An AI intelligent training response system and method applicable to general electric equipment

Abstract

The present invention provides an AI intelligent training response system and method applicable to general electric equipment, belonging to the technical field of intelligent response. The AI intelligent training response method applicable to general electric equipment includes: constructing an intelligent algorithm library; collecting and obtaining the instruction text of a user in real time; using the intelligent algorithm library to identify, judge and response-identify the instruction text of the user, and outputting the identification and judgment result corresponding to the response identification; determining a result output mode according to different output forms of the identification and judgment result, and inputting the identification and judgment result into a controlled device according to the result output mode; and the controlled device performing an execution operation according to the result output mode of the identification and judgment result. The system includes modules corresponding to the method steps.

Description

Technical Field

[0001] The present invention provides an AI intelligent training response system and method applicable to general electric equipment, belonging to the technical field of intelligent response. Background Art

[0002] In recent years, with the increase in computing power, the popularization of AI intelligent algorithms has been greatly applied, especially in artificial response (AI customer service) and intelligent control (such as intelligent speakers or TVs, which can control devices through voice), greatly changing people's lifestyles. However, those who have used these intelligent controls should all have a feeling that most speakers and TVs can answer only a few questions or perform only a few operations, or the degree of fit with their own industries is not high, and they cannot add or modify corresponding response contents or operation instructions very well by themselves.

[0003] At the same time, it will be found that most household appliances currently cannot answer many questions or perform operations. This is not because the manufacturers of these household appliances do not want to solve the problems. There are mainly the following reasons:

[0004] First, the pre-trained algorithm modules purchased externally cannot be modified;

[0005] Second, the algorithms are complex and require professional personnel to modify and improve the algorithms;

[0006] Third, the computing power and labor costs required for re-training the algorithms are too high Summary of the Invention

[0007] The present invention provides an AI intelligent training response system and method applicable to general electric equipment to solve the problems of poor industry fit and high training costs in the current AI intelligent response of various large and small household appliances or other operating devices. The technical solutions adopted are as follows:

[0008] An AI intelligent training response method applicable to general electric equipment, the AI intelligent training response method applicable to general electric equipment includes:

[0009] Construct an intelligent algorithm library;

[0010] Collect and obtain the user's instruction text in real time;

[0011] Use the intelligent algorithm library to identify, judge and respond to the user's instruction text, and output the identification and judgment results corresponding to the response identification;

[0012] Determine the result output method according to different output forms of the identification and judgment results, and input the identification and judgment results into the controlled device according to the result output method;

[0013] Perform an execution operation according to the result output method of the controlled device based on the recognition judgment result.

[0014] Furthermore, construct an intelligent algorithm library, including:

[0015] Construct the architecture of the intelligent algorithm library; wherein, there are at least two types of tables in the intelligent algorithm library, and one type of table is a parameter list, and the other type of table is a response content list;

[0016] Establish training parameters corresponding to the product characteristics of the user's industry company and corresponding response methods in the intelligent algorithm library;

[0017] Perform intelligent training on the intelligent algorithm library to obtain an intelligent algorithm library that has completed intelligent training.

[0018] Furthermore, the response method includes a first-level response and a second-level response, and the first-level response is a super-simple dialogue response without parameter information, and the second-level response is a dialogue response with parameters.

[0019] Furthermore, use the intelligent algorithm library to perform recognition judgment and response recognition on the user's instruction text, and output the recognition judgment result corresponding to the response recognition, including:

[0020] Perform recognition judgment on the user's instruction text to obtain the text content in the instruction text;

[0021] Use the intelligent algorithm library to perform response recognition on the text content corresponding to the instruction text according to the first-level response and / or the second-level response, and obtain a response recognition result;

[0022] Output the response recognition result as the recognition judgment result, wherein the output form of the recognition judgment result includes a reply text and an operation instruction.

[0023] Furthermore, use the intelligent algorithm library to perform response recognition on the text content corresponding to the instruction text according to the first-level response and / or the second-level response, and obtain a response recognition result, including:

[0024] When it is determined that the response method for the user's instruction text is the first-level response, obtain the response recognition result according to the response recognition method corresponding to the first-level response;

[0025] When it is determined that the response method for the user's instruction text is the second-level response, obtain the response recognition result according to the response recognition method corresponding to the second-level response.

[0026] Furthermore, the response recognition method corresponding to the first-level response includes:

[0027] Receive the text content corresponding to the user's instruction text;

[0028] Automatically search for one or more reply contents corresponding to the text content from the corresponding list of reply contents according to the text content corresponding to the instruction text;

[0029] When the number of the reply contents is multiple, randomly select one reply content from the multiple reply contents as the target reply content;

[0030] Arrange and merge one or more reply sub - contents corresponding to the target reply content in the order of the serial numbers of the reply sub - contents to form the arranged target reply content;

[0031] Scan each reply sub - content included in the arranged target reply content to determine whether there is a preset specific symbol in the reply sub - content;

[0032] When there is the specific symbol in the reply sub - content, use the specific symbol as the delimiter;

[0033] Break the text in the arranged target reply content according to the delimiter to form the response recognition result.

[0034] Further, the response recognition method corresponding to the secondary response includes:

[0035] Receive the text content corresponding to the instruction text of the user;

[0036] Split the text content corresponding to the instruction text into words to obtain multiple split words corresponding to the text content;

[0037] Compare the multiple split words with each entry in the parameter list in turn, and obtain the entry in the parameter list that contains the most split words as the target entry;

[0038] Search for the reply strategy corresponding to the target entry, and obtain the parameter row corresponding to the target entry, as well as the first reply instruction parameter, the second reply instruction parameter, and the third reply instruction parameter included in the parameter row;

[0039] Split the first reply instruction parameter to obtain a first parameter part, a second parameter part, and a third parameter part corresponding to the reply instruction parameter; wherein, the first parameter part includes the main body frame text of the response reply content and a position parameter; the second parameter part includes a separator, the number of parameters, and the label corresponding to the retrieved parameter, and the second parameter part is used to retrieve parameters from the second reply instruction parameter; the third parameter part includes numbers, and the numbers included in the third parameter part indicate the key parameter positions of the second parameter part, and the key parameter positions are used to retrieve key reply words from the third reply instruction parameter;

[0040] Use the first parameter part, the second parameter part, and the third parameter part to combine with the text content stored in the parameter list to obtain a response recognition result corresponding to the user's instruction text;

[0041] Meanwhile, the step of using the first parameter part, the second parameter part, and the third parameter part to combine with the text content stored in the parameter list to obtain a response recognition result corresponding to the user's instruction text includes:

[0042] Retrieve the main body frame text of the response reply content included in the first parameter part, identify the position parameter, retrieve the word corresponding to the position parameter from the second reply instruction parameter, and supplement the word corresponding to the position parameter to the main body frame text of the response reply content, and replace the position corresponding to the original position parameter to form an initial short phrase;

[0043] Compare the first part of the text in the second reply instruction parameter with the first part of the third reply instruction parameter to determine the attribute corresponding to the array corresponding word, retrieve the target word with the same attribute as the attribute in the third reply instruction parameter according to the attribute corresponding to the array corresponding word, and merge the target word with the initial short phrase to generate a response recognition result corresponding to the user's instruction text.

[0044] Further, determine the result output method according to different output forms of the recognition and judgment result, and input the recognition and judgment result into the controlled device according to the result output method, including:

[0045] Extract the output form of the recognition and judgment result;

[0046] When the output form of the recognition and judgment result is a reply text, directly display the reply text or convert the reply text into an audio through a speech recognition tool, and send the audio to the controlled device; wherein, the speech recognition tool includes various mainstream speech recognition modules or APIs currently;

[0047] When the output form of the recognition and judgment result is an operation instruction, the operation instruction is sent to the controlled device.

[0048] Further, the execution operation is performed according to the result output mode of the controlled device based on the recognition and judgment result, including:

[0049] When the output form of the recognition and judgment result is a reply text, after receiving the audio corresponding to the recognition and judgment result, the controlled device performs voice output on the audio corresponding to the recognition and judgment result;

[0050] When the output form of the recognition and judgment result is an operation instruction, after receiving the operation instruction corresponding to the recognition and judgment result, the controlled device executes the instruction content corresponding to the operation instruction.

[0051] A system for implementing the AI intelligent training response method applicable to general electric equipment, the system includes:

[0052] An algorithm library construction module for constructing an intelligent algorithm library;

[0053] An instruction text acquisition module for real-time collecting and obtaining the instruction text of the user;

[0054] A response recognition module for using the intelligent algorithm library to perform recognition judgment and response recognition on the instruction text of the user, and outputting the recognition judgment result corresponding to the response recognition;

[0055] A result output module for determining the result output mode according to the different output forms of the recognition and judgment result, and inputting the recognition and judgment result into the controlled device according to the result output mode;

[0056] A result execution module for performing an execution operation according to the result output mode of the controlled device based on the recognition and judgment result.

[0057] Advantages of the present invention:

[0058] The present invention proposes an AI intelligent training response method and system applicable to general electric equipment. To solve this problem, the present invention combines the intelligent speech recognition in AI technology (mainly the mutual conversion of text and speech), the database and the corresponding mechanism of the hardware to propose a simple, low-cost and effective AI artificial intelligence training method, so that each operation device can recognize the voice or text command of the user and perform corresponding responses and operations. Description of the drawings

[0059] Figure 1 One of the interface schematic diagrams of the method described in the present invention;

[0060] Figure 2 The second interface schematic diagram of the method of the present invention;

[0061] Figure 3 The third interface schematic diagram of the method of the present invention;

[0062] Figure 4 The fourth interface schematic diagram of the method of the present invention;

[0063] Figure 5 The fifth interface schematic diagram of the method of the present invention;

[0064] Figure 6 The sixth interface schematic diagram of the method of the present invention;

[0065] Figure 7 The seventh interface schematic diagram of the method of the present invention;

[0066] Figure 8 The eighth interface schematic diagram of the method of the present invention;

[0067] Figure 9 The ninth interface schematic diagram of the method of the present invention;

[0068] Figure 10 The tenth interface schematic diagram of the method of the present invention;

[0069] Figure 11 The eleventh interface schematic diagram of the method of the present invention;

[0070] Figure 12 The flowchart of the method of the present invention;

[0071] Figure 13 The system block diagram of the system of the present invention. Detailed implementation manners

[0072] The preferred embodiments of the present invention will be described below with reference to the accompanying drawings. It should be understood that the preferred embodiments described herein are only for the purpose of illustrating and explaining the present invention, and are not used to limit the present invention.

[0073] This embodiment proposes an AI intelligent training response method applicable to general electric equipment, as Figure 12 shown. The AI intelligent training response method applicable to general electric equipment includes:

[0074] S1. Construct an intelligent algorithm library;

[0075] S2. Real-time collect and obtain the instruction text of the user. Among them, the obtaining method of the instruction text of the user is: obtaining the instruction text of the user through the current mainstream speech recognition modules, APIs (converting speech into text), or the active text input of the user;

[0076] S3. Use the intelligent algorithm library to identify, judge, and respond to the instruction text of the user, and output the identification and judgment results corresponding to the response identification;

[0077] S4. Determine the result output method according to the different output forms of the identification and judgment results, and input the identification and judgment results into the controlled device according to the result output method;

[0078] S5. The controlled device performs an execution operation according to the result output method of the identification and judgment results.

[0079] Specifically, the instruction text of the user is obtained through current mainstream speech recognition modules or APIs (converting speech to text) or text input methods. Through the recognition and judgment of this system, and then making a response, the response is divided into two types: a reply text or an operation instruction to be performed on the device. According to the result of the previous step, it is processed as follows: if it is a reply text, it is directly displayed or the reply text is converted into audio output through current mainstream speech recognition modules or APIs; if it is an operation on the device, the corresponding instruction is sent to the module; after the device obtains the corresponding audio or instruction, it performs voice output or executes relevant commands.

[0080] This embodiment proposes an AI intelligent training response method applicable to general electric equipment, which combines the intelligent speech recognition in AI technology (mainly the mutual conversion of text and speech), the database, and the corresponding mechanism of the hardware, and proposes a simple, low-cost, and effective AI artificial intelligence training method, enabling each operation device to recognize and respond to the commands in the form of voice or text of the user and perform corresponding operations.

[0081] An embodiment of the present invention constructs an intelligent algorithm library, including:

[0082] S101. Construct the architecture of the intelligent algorithm library; wherein, at least two types of tables are included in the intelligent algorithm library, and one type of table is a parameter list, and the other type of table is a response content list;

[0083] S102. Establish training parameters corresponding to the product characteristics of the user's industry company and corresponding response methods in the intelligent algorithm library;

[0084] S103. Perform intelligent training on the intelligent algorithm library to obtain an intelligent algorithm library that has completed intelligent training.

[0085] Among them, the response method includes a primary response and a secondary response, and the primary response is a super-simple dialogue response without parameter information, and the secondary response is a dialogue response with parameters.

[0086] Specifically, it is necessary to design and construct the architecture of an intelligent algorithm library. This architecture should include at least two types of tables: a parameter list and a response content list. The parameter list is used to store various training parameters, while the response content list is used to store the corresponding response methods for these parameters.

[0087] In the intelligent algorithm library, corresponding training parameters and corresponding response methods are established according to the industry, company, and product characteristics of the user. The algorithm library will customize its training data and response strategies according to the product characteristics of different industries and companies.

[0088] The established training parameters are used to perform intelligent training on the intelligent algorithm library. The training process may include algorithms such as machine learning and deep learning, enabling the algorithm library to intelligently select the corresponding response method according to the input parameters.

[0089] Through the above technical solution of this embodiment, by establishing training parameters and response methods according to the industry and product characteristics of the user, this intelligent algorithm library has high flexibility and customizability. It can be adjusted and optimized according to different requirements and application scenarios; by introducing the concepts of first-level response and second-level response, the algorithm library can process different types of inputs more accurately. The first-level response is applicable to simple conversations without parameter information, while the second-level response can handle complex conversations with parameters, thereby improving the quality and efficiency of the response.

[0090] At the same time, through the intelligent training process, the algorithm library can self-learn and optimize. It can adjust its parameters and response strategies according to the feedback and data in actual applications, further improving its performance and accuracy. Since the design of the algorithm library adopts a modular and structured approach, it has good scalability and maintainability. In the future, new functions or modules can be added according to needs, and at the same time, the existing functions can be conveniently maintained and updated.

[0091] In summary, the above technical solution of this embodiment aims to provide a flexible, customizable, efficient, and intelligent solution to meet the needs of different industries and products.

[0092] In an embodiment of the present invention, the intelligent algorithm library is used to identify, judge, and answer the user's instruction text, and the recognition and judgment results corresponding to the answer recognition are output, including:

[0093] S301. Identify and judge the user's instruction text to obtain the text content in the instruction text;

[0094] S302. Use the intelligent algorithm library to perform answer recognition on the text content corresponding to the instruction text according to the first-level response and / or the second-level response, and obtain the answer recognition result;

[0095] S303. Output the response recognition result as the recognition judgment result, where the output form of the recognition judgment result includes a reply text and an operation instruction.

[0096] Among them, using the intelligent algorithm library to perform response recognition on the text content corresponding to the instruction text according to the first-level response and / or the second-level response, and obtaining the response recognition result, including:

[0097] S3021. When it is determined that the response method to the instruction text of the user is the first-level response, obtain the response recognition result according to the response recognition method corresponding to the first-level response;

[0098] S3022. When it is determined that the response method to the instruction text of the user is the second-level response, obtain the response recognition result according to the response recognition method corresponding to the second-level response.

[0099] Among them, the response recognition method corresponding to the first-level response includes:

[0100] Step 1. Receive the text content corresponding to the instruction text of the user;

[0101] Step 2. Automatically search for one or more reply contents corresponding to the text content from the corresponding response content list according to the text content corresponding to the instruction text;

[0102] Step 3. When the number of the reply contents is multiple, randomly select one reply content from the multiple reply contents as the target reply content;

[0103] Step 4. Arrange and merge one or more reply sub-contents corresponding to the target reply content in the order of the serial numbers of the reply sub-contents to form the arranged target reply content;

[0104] Step 5. Scan each reply sub-content included in the arranged target reply content to determine whether there is a preset specific symbol in the reply sub-content; where the specific symbol in this embodiment is "-"

[0105] Step 6. When there is the specific symbol in the reply sub-content, use the specific symbol as the delimiter;

[0106] Step 7. Break the text in the arranged target reply content according to the delimiter to form the response recognition result.

[0107] Specifically, as Figure 1 shown, when the command "Have you eaten?" is input, the corresponding reply content will be automatically found from the corresponding response content table. If there are multiple contents, randomly select from the reply contents. The selection situation is as Figure 2As shown, the reply to the command "Have you eaten?" will randomly select a reply from the replies with IDs 2 or 3 in the figure above. The reply method is words1 + words2 + words3. If there is a "-" character in the corresponding words, then the words will be divided into multiple words with "-" as the separator, and then a random one will be selected. For example, if the random result is the reply of ID3, then the reply content will be "Alas, not yet. I'm just about to go and have something to eat" or "I got up late and haven't eaten yet. How about you treating me to something?" etc. For this type of parameter, the reply content is added through the corresponding program or page, specifically as Figure 3 shown.

[0108] Meanwhile, the reply recognition method corresponding to the secondary response includes:

[0109] Step 1: Receive the text content corresponding to the instruction text of the user;

[0110] Step 2: Split the words from the text content corresponding to the instruction text to obtain multiple split words corresponding to the text content;

[0111] Step 3: Compare each entry in the parameter list with the multiple split words in turn, and obtain the entry in the parameter list that contains the most split words as the target entry;

[0112] Step 4: Search for the reply strategy corresponding to the target entry, and obtain the parameter row corresponding to the target entry, as well as the first reply instruction parameter, the second reply instruction parameter, and the third reply instruction parameter included in the parameter row;

[0113] Step 5: Split the first reply instruction parameter to obtain the first parameter part, the second parameter part, and the third parameter part corresponding to the reply instruction parameter; wherein, the first parameter part contains the main frame text and position parameters of the reply content; the second parameter part includes a separator, the number of parameters, and the label corresponding to the retrieved parameter, and the second parameter part is used to retrieve parameters from the second reply instruction parameter; the third parameter part contains numbers, and the numbers included in the third parameter part represent the key parameter positions of the second parameter part. At the same time, the key parameter positions are used to retrieve the key reply words from the third reply instruction parameter;

[0114] Step 6: Use the first parameter part, the second parameter part, and the third parameter part to combine the text content stored in the parameter list to obtain the reply recognition result corresponding to the instruction text of the user;

[0115] Meanwhile, obtaining a response recognition result corresponding to the user's instruction text by using the first parameter part, the second parameter part, and the third parameter part in combination with the text content stored in the parameter list includes:

[0116] Step 601: Retrieve the text of the main framework of the response reply content included in the first parameter part, identify the position parameter, retrieve the word corresponding to the position parameter from the second reply instruction parameter, and supplement the word corresponding to the position parameter to the text of the main framework of the response reply content, and replace the position corresponding to the original position parameter to form an initial short phrase;

[0117] Step 602: Compare the first part of the text in the second reply instruction parameter with the first part of the third reply instruction parameter, determine the attribute corresponding to the array corresponding word, retrieve the target word with the same attribute as the attribute from the third reply instruction parameter according to the attribute corresponding to the array corresponding word, and merge the target word with the initial short phrase to generate a response recognition result corresponding to the user's instruction text.

[0118] Specifically, for commands such as "What's your favorite book?" or "What do you like to do the most?", the corresponding parameters will all be Figure 4 the parameter with ID number 13 in Figure 5 "(What do you like the most|What do you like the most)(.*)(like|hate|dislike|don't love|love|don't hate|want|don't want)(.*)(What is|Which is|That is)", and the corresponding match is for convenience. The debugging content is directly screenshot as shown in

[0119] The code explanation for the above reply content is as follows:

[0120]

[0121]

[0122]

[0123] Therefore, when the obtained command line is "What's your favorite book?", the system will match all parameter lines, and finally will obtain as Figure 6The reply strategy corresponding to the content shown, according to the above code part and description, can be obtained that the content to be answered is "My most $3$4 is |5|2". This content is temporarily called "reply instruction parameter", and the specific interpretation is:

[0124] Use "|" as the separator of "Reply Command Parameters". You will find that there are three parts: 0 (my favorite $3$4), 1 (5), and 2 (2). The value of part 1 is 5, which means that you only need to take 5 parameters of the command line. According to the figure below, you can see that the position 1 to position 5 of several parameters are $1, $2, $3, $4, and $5 respectively. At this time, put these parameter values into 0 of "Reply Command Parameters", and you will find that "my favorite $3$4" becomes "my favorite book is", because $3 = like, $4 = book,

[0125] The third parameter of "Reply Instruction Parameter" is 2. According to the yellow part in the above code, we can know that 2 here represents taking the last character. What is taken? It is words2. words2=$4###的|地. The action is to use the "###" symbol as a separator to get two parts. The first part is "$4", and the second part is "的|地". It means to take the character after "的" or "地" in $4, that is, the string "书". In the string "书", the character after "的" is "书", that is, the character string taken out is "书". This is taken out for use in words3.

[0126] Words3 mentioned above, the complete text is: "person|book|Four Great Classical Novels|event|sports|dreamyou-my dad-my mom-name:)-Zhu Xian|The Classic of Mountains and Seas|any work by Gu Long|Water Margin|Romance of the Three Kingdoms|Journey to the West|Dream of the Red Chamber|sleeping|playing games|staying with you in a daze|taking you on a trip|playing mahjong|swimming|playing table tennis|playing badminton|running. I don't like sports|seeing that you are very excellent". Looking closely, it can be seen that words3 is separated into two parts by "", namely the first part "person|book|Four Great Classical Novels|event|sports|dream" and the second part "you-my dad-my mom-name:)-Zhu Xian|The Classic of Mountains and Seas|any work by Gu Long|Water Margin|Romance of the Three Kingdoms|Journey to the West|Dream of the Red Chamber|sleeping|playing games|staying with you in a daze|taking you on a trip|playing mahjong|swimming|playing table tennis|playing badminton|running. I don't like sports|seeing that you are very excellent". Please take a closer look at these two parts and find that both parts can actually be further separated into six parts using the "|" symbol. They are two arrays of the same length. The array separated from the first part will be (person, book, Four Great Classical Novels, event, sports, dream), and the array separated from the second part will be (you-my dad-my mom-name:)-Zhu Xian|The Classic of Mountains and Seas|any work by Gu Long, Water Margin|Romance of the Three Kingdoms|Journey to the West|Dream of the Red Chamber, sleeping|playing games|staying with you in a daze|taking you on a trip, playing mahjong|swimming|playing table tennis|playing badminton|running, seeing that you are very excellent).

[0127] The string extracted from words2 above is the character "book". Then, when comparing with the first part of words3, it is found that the one that matches is the "book" at position 1 in the array of the first part. The value corresponding to position 1 in the second part array of words3 is "Zhu Xian|The Classic of Mountains and Seas|any work by Gu Long". This value is the main content to be returned. This is also an array, separated by the "-" symbol. As long as the number of separated contents is greater than 1, then randomly select. Suppose "Zhu Xian" is randomly selected, then the final parsing result of the command line is "My favorite book is Zhu Xian".

[0128] If your command is "Who is your favorite person?" or "Is your favorite person Xiao Li?", then it will not correspond to the parameter with id 13 and will be judged as not matching, as Figure 8 shown. Because these two commands do not match all of the ID13 parameters. The above has solved the pure text processing commands. Now let's specifically analyze the part of executing mechanical operations according to the command. First, add a parameter entry such as ID21 in the above figure. The complete content of its str1 is "(.*)(open|start)(.*)(light|TV|air conditioner|refrigerator|washing machine|speaker|stereo|window|curtain|gate|water heater|car|motorcycle|fan)(.*)", and at the same time add the corresponding response content for the access entry as "Okay, now I will $2$4 for you.@@Execute: start - $4@@|5|0". Specifically, asFigure 9 , Figure 10 and Figure 11 As shown in Figure 10 and Figure 11 , after creating this parameter, as long as the user enters a command line such as "Hello, Little T, please help me turn on the air conditioner, thank you" or "It's so hot. Hurry up and turn on the air conditioner".

[0129] It will match the newly created parameter entry and get a reply with the execution content of "Okay, now for you $2$4.@@Execute: Start - $4@@|5|0". According to the explanation in the previous document, similarly, it is still separated by the "|" symbol. From the parameter 5 in the second part, it can be understood that the first 5 parts of the command splitting and interpretation are taken, that is, $1 (position 1) to $5 (position 5). So the content of the first part can be interpreted as "Okay, now for you to turn on the air conditioner.@@Execute: Start - air conditioner@@". Since words2 is empty and words3 is also empty, the interpretation of the parameter 0 in the third part can be ignored. At this time, the final reply content to be executed is "Okay, now for you to turn on the air conditioner.@@Execute: Start - air conditioner@@". Before the actual final output, there will be a function getorderstr to determine whether the command to be executed is as follows:

[0130]

[0131]

[0132] It can be found through the code that there is a device operation instruction "Start - air conditioner" in this reply. When a device operation instruction is found, relevant operations will be performed according to the operation instruction "Start - air conditioner". It is also separated by "-". The first part "Start" will send the corresponding start instruction (such as a switch or power button instruction) to the second part "air conditioner" of this device or module. The specific code is not written here because different small household appliances and device manufacturers use different commands and modules. Just send and call the specific commands for device operations according to the actual situation.

[0133] Finally, after interpreting and pairing the command "It's so hot. Hurry up and turn on the air conditioner", the text response given is "Okay, now for you to turn on the air conditioner.", and while displaying the text, a "Start" instruction is sent to the "air conditioner" module.

[0134] If the response text content needs to be converted into voice, it can be converted into voice and played through the existing voice text conversion modules or APIs.

[0135] An embodiment of the present invention determines the result output method according to different output forms of the recognition and judgment result, and inputs the recognition and judgment result into the controlled device according to the result output method, including:

[0136] S401, extracting the output form of the recognition result;

[0137] S402: When the output form of the recognition result is a reply text, the reply text is directly displayed or converted into audio through a speech recognition tool, and the audio is sent to the controlled device; wherein the speech recognition tool includes currently mainstream speech recognition modules or APIs;

[0138] S403: When the output form of the recognition result is an operation instruction, the operation instruction is sent to the controlled device.

[0139] Specifically, the system needs to extract the output form of the recognition result, which may be a text reply, an operation instruction or other forms of data.

[0140] If the output form of the recognition result is a reply text, the system has two processing methods:

[0141] Display reply text directly: This is usually applicable to user interfaces such as mobile applications, web pages, etc., where the text can be displayed directly to the user.

[0142] Convert reply text to audio: By using speech recognition tools (such as mainstream speech recognition modules or APIs), the system can convert text to audio. This conversion is particularly suitable for scenarios where text cannot be read directly, such as when driving, blind users, etc. The converted audio can be sent to the controlled device (such as smart speakers, car systems, etc.) and played to the user by the device.

[0143] If the output form of the recognition result is an operation instruction, the system directly sends the operation instruction to the controlled device. These operation instructions may include starting an application, adjusting device settings, executing a task, etc. After receiving the instruction, the controlled device will perform the corresponding operation according to the content of the instruction.

[0144] The above-mentioned technical scheme of this embodiment can adapt to different output forms of recognition and judgment results, and both text replies and operation instructions can be properly processed. For the reply text, in addition to direct display, it can also be converted into audio, which is very friendly to users who are inconvenient to read or want to listen to information. By sending the operation instructions directly to the controlled device, the device can be quickly and accurately controlled to perform the corresponding operation, thereby improving the efficiency of the device. The technical scheme of this embodiment integrates mainstream voice recognition modules or APIs, so that the system can easily interact with other services or devices, and also provides the possibility of introducing new recognition tools or functions in the future.

[0145] In general, the technical solution of this embodiment improves the system usability and user experience by flexibly processing the output form of the recognition and judgment results and effectively inputting the results into the controlled device, while also providing an efficient way for device control.

[0146] In one embodiment of the present invention, the controlled device performs an operation according to the output mode of the result of the identification and judgment result, including:

[0147] S501, when the output form of the recognition result is a reply text, the controlled device performs voice output of the audio corresponding to the recognition result after receiving the audio corresponding to the recognition result;

[0148] S502: When the output form of the recognition and judgment result is an operation instruction, after receiving the operation instruction corresponding to the recognition and judgment result, the controlled device executes the instruction content corresponding to the operation instruction.

[0149] Specifically, when the output form of the recognition result is a reply text, the controlled device first receives the audio corresponding to the recognition result. The controlled device plays or outputs the audio. This usually involves an audio playback function, which may be done through a built-in speaker, headphones or other audio output device. In this way, the user can obtain the recognition result by listening, which is very useful for those users who are not convenient to view the screen or prefer to listen to the information.

[0150] When the output of the recognition judgment result is in the form of operation instructions, the controlled device will receive these instructions. The controlled device will parse these instructions and understand the content and purpose of the instructions. This usually involves decoding the instructions to ensure that the device knows what actions or tasks need to be performed. Once the instructions are parsed and understood, the controlled device will perform the corresponding operations according to the content of the instructions. This may include launching an application, adjusting device settings, performing an action or task, etc. After the operation is performed, the controlled device may provide feedback to the user to confirm that the instruction has been executed or provide information on the execution results.

[0151] By replying to text via audio output, users can get information without looking at the screen, which is particularly useful for users in moving vehicles, busy work environments, or with limited vision. Direct execution of operation instructions can simplify the interaction process between users and devices, reduce the need for users to operate manually, and improve the convenience and efficiency of device use. The controlled device can understand and execute operation instructions, making the device more intelligent and automated. This helps reduce user intervention in device operation and improves the autonomy and adaptability of the device.

[0152] In summary, the above technical solution of this embodiment is applicable to various application scenarios, such as smart home, in-vehicle systems, intelligent robots, etc. Through audio output and execution of operation instructions, the controlled device can provide more flexible and convenient services for users in different environments. The technical solution of this embodiment improves the interaction experience between the user and the device and the intelligent level of the device by outputting a reply text in audio and executing operation instructions, and at the same time expands the application scenarios of the device in different scenarios.

[0153] An embodiment of the present invention proposes a system for implementing the AI intelligent training response method applicable to general electrical equipment, such as Figure 12 as shown, the system includes:

[0154] An algorithm library construction module for constructing an intelligent algorithm library;

[0155] An instruction text acquisition module for real-time acquisition and obtaining of the user's instruction text, wherein the acquisition method of the user's instruction text is: obtaining the user's instruction text through current mainstream speech recognition modules, APIs (converting speech to text), or the user's active text input;

[0156] A response recognition module for using the intelligent algorithm library to perform recognition judgment and response recognition on the user's instruction text, and outputting the recognition judgment result corresponding to the response recognition;

[0157] A result output module for determining the result output method according to different output forms of the recognition judgment result, and inputting the recognition judgment result into the controlled device according to the result output method;

[0158] A result execution module for performing an execution operation according to the result output method of the controlled device according to the recognition judgment result.

[0159] Specifically, the user's instruction text is obtained through current mainstream speech recognition modules or APIs (converting speech to text) or text input. Through the recognition and judgment of this system, and then making a response, the response is divided into two types: a reply text or an operation instruction to be performed on the device. According to the result of the previous step, processing is carried out: if it is a reply text, it is directly displayed or the reply text is converted into audio output through current mainstream speech recognition modules or APIs; if it is an operation on the device, the corresponding instruction is sent to the module; after the device obtains the corresponding audio or instruction, it performs voice output or executes relevant commands.

[0160] This embodiment proposes an AI intelligent training and response system applicable to general electric equipment. Combining the mechanisms of intelligent speech recognition (mainly the mutual conversion between text and speech), database, and hardware in AI technology, a simple, low-cost, and effective AI artificial intelligence training method is proposed, enabling each operating device to recognize commands in the form of speech or text from users and perform corresponding responses and operations.

[0161] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and its equivalent technologies, the present invention also intends to include these changes and modifications.

Claims

1. An AI intelligent training response method applicable to general electric equipment, characterized in that, The AI intelligent training response method applicable to general electric equipment includes: Construct an intelligent algorithm library; Collect and obtain the user's instruction text in real time; Use the intelligent algorithm library to identify, judge and respond to the user's instruction text, and output the identification and judgment results corresponding to the response identification; Determine the result output method according to the different output forms of the identification and judgment results, and input the identification and judgment results into the controlled device according to the result output method; The controlled device performs an execution operation according to the result output method of the identification and judgment results; Among them, constructing the intelligent algorithm library includes: Construct the architecture of the intelligent algorithm library; among them, at least two types of tables are included in the intelligent algorithm library, and one table is a parameter list, and the other table is a response content list; Establish training parameters corresponding to the product characteristics of the user's industry company in the intelligent algorithm library and corresponding response methods; Perform intelligent training on the intelligent algorithm library to obtain an intelligent algorithm library that has completed intelligent training; Among them, the response method includes a first-level response and a second-level response, and the first-level response is a super-simple dialogue response without parameter information, and the second-level response is a dialogue response with parameters; Among them, the response identification method corresponding to the second-level response includes: Receive the text content corresponding to the user's instruction text; Split the words from the text content corresponding to the instruction text to obtain multiple split words corresponding to the text content; Compare the multiple split words with each entry in the parameter list in turn to obtain the entry in the parameter list that contains the most split words as the target entry; Search for the reply strategy corresponding to the target entry to obtain the parameter row corresponding to the target entry, and the first reply instruction parameter, the second reply instruction parameter and the third reply instruction parameter included in the parameter row; Split the first reply instruction parameter to obtain the first parameter part, the second parameter part and the third parameter part corresponding to the reply instruction parameter; among them, the first parameter part contains the main frame text and position parameters of the reply content; the second parameter part includes a separator, the number of parameters and the label corresponding to the retrieved parameter, and the second parameter part is used to retrieve parameters from the second reply instruction parameter; the third parameter part contains numbers, and the numbers included in the third parameter part represent the key parameter positions of the second parameter part, and the key parameter positions are used to retrieve the key reply words from the third reply instruction parameter; Use the first parameter part, the second parameter part and the third parameter part to combine the text content stored in the parameter list to obtain the response identification result corresponding to the user's instruction text; At the same time, the using the first parameter part, the second parameter part and the third parameter part to combine the text content stored in the parameter list to obtain the response identification result corresponding to the user's instruction text includes: Retrieve the text of the main framework of the response content included in the first parameter part, identify the position parameter, retrieve the word corresponding to the position parameter from the second response instruction parameter, and supplement the word corresponding to the position parameter to the text of the main framework of the response content, and replace the position corresponding to the original position parameter to form an initial short phrase; Compare the first part of the text in the second response instruction parameter with the first part of the third response instruction parameter, determine the attribute corresponding to the array corresponding word, retrieve the target word with the same attribute as the attribute in the third response instruction parameter according to the attribute corresponding to the array corresponding word, and merge the target word with the initial short phrase to generate the response recognition result corresponding to the user's instruction text.

2. The AI intelligent training response method applicable to General Electric equipment according to claim 1, characterized in that Use the intelligent algorithm library to identify and judge the user's instruction text and perform response recognition, and output the recognition and judgment result corresponding to the response recognition, including: Identify and judge the user's instruction text to obtain the text content in the instruction text; Use the intelligent algorithm library to perform response recognition on the text content corresponding to the instruction text according to the first-level response and / or the second-level response, and obtain the response recognition result; Output the response recognition result as the recognition and judgment result, wherein the output form of the recognition and judgment result includes a reply text and an operation instruction.

3. The AI intelligent training response method applicable to General Electric equipment according to claim 2, characterized in that, Use the intelligent algorithm library to perform response recognition on the text content corresponding to the instruction text according to the first-level response and / or the second-level response, and obtain the response recognition result, including: When it is determined that the response mode to the user's instruction text is the first-level response, obtain the response recognition result according to the response recognition mode corresponding to the first-level response; When it is determined that the response mode to the user's instruction text is the second-level response, obtain the response recognition result according to the response recognition mode corresponding to the second-level response.

4. The AI intelligent training response method applicable to General Electric equipment according to claim 3, wherein The response recognition mode corresponding to the first-level response includes: Receive the text content corresponding to the user's instruction text; Automatically search for one or more reply contents corresponding to the text content from the corresponding response content list according to the text content corresponding to the instruction text; When the number of the reply contents is multiple, randomly select one reply content as the target reply content from the multiple reply contents; Arrange and merge one or more reply sub-contents corresponding to the target reply content in the order of the serial numbers of the reply sub-contents to form the arranged target reply content; Scan each reply sub-content included in the arranged target reply content to determine whether there is a preset specific symbol in the reply sub-content; When there is the specific symbol in the reply sub-content, use the specific symbol as the delimiter; Break the text in the arranged target reply content according to the delimiter to form the response recognition result.

5. The AI intelligent training response method applicable to General Electric equipment according to claim 1, characterized in that Determine the result output mode according to the different output forms of the recognition and judgment result, and input the recognition and judgment result into the controlled device according to the result output mode, including: Extract the output form of the recognition and judgment result; When the output form of the recognition and judgment result is a reply text, the reply text is directly displayed or converted into audio by a speech recognition tool, and the audio is sent to the controlled device; When the output form of the recognition and judgment result is an operation instruction, the operation instruction is sent to the controlled device.

6. The AI intelligent training response method applicable to General Electric equipment according to claim 1, wherein The execution operation is performed according to the result output method of the recognition and judgment result by the controlled device, including: When the output form of the recognition and judgment result is a reply text, after receiving the audio corresponding to the recognition and judgment result, the controlled device performs voice output on the audio corresponding to the recognition and judgment result; When the output form of the recognition and judgment result is an operation instruction, after receiving the operation instruction corresponding to the recognition and judgment result, the controlled device executes the instruction content corresponding to the operation instruction.

7. A system for implementing the AI intelligent training response method applicable to general electric equipment described in claim 1, characterized in that, The system includes: An algorithm library construction module for constructing an intelligent algorithm library; An instruction text acquisition module for real-time acquisition and obtaining the instruction text of the user; A response recognition module for using the intelligent algorithm library to perform recognition judgment and response recognition on the instruction text of the user, and outputting the recognition judgment result corresponding to the response recognition; A result output module for determining the result output method according to the different output forms of the recognition and judgment result, and inputting the recognition and judgment result into the controlled device according to the result output method; A result execution module for performing an execution operation according to the result output method of the recognition and judgment result by the controlled device.

Images