Implementation method and system of intelligent voice assistant supporting multi-scene access

Abstract

The invention relates to the technical field of intelligent assistant software, in particular to an implementation method and system for an intelligent voice assistant supporting multi-scene access, and the method comprises the following steps: semantic protocol definition, intention analysis, instruction implementation, scene analysis and implementation, and third-party application access. The method and the system for realizing the intelligent voice assistant supporting the multi-scene access have the beneficial effects that the third-party application is supported to access the voice recognition capability in a scene mode, and full-platform operation of equipment such as a television, a mobile phone and the like is realized through voice recognition; inconvenience caused by key operation or button operation is reduced to the maximum extent, and two hands are released.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent assistant software, and specifically to a method and system for implementing an intelligent voice assistant that supports multi-scenario access. Background Art

[0002] In the past decade, the Internet has developed rapidly, network infrastructure has been improved, and voice recognition technology has begun to attract more and more attention. The first to kick off was Apple's personal voice assistant Siri, which first appeared on the iPhone 4S. It supports natural language input and can call the system's built-in weather forecast, schedule and other applications, giving users a new experience beyond touch screens and buttons. Subsequently, Google integrated Google Now (Google's voice assistant service) on Android 4.1, which supports proactive reminders for users, further promoting the application of voice recognition technology on smart terminals. In China, iFlytek, which focuses on voice recognition, has launched iFlytek voice input method, etc. In addition, iFlytek, Baidu, Sogou, and Yunzhisheng have also launched voice cloud platforms to provide voice technology services for developers and enterprises.

[0003] Voice commands are the implementation of a specific operation, but in actual application environments, many applications expect to achieve similar functions at a very low cost. In addition, the same voice command also requires different command execution implementations in different interfaces. This invention refers to the functions provided by current mainstream smart manufacturers. The requirements in this regard have not yet been implemented or the implementation is too simple and the functions are single. Summary of the invention

[0004] The purpose of the present invention is to provide a method and system for implementing an intelligent voice assistant that supports multi-scenario access, so as to solve the problems raised in the above-mentioned background technology.

[0005] To achieve the above object, the present invention provides the following technical solution: a method for implementing an intelligent voice assistant supporting multi-scenario access, the method comprising the following steps:

[0006] Semantic protocol definition: Design a set of semantic protocols to standardize the format of speech recognition results. The protocol includes four parts: intent name, natural language processing results, intent parameters, and intent score. The intent name is divided into intent category and operation. The natural language processing results contain auxiliary data required to execute instructions. The intent parameters contain key content required to execute instructions, which are divided into necessary parameters and non-essential parameters.

[0007] Intent parsing: Based on the above semantic protocol, parse the JSON format data returned by the speech recognition service, extract the intent name, natural language processing results and intent parameters, and support processing multiple intent results at the same time;

[0008] Instruction implementation: Convert the parsed intent results into executable instructions. Each instruction contains specific actions and required parameters to achieve the mapping from intent to specific operations.

[0009] Scenario analysis and implementation: Define scenario protocols, including scenario events and scenario intent data descriptions, allowing different interfaces to implement the same scenario operations based on specific businesses;

[0010] Third-party application access: Provides registration interfaces and lifecycle management methods, allowing third-party applications to register their own scene elements at startup, and dynamically update scene information and execution result replies when the scene switches, to achieve multi-scene access and intelligent voice interaction.

[0011] Preferably, the intent name part in the semantic protocol is further subdivided into intent categories and operations, wherein the intent categories include but are not limited to time, music, application, news, and the operations are specifically defined according to the intent categories, such as query, play, and close.

[0012] Preferably, the natural language processing result part includes but is not limited to the text after voice recognition, picture URL, response word after command execution, voice broadcast text, voiceprint recognition result and age recognition result, providing necessary auxiliary information for command execution.

[0013] Preferably, in the third-party application access step, the application is registered through the registerApp interface, the life cycle of the scene is managed through the onResumeScene and onPauseScene methods, the scene prompt words and execution result replies are set through the setScenePrompt and addSceneAnswer methods, and the scenario information is dynamically added through the addForScene method.

[0014] Preferably, the method further includes error handling and exception management, which is used to detect and handle errors during speech recognition, intent parsing, command execution or scene switching to ensure the stability of the intelligent voice assistant and user experience.

[0015] A system for implementing an intelligent voice assistant supporting multi-scenario access, the system comprising:

[0016] Semantic parsing module: used to parse the intent results in JSON format returned by the speech recognition service. The intent results include intent name, natural language processing results, intent parameters, and intent score. The intent name is divided into intent category and operation. The natural language processing results include auxiliary data required for command execution. The intent parameters include necessary parameters and non-essential parameters to support flexible and scalable voice command processing.

[0017] Instruction implementation module: connected to the semantic parsing module, used to convert the parsed intent results into executable instructions. The instructions contain specific actions and required parameters, and form a one-to-many relationship with the intent name to support the execution of multiple intents corresponding to a single instruction;

[0018] Scenario management module: defines scenario protocols, including scenario events and scenario intent data descriptions, and is used to manage and maintain operation sets in different scenarios, allowing interfaces or applications to implement different scenario operations based on specific businesses;

[0019] Third-party application access module: provides a series of registration interfaces and lifecycle management methods, including methods for registering applications, entering and exiting scenes, setting scenario prompts, adding reply text for scenario execution results, and adding scenario information. It supports third-party applications to integrate their own scenario elements with intelligent voice assistants to achieve multi-scenario access and intelligent voice interaction.

[0020] Preferably, the semantic parsing module is capable of processing the JSON string containing multiple intent results returned by the speech recognition service, thereby improving the system's flexibility and understanding accuracy of user voice input.

[0021] Preferably, the instruction implementation module dynamically constructs execution instructions through parsed intent parameters, supporting complex business logic and diverse instruction requirements.

[0022] Preferably, the scenario management module represents scenario events and scenario intent data descriptions by enumeration, allowing the same scenario definition to be introduced in different interfaces or applications and customized according to specific business scenarios.

[0023] Preferably, the third-party application access module also includes an error handling and exception management mechanism to ensure that potential errors and exceptions are handled during the access to the third-party application, thereby improving the stability and reliability of the system.

[0024] Compared with the prior art, the present invention has the following beneficial effects:

[0025] The implementation method and system of the intelligent voice assistant supporting multi-scenario access proposed in the present invention support the ability of third-party applications to access voice recognition in a scenario-based manner, and realize full-platform operation of devices such as televisions and mobile phones through voice recognition, thereby minimizing the inconvenience caused by key operations or button operations and freeing both hands. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] Figure 1 is a flow chart of the method of the present invention;

[0027] Figure 2 It is a flow chart for realizing the instructions of the present invention;

[0028] Figure 3 Analyze and implement the flow chart of the scenario of the present invention;

[0029] Figure 4 This is a flowchart of third-party application access in the present invention. DETAILED DESCRIPTION

[0030] In order to make the purpose and technical solution of the present invention clearly and completely described, and the advantages more clearly understood, the embodiments of the present invention are further described in detail with reference to the accompanying drawings. It should be understood that the specific embodiments described herein are part of the embodiments of the present invention, rather than all of the embodiments, and are only used to explain the embodiments of the present invention, and are not used to limit the embodiments of the present invention. All other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0031] For example, see Figures 1 to 4 The present invention provides a technical solution: a method for implementing an intelligent voice assistant supporting multi-scenario access, the method comprising the following steps:

[0032] Step 1: Implement intent analysis of speech recognition results

[0033] The current mainstream speech recognition results are implemented in the form of category division and fixed parameters, which greatly reduces the flexibility and scalability of commands to a certain extent. In order to improve the flexibility of speech recognition result processing, this system analyzes the operation of smart TVs and designs a complete set of semantic protocols to support the implementation of various voice commands.

[0034] The semantic protocol stipulates that the speech recognition result contains three main parts: intent name, natural language processing result, intent parameters and intent score:

[0035] (1) Intent name:

[0036] The intent name is divided into two parts. The first part is the intent category, such as time, music, application, news, etc. The second part is the operation under this category, such as querying the current time, querying festivals, playing music, finding music, closing applications, etc.

[0037] (2) Natural language processing results:

[0038] The result mainly includes data related to the intent, such as the text after voice recognition, image URL, response words after command execution, voice broadcast text that needs to be implemented through TTS, voiceprint recognition results, age recognition results, etc. This part is mainly used as auxiliary data when executing commands.

[0039] (3) Intent parameters:

[0040] The intent parameters contain speech recognition keywords, which are the key content contained in the text after speech recognition and serve as the main parameters for command execution. In addition, the parameters may also contain multiple commands and the parameters required for command execution. Parameters are divided into necessary parameters and non-necessary parameters. Necessary parameters must exist during the intent parsing process, otherwise, the intent parsing fails and the command cannot be executed correctly. Non-necessary parameters may or may not exist depending on the specific circumstances of the command.

[0041] The intent analysis process is based on the semantic protocol above to analyze the intent results. The intent result is the semantic protocol-based data returned to the client by the speech recognition service, which is a string based on the JSON format. According to the semantic protocol, the user's input voice may have multiple intents, so in actual execution, the server can return multiple intent results, so the overall data structure is expressed in JSON Array: [

[0043] {

[0044] "intent":{

[0045] "text":"Speech recognition text",

[0046] "param1":"Parameter 1",

[0047] "param2":"Parameter 2", ...

[0049] },

[0050] "nlg_result":{

[0051] "displayTime":"Display time",

[0052] "imgUrl":"Related image link",

[0053] "replyText":"Response word",

[0054] "ttsText":"Voice broadcast text",

[0055] …

[0056] },

[0057] "intent_name":"<intent category name>",

[0058] "score": score

[0059] },

[0060] {…}, ... ]

[0063] The meanings of the key fields involved in the above data are as follows:

[0064]

[0065]

[0066]

[0067] Step 2: Instruction Implementation

[0068] A command is a set of specific actions and required parameters that can be parsed and executed by the voice assistant. It is a data structure containing the parameters required for execution after the intent result is processed through intent parsing. It has a one-to-many relationship with the intent name, which means that each command can have multiple intents, which means that each command can have multiple data parsing descriptions.

[0069] The instruction is parsed from the intent result. The basic processing flow is as follows: Figure 2 shown.

[0070] Step 3: Scenario Analysis and Implementation

[0071] A scenario can be viewed as a collection of operations supported by an interface. A scenario only defines the category of operations and operation parameter information, but does not contain the implementation of the operations itself. The specific implementation is handled by the interface (or application) that introduces the scenario. That is, the same scenario introduced in different interfaces can show completely different implementations depending on the specific business.

[0072] The scene protocol is a specific implementation that defines scene events. It can be represented in an enumerated manner in the program. It includes two parts: scene events and scene intent data descriptions. Each scene protocol can contain multiple scene intent data descriptions. The scene implementation process is as follows: Figure 3 shown.

[0073] Step 4: Third-party application access

[0074] After the application is started, it connects to the voice recognition assistant and registers the scene elements of the current page, such as instructions, parameters, response words, etc. to inform the voice assistant.

[0075] Several core registration methods:

[0076] registerApp: The application that needs to connect to the voice recognition function calls this interface to register when it starts.

[0077] onResumeScene: method implementation for entering a scene, called during the onResume lifecycle in an Activity, and supports the settings of scene type, scene command words, and scene callbacks.

[0078] onPauseScene: The method implementation for exiting the scene. It is called in the onPause lifecycle in the Activity. It internally implements the release and cleanup of resources when the scene exits to prevent memory leaks and invalid resources occupying memory space.

[0079] setScenePromp: Sets the scenario-based prompt word, which is used to prompt the user for the voice commands that can be entered on the current page. The scenario-based prompt word can be replaced with the scenario command word.

[0080] addSceneAnswer: Adds the reply text of the scene execution result. After the scene execution is completed, it is broadcast to the user through TTS voice.

[0081] addForScene: Add scenario information. This method supports dynamic scenario words and scenario types.

[0082] The implementation process is as follows Figure 4 shown.

[0083] Embodiment 2, based on embodiment 1, proposes a system for implementing an intelligent voice assistant supporting multi-scenario access, the system comprising:

[0084] Semantic parsing module: used to parse the JSON-formatted intent results returned by the speech recognition service. The intent results include intent name, natural language processing results, intent parameters, and intent score. The intent name is divided into intent category and operation, the natural language processing results include auxiliary data required for command execution, and the intent parameters include necessary parameters and non-essential parameters to support flexible and scalable voice command processing. The semantic parsing module processes the JSON string containing multiple intent results returned by the speech recognition service to improve the system's flexibility and understanding accuracy of user voice input.

[0085] Instruction implementation module: connected to the semantic parsing module, it is used to convert the parsed intent results into executable instructions. The instructions contain specific actions and required parameters, and form a one-to-many relationship with the intent name to support the execution of multiple intents corresponding to a single instruction; the instruction implementation module dynamically constructs execution instructions through the parsed intent parameters to support complex business logic and diverse instruction requirements.

[0086] Scenario management module: defines the scenario protocol, including scenario events and scenario intent data descriptions, which is used to manage and maintain the operation sets in different scenarios, allowing the interface or application to implement different scenario operations according to specific business; the scenario management module represents the scenario events and scenario intent data descriptions in an enumerated manner, allowing the same scenario definition to be introduced in different interfaces or applications, and customized implementation according to specific business scenarios.

[0087] Third-party application access module: provides a series of registration interfaces and lifecycle management methods, including methods for registering applications, entering and exiting scenes, setting scenario-based prompts, adding reply text for scenario execution results, and adding scenario-based information. It supports third-party applications to integrate their own scenario elements with intelligent voice assistants to achieve multi-scenario access and intelligent voice interaction. The third-party application access module also includes error handling and exception management mechanisms to ensure that potential errors and exceptions are handled during the access to third-party applications, thereby improving the stability and reliability of the system.

[0088] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A method for implementing an intelligent voice assistant supporting multi-scenario access, characterized in that: The method comprises the following steps: Semantic protocol definition: Design a set of semantic protocols to standardize the format of speech recognition results. The protocol includes four parts: intent name, natural language processing results, intent parameters, and intent score. The intent name is divided into intent category and operation. The natural language processing results contain auxiliary data required to execute instructions. The intent parameters contain key content required to execute instructions, which are divided into necessary parameters and non-essential parameters. Intent parsing: Based on the above semantic protocol, parse the JSON format data returned by the speech recognition service, extract the intent name, natural language processing results and intent parameters, and support processing multiple intent results at the same time; Instruction implementation: Convert the parsed intent results into executable instructions. Each instruction contains specific actions and required parameters to achieve the mapping from intent to specific operations. Scenario analysis and implementation: Define scenario protocols, including scenario events and scenario intent data descriptions, allowing different interfaces to implement the same scenario operations based on specific businesses; Third-party application access: Provides registration interfaces and lifecycle management methods, allowing third-party applications to register their own scene elements at startup, and dynamically update scene information and execution result replies when the scene switches, to achieve multi-scene access and intelligent voice interaction.

2. According to claim 1, a method for implementing an intelligent voice assistant supporting multi-scenario access is characterized in that: The intent name part in the semantic protocol is further subdivided into intent categories and operations. Intent categories include but are not limited to time, music, application, and news. Operations are specifically defined according to intent categories, such as query, play, and close.

3. The method for implementing an intelligent voice assistant supporting multi-scenario access according to claim 1, characterized in that: The natural language processing result part includes but is not limited to the text after voice recognition, image URL, response words after command execution, voice broadcast text, voiceprint recognition results and age recognition results, providing necessary auxiliary information for command execution.

4. The method for implementing an intelligent voice assistant supporting multi-scenario access according to claim 1, characterized in that: In the third-party application access step, register the application through the registerApp interface, manage the life cycle of the scene through the onResumeScene and onPauseScene methods, set the scene prompt words and execution result replies through the setScenePrompt and addSceneAnswer methods, and dynamically add scenario information through the addForScene method.

5. The method for implementing an intelligent voice assistant supporting multi-scenario access according to claim 1, characterized in that: The method further includes error handling and exception management, which is used to detect and handle errors during speech recognition, intent parsing, command execution or scene switching to ensure the stability and user experience of the intelligent voice assistant.

6. A system for implementing an intelligent voice assistant supporting multi-scenario access according to the method for implementing an intelligent voice assistant supporting multi-scenario access according to any one of claims 1 to 5, characterized in that: The system comprises: Semantic parsing module: used to parse the intent results in JSON format returned by the speech recognition service. The intent results include intent name, natural language processing results, intent parameters, and intent score. The intent name is divided into intent category and operation. The natural language processing results include auxiliary data required for command execution. The intent parameters include necessary parameters and non-essential parameters to support flexible and scalable voice command processing. Instruction implementation module: connected to the semantic parsing module, used to convert the parsed intent results into executable instructions. The instructions contain specific actions and required parameters, and form a one-to-many relationship with the intent name to support the execution of multiple intents corresponding to a single instruction; Scenario management module: defines scenario protocols, including scenario events and scenario intent data descriptions, and is used to manage and maintain operation sets in different scenarios, allowing interfaces or applications to implement different scenario operations based on specific businesses; Third-party application access module: provides a series of registration interfaces and lifecycle management methods, including methods for registering applications, entering and exiting scenes, setting scenario prompts, adding reply text for scenario execution results, and adding scenario information. It supports third-party applications to integrate their own scenario elements with intelligent voice assistants to achieve multi-scenario access and intelligent voice interaction.

7. According to claim 6, a system for implementing an intelligent voice assistant supporting multi-scenario access, characterized in that: The semantic parsing module processes the JSON string containing multiple intent results returned by the speech recognition service, improving the system's flexibility and understanding accuracy of user voice input.

8. The system for implementing an intelligent voice assistant supporting multi-scenario access according to claim 6, characterized in that: The instruction implementation module dynamically builds execution instructions through parsed intent parameters, supporting complex business logic and diverse instruction requirements.

9. The implementation system of an intelligent voice assistant supporting multi-scenario access according to claim 6, characterized in that: The scenario management module represents scenario events and scenario intent data descriptions through enumeration, allowing the same scenario definition to be introduced in different interfaces or applications and customized according to specific business scenarios.

10. The implementation system of an intelligent voice assistant supporting multi-scenario access according to claim 6, characterized in that: The third-party application access module also includes error handling and exception management mechanisms to ensure that potential errors and exceptions are handled during the access process of third-party applications, thereby improving the stability and reliability of the system.

Images