Television interaction method and system based on application scene large model intelligent decision

Abstract

The invention discloses a television interaction method and system based on application scene large model intelligent decision, and the method comprises the steps: receiving an interaction instruction inputted by a user through a smart television, and carrying out the intention recognition reasoning of the interaction instruction, and obtaining an intention set; performing intention classification on the intention set to obtain an intention label corresponding to each intention in the intention set; based on the intention label, identifying a current application scene through AI large model service, and performing intelligent decision according to the intention label and the current application scene to obtain an application scene large model intelligent decision instruction set; and issuing an instruction of the application scene large-scale model intelligent decision instruction set to a corresponding television system or application, and controlling the television system or application to respond to the instruction. According to the invention, user real intention reasoning based on the application scene is realized, and more accurate AI large model service is provided for the user by fusing the system capability and the application capability of the smart television through the association relationship between the user interaction instruction and the application scene.

Description

Technical Field

[0001] The present invention relates to the field of television interaction technology, and in particular to a television interaction method, system, terminal and computer-readable storage medium based on intelligent decision-making of a large application scenario model. Background Art

[0002] With the advancement of science and technology, AI big models have gradually begun to enter the process of product application, empowering more and more products. At present, AI big models have begun to be productized for the empowerment scenarios of smart TVs. New products include voice wizard + AI big model, single APP products of AI voice big model and single APP products of AI image big model. These products have different forms, but are usually controlled and interacted by voice. Usually, because the interaction logic of voice commands between the system and APP, and between APPs are independent, if a user issues a voice command to a certain APP, then this voice command will only be processed inside the APP that receives the voice command, which will cause the processing result to be not the user's true intention.

[0003] Therefore, the prior art still needs to be improved and developed. Summary of the invention

[0004] The main purpose of the present invention is to provide a TV interaction method and system based on intelligent decision-making of a large model of application scenarios, aiming to solve the problem in existing TV interactions that interaction instructions can only be processed within the application that receives the interaction instructions, and it is impossible to perform inference operations on the extracted interaction instructions according to the application scenario within the application that receives the interaction instructions, and give different intentions to the system or other applications for processing.

[0005] To achieve the above-mentioned object of the invention, the present invention provides a television interaction method based on intelligent decision-making of a large application scenario model, and the television interaction method based on intelligent decision-making of a large application scenario model comprises:

[0006] Receive interaction instructions input by the user, and perform intent recognition reasoning on the interaction instructions to obtain an intent set;

[0007] Classifying the intent set to obtain an intent label corresponding to each intent in the intent set;

[0008] Based on the intent tag, the current application scenario is identified through the AI ​​big model service, and intelligent decision is made according to the intent tag and the current application scenario to obtain an application scenario big model intelligent decision instruction set;

[0009] The instructions of the application scenario large model intelligent decision instruction set are sent to the corresponding television system or application to control the television system or application to respond to the instructions.

[0010] Optionally, the receiving an interaction instruction input by a user and performing intent recognition reasoning on the interaction instruction to obtain an intent set specifically includes:

[0011] Receiving an interaction instruction input by a user through the AI ​​big model application, wherein the interaction instruction includes a voice interaction instruction and / or a text interaction instruction;

[0012] The AI ​​big model service performs intent recognition and reasoning on the voice interaction instructions and / or text interaction instructions to obtain an intent set containing one or more intents.

[0013] Optionally, the performing intent classification on the intent set to obtain an intent label corresponding to each intent in the intent set specifically includes:

[0014] Obtain the vertical field intention label of the AI ​​big model service;

[0015] According to the vertical field intention labels, one or more intentions in the intention set are respectively classified to obtain intention labels corresponding to the one or more intentions in the intention set;

[0016] The intent tags include system service intent tags and application service intent tags.

[0017] Optionally, based on the intent tag, identifying the current application scenario through the AI ​​big model service, and making an intelligent decision according to the intent tag and the current application scenario to obtain an application scenario big model intelligent decision instruction set specifically includes:

[0018] Based on the intent tags corresponding to one or more intents in the intent set, identify the current application scenario when receiving the interaction instruction input by the user through the AI ​​big model service;

[0019] Based on the intent labels corresponding to one or more intents in the intent set, the AI ​​big model service is used to analyze the association between the intent corresponding to the intent label and the current application scenario, and the application scenario big model intelligent decision instruction set is constructed according to the association.

[0020] Optionally, the identifying, based on the intent tags corresponding to one or more intents in the intent set, the current application scenario when receiving the interaction instruction input by the user through the AI ​​big model service specifically includes:

[0021] When the intent tag includes a system service intent tag, the state of the current system when receiving the interactive instruction input by the user is identified through the AI ​​big model service;

[0022] When the intent tag includes an application service intent tag, the service capability of the current application when receiving the interactive instruction input by the user is identified through the AI ​​big model service.

[0023] Optionally, based on the intent labels corresponding to one or more intents in the intent set, analyzing the association between the intent corresponding to the intent label and the current application scenario through the AI ​​big model service, and constructing the application scenario big model intelligent decision instruction set according to the association, specifically includes:

[0024] When the intent tag includes a system service intent tag, the AI ​​big model service is used to analyze whether the current system state meets the requirements embodied by the intent corresponding to the system service intent tag. If so, the intent corresponding to the system service intent tag is considered to be the user's true intent, and a single instruction corresponding to the precise intent is generated; if not, the intent corresponding to the system service intent tag is considered not to be the user's true intent, and multiple instructions corresponding to the fuzzy intent are generated, and the user's true intent is determined through multiple rounds of inquiries;

[0025] When the intent tag includes an application service intent tag, the AI ​​big model service is used to analyze whether the service capability of the current application meets the requirements embodied by the intent corresponding to the application service intent tag. If so, the intent corresponding to the application service intent tag is considered to be the user's true intention, and a single instruction corresponding to the precise intent is generated; if not, the intent corresponding to the application service intent tag is considered not to be the user's true intention, and multiple instructions corresponding to the fuzzy intent are generated, and the user's true intention is determined through multiple rounds of questioning.

[0026] Optionally, the TV interaction method based on intelligent decision-making of the application scenario big model also includes: tracking user satisfaction with the AI ​​big model service, and optimizing the intention recognition reasoning process according to the satisfaction.

[0027] To achieve the above-mentioned purpose, the present invention further provides a television interaction system based on intelligent decision-making of a large application scenario model, the television interaction system based on intelligent decision-making of a large application scenario model comprising:

[0028] Intention recognition and reasoning module: used to receive the interaction instructions input by the user, and perform intention recognition and reasoning on the interaction instructions to obtain an intention set;

[0029] Intent classification module: used to classify the intent set and obtain the intent label corresponding to each intent in the intent set;

[0030] Intelligent decision-making module: used to identify the current application scenario through the AI ​​big model service based on the intent label, and make intelligent decisions according to the intent label and the current application scenario to obtain the application scenario big model intelligent decision instruction set;

[0031] Instruction issuing module: used to issue the instructions of the application scenario large model intelligent decision-making instruction set to the corresponding television system or application, and control the television system or application to respond to the instructions.

[0032] To achieve the above-mentioned purpose of the invention, the present invention also provides a smart TV, which includes: a memory, a processor, and a TV interaction program based on intelligent decision-making of application scenario big model stored in the memory and run on the processor, and when the TV interaction program based on intelligent decision-making of application scenario big model is executed by the processor, the steps of the TV interaction method based on intelligent decision-making of application scenario big model as described above are implemented.

[0033] To achieve the above-mentioned purpose of the invention, the present invention also provides a computer-readable storage medium, which stores a TV interaction program based on intelligent decision-making of application scenario big models. When the TV interaction program based on intelligent decision-making of application scenario big models is executed by a processor, the steps of the TV interaction method based on intelligent decision-making of application scenario big models as described above are implemented.

[0034] In the present invention, the smart TV receives the interaction instruction input by the user, and performs intent recognition reasoning on the interaction instruction to obtain an intent set; the intent set is classified according to the vertical field intent label of the AI ​​big model service to obtain the intent label corresponding to each intent in the intent set; based on the intent label corresponding to each intent in the intent set, the current application scenario is identified through the AI ​​big model service, and intelligent decision is made according to the intent label and the current application scenario to obtain the application scenario big model intelligent decision instruction set; the AI ​​big model service sends the instructions of the application scenario big model intelligent decision instruction set to the corresponding TV system or application, and controls the TV system or application to respond to the instructions of the application scenario big model intelligent decision instruction set. The present invention realizes the reasoning of the user's real intention based on the application scenario, and through the association between the user interaction instruction and the application scenario, the application that receives the user interaction instruction can be jumped out, and the system capability and application capability of the smart TV are integrated to provide users with more accurate AI big model services, optimize the current AI big model service intention reasoning logic of the smart TV, and provide users with AI big model interaction services that are more in line with the user's real intentions, effectively ensuring the consistency of the experience of the interaction between the TV and the user and the continuity of the control experience. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] Figure 1 It is a flow chart of a preferred embodiment of the television interaction method based on intelligent decision-making of a large application scenario model of the present invention;

[0036] Figure 2It is another flow chart of a preferred embodiment of the television interaction method based on intelligent decision-making of a large application scenario model of the present invention;

[0037] Figure 3 It is a flowchart of a preferred embodiment of the AI ​​big model service of the present invention performing big model intelligent decision-making based on the intent tag according to the current application scenario;

[0038] Figure 4 It is a structural diagram of a preferred embodiment of the television interactive system based on intelligent decision-making of a large application scenario model of the present invention;

[0039] Figure 5 It is a structural diagram of a preferred embodiment of the smart television of the present invention. DETAILED DESCRIPTION

[0040] In order to make the purpose, technical solution and advantages of the present invention clearer and more specific, the present invention is further described in detail below with reference to the accompanying drawings and examples. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0041] With the advancement of science and technology, AI big models have gradually begun to enter the process of product application, empowering more and more products. At present, AI big models have begun to be productized for the empowerment scenarios of smart TVs. New products include voice wizard + AI big model, single APP products of AI voice big model and single APP products of AI image big model. These products have different forms, but are usually controlled and interacted by voice. Usually, because the interaction logic of voice commands between the system and APP, and between APPs are independent, if a user issues a voice command to a certain APP, then this voice command will only be processed inside the APP that receives the voice command, which will cause the processing result to be not the user's true intention.

[0042] The current TV system can install and run multiple large model applications at the same time, such as: voice wizard, AI assistant, AI painting and other voice and large model product applications and services, they can all use voice input as the interaction entrance. In so many voice interaction and AI large model + N application scenarios, how to ensure the consistency of voice interaction experience and the continuity of control experience faces some problems, such as: in the scenario of a single APP product of the AI ​​voice large model, when the user uses voice as input, whether the user input content needs to interact with the single APP product of the AI ​​voice large model, or is it just the interaction between voice and the TV system, who is the instruction given, what kind of instruction is executed, and how to ensure that the intention of the voice input instruction is correctly understood and matches the user's real intention under the same TV system.

[0043] For example, in the big model application of AI assistant, the user issues an interactive command through voice, and the AI ​​assistant outputs a result and voice broadcast through the big model, but the user finds that the voice broadcast of the TV is too low, and the user issues another voice interactive command "the sound is too low, I can't hear it". At this time, what kind of result will "the sound is too low, I can't hear it" execute, is it given to the AI ​​assistant or the TV system, how to judge, what criteria to use to make decisions, and how to build this decision-making process and reasoning process. Under normal circumstances, this command must be given to the AI ​​assistant, but the user's actual intention is to let the TV system turn up the volume of the TV so that the voice broadcast can be heard. The control of this sound needs to be executed by receiving the command through the TV system. In this scenario, who to give the voice command to and how to make the decision have become the problems faced by the current TV system and all voice interactive systems.

[0044] In order to solve the above technical problems, the present invention provides a TV interaction method based on intelligent decision-making of an application scenario big model. The AI ​​big model service of the smart TV receives the interaction instructions input by the user and performs intent recognition and inference on the interaction instructions to obtain an intent set; the intent set is classified according to the vertical field intention labels of the AI ​​big model service to obtain the intent labels corresponding to each intent in the intent set; based on the intent labels corresponding to each intent in the intent set, the current application scenario is identified through the AI ​​big model service, and intelligent decision is made according to the intent labels and the current application scenario to obtain the application scenario big model intelligent decision instruction set; the AI ​​big model service sends the instructions of the application scenario big model intelligent decision instruction set to the corresponding TV system or application, and controls the TV system or application to respond to the instructions of the application scenario big model intelligent decision instruction set. The present invention realizes the reasoning of the user's real intention based on the application scenario. Through the association between the user interaction instruction and the application scenario, it can jump out of the application that receives the user interaction instruction, and give different intentions to the system or other applications for processing. It integrates the system capabilities and application capabilities of the smart TV to provide users with more accurate AI big model services, optimizes the current smart TV's AI big model service intention reasoning logic, and provides users with AI big model interaction services that are more in line with the user's real intention, effectively ensuring the consistency of the experience of interaction between the TV and the user and the continuity of the control experience.

[0045] The application content is further explained below through the description of embodiments in conjunction with the accompanying drawings.

[0046] A preferred embodiment of the television interaction method based on intelligent decision-making of a large application scenario model of the present invention is as follows: Figure 1 and Figure 2 As shown, specifically including:

[0047] S1. Receive interaction instructions input by the user, perform intent recognition and reasoning on the interaction instructions, and obtain an intent set.

[0048] In an implementation of this embodiment, receiving an interaction instruction input by a user, and performing intent recognition reasoning on the interaction instruction to obtain an intent set specifically includes:

[0049] S11. Receive an interaction instruction input by a user through an AI big model application, where the interaction instruction includes a voice interaction instruction and / or a text interaction instruction;

[0050] S12. Perform intent recognition and reasoning on voice interaction instructions and / or text interaction instructions through the AI ​​big model service to obtain an intent set containing one or more intents.

[0051] Specifically, the smart TV system deploys AI big model service (AI big model service is a built-in service of the TV system). Users install applications with AI big model, namely AI big model applications, according to their needs, such as voice wizard application, AI assistant application and AI painting application. Voice wizard application is usually a built-in application of the TV system, supports far-field and near-field voice interaction, supports AI voice / text big model, and provides TV system control, system service call, application service call and other services; AI assistant application supports AI voice / text big model, and provides services such as intelligent question and answer; AI painting application supports AI voice / text big model, and provides services such as picture creation.

[0052] After installing the AI ​​big model application, register the service tag in the AI ​​big model service of the smart TV system. Register the service tag according to the AI ​​big model attributes and service capabilities provided by the AI ​​big model application, such as: the service tags of the voice wizard application include system control, full network search, application control, service control, voice chat, smart home control, etc.; the service tags of the AI ​​assistant application include intelligent question and answer, AI text creation, intelligent writing, English oral training, expert Q&A, etc.; the service tags of the AI ​​painting application include AI picture creation, picture optimization, picture generation, text generation, etc. The service tag indicates what kind of service capabilities the installed AI big model application can provide to users. The service capability is used when the AI ​​big model service performs intent recognition reasoning. The service tag can be used to synchronize the TV system upgrade in the cloud according to the service capabilities provided by the big model application.

[0053] After the smart TV is turned on, the AI ​​large model service starts and waits for the user's interactive command input, such as the user inputting voice commands through the voice wizard after turning on the TV or opening an application; the user turns on the AI ​​assistant after turning on the TV and inputs voice commands through the AI ​​assistant; the user turns on AI painting after turning on the TV and inputs voice commands through AI painting.

[0054] After receiving the interaction command input by the user, the AI ​​big model service obtains the intent set corresponding to the interaction command through the intent recognition reasoning of the AI ​​big model service. A sentence input by the user may contain one or more intents. For example, when the user issues a voice interaction command "The voice is too soft, I can't hear it clearly" in the AI ​​assistant application, the intent recognition reasoning of this voice interaction command may include: Figure 1 "Turn up the volume on the TV" Figure 2 "Through the AI ​​assistant's intelligent question-and-answer capabilities, it can tell you how to adjust the volume of related devices if the sound is too low," etc., and output the recognition results to the smart TV AI big model service vertical field intent classification.

[0055] S2. Classify the intent set and obtain the intent label corresponding to each intent in the intent set.

[0056] In an implementation of this embodiment, the intent set is classified to obtain an intent label corresponding to each intent in the intent set, specifically including:

[0057] S21. Obtain vertical field intent labels for AI big model services;

[0058] S22, classifying one or more intents in the intent set according to the vertical field intent labels, and obtaining intent labels corresponding to one or more intents in the intent set;

[0059] Among them, intent tags include system service intent tags and application service intent tags.

[0060] Specifically, the above intention recognition reasoning result - the intention in the intention set is classified according to the vertical field intention label of the smart TV AI big model service. In the example above, the intention Figure 1 The intent tag corresponding to "Turn up the TV volume" is: system service-TV system control; Figure 2 The intent tag corresponding to "Using the intelligent question-answering capability of the AI ​​assistant, if the sound is too low, how to adjust the volume of the relevant device" is: Application Service AI Assistant-Intelligent Question and Answer. It can be understood that the intent tag actually corresponds to the service capability that the TV system or application can provide. The present invention classifies the intent recognition reasoning results through the AI ​​big model service to provide the big model service capability more accurately.

[0061] S3. Based on the intent label, the current application scenario is identified through the AI ​​big model service, and intelligent decisions are made based on the intent label and the current application scenario to obtain the application scenario big model intelligent decision instruction set.

[0062] In one implementation of this embodiment, based on the intent tag, the current application scenario is identified through the AI ​​big model service, and intelligent decision-making is performed according to the intent tag and the current application scenario to obtain the application scenario big model intelligent decision instruction set, which specifically includes:

[0063] S31, based on the intent tags corresponding to one or more intents in the intent set, identifying the current application scenario when receiving the interaction instruction input by the user through the AI ​​big model service;

[0064] S32. Based on the intent tags corresponding to one or more intents in the intent set, the AI ​​big model service is used to analyze the association between the intent corresponding to the intent tag and the current application scenario, and a big model intelligent decision instruction set for the application scenario is constructed according to the association.

[0065] In one implementation of this embodiment, based on the intent tags corresponding to one or more intents in the intent set, the current application scenario when receiving the interaction instruction input by the user is identified through the AI ​​big model service, specifically including:

[0066] S311, when the intent tag includes a system service intent tag, identifying the current system state when receiving the interactive instruction input by the user through the AI ​​big model service;

[0067] S312. When the intent tag includes an application service intent tag, the service capability of the current application when receiving the interactive instruction input by the user is identified through the AI ​​big model service.

[0068] In one implementation of this embodiment, based on the intent tags corresponding to one or more intents in the intent set, the AI ​​big model service analyzes the association between the intent corresponding to the intent tag and the current application scenario, and constructs an application scenario big model intelligent decision instruction set according to the association, specifically including:

[0069] S321. When the intent tag includes a system service intent tag, the AI ​​big model service is used to analyze whether the current system state meets the requirements embodied by the intent corresponding to the system service intent tag. If so, the intent corresponding to the system service intent tag is considered to be the user's true intent, and a single instruction corresponding to the precise intent is generated; if not, the intent corresponding to the system service intent tag is considered not to be the user's true intent, and multiple instructions corresponding to the fuzzy intent are generated, and the user's true intent is determined through multiple rounds of inquiries;

[0070] S322. When the intent tag includes an application service intent tag, the AI ​​big model service is used to analyze whether the service capabilities of the current application meet the requirements embodied by the intent corresponding to the application service intent tag. If so, the intent corresponding to the application service intent tag is considered to be the user's true intent, and a single instruction corresponding to the precise intent is generated; if not, the intent corresponding to the application service intent tag is considered not to be the user's true intent, and multiple instructions corresponding to the fuzzy intent are generated, and the user's true intent is determined through multiple rounds of inquiries.

[0071] Specifically, based on the intent tag, a large-model intelligent decision is made according to the application scenario of the interactive instruction input by the user. The intelligent decision includes identifying the application scenario of the received instruction through AI scene recognition and analyzing the association between the intent and the application scenario. The association includes but is not limited to whether the received instruction needs to be executed by other applications - which application executes - what function to execute, whether the TV system needs to execute, whether only the application currently receiving the instruction needs to execute, etc. Among them, based on the intent tag, the application scenario of the received instruction is identified through AI scene recognition and the association between the intent and the application scenario is analyzed. If the intent tag includes a system service intent tag, the current system state is identified, and the requirements expressed by the intent corresponding to the intent tag are analyzed to see whether they are consistent with the current system state; if the intent tag includes an application service intent tag, the service capabilities that the current application can provide are identified, and the requirements expressed by the intent corresponding to the intent tag are analyzed to see whether they are consistent with the service capabilities that the current application can provide.

[0072] In the above example, the intention recognition inference results of the user input voice interaction command "The sound is too soft, I can't hear it clearly" include: Figure 1"Turn up the TV volume", the corresponding intent tag is system service-TV system control, that is, it includes the system service tag, then AI scene recognition obtains the current TV system volume (current system status), if it is status 1 "the current TV volume is already the optimal volume (such as 35) or the maximum volume (100)", it is believed that the user's real intention may not be to call the system's volume adjustment to adjust the TV volume to the maximum, then the instruction set returns the voice broadcast and the next round of dialogue reminder, such as the dialogue "the current TV volume is already the optimal volume (not the maximum volume), you want to ask how to adjust the volume of other devices, right, you can say the specific device name, we can answer you according to the device you specified", if the user chooses no, then turn up the volume (such as using a gradient strategy to adjust the volume to the optimal volume, the gradient strategy can be 50+5, 60+3, 70+2, etc.); if the user chooses yes, then proceed to the next round of dialogue to obtain the user's device information or output general device volume adjustment knowledge items. In summary, there may be two intents corresponding to state 1, turning up the TV volume (corresponding intent tag: system service-TV system control) and giving instructions on how to adjust the volume of related devices (corresponding intent tag: application service AI assistant-intelligent question and answer). Multiple instructions corresponding to the fuzzy intent are issued, including instruction 1 "adjust the TV volume to the user-specified value or adjust the TV volume according to the gradient strategy (obtain the application service with TV system control from the large model service tag: voice wizard-system control)"; instruction 2 "AI intelligent question and answer-answer how to adjust the device volume (obtain the application service with AI intelligent question and answer from the large model service tag: AI assistant-intelligent question and answer)". If it is state 2 "the current TV volume is not the optimal volume, nor the maximum volume", it is considered that the user's real intention may be to turn up the TV volume (corresponding intent tag: system service-TV system control), and a single instruction of the precise intent is issued to adjust the TV volume to the optimal volume (obtain the application service with TV system control from the large model service tag: voice wizard-system control). Among them, the voice wizard application is a built-in application of the TV system, which provides system control and other services, so the instruction corresponding to the "system service-TV system control" intent label is sent to the voice wizard application for processing.

[0073] In addition, analyzing the relationship between intent and application scenarios also includes: analyzing the relationship between intent and the service capabilities of the application. When the intent requirement meets the service capabilities of the current application, the instruction will be sent to the application for processing; otherwise, the instruction will be sent to other applications with corresponding service capabilities for processing. For example, a user issues the instruction "How to perform AI creation" through the voice wizard on the TV system desktop, and the current voice wizard's service tag does not support smart Q&A and AI creation, so it is judged that the user needs to open other applications or services, and further confirm which applications can support the user's intent through the service tags of other applications, and then perform AI application scenario inference. In fact, after the AI ​​large model service performs intent recognition and inference on the interactive instruction "How to perform AI creation", it will theoretically produce multiple results, such as: Figure 1 "AI Assistant - Intelligent Question and Answer" Figure 2 "AI painting-AI picture creation"; meaning Figure 3 "AI Assistant - AI Text Creation", etc. If the current intent meets the scenario of multiple applications with fuzzy intent and multiple instructions, multiple application windows will be opened at the same time to respond. For example: the user's instruction is how to create AI. If the user's real intention is to create AI pictures while opening AI Assistant - Smart Question and Answer, then while opening the window of AI Assistant's AI Painting application guidance video, another window will be opened to create AI pictures with AI Painting, that is, to create pictures under the guidance video.

[0074] After making an intelligent decision on the intent based on AI scenario recognition, a relevant instruction set is formed. This instruction set can be an instruction set for precise intent (the user's intent is clearly inferred and the interaction is completed immediately) or an instruction set for fuzzy intent (multiple possible user intentions are inferred and user interaction is completed through multiple rounds of questioning).

[0075] To further summarize, Figure 3As shown, based on the intent labels corresponding to each intent in the intent set, the AI ​​big model service performs big model intelligent decision-making according to the current application scenario, including: if it is a system service intent label, identify and obtain the current system state. If the current system state does not meet the user's intent requirements, it is inferred that this intent may be the user's true intention, and a policy instruction (a single instruction for precise intent) is issued; if the current system state meets the user's intent requirements, it is inferred that this intent may not be the user's true intention, and the user's true intention is confirmed through multiple rounds of inquiries, and multiple instructions for fuzzy intentions are issued; if it is an application service intent label, identify and obtain the service capabilities of the current application. If the current application service capabilities meet the user's intent requirements, it is inferred that this intent may be the user's true intention, and a policy instruction (a single instruction for precise intent) is issued; if the current application service capabilities do not meet the user's intent requirements, it is inferred that this intent may not be the user's true intention, and the user's true intention is confirmed through multiple rounds of inquiries, and multiple instructions for fuzzy intentions are issued.

[0076] S4. Send the instructions of the application scenario large model intelligent decision instruction set to the corresponding TV system or application, and control the TV system or application to respond to the instructions.

[0077] Specifically, the instructions in the application scenario large model intelligent decision-making instruction set include the execution objects and specific execution operations of the instructions. According to the service tags of the applications, it is determined which applications can realize the user's intentions, and the instructions are sent to the corresponding applications. The instructions corresponding to the system service intention tags are generally sent to the voice wizard application, and the voice wizard application implements system control.

[0078] S5. Track user satisfaction with AI big model services and optimize the intent recognition reasoning process based on satisfaction.

[0079] Through the interaction between the application scenario big model intelligent decision instruction set and the user's instruction issuance, the user's satisfaction with the instruction service is tracked, the logic of the current TV system AI big model service intention reasoning is optimized, and the user is provided with AI big model interactive services that are more in line with the user's intentions.

[0080] In summary, the present invention provides a TV interaction method based on application scenario big model intelligent decision-making, and registers the service tag of the AI ​​big model application in the AI ​​big model service of the smart TV system. The service tag indicates the service capability of the installed AI big model application, and the service capability is used when the AI ​​big model service performs intent recognition reasoning. After receiving the user's interaction instruction, the AI ​​big model service performs intent recognition reasoning through the AI ​​big model service, and then classifies the intent according to the vertical field intention tag of the smart TV AI big model service, and then performs big model intelligent decision-making according to the application scenario of the interaction instruction input by the user, and obtains the application scenario big model intelligent decision instruction set. The instructions in the instruction set include the execution object of the instruction and the specific execution operation. The instruction execution object can be the application that receives the user interaction instruction, or it can be the TV system, or it can be other applications other than the application that receives the user interaction instruction.

[0081] In addition, based on the above-mentioned TV interaction method based on application scenario large model intelligent decision, the present invention also provides a TV interaction system based on application scenario large model intelligent decision, wherein the preferred embodiment of the TV interaction system based on application scenario large model intelligent decision is as follows: Figure 4 As shown, specifically including:

[0082] Intention recognition and reasoning module 01: used to receive the interaction instructions input by the user, and perform intention recognition and reasoning on the interaction instructions to obtain an intention set;

[0083] Intent classification module 02: used to classify the intent set and obtain the intent label corresponding to each intent in the intent set;

[0084] Intelligent decision module 03: used to identify the current application scenario through the AI ​​big model service based on the intent label, make intelligent decisions based on the intent label and the current application scenario, and obtain the application scenario big model intelligent decision instruction set;

[0085] Instruction sending module 04: used to send the instructions of the application scenario large model intelligent decision instruction set to the corresponding TV system or application, and control the TV system or application to respond to the instructions.

[0086] Specifically, after receiving the interactive instruction input by the user, the intention recognition and reasoning module 01 of the AI ​​big model service of the present invention obtains the intention set corresponding to the interactive instruction through the intention recognition and reasoning of the AI ​​big model; the intention classification module 02 classifies the intent according to the vertical field intention label of the smart TV AI big model service, so as to provide the big model service capability more accurately; the intelligent decision-making module 03 makes a big model intelligent decision based on the intention label according to the application scenario of the interactive instruction input by the user, identifies the application scenario of the received instruction through AI scene recognition and analyzes the association between the intention and the application scenario. If the intention label includes a system service intention label, the current system state is identified, and the requirements expressed by the intention corresponding to the intention label are analyzed to be consistent with the current system state. If the intention label includes an application service intention label, the service capabilities that the current application can provide are identified, and the requirements expressed by the intention corresponding to the intention label are analyzed to be consistent with the service capabilities that the current application can provide. The association includes but is not limited to whether the received instruction needs to be executed by other applications-which application executes-what function is executed, whether the TV system needs to execute, whether only the application currently receiving the instruction needs to execute, etc. After making intelligent decisions on intent based on AI scene recognition, a relevant instruction set is formed, which can be an instruction set of precise intent (clearly inferring the user's intent and executing immediately to complete the interaction) or an instruction set of fuzzy intent (inferring multiple possible user intentions and completing user interaction through multiple rounds of inquiries). Instruction delivery module 04 delivers relevant instructions to a TV system or application with corresponding service capabilities based on the service tag of the large model service, which can be a TV system or an application that receives user interaction instructions or other applications.

[0087] In addition, based on the above-mentioned TV interaction method and system based on intelligent decision-making of application scenario large model, the present invention also provides a smart TV accordingly, wherein a preferred embodiment of the smart TV is as follows: Figure 5 As shown, it specifically includes a processor 10, a memory 20 and a display 30. Figure 5 Only some components of the smart TV are shown, but it should be understood that it is not required to implement all the components shown, and more or fewer components may be implemented instead.

[0088] In some embodiments, the memory 20 may be an internal storage unit of a smart TV, such as a hard disk or memory of a smart TV. In other embodiments, the memory 20 may also be an external storage device of a smart TV, such as a plug-in hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (Secure Digital, SD) card, and a flash card (Flash Card) equipped on a smart TV. Further, the memory 20 may also include both an internal storage unit and an external storage device of a smart TV. The memory 20 is used to store application software and various types of data installed on the smart TV, such as storing program codes of the smart TV. The memory 20 may also be used to temporarily store data that has been output or is to be output. In one embodiment, a TV interaction program 40 based on intelligent decision-making of a large model of an application scenario is stored on the memory 20, and the TV interaction program 40 based on intelligent decision-making of a large model of an application scenario can be executed by the processor 10, thereby realizing the steps of the TV interaction method based on intelligent decision-making of a large model of an application scenario in the present application.

[0089] In some embodiments, the processor 10 can be a central processing unit (CPU), a microprocessor or other data processing chip, used to run the program code stored in the memory 20 or process data, such as executing a TV interactive program 40 based on intelligent decision-making based on a large model of application scenarios.

[0090] In some embodiments, the display 30 may be an LED display, a liquid crystal display, a touch-sensitive liquid crystal display, an OLED (Organic Light-Emitting Diode) touch device, etc. The display 30 is used to display information on the smart TV and to display a visual user interface. The components 10-30 of the smart TV communicate with each other through a system bus.

[0091] In one embodiment, when the processor 10 executes the TV interaction program 40 based on intelligent decision-making of the large model of application scenarios in the memory 20, the steps of the TV interaction method based on intelligent decision-making of the large model of application scenarios as described above are implemented.

[0092] The present invention also provides a computer-readable storage medium accordingly, wherein the computer-readable storage medium stores a television interaction program based on intelligent decision-making based on a large model of application scenarios, and when the television interaction program based on intelligent decision-making based on a large model of application scenarios is executed by a processor, the steps of the television interaction method based on intelligent decision-making based on a large model of application scenarios as described above are implemented.

[0093] It should be noted that, in this article, the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, article or smart TV including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or smart TV. In the absence of further restrictions, an element defined by the sentence "includes a ..." does not exclude the existence of other identical elements in the process, method, article or smart TV including the element.

[0094] Of course, those skilled in the art can understand that all or part of the processes in the above-mentioned embodiments can be implemented by instructing related hardware (such as a processor, a controller, etc.) through a computer program, and the program can be stored in a computer-readable storage medium that can be read by a computer, and the program can include the processes of the above-mentioned method embodiments when executed. The computer-readable storage medium can be a memory, a disk, an optical disk, etc.

[0095] It should be understood that the application of the present invention is not limited to the above examples. For ordinary technicians in this field, improvements or changes can be made based on the above description. All these improvements and changes should fall within the scope of protection of the claims attached to the present invention.

Claims

1. A television interaction method based on intelligent decision-making of a large application scenario model, characterized in that: The television interaction method based on intelligent decision-making of the application scenario large model includes: Receive interaction instructions input by the user, and perform intent recognition reasoning on the interaction instructions to obtain an intent set; Classifying the intent set to obtain an intent label corresponding to each intent in the intent set; Based on the intent tag, the current application scenario is identified through the AI ​​big model service, and intelligent decision is made according to the intent tag and the current application scenario to obtain an application scenario big model intelligent decision instruction set; The instructions of the application scenario large model intelligent decision instruction set are sent to the corresponding television system or application to control the television system or application to respond to the instructions.

2. The television interaction method based on intelligent decision-making of application scenario large model according to claim 1 is characterized in that: The receiving of the interaction instruction input by the user and performing intention recognition reasoning on the interaction instruction to obtain the intention set specifically includes: Receiving an interaction instruction input by a user through the AI ​​big model application, wherein the interaction instruction includes a voice interaction instruction and / or a text interaction instruction; The AI ​​big model service performs intent recognition and reasoning on the voice interaction instructions and / or text interaction instructions to obtain an intent set containing one or more intents.

3. The television interaction method based on intelligent decision-making of application scenario large model according to claim 2 is characterized in that: The performing intent classification on the intent set to obtain an intent label corresponding to each intent in the intent set specifically includes: Obtain the vertical field intention label of the AI ​​big model service; According to the vertical field intention labels, one or more intentions in the intention set are respectively classified to obtain intention labels corresponding to the one or more intentions in the intention set; The intent tags include system service intent tags and application service intent tags.

4. The television interaction method based on intelligent decision-making of application scenario large model according to claim 2 or 3 is characterized in that: Based on the intent tag, the current application scenario is identified through the AI ​​big model service, and intelligent decision-making is performed according to the intent tag and the current application scenario to obtain the application scenario big model intelligent decision instruction set, which specifically includes: Based on the intent tags corresponding to one or more intents in the intent set, identify the current application scenario when receiving the interaction instruction input by the user through the AI ​​big model service; Based on the intent labels corresponding to one or more intents in the intent set, the AI ​​big model service is used to analyze the association between the intent corresponding to the intent label and the current application scenario, and the application scenario big model intelligent decision instruction set is constructed according to the association.

5. The television interaction method based on intelligent decision-making of application scenario large model according to claim 4 is characterized in that: The identifying, based on the intent tags corresponding to one or more intents in the intent set, the current application scenario when receiving the interaction instruction input by the user through the AI ​​big model service specifically includes: When the intent tag includes a system service intent tag, the state of the current system when receiving the interactive instruction input by the user is identified through the AI ​​big model service; When the intent tag includes an application service intent tag, the service capability of the current application when receiving the interactive instruction input by the user is identified through the AI ​​big model service.

6. The television interaction method based on intelligent decision-making of application scenario large model according to claim 5 is characterized in that: Based on the intent labels corresponding to one or more intents in the intent set, analyzing the association between the intent corresponding to the intent label and the current application scenario through the AI ​​big model service, and constructing the application scenario big model intelligent decision instruction set according to the association, specifically includes: When the intent tag includes a system service intent tag, the AI ​​big model service is used to analyze whether the current system state meets the requirements embodied by the intent corresponding to the system service intent tag. If so, the intent corresponding to the system service intent tag is considered to be the user's true intent, and a single instruction corresponding to the precise intent is generated; if not, the intent corresponding to the system service intent tag is considered not to be the user's true intent, and multiple instructions corresponding to the fuzzy intent are generated, and the user's true intent is determined through multiple rounds of inquiries; When the intent tag includes an application service intent tag, the AI ​​big model service is used to analyze whether the service capability of the current application meets the requirements embodied by the intent corresponding to the application service intent tag. If so, the intent corresponding to the application service intent tag is considered to be the user's true intention, and a single instruction corresponding to the precise intent is generated; if not, the intent corresponding to the application service intent tag is considered not to be the user's true intention, and multiple instructions corresponding to the fuzzy intent are generated, and the user's true intention is determined through multiple rounds of questioning.

7. The television interaction method based on intelligent decision-making of application scenario large model according to claim 1 is characterized in that: The television interaction method based on intelligent decision-making of the application scenario big model also includes: tracking user satisfaction with the AI ​​big model service, and optimizing the intention recognition reasoning process according to the satisfaction.

8. A television interactive system based on intelligent decision-making of a large application scenario model, characterized in that: The television interactive system based on intelligent decision-making of the application scenario large model includes: Intention recognition and reasoning module: used to receive the interaction instructions input by the user, and perform intention recognition and reasoning on the interaction instructions to obtain an intention set; Intent classification module: used to classify the intent set and obtain the intent label corresponding to each intent in the intent set; Intelligent decision-making module: used to identify the current application scenario through the AI ​​big model service based on the intent label, and make intelligent decisions according to the intent label and the current application scenario to obtain the application scenario big model intelligent decision instruction set; Instruction issuing module: used to issue the instructions of the application scenario large model intelligent decision-making instruction set to the corresponding television system or application, and control the television system or application to respond to the instructions.

9. A smart TV, characterized in that: The smart TV includes: a memory, a processor, and a TV interaction program based on intelligent decision-making based on a large model of application scenarios, which is stored in the memory and can be run on the processor. When the TV interaction program based on intelligent decision-making based on a large model of application scenarios is executed by the processor, the steps of the TV interaction method based on intelligent decision-making based on a large model of application scenarios are implemented as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a TV interaction program based on intelligent decision-making of a large model of application scenarios. When the TV interaction program based on intelligent decision-making of a large model of application scenarios is executed by a processor, the steps of the TV interaction method based on intelligent decision-making of a large model of application scenarios as described in any one of claims 1-7 are implemented.

Images