Voice recognition system and construction method thereof

• US 10,224,034 B2
• Filed: 03/03/2018
• Issued: 03/05/2019
• Est. Priority Date: 02/03/2016
• Status: Active Grant

• Current Assignee: Hua Xu
• Original Assignee: Hua Xu
• Inventors: Xu, Hua
• Primary Examiner(s): Pullias, Jesse S
• Application Number:
  US15/911,114
• Publication Number:
  US 20180190292A1
• Time in Patent Office: 367 Days
• Field of Search: 704231-257, 704270-275
• CPC Class Codes: G10L 15/1815 Semantic context, e.g. disambiguation of the recognition hypotheses based on word meaning
• CPC Class Codes: G10L 15/22 Procedures used during a speech recognition process, e.g. man-machine dialogue
• CPC Class Codes: G10L 15/28 Constructional details of speech recognition systems
• CPC Class Codes: G10L 2015/223 Execution procedure of a spoken command
• CPC Class Codes: G10L 25/51 for comparison or discrimination

Abstract

Disclosed are a voice recognition system and a construction method for the voice recognition system. By way of layering the system, a general semantic recognition operation for the system is separated from a specific semantic recognition operation for an application program; and by way of classifying the application programs and abstracting out a common performance function, the system can find the application program matching the voice content semantics very efficiently and a third-party program is easily added into the existing voice recognition system. The present invention maps many performance functions to a preset expression with semantic variables, so that the system can recognize more semantic expression manners with optimization of the semantic recognition. Therefore, the system can show more humanized characteristics.

Claims

1. A voice recognition system, comprising:

a registration center storing information of application programs installed on the voice recognition system;

wherein the registration center comprises;

class nodes classified by intent, each class node storing preset expressions with semantic variables which describe common behaviors of a series of application programs, and the preset expressions are mapped to a first public function set;

the voice recognition system further comprising;

the application programs, each of the application programs joining the registration center by declaring a first class node which is one of the class nodes to which each of the application programs belongs and inheriting a second public function set of the first class node;

each of the application programs is a third-party application program downloaded from an application store; and

a voice recognition module converting first voice input by a user into first text content;

the voice recognition system selects a selected application program which is one of the class nodes to process information obtained from the first text content output by the voice recognition module;

when the first text content output by the voice recognition module matches with a first preset expression with semantic variables of a matching class node which is one of the class nodes in the registration center, the voice recognition system selects the selected application program belonging to the matching class node and writes a first intent name and first intent variables into a private directory of the selected application program, the selected application program selects and executes a third public function inherited from the matching class node based on the first intent name and the first variables after the selected application program is started;

the first intent name is bound to the first preset expression with semantic variables and corresponds to a constant in the third public function inherited from the matching class node; and

the first intent variables are extracted from the first text content output by the voice recognition module and correspond to variables in the third public function inherited from the matching class node;

wherein when the first text content output by the voice recognition module matches with the first preset expression with semantic variables of the matching class node in the registration center, manners in which the voice recognition system selects the selected application program include;

if the first preset expression with semantic variables specifies an application name, the voice recognition system selects the selected application program according to the application name.

2. The voice recognition system according to claim 1, wherein when the first text content matches with the first preset expression with semantic variables of the matching class node in the registration center, the manners in which the voice recognition system selects the selected application program further comprises:

if the first preset expression with semantic variables does not specify the application name, the voice recognition system selects a default application program that belongs to the matching class node as the selected application program.

3. The voice recognition system according to claim 2, the registration center further comprises application program nodes;

and each of the application program nodes represents one of the application programs and stores preset expressions with semantic variables that are not used in the class nodes of the registration center, and the preset expressions are mapped to a first specific function set inside the application programs, if the first text content obtained from the voice recognition module matches with a second preset expression with semantic variables of one of the application program nodes, one of the application programs represented by the application program node is started by the voice recognition system and calls a second specific function to which the second preset expression is mapped.

4. The voice recognition system according to claim 3, wherein the registration center further comprises application program internal nodes;

and each of the application program internal nodes stores preset expressions with semantic variables that are used in one of the application programs and corresponding function calls, after entering the selected application program, the voice recognition system processes the user'"'"'s voice input in following manner;

the voice recognition module converts second voice into second text content, the voice recognition system validates whether the second text content matches with a third preset expression with semantic variables of one of the application program internal nodes, if they match, the voice recognition system executes a corresponding function call;

otherwise, the second text content is passed to all the class nodes, if the second text content matches with a third preset expression with semantic variables of a first class node which is one of the class nodes, the voice recognition system selects a second application program which is one of the class nodes that belongs to the first class node to be started and passes a second intent name and second intent variables to the second application program, the second application program calls a fourth public function inherited from the first class node based on the second intent name after the second application is started, wherein, the second intent name is bound to the third preset expression and corresponds to a constant in the fourth public function inherited from the first class node, and the second intent variables are extracted from the second text content output by the voice recognition module and correspond to variables in the fourth public function inherited from the first class node.

5. The voice recognition system according to claim 1, wherein when the first text content matches with the first preset expression with semantic variables of the matching class node in the registration center, the manners in which the voice recognition system selects the selected application program further comprises:

if the first preset expression with semantic variables does not specify the application name, the voice recognition system selects each of the application programs that belongs to the matching class node as the selected application program.

6. The voice recognition system according to claim 1, wherein the class nodes of the registration center are related to geographical location, and each of the application programs has a priority, the voice recognition system records the number of times each of the application programs is used and the geographical location where the user is located when each of the application programs is used, a first application program which is one of the application programs that is used most frequently in a certain area obtains the highest priority in this area;

the manners in which the voice recognition system selects the selected application program further comprises;

if the first text content obtained from the voice recognition module matches with the first preset expression of the matching class node, but the first preset expression with semantic variables does not specify the application name, the voice recognition system selects the first application program as the selected application program.

7. A method for constructing a voice recognition system, the voice recognition system comprises a voice recognition module and a registration center, the registration center is configured to store information of application programs installed in the voice recognition system, the application programs are third-party applications downloaded from an application store, the registration center further comprising:

class nodes classified by intent, each class node stores preset expressions with semantic variables which describe common behaviors of a series of the application programs, and the preset expressions are mapped to a first public function set;

wherein the construction method comprising;

S701, each of the application programs that belongs to a first class node which is one of the class nodes and inherits a second public function set of the first class node submitting information to the registration center;

S702, the voice recognition system receiving first voice input by a user, and the voice recognition module converting voice into first text content;

S703, transmitting the first text content obtained from the voice recognition module to all the class nodes, if the first text content matches with a first preset expressions with semantic variables of a matching class node which is one of the class nodes in the registration center, S704 is performed;

S704, the voice recognition system selecting a selected application program which is one of the application programs and writing a first intent name and first intent variables corresponding to the first preset expression with semantic variables into a private directory of the selected application program installed on the voice recognition system, wherein, the first intent name is bound to the first preset expression and corresponds to a constant in a third public function inherited from the matching class node, and the first intent variables are extracted from the first text content output by the voice recognition module and correspond to variables in the third public function inherited from the matching class node;

S705, the voice recognition system starting the selected application program, the selected application program calls the third public function inherited from the matching class node according to the first intent name and the first intent variables stored in the private directory of the selected application program;

wherein

manners for selecting the selected application program in step S704 include;

If the first preset expression with semantic variables specifies an application name, the voice recognition system selects the selected application program according to the application name.

8. The method for constructing the voice recognition system according to claim 7, wherein the information submitted in step S701 includes a name and an identifier of each of the application programs, class information of each of the application programs.

9. The method for constructing the voice recognition system according to claim 8, wherein the manners for selecting the selected application program in step S704 further comprises:

if the first preset expression with semantic variables does not specify the application name, the voice recognition system selects a default application program that belongs to the matching class node as the selected application program.

10. The method for constructing the voice recognition system according to claim 9, wherein the registration center further comprises:

application program internal nodes;

wherein each of the application program internal nodes stores preset expressions with semantic variables that are used in one of the application programs and corresponding function calls.

11. The method for constructing the voice recognition system according to claim 10, wherein after entering the selected application program in step S705, the voice recognition system processes the user'"'"'s voice input in following manner:

the voice recognition module converts second voice into second text content, the voice recognition system validates whether the second text content matches with a second preset expression with semantic variables of one of the application program internal nodes, if they match, the voice recognition system executes a corresponding function call, otherwise the second text content is passed to all the class nodes, if the second text content matches with a third preset expression with semantic variables of a second class node which is one of the class nodes, the voice recognition system selects a second application program which is one of the application programs that belongs to the second class node to be started and passes an second intent name and second intent variables corresponding to the third preset expression with semantic variables to the second application program, the second application program calls a fourth public function inherited from the second class node based on the second intent name after the second application is started, wherein, the second intent name is bound to the third preset expression and corresponds to a constant in the fourth public function inherited from the second class node, and the second intent variables are extracted from the second text content output by the voice recognition module and correspond to variables in the fourth public function inherited from the second class node.

12. The method for constructing the voice recognition system according to claim 8, wherein the manners for selecting the selected application program in step S704 further comprises:

if the first preset expression with semantic variables does not specify the application name, the voice recognition system selects each of the application programs that belongs to the matching class node as the selected application program.

13. A voice recognition system comprising:

a voice recognition module, converting first voice into first text content; and

class nodes classified by intent, each class node storing preset expressions with semantic variables which describe common behaviors of a series of application programs, the preset expressions are mapped to a first public function set; and

the application programs, each of the application programs joining the voice recognition system by declaring a first class node which is one of the class nodes to which each of the application programs belongs and inheriting a second public function set of the first class node;

wherein the application programs are third-party application programs downloaded from an application store;

manners in which the application programs interact with a user through the voice recognition system includes;

after voice input by the user is converted into first text content by the voice recognition module, when the first text content matches with a first preset expression with semantic variables of a matching class node which is one of the class nodes, the voice recognition system selects a selected application program which is one of the application programs according to an application name and writes an first intent name and first intent variables into a private directory of the selected application program and starts the selected application program, the selected application program loads the first intent name and first intent variables from the private directory and calls a second public function inherited from the matching class node;

wherein, the first intent name is bound to the first preset expression and corresponds to a constant in the second public function inherited from the matching class node, and the first intent variables are extracted from the first text content output by the voice recognition module and correspond to variables in the second public function inherited from the matching class node.

14. The voice recognition system according to claim 13, wherein, the manners in which the application programs interact with the user through the voice recognition system further include:

after the first voice input by the user is converted into the first text content by the voice recognition module, when the first text content matches with a second preset expression with semantic variables specific to a second application program which is one of the application programs, the second application program is selected and passed into a second intent name and second intent variables, after the second application program is started, the second application program calls a third function according to the second intent name and second intent variables.

15. The voice recognition system according to claim 14, wherein, after the selected application program is started, the manners in which the application programs interact with the user through the voice recognition system further include:

after second voice input by the user is converted into second text content by the voice recognition module, the selected application program executes a fourth function call if the second text content matches with a third preset expression with semantic variables used in the selected application program.