Method and system for building speech recognizer, and speech recognition method and system
View Patent ↗A method and system for building a speech recognizer, and a speech recognition method and system are proposed. The method for building a speech recognizer includes: reading and parsing each grammar file, and building a network of each grammar; reading an acoustic syllable mapping relationship table, and deploying the network of each grammar as a syllable network; performing a merge minimization operation for each syllable network to form a sound element decoding network; forming the speech recognizer by using the sound element decoding network and a language model. The technical solutions of the present disclosure may be applied to exhibit strong extensibility, support an N-Gram language model, support a class model, present flexible use, and adapt for an embedded recognizer in a vehicle-mounted environment.
1. A method for building a decoding network for speech recognition, comprising:
obtaining a plurality of grammar files;
reading and parsing each grammar file of the plurality of grammar files, and building a network of grammar for each grammar file;
reading an acoustic syllable mapping relationship table, and deploying the network of grammar for each grammar file as a syllable network; and
performing a merge minimization operation for all syllable networks to form the decoding network,
wherein the acoustic syllable mapping relationship table is a pre-trained Deep Neural Network (DNN) model, and
wherein deploying the network of grammar for each grammar file as the syllable network comprises:
replacing each word in the network of grammar for each grammar file with the DNN model of the word, and the DNN model containing a transition probability between syllables constituting the word.
2. The method according to claim 1 , wherein the reading and parsing each grammar file comprises:
parsing each grammar file to obtain dictionary information and a connection relationship in each grammar file.
3. The method according to claim 1 , wherein the syllable is a pronunciation of each word.
4. The method according to claim 1 , wherein the performing a merge minimization operation for each syllable network comprises:
optimizing combinations of the respective syllable networks by using a computer graph theory method.
5. A speech recognition method, comprising:
obtaining target audio data;
extracting a target filter bank feature in the target audio data;
taking the target filter bank feature in the target audio data as input data of a speech recognizer, and obtaining target word sequence data corresponding to the target audio data,
wherein the speech recognizer comprises an acoustic model, a language model and a decoding network, wherein the decoding network is built by:
obtaining a plurality of grammar files;
reading and parsing each grammar file of the plurality of grammar files, and building a network of grammar for each grammar file;
reading an acoustic syllable mapping relationship table, and deploying the network of grammar for each grammar file as a syllable network; and
performing a merge minimization operation for all syllable networks to form the decoding network,
wherein the acoustic syllable mapping relationship table is a pre-trained Deep Neural Network (DNN) model, and
wherein deploying the network of grammar for each grammar file as the syllable network comprises:
replacing each word in the network of grammar for each grammar file with the DNN model of the word, and the DNN model containing a transition probability between syllables constituting the word.
6. The method according to claim 5 , wherein the reading and parsing each grammar file comprises:
parsing each grammar file to obtain dictionary information and a connection relationship in each grammar file.
7. The method according to claim 5 , wherein the syllable is a pronunciation of each word.
8. The method according to claim 5 , wherein the performing a merge minimization operation for each syllable network comprises:
optimizing combinations of the respective syllable networks by using a computer graph theory method.
9. A computer device, comprising a memory, a processor and a computer program which is stored on the memory and runs on the processor, wherein the processor, upon executing the program, implements the method for building a decoding network for speech recognition, wherein the method comprises:
obtaining a plurality of grammar files;
reading and parsing each grammar file of the plurality of grammar files, and building a network of grammar for each grammar file;
reading an acoustic syllable mapping relationship table, and deploying the network of grammar for each grammar file as a syllable network; and
performing a merge minimization operation for all syllable networks to form the decoding network,
wherein the acoustic syllable mapping relationship table is a pre-trained Deep Neural Network (DNN) model, and
wherein deploying the network of grammar for each grammar file as the syllable network comprises:
replacing each word in the network of grammar for each grammar file with the DNN model of the word, and the DNN model containing a transition probability between syllables constituting the word.
10. The computer device according to claim 9 , wherein the reading and parsing each grammar file comprises:
parsing each grammar file to obtain dictionary information and a connection relationship in each grammar file.
11. The computer device according to claim 9 , wherein the syllable is a pronunciation of each word.
12. The computer device according to claim 9 , wherein the performing a merge minimization operation for each syllable network comprises:
optimizing combinations of the respective syllable networks by using a computer graph theory method.
13. A non-transitory computer-readable storage medium on which a computer program is stored, wherein the program, when executed by a processor, implements the method for building a decoding network for speech recognition, wherein the method comprises:
obtaining a plurality of grammar files;
reading and parsing each grammar file of the plurality of grammar files, and building a network of grammar for each grammar file;
reading an acoustic syllable mapping relationship table, and deploying the network of grammar for each grammar file as a syllable network; and
performing a merge minimization operation for all syllable networks to form the decoding network,
wherein the acoustic syllable mapping relationship table is a pre-trained Deep Neural Network (DNN) model, and
wherein deploying the network of grammar for each grammar file as the syllable network comprises:
replacing each word in the network of grammar for each grammar file with the DNN model of the word, and the DNN model containing a transition probability between syllables constituting the word.
14. The non-transitory computer-readable storage medium according to claim 13 , wherein the reading and parsing each grammar file comprises:
parsing each grammar file to obtain dictionary information and a connection relationship in each grammar file.
15. The non-transitory computer-readable storage medium according to claim 13 , wherein the syllable is a pronunciation of each word.
16. The non-transitory computer-readable storage medium according to claim 13 , wherein the performing a merge minimization operation for each syllable network comprises:
optimizing combinations of the respective syllable networks by using a computer graph theory method.