Speech interaction method, and apparatus, computer readable storage medium, and electronic device

A non-transitory computer readable storage medium, storing a computer program which, when executed, implements a speech interaction method, the method including:

• obtaining at least one channel of audio signal;
• recognizing the at least one channel of audio signal by using a preset speech recognition model, to obtain a first type of recognition result;
• determining stored recognition data about the at least one channel of audio signal from a buffer;
• generating a second type of recognition result based on the stored recognition data;
• decoding the first type of recognition result and the second type of recognition result by using a decoding network included in the speech recognition model, to obtain at least one statement recognition result corresponding to the at least one channel of audio signal;
• performing semantic parsing on the at least one statement recognition result, to obtain at least one parsing result;
• and
• generating, based on the at least one parsing result, an instruction for controlling a speech interaction device to perform a corresponding function,
• wherein the recognizing the at least one channel of audio signal by using a preset speech recognition model, to obtain a first type of recognition result includes:
• determining a syllable set respectively corresponding to the at least one channel of audio signal and a first probability score corresponding to a syllable in the syllable set by using an acoustic submodel included in the speech recognition model;
• determining a word set respectively corresponding to the at least one channel of audio signal by using a language submodel included in the speech recognition model;
• determining, based on a word in the word set, whether a second probability score corresponding to the word exists in the buffer;
• if the second probability score corresponding to the word does not exist in the buffer, determining the second probability score corresponding to the word by using the language submodel;
• and
• determining the first probability score and the second probability score that is determined by the language submodel as the first type of recognition result;
• wherein the determining stored recognition data about the at least one channel of audio signal from the buffer, generating a second type of recognition result based on the stored recognition data, comprises:
• if the second probability score corresponding to the word exists in the buffer, determining the second probability score in the buffer as the second probability score of the word;
• and
• determining the first probability score and the second probability score that is determined from the buffer as the second type of recognition result.

An electronic device, wherein the electronic device includes:

• one or more processors;
• and
• a memory, configured to store processor-executable instructions, wherein
• the processor is configured to read the executable instructions from the memory, and execute the instructions to implement a speech interaction method which includes:
• obtaining at least one channel of audio signal;
• recognizing the at least one channel of audio signal by using a preset speech recognition model, to obtain a first type of recognition result;
• determining stored recognition data about the at least one channel of audio signal from a buffer;
• generating a second type of recognition result based on the stored recognition data;
• decoding the first type of recognition result and the second type of recognition result by using a decoding network included in the speech recognition model, to obtain at least one statement recognition result corresponding to the at least one channel of audio signal;
• performing semantic parsing on the at least one statement recognition result, to obtain at least one parsing result;
• and
• generating, based on the at least one parsing result, an instruction for controlling a speech interaction device to perform a corresponding function;
• wherein the recognizing the at least one channel of audio signal by using a preset speech recognition model, to obtain a first type of recognition result includes:
• determining a syllable set respectively corresponding to the at least one channel of audio signal and a first probability score corresponding to a syllable in the syllable set by using an acoustic submodel included in the speech recognition model;
• determining a word set respectively corresponding to the at least one channel of audio signal by using a language submodel included in the speech recognition model;
• determining, based on a word in the word set, whether a second probability score corresponding to the word exists in the buffer;
• if the second probability score corresponding to the word does not exist in the buffer,
• determining the second probability score corresponding to the word by using the language submodel;
• and
• determining the first probability score and the second probability score that is determined by the language submodel as the first type of recognition result;
• wherein the determining stored recognition data about the at least one channel of audio signal from the buffer, generating a second type of recognition result based on the stored recognition data, comprises:
• if the second probability score corresponding to the word exists in the buffer, determining the second probability score in the buffer as the second probability score of the word;
• and
• determining the first probability score and the second probability score that is determined from the buffer as the second type of recognition result.