Automatic training of a syntactic and semantic parser using a genetic algorithm
View Patent ↗Disclosed are methods, systems, and computer-readable mediums for automatic training of a syntactic and semantic parser using a genetic algorithm. An initial population is created, where the initial population comprises a vector of parameters for elements of syntactic and semantic descriptions of a source sentence. A natural language compiler (NLC) system is used to translate the sentence from the source language into a target language based on the syntactic and semantic descriptions of the source sentence. A vector of quality ratings is generated where each quality rating in the vector of quality ratings is of a corresponding parameter in the vector of parameters. Quality ratings are evaluated according to specific criterion, which comprise parameters such as a BLEU score and a number of emergency sentences. A number of parameters in the vector of parameters are replaced with adjusted parameters.
1. A method comprising:
generating, by a processing device, a set of vectors of parameters of a translation system utilizing syntactic and semantic descriptions;
translating, by the translation system using the vectors of parameters, a natural language text to produce a first plurality of translated texts;
determining a first vector of translation quality ratings reflecting translation quality of the first plurality of translated texts;
modifying at least one parameter of each vector of parameters to produce a plurality of vectors of modified parameters;
translating, by the translation system using the vectors of modified parameters, the natural language text to produce a second plurality of translated texts;
determining a second vector of translation quality ratings reflecting translation quality of the second plurality of translated texts; and
responsive to determining that an element of the second vector of translation quality ratings exceeds a corresponding element of the first vector of translation quality ratings, modifying a vector of parameters based on a corresponding vector of modified parameters.
2. The method of claim 1 , wherein modifying the vector of parameters comprises:
identifying an ordinal number of the element of the second vector of translation quality ratings that exceeds the corresponding element of the first vector of translation quality ratings;
identifying the vector of parameters as having the ordinal number in the set of vectors of parameters;
identifying an element of the vector of parameters; and
replacing the element of the vector of parameters with a corresponding element of the vector of modified parameters.
3. The method of claim 1 , wherein determining a first vector of translation quality ratings comprises determining a plurality of Bilingual Evaluation Understudy (BLEU) scores reflecting translation quality of the first plurality of translated texts.
4. The method of claim 1 , further comprising:
repeating the translating operation using the vector of parameters responsive to evaluating a terminating condition.
5. The method of claim 4 , wherein evaluating the terminating condition comprises determining that at least one element of the second vector of translation quality ratings exceeds a pre-defined quality rating value.
6. The method of claim 4 , wherein evaluating the terminating condition comprises determining that the vectors of modified parameters are equal to each other.
7. The method of claim 1 , wherein the translation system comprises a natural language compiler (NLC).
8. A system comprising:
a memory; and
a processing device coupled to the memory, the processing device configured to:
generate a set of vectors of parameters of a translation system that uses syntactic and semantic descriptions;
translate, by the translation system using the vectors of parameters, a natural language text to produce a first plurality of translated texts;
determine a first vector of translation quality ratings reflecting translation quality of the first plurality of translated texts;
modify at least one parameter of each vector of parameters to produce a plurality of vectors of modified parameters;
translate, by the translation system using the vectors of modified parameters, the natural language text to produce a second plurality of translated texts;
determine a second vector of translation quality ratings reflecting translation quality of the second plurality of translated texts; and
responsive to determining that an element of the second vector of translation quality ratings exceeds a corresponding element of the first vector of translation quality ratings, modify a vector of parameters based on a corresponding vector of modified parameters.
9. The system of claim 8 , wherein modifying the vector of parameters comprises:
identifying an ordinal number of the element of the second vector of translation quality ratings that exceeds the corresponding element of the first vector of translation quality ratings;
identifying the vector of parameters as having the ordinal number in the set of vectors of parameters;
identifying an element of the vector of parameters; and
replacing the element of the vector of parameters with a corresponding element of the vector of modified parameters.
10. The system of claim 8 , wherein determining a first vector of translation quality ratings comprises determining a plurality of Bilingual Evaluation Understudy (BLEU) scores reflecting translation quality of the first plurality of translated texts.
11. The system of claim 8 , wherein the processing device is further configured to:
repeat the translating operation using the vector of parameters responsive to evaluating a terminating condition.
12. The system of claim 11 , wherein evaluating the terminating condition comprises determining that at least one element of the second vector of translation quality ratings exceeds a pre-defined quality rating value.
13. The system of claim 11 , wherein evaluating the terminating condition comprises determining that the vectors of modified parameters are equal to each other.
14. The system of claim 8 , wherein the translation system comprises a natural language compiler (NLC).
15. A computer-readable non-transitory storage medium comprising executable instructions that, when executed by a computing device, cause the computing device to:
generate a set of vectors of parameters of a translation system that uses syntactic and semantic descriptions;
translate, by the translation system using the vectors of parameters, a natural language text to produce a first plurality of translated texts;
determine a first vector of translation quality ratings reflecting translation quality of the first plurality of translated texts;
modify at least one parameter of each vector of parameters to produce a plurality of vectors of modified parameters;
translate, by the translation system using the vectors of modified parameters, the natural language text to produce a second plurality of translated texts;
determine a second vector of translation quality ratings reflecting translation quality of the second plurality of translated texts; and
responsive to determining that an element of the second vector of translation quality ratings exceeds a corresponding element of the first vector of translation quality ratings, modify a vector of parameters based on a corresponding vector of modified parameters.
16. The computer-readable non-transitory storage medium of claim 15 , wherein modifying the vector of parameters comprises:
identifying an ordinal number of the element of the second vector of translation quality ratings that exceeds the corresponding element of the first vector of translation quality ratings;
identifying the vector of parameters as having the ordinal number in the set of vectors of parameters;
identifying an element of the vector of parameters; and
replacing the element of the vector of parameters with a corresponding element of the vector of modified parameters.
17. The computer-readable non-transitory storage medium of claim 15 , wherein determining a first vector of translation quality ratings comprises determining a plurality of Bilingual Evaluation Understudy (BLEU) scores reflecting translation quality of the first plurality of translated texts.
18. The computer-readable non-transitory storage medium of claim 15 , further comprising executable instructions causing the processing device to:
repeat the translating operation using the vector of parameters responsive to evaluating a terminating condition.
19. The computer-readable non-transitory storage medium of claim 18 , wherein evaluating the terminating condition comprises determining that at least one element of the second vector of translation quality ratings exceeds a pre-defined quality rating value.
20. The computer-readable non-transitory storage medium of claim 18 , wherein evaluating the terminating condition comprises determining that the vectors of modified parameters are equal to each other.