar.