Control method and device of intelligent toy, electronic equipment and storage medium

Abstract

The application provides a control method of an intelligent toy, which comprises: when the intelligent toy is woken up by a user, intention detection is performed; if feeding intention is detected, feeding data is acquired, and the growth value added by the intelligent toy is determined based on the feeding data; if question and answer intention is detected, question and answer data and a target knowledge base are acquired, the target knowledge base has a level corresponding to the current growth value of the intelligent toy, and the target knowledge bases of different levels have knowledge data corresponding to different user ages; the question and answer data are processed based on the target knowledge base to obtain answer data, and the intelligent toy is controlled to output the answer data. The application solves the problem that the existing intelligent toy has a single interaction mode, is mainly based on passive answering, lacks multi-modal interaction, and thus leads to the quick loss of the interest of the user in the intelligent toy and poor accompanying experience, and the interaction experience of the user is enhanced through the pseudo-feeding and the setting of the knowledge bases of different levels.

Description

Technical Field

[0001] This invention relates to the field of artificial intelligence, and more particularly to a control method, device, electronic device, and storage medium for an intelligent toy. Background Technology

[0002] Current smart toys offer children storytelling abilities and simple command-based interaction, imparting knowledge during these interactions. However, because these toys rely primarily on simple command-based interaction and passively answer children's questions, they fail to grow alongside the child's development. In other words, these smart toys act more as databases providing knowledge than considering multimodal interaction. Consequently, children quickly lose interest, resulting in a poor experience of companionship. Therefore, existing smart toys suffer from a lack of multimodal interaction, leading to a rapid loss of user interest and a poor overall experience. Summary of the Invention

[0003] This invention provides a control method for smart toys, aiming to address the problems of existing smart toys having a single interaction mode, mainly based on passive responses, lacking multimodal interaction, and failing to grow together with children, leading to a rapid loss of interest and a poor companionship experience. When the smart toy is activated by the user, intent detection is performed. Upon detecting a feeding intent, feeding data is acquired, and the growth value added to the smart toy is determined based on this data. Upon detecting a question-and-answer intent, question-and-answer data and a target knowledge base are acquired, and the question-and-answer data is processed using the target knowledge base to obtain answer data. This allows the smart toy to output the answer data. This invention increases the interactivity between the user and the companion toy through feeding interaction events, and matches the corresponding knowledge base according to the growth value, increasing the fun of the companionship process. It solves the problems of existing smart toys having a single interaction mode, mainly based on passive responses, lacking multimodal interaction, and causing users to quickly lose interest and resulting in a poor companionship experience. Through simulated feeding and setting different levels of knowledge bases, the user's interactive experience is enhanced.

[0004] In a first aspect, embodiments of the present invention provide a control method for an intelligent toy, the method comprising the following steps:

[0005] When the smart toy is activated by the user, intent detection is performed;

[0006] If a feeding intention is detected, feeding data is acquired, and the growth value increased by the smart toy is determined based on the feeding data.

[0007] If a question-and-answer intent is detected, question-and-answer data and a target knowledge base are obtained. The level of the target knowledge base corresponds to the current growth value of the smart toy. Different levels of the target knowledge base have knowledge data corresponding to different user ages.

[0008] Based on the target knowledge base, the question-and-answer data is processed to obtain answer data, and the smart toy is controlled to output the answer data.

[0009] Optionally, the step of performing intent detection when the smart toy is awakened by the user includes:

[0010] When the smart toy is activated by the user, it acquires the user's voice data;

[0011] The user's intent is obtained by performing intent recognition on the voice data.

[0012] Optionally, acquiring feeding data and determining the increased growth value of the smart toy based on the feeding data includes:

[0013] If the user's intention is to feed the user, then control the smart toy to capture the current image;

[0014] The current image is subjected to image recognition to identify the type of feeding food in the current image; based on the type of feeding food, the growth value corresponding to the feeding food is determined;

[0015] Based on the growth value corresponding to the feeding food, the growth value increased by the smart toy is determined.

[0016] Optionally, obtaining question-and-answer data includes:

[0017] The speech data is sequentially subjected to speech recognition and semantic recognition to obtain the semantic recognition result corresponding to the speech data;

[0018] Based on the semantic recognition results, the user's question and answer data is determined.

[0019] Optionally, acquiring the target knowledge base includes:

[0020] Determine the current growth value of the intelligent toy;

[0021] Based on the correspondence between the growth value stage and the knowledge base level, the target knowledge base corresponding to the current growth value is determined, and different growth value stages correspond to different knowledge base levels.

[0022] Optionally, the step of performing question-and-answer processing on the question-and-answer data based on the target knowledge base to obtain answer data includes:

[0023] Based on the question-and-answer data and the target knowledge base, prompt words for the large language model are constructed. The prompt words are used to instruct the large language model to call the target knowledge base to perform question-and-answer processing on the question-and-answer data.

[0024] The prompt words are input into the large language model for question-and-answer processing to obtain answer data.

[0025] Optionally, controlling the smart toy to output the answer data includes:

[0026] Based on the semantics of the answer data, the output model of the smart toy is determined to be a voice mode and / or a motion mode;

[0027] Based on the voice pattern and / or the action pattern, the smart toy is controlled to output the answer data via voice and / or action.

[0028] Optionally, after performing intent detection when the smart toy is activated by the user, the method further includes:

[0029] If a functional intent is detected, search data and the functional database are retrieved;

[0030] Based on the search data, the target function data is found in the function database, and the smart toy is controlled to output the target function data.

[0031] Secondly, embodiments of the present invention also provide a control device for a smart toy, the control device for the smart toy comprising:

[0032] An intent detection module is used to detect intent when the smart toy is woken up by the user;

[0033] The determination module is used to acquire feeding data if a feeding intention is detected, and to determine the growth value increased by the smart toy based on the feeding data;

[0034] The first acquisition module is used to acquire question and answer data and a target knowledge base if a question and answer intent is detected. The level of the target knowledge base corresponds to the current growth value of the smart toy, and different levels of the target knowledge base have knowledge data corresponding to different user ages.

[0035] The first control module is used to perform question-and-answer processing on the question-and-answer data based on the target knowledge base to obtain answer data, and to control the smart toy to output the answer data.

[0036] Thirdly, embodiments of the present invention provide an electronic device, including: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps in the control method for the intelligent toy provided in the embodiments of the present invention.

[0037] Fourthly, embodiments of the present invention provide a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the steps in the control method for the intelligent toy provided in the embodiments of the present invention.

[0038] In this embodiment of the invention, when the smart toy is awakened by the user, intent detection is performed; if a feeding intent is detected, feeding data is acquired, and the growth value of the smart toy is determined based on the feeding data; if a question-and-answer intent is detected, question-and-answer data and a target knowledge base are acquired, the level of the target knowledge base corresponds to the current growth value of the smart toy, and different levels of the target knowledge base have knowledge data corresponding to different user ages; based on the target knowledge base, question-and-answer processing is performed on the question-and-answer data to obtain answer data, and the smart toy is controlled to output the answer data. This invention increases the interactivity between users and companion toys through feeding interactions. It detects intentions when a user wakes up a smart toy. Upon detecting a feeding intention, it acquires feeding data and determines the increase in growth value based on this data. Upon detecting a question-and-answer intention, it acquires question-and-answer data and a target knowledge base, processes the data using the knowledge base to obtain the answer data, and controls the smart toy to output the answer data. Simultaneously, it matches the growth value to the corresponding knowledge base, increasing the fun of the companionship process. This solves the problem of existing smart toys having a single interaction method, mainly based on passive answers, lacking multimodal interaction, leading to a rapid loss of user interest and a poor companionship experience. By simulating feeding and setting different levels of knowledge bases, the invention enhances the user's interactive experience. Attached Figure Description

[0039] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0040] Figure 1 This is a flowchart of a control method for an intelligent toy provided in an embodiment of the present invention;

[0041] Figure 2 This is a schematic diagram of the structure of a control device for an intelligent toy provided in an embodiment of the present invention;

[0042] Figure 3 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present invention. Detailed Implementation

[0043] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0044] like Figure 1 As shown, Figure 1 This is a flowchart of a control method for an intelligent toy provided in an embodiment of the present invention. The control method for the intelligent toy includes the following steps:

[0045] 101. When the smart toy is woken up by the user, perform intent detection.

[0046] In this embodiment of the invention, the control method for the aforementioned smart toy can be applied to a server. The server and the smart toy communicate with each other. The smart toy includes a voice output device, a motion output device, a voice acquisition device, and an image acquisition device. The voice output device outputs voice data, the motion output device outputs motion data, the voice acquisition device acquires voice data, and the image acquisition device acquires image data. Upon initial use of the smart toy, the user can set user information, including name, age, etc., and can also set a wake-up word, such as "Hello, Little One" or "Hello, Little Bean." During setup, the smart toy can set a main user and multiple other user information, such as information about the user's family and friends, or a special title, such as "Little Master." When the main user interacts with the smart toy, the smart toy can address the main user as "Little Master." The aforementioned user information includes name, age, hobbies, etc.

[0047] The users mentioned above can be understood as users bound to the smart toy.

[0048] The aforementioned wake-up can be understood as the process by which a user activates a smart toy in a specific way, such as through voice wake-up commands, touch wake-up operations, etc.

[0049] The intent detection described above can be understood as a process of analyzing the user's voice data and / or image data to determine the user's intent. The voice data is acquired in real-time by the voice acquisition device of the smart toy, and the image data is acquired in real-time by the image acquisition device of the smart toy.

[0050] Furthermore, after performing intent detection, the user's intent detection results are obtained. These intent detection results can be feeding intent, question-and-answer intent, or functional intent, etc.

[0051] Specifically, when a smart toy is activated by a user, the user's voice data can be collected through the toy's voice acquisition device. This voice data is then processed sequentially for both speech and semantic recognition to determine the user's intent. This speech recognition can be understood as the process of converting speech data into text data. A speech recognition model can be used to process the speech data of both the target and the user. This speech recognition model learns speech-to-text conversion rules through training and can recognize and understand diverse speech signals. This speech recognition model can be based on deep learning and machine learning, such as Long Short-Term Memory (LSTM) networks or Gated Recurrent Units (GRUs). Specifically, a large dataset of historical speech-to-text conversions is collected and input into an untrained speech recognition model for training. During training, the parameters of the speech recognition model are adjusted using a minimum loss function. Once training is complete, a trained speech recognition model is obtained. This training can be supervised training. Supervised training uses a set of data with known labels to train the model. By optimizing the model parameters, the model can predict the labels of new data or make decisions based on the characteristics of existing data. The loss function described above is used to measure the difference between the model's predicted value and the true value, and is used to evaluate and optimize model performance. This loss function can be the mean squared error loss function, cross-entropy loss function, etc. The semantic recognition processing described above can be understood as the process of analyzing word meanings, semantic relationships, and roles in text data after speech conversion to understand the meaning of the text. This semantic recognition can be performed on text data using natural language processing techniques.

[0052] In one possible embodiment, when a user interacts with the smart toy, they can wake it up using a pre-set voice wake-up phrase, such as "Hello, Little One" or "Hello, Little Bean," or by touching a button. When the smart toy is woken up, it can collect the user's voice data through a voice acquisition device and perform intent detection on the user's voice data.

[0053] 102. If a feeding intention is detected, the feeding data is acquired, and the growth value increased by the smart toy is determined based on the feeding data.

[0054] In this embodiment of the invention, upon detecting a feeding intention, the image acquisition device of the smart toy acquires the user's current image data in real time, performs image recognition on the current image data, identifies the type of food being fed in the current image, determines the corresponding growth value based on the type of food, and determines the increase in growth value of the smart toy based on the corresponding growth value of the food. The aforementioned image recognition can be understood as the process of identifying and analyzing image data to identify the type of food being fed in the current image. Image recognition can be performed on the user's current image data using an image recognition model. This image recognition model is obtained by training an untrained image recognition model with a large number of images and corresponding food annotation data. The untrained image recognition model can be an image recognition model built based on deep learning or machine learning, such as ResNet, AlexNet, etc. Specifically, the untrained image recognition model is trained with a large number of images and corresponding food annotation data. During the training process, the parameters of the image recognition model are adjusted using a minimum loss function to obtain a trained image recognition model. The aforementioned food annotation data can be information that classifies and describes the food, such as food type. The training described above can be supervised training. Supervised training is a method of training a model using a set of data with known labels, and optimizing the model parameters so that the model can predict the labels of new data or make decisions based on the characteristics of existing data.

[0055] The types of food mentioned above include fruits and vegetables. The growth value corresponding to these foods can be increased based on their nutritional value.

[0056] The aforementioned feeding intent can be understood as the user's interactive intention to feed the smart toy.

[0057] The feeding data mentioned above can be understood as the data related to the food fed by the smart toy, such as the type of food fed and the corresponding growth value for that type of food. For example, when the type of food fed is fruit, the growth value corresponding to the type of fruit recorded by the smart toy is obtained; when the type of food fed is vegetables, the growth value corresponding to the type of vegetables recorded by the smart toy is obtained.

[0058] It should be noted that the growth value increased by the smart toy can be determined by the type of food fed to it and the corresponding growth value of the food type.

[0059] In one possible embodiment, when it is identified that the type of food being fed in the current image is fruit, the growth value corresponding to the fruit is obtained, and the growth value of the smart toy is increased accordingly based on the nutritional value of the fruit.

[0060] It should be noted that when the growth value of the smart toy reaches the preset growth value for level upgrade, the smart toy's level can be upgraded. The growth value corresponds to the smart toy's level. During each level upgrade, the smart toy can announce via voice, "Little master, I've grown a little more, and now I've learned XX skill." For example, when the smart toy's level is 0 and it upgrades to 1, it announces, "Little master, I've grown a little more, and now I can eat fruit. Please talk to me so I can learn more about you." When the smart toy's level is 1 and it upgrades to 2, it announces, "Little master, I've grown a little more, and now I can eat vegetables. I've already met my family members," and so on. This invention allows the growth value to increase with user interaction with the smart toy, enhancing its fun factor.

[0061] 103. If question-and-answer intent is detected, obtain the question-and-answer data and the target knowledge base.

[0062] In this embodiment of the invention, the level of the target knowledge base corresponds to the current growth value of the smart toy, and different levels of the target knowledge base have knowledge data corresponding to different user ages.

[0063] The above question-and-answer intent can be understood as the user's intention to interact with the smart toy through question-and-answer.

[0064] The above Q&A data can be understood as data containing questions. For example, Q&A data such as "What day of the week is it today?" or "What color is an apple?".

[0065] The aforementioned target knowledge base can be understood as the knowledge base corresponding to the current growth value of the smart toy. It's important to understand that the growth value of a smart toy corresponds to its level.

[0066] Furthermore, upon detecting a question-and-answer intent, the system acquires the user's question-and-answer data and determines the current growth value of the smart toy. Based on the correspondence between the growth value's stages and the knowledge base's levels, it determines the target knowledge base corresponding to the current growth value. Different stages of the growth value correspond to different levels of the knowledge base. The current growth value of the smart toy can be understood as the current growth value of the smart toy in its interaction with the user. The correspondence between the growth value's stages and the knowledge base's levels is a pre-set relationship by the system. It can be understood that the growth value's stages can be understood as the smart toy's levels, and the smart toy's levels correspond to the knowledge base's levels; for example, the smart toy's first level corresponds to the first level of the knowledge base, the smart toy's second level corresponds to the second level of the knowledge base, and so on.

[0067] The aforementioned knowledge base is a pre-set knowledge base of the system, including an ontology self-awareness knowledge base, a habit formation knowledge base, an emotion guidance knowledge base, a user knowledge base, and a hierarchical communication knowledge base. The self-awareness knowledge base can be understood as a knowledge base containing the smart toy's basic information and cognitive data; the habit formation knowledge base can be understood as a knowledge base focused on collecting and analyzing information and methods that help form good habits; the emotion guidance knowledge base can be understood as a knowledge base for managing and regulating the user's emotional state data; the user knowledge base can be understood as a knowledge base containing user-related information, including personal information, hobbies, etc.; through the hierarchical communication knowledge base, the smart toy and the user can interact and communicate at different knowledge levels.

[0068] The aforementioned knowledge bases have different levels. For example, they could be an initial level knowledge base, a first level knowledge base, a second level knowledge base, a third level knowledge base, etc. The initial level knowledge base could correspond to knowledge data for 2-year-old children; the first level knowledge base could correspond to knowledge data for 2-3-year-old children; the second level knowledge base could correspond to knowledge data for 3-4-year-old children; and the third level knowledge base could correspond to knowledge data for 4-5-year-old children, etc.

[0069] In one possible embodiment, when the smart toy's level is first level, the target knowledge base corresponding to the current growth value is determined to be a first-level knowledge base based on the correspondence between the growth value stage and the knowledge base level; when the smart toy's level is second level, the target knowledge base corresponding to the current growth value is determined to be a second-level knowledge base based on the correspondence between the growth value stage and the knowledge base level; and when the smart toy's level is third level, the target knowledge base corresponding to the current growth value is determined to be a third-level knowledge base based on the correspondence between the growth value stage and the knowledge base level.

[0070] In another possible embodiment, when using the smart toy for the first time, the user sets their age to M, where M can be 3, 4, 5, etc. It is understood that the smart toy matches a corresponding level of knowledge base to the user's set age. Different levels of the knowledge base correspond to knowledge data for different ages. The smart toy can interact with the user through the knowledge data in the corresponding level knowledge base, effectively improving the companionship experience of the smart toy.

[0071] 104. Based on the target knowledge base, perform question-and-answer processing on the question-and-answer data to obtain answer data, and control the smart toy to output the answer data.

[0072] In this embodiment of the invention, prompt words for a large language model can be constructed based on a target knowledge base and question-and-answer data. These prompt words are then input into the large language model for question-and-answer processing to obtain answer data. The prompt words instruct the large language model to call the target knowledge base to process the question-and-answer data. The large language model can be a large language model (LLM) built based on deep learning or machine learning, such as GPT, BERT, or LLaMA.

[0073] The above question-and-answer processing can be understood as the process of finding the most relevant answer to the question-and-answer data from the target knowledge base.

[0074] The above answer data can be understood as the most relevant answer data obtained by analyzing question and answer data.

[0075] In one possible implementation, when the target knowledge base is a third-level knowledge base, if the question-and-answer data is "Does a calf eat grass?", and the target knowledge base includes information about animals, a prompt word for the large language model is constructed using the question-and-answer data and the target knowledge base. The prompt word could be "Please use information from the target knowledge base to answer 'Does a calf eat grass?'". Then, the prompt word "Please use information from the target knowledge base to answer 'Does a calf eat grass?'" is input into the large language model. The large language model will then obtain the answer data 'Does a calf eat grass?' from the target knowledge base based on "Please use information from the target knowledge base to answer 'Does a calf eat grass?'".

[0076] In another possible embodiment, when the target knowledge base is a third-level knowledge base, if the question-and-answer data is "What color is an apple?", and the target knowledge base includes information about fruits, a prompt word is constructed using the question-and-answer data and the target knowledge base. The prompt word could be "Please use information from the target knowledge base to answer 'What color is an apple?'". Then, the prompt word "Please use information from the target knowledge base to answer 'What color is an apple?'" is input into the large language model. The large language model will then retrieve the answer data "The apple is red" from the target knowledge base based on "Please use information from the target knowledge base to answer 'What color is an apple?'".

[0077] Furthermore, after obtaining the answer data, semantic recognition can be performed on the answer data to identify its semantics and determine whether the smart toy's output mode is a voice mode and / or a motion mode. The semantics of the answer data can be understood as the meaning and content expressed by the answer data. Semantic recognition can be understood as the process of analyzing the word meaning and semantics of the answer data using natural language processing technology to understand its meaning and content. The voice mode can be understood as a mode of conveying information through speech, such as the smart toy answering questions through voice. The motion mode can be understood as a mode of expressing information through physical actions, such as the smart toy answering questions by moving its eyes, arms, or other parts. The voice output can be understood as the smart toy converting the answer data into audible voice information. The motion output can be understood as the smart toy performing corresponding physical actions based on the answer data, such as moving its eyes, rotating its arms, or moving other parts. Through motion output, users can obtain answers in a direct or interactive way.

[0078] It should be noted that when the smart toy's output mode is voice mode or action mode, the answer data is converted into audible voice information and output through the voice output device. At the same time, the answer data is output through the action output device by the smart toy's physical actions, such as moving the eyes, rotating the arms, or moving other parts. When the smart toy's output mode is voice mode, the answer data is converted into audible voice information and output through the voice output device. When the smart toy's output mode is action mode, the answer data is output through the action output device by the smart toy's physical actions, such as moving the eyes, rotating the arms, or moving other parts.

[0079] In this embodiment of the invention, when the smart toy is awakened by the user, intent detection is performed. If a feeding intent is detected, feeding data is acquired, and the growth value of the smart toy is determined based on the feeding data. If a question-and-answer intent is detected, question-and-answer data and a target knowledge base are acquired. The level of the target knowledge base corresponds to the current growth value of the smart toy, and different levels of the target knowledge base have knowledge data corresponding to different user ages. Based on the target knowledge base, question-and-answer processing is performed on the question-and-answer data to obtain answer data, and the smart toy is controlled to output the answer data. By increasing the interactivity between the user and the companion toy through feeding interaction events, and matching the corresponding knowledge base according to the growth value, the fun of the companionship process is increased. This invention solves the problem that existing smart toys have a single interaction mode, mainly based on passive answering, lacking multimodal interaction, which leads to the user's interest in the smart toy being quickly lost and the companionship experience being poor. By simulating feeding and setting different levels of knowledge base, the user's interactive experience is enhanced.

[0080] It is understood that in the specific implementation of this application, data such as voice data, image data, knowledge data, and user data are involved. When the embodiments in this application are applied to specific products or technologies, user permission or consent is required. Furthermore, the collection, use, and processing of related data, as well as the training, deployment, and invocation of algorithm models, must comply with the relevant laws, regulations, and standards of the relevant countries and regions.

[0081] Optionally, in the step of performing intent detection when the smart toy is woken up by the user, the user's voice data is acquired when the smart toy is woken up by the user; based on the voice data, intent recognition is performed on the voice data to obtain the user's intent.

[0082] In this embodiment of the invention, when the above-mentioned smart toy is used for the first time, the user can set a wake word, such as "Hello, Little Dot" or "Hello, Little Bean". The smart toy can set one main user information and multiple user information, such as the user's family and friends information.

[0083] The user's voice data could include phrases like "Hello, Little Bean, I want to feed you something" or "Hello, Little Bean, let's play a game together." This voice data is collected in real-time by the smart toy's voice acquisition device.

[0084] The users mentioned above can be understood as users bound to the smart toy.

[0085] The aforementioned intent recognition can be understood as the process of judging the purpose or intent of a user's voice data.

[0086] The user intent mentioned above can be understood as the purpose the user wants to express or the action they want to perform. User intent can be feeding intent, question-and-answer intent, etc.

[0087] Furthermore, when the smart toy is activated by the user's wake word, the user's voice data can be used to identify the user's intent.

[0088] In one possible embodiment, when the wake word of the smart toy is "Hello, little one", when the user says to the smart toy, "Hello, little one, I want to feed you something", the smart toy is awakened by the wake word "Hello, little one". The smart toy collects the user's voice data, performs intent recognition on the user's voice data "Hello, little one, I want to feed you something", and identifies the user's intent as a feeding intent.

[0089] In another possible embodiment, when the wake word of the smart toy is "Hello, Little Bean", when the user says to the smart toy, "Hello, Little Bean, what day of the week is it today", the smart toy is woken up by the wake word "Hello, Little Bean", and then the user's voice data "Hello, Little Bean, what day of the week is it today" is used to perform intent recognition, and the user's intent is identified as a question-and-answer intent.

[0090] Optionally, in the steps of acquiring feeding data and determining the growth value added by the smart toy based on the feeding data, if the user's intention is to feed, the smart toy is controlled to acquire the current image; image recognition is performed on the current image to identify the type of feeding food in the current image; the growth value corresponding to the feeding food is determined based on the type of feeding food; and the growth value added by the smart toy is determined based on the growth value corresponding to the feeding food.

[0091] In this embodiment of the invention, when the user's intention is to feed the food, the current image data of the user can be collected by the image acquisition device of the smart toy, and image recognition can be performed on the current image data to identify the type of food being fed in the current image.

[0092] The image recognition described above can be understood as the process of analyzing image data and identifying the type of food being fed in the image. Image recognition models can be used to perform image recognition on a user's current image data. These image recognition models are obtained by training an untrained image recognition model using an image dataset. The image dataset includes a large number of food images and corresponding food annotations. The food annotations can be information that classifies and describes the food, such as food type and food knowledge. The untrained image recognition model can be an image recognition model built based on deep learning or machine learning, such as ResNet or AlexNet. Specifically, the untrained image recognition model is trained using a large number of images and corresponding food annotations. During training, the parameters of the image recognition model are adjusted using a minimum loss function. After training, a trained image recognition model is obtained. The loss function measures the difference between the model's prediction and the actual result; it can be a mean squared error loss function, a cross-entropy loss function, etc.

[0093] The types of food mentioned above include any one of the following food types: fruits, vegetables, staple foods, etc.

[0094] The growth value corresponding to the above-mentioned feeding foods can be the growth value that increases according to the nutritional value of the feeding foods.

[0095] In one possible embodiment, if the type of food fed is fruit, and the nutritional value of the fruit is m, then the growth value corresponding to the food fed is +m; if the type of food fed is vegetables, and the nutritional value of the vegetables is n, then the growth value corresponding to the food fed is +n, and so on. Furthermore, if the growth value corresponding to the food fed is +a, then the growth value added by the smart toy is determined to be +a; if the growth value corresponding to the food fed is +b, then the growth value added by the smart toy is determined to be +b.

[0096] It should be noted that the growth value corresponding to feeding food corresponds to the growth value of smart toys, and the growth value increased by feeding food corresponds to the growth value increased by smart toys.

[0097] Optionally, in the step of obtaining question-and-answer data, speech recognition and semantic recognition can be performed on the speech data in sequence to obtain the semantic recognition result corresponding to the speech data; based on the semantic recognition result, the user's question-and-answer data can be determined.

[0098] In this embodiment of the invention, the aforementioned voice data is obtained in real time by the voice acquisition device of the smart toy from the user's voice data.

[0099] The aforementioned speech recognition can be understood as the process of converting speech data into text data. This speech recognition can be performed on user speech data using a speech recognition model. This model is trained on an untrained speech recognition model using a large number of historical speech-to-text conversion samples. The resulting speech recognition model can recognize and understand diverse speech signals. Specifically, firstly, the collected large dataset of historical speech-to-text conversions is preprocessed to obtain a processed dataset. Then, the processed dataset is input into the trained speech recognition model. During training, the parameters of the speech recognition model are adjusted using a minimum loss function. After training, a trained speech recognition model is obtained. The aforementioned data preprocessing can be understood as the process of cleaning and normalizing the historical speech-to-text conversion dataset. The aforementioned training can be supervised training, which uses a set of data with known labels to train the model. By optimizing the model parameters, the model can predict the labels of new data or make decisions based on the characteristics of existing data. The aforementioned loss function can be the mean squared error loss function, cross-entropy loss function, etc. The aforementioned untrained speech recognition model can be built based on machine learning or deep learning models, such as Long Short-Term Memory Network (LSTM), Gated Recurrent Unit (GRU), etc.

[0100] The aforementioned semantic recognition can be understood as the process of semantic recognition of text data obtained through speech recognition. This semantic recognition can be achieved through natural language processing (NLP) technology. NLP technology aims to enable machines to understand, interpret, and generate human language, achieving effective communication between humans and machines, and allowing computers to perform tasks such as language translation, sentiment analysis, and text summarization.

[0101] The above semantic recognition results were obtained by sequentially performing speech recognition and semantic recognition processing on the speech data.

[0102] Specifically, speech recognition and semantic recognition can be performed on the user's voice data in sequence to identify the user's semantic recognition results, and the user's question and answer data can be determined based on the semantic recognition results.

[0103] The above Q&A data can be understood as data containing questions.

[0104] In one possible embodiment, when the user's voice data is "Hello, Little Bean, what day of the week is it today", speech recognition and semantic recognition are performed on the voice data "Hello, Little Bean, what day of the week is it today" in sequence to obtain the semantic recognition result corresponding to the voice data, and the user's question and answer data is determined to be "what day of the week is it today" based on the semantic recognition result.

[0105] In another possible embodiment, when the user's voice data is "Hello, little one, what color is an apple?", speech recognition and semantic recognition are performed on the voice data "Hello, little one, what color is an apple?" in sequence to obtain the semantic recognition result corresponding to the voice data "Hello, little one, what color is an apple?", and based on the semantic recognition result, the user's question and answer data is determined to be "What color is an apple?".

[0106] Optionally, in the step of acquiring the target knowledge base, the current growth value of the smart toy can be determined; based on the correspondence between the stage of the growth value and the level of the knowledge base, the target knowledge base corresponding to the current growth value can be determined.

[0107] In this embodiment of the invention, different stages of growth value correspond to different levels of the knowledge base.

[0108] The current growth value of the aforementioned smart toy can be understood as the current growth value of the smart toy's interaction with the user. It should be noted that the growth value of the smart toy corresponds to the level of the smart toy.

[0109] Furthermore, after determining the current growth value of the smart toy, the target knowledge base corresponding to the current growth value can be determined based on the correspondence between the growth value stage and the knowledge base level.

[0110] The correspondence between the growth value stages and the knowledge base levels mentioned above is a pre-set correspondence by the system. It can be understood that the first stage of growth value corresponds to the first level of the knowledge base, the second stage of growth value corresponds to the second level of the knowledge base, the third stage of growth value corresponds to the third level of the knowledge base, and so on.

[0111] It should be noted that the knowledge bases at different levels mentioned above contain knowledge data corresponding to the different ages of users.

[0112] The aforementioned knowledge base is a pre-set knowledge base of the system. It has different levels, each with knowledge data corresponding to different age groups. This knowledge base includes an initial level knowledge base, a first-level knowledge base, a second-level knowledge base, and a third-level knowledge base. The initial level knowledge base corresponds to knowledge data for a 2-year-old child. Specifically, it can be a Level 1 self-awareness knowledge base, without using the habit formation knowledge base, emotion guidance knowledge base, or user knowledge base. For questions that a 2-year-old child cannot understand, the response should be: "I haven't learned this yet." The first-level knowledge base corresponds to knowledge data for 2-3-year-old children. Specifically, it can be a Level 1 self-awareness knowledge base, without using the habit formation knowledge base, emotion guidance knowledge base, or user knowledge base. It can explain user-posed questions from the cognitive perspective of a 2-3-year-old child. For questions that a 2-3-year-old child cannot understand, the response should be: "I haven't learned this yet." The second-level knowledge base mentioned above corresponds to knowledge data for children aged 3-4. Specifically, the second-level knowledge base can be Level 2 of the ontology self-cognition knowledge base, using the user knowledge base but not the habit formation knowledge base or the emotion guidance knowledge base. It can explain the questions raised by users from the cognitive perspective of 3-4 year olds. For questions that 3-4 year olds cannot understand, the response should be: "I haven't learned this question yet." The third-level knowledge base mentioned above corresponds to knowledge data for children aged 4-5. Specifically, the third-level knowledge base can be Level 2 of the ontology self-cognition knowledge base, using the user knowledge base but not the habit formation knowledge base or the emotion guidance knowledge base. It can explain the questions raised by users from the cognitive perspective of 4-5 year olds, and can master simple English words (such as common fruits, daily greetings, common animals). For questions that 4-5 year olds cannot understand, the response should be: "I haven't learned this question yet."

[0113] The aforementioned knowledge base also includes Level 4, Level 5, Level 6, and Level 7 knowledge bases. The Level 4 knowledge base corresponds to knowledge data for 5-year-old children. Specifically, it can be Level 2 of the ontology self-awareness knowledge base, using the user knowledge base and habit-forming knowledge base, but not the emotion-guided knowledge base. It should be able to explain user-posed questions from a 5-year-old's cognitive perspective, and should have a grasp of simple English vocabulary (such as common fruits, daily greetings, and common animals). For questions that a 5-year-old cannot understand, the response should be: "I haven't learned this yet." The Level 5 knowledge base corresponds to knowledge data for 5-year-old children. Specifically, it can be Level 3 of the ontology self-awareness knowledge base, using the user knowledge base and habit-forming knowledge base, but not the emotion-guided knowledge base. It should be able to explain user-posed questions from a 5-year-old's cognitive perspective, and should have a grasp of simple English vocabulary (such as common fruits, daily greetings, and common animals), and encyclopedic knowledge of plants, animals, and natural phenomena. For questions that children older than 5 cannot understand, the response should be: "I haven't learned this yet." The aforementioned Level 6 knowledge base corresponds to the knowledge data of a 6-year-old child. Specifically, the Level 6 knowledge base can be Level 3 of the ontology self-cognition knowledge base. It uses the user knowledge base, habit formation knowledge base, and emotion guidance knowledge base to explain the questions raised by users from the cognitive perspective of a 5-year-old child. It can master simple English words (such as common fruits, daily greetings, and common animals), master encyclopedic knowledge of plants, animals, and natural phenomena, and master addition, subtraction, multiplication, and division methods within 100. For questions beyond this range, it should reply: "I haven't learned this question yet." The aforementioned Level 7 knowledge base corresponds to the knowledge data of a 7-year-old child. Specifically, Level 7 knowledge base can be Level 3 of the ontology self-cognition knowledge base. It uses the user knowledge base, habit formation knowledge base, and emotion guidance knowledge base to explain the questions raised by users from the cognitive perspective of a 7-year-old child. It can master simple English words (such as common fruits, daily greetings, and common animals), master encyclopedic knowledge of plants, animals, and natural phenomena, master addition, subtraction, multiplication, and division methods within 100, and master historical knowledge, historical figures, myths, and idioms. For questions beyond this scope, the response should be: "I haven't learned this question yet."

[0114] In one possible embodiment, when the smart toy is at level three, based on the correspondence between the growth value stage and the knowledge base level, the target knowledge base corresponding to the current growth value is determined to be the level three knowledge base. The level three knowledge base corresponds to knowledge data for 3-4 year old children and can explain the questions raised by users from the cognitive perspective of 3-4 year old children. When the smart toy is at level four, based on the correspondence between the growth value stage and the knowledge base level, the target knowledge base corresponding to the current growth value is determined to be the level four knowledge base. The level four knowledge base corresponds to knowledge data for 5-year-old children and can explain the questions raised by users from the cognitive perspective of 5-year-old children. When the smart toy is at level five, based on the correspondence between the growth value stage and the knowledge base level, the target knowledge base corresponding to the current growth value is determined to be the level five knowledge base. The level five knowledge base corresponds to knowledge data for 6-year-old children and can explain the questions raised by users from the cognitive perspective of 6-year-old children.

[0115] Optionally, in the step of processing the question-and-answer data based on the target knowledge base to obtain the answer data, prompt words for the large language model can be constructed based on the question-and-answer data and the target knowledge base; the prompt words are then input into the large language model for question-and-answer processing to obtain the answer data.

[0116] In this embodiment of the invention, the aforementioned prompt words are used to instruct the large language model to call the target knowledge base to perform question-and-answer processing on the question-and-answer data. The aforementioned large language model can be a large language model built based on deep learning or machine learning, such as GPT, BERT, LLaMA, etc.

[0117] The aforementioned target knowledge base refers to the knowledge base corresponding to the current growth value of the smart toy. It's important to understand that the growth value of a smart toy corresponds to its level.

[0118] The above question-and-answer processing can be understood as the process of finding the most relevant answer to the question-and-answer data from the target knowledge base.

[0119] The above answer data can be understood as the most relevant answer data obtained by analyzing question and answer data.

[0120] In one possible implementation, if the question-and-answer data is "Does the calf eat grass?", and the target knowledge base includes information about the days of the week, a prompt word is constructed for the large language model using the question-and-answer data and the target knowledge base. The prompt word could be "Please use information from the target knowledge base to answer 'Does the calf eat grass?'". Then, the prompt word "Please use information from the target knowledge base to answer 'Does the calf eat grass?'" is input into the large language model, which then retrieves the answer data 'The calf does eat grass' from the target knowledge base based on "Please use information from the target knowledge base to answer 'Does the calf eat grass?'".

[0121] In another possible implementation, if the question-and-answer data is "What color is an apple?", and the target knowledge base includes information about fruits, a prompt word for the large language model is constructed using the question-and-answer data and the target knowledge base. The prompt word could be "Please use information from the target knowledge base to answer 'What color is an apple?'". Then, the prompt word "Please use information from the target knowledge base to answer 'What color is an apple?'" is input into the large language model. The large language model will then retrieve the answer data "The apple is red" from the target knowledge base based on "Please use information from the target knowledge base to answer 'What color is an apple?'".

[0122] Optionally, in the step of controlling the smart toy to output answer data, the output mode of the smart toy can be determined to be a voice mode and / or an action mode based on the semantics of the answer data; based on the voice mode and / or action mode, the smart toy can be controlled to output the answer data via voice and / or action.

[0123] In this embodiment of the invention, after obtaining the answer data, semantic recognition can be performed on the answer data to identify its semantics and determine that the output mode of the smart toy is a voice mode and / or a motion mode. The semantics of the answer data can be understood as the meaning and content expressed by the answer data. Semantic recognition can be understood as the process of analyzing the word meaning and semantics of the answer data using natural language processing technology to understand the meaning and content within the answer data.

[0124] The aforementioned voice mode can be understood as a mode of conveying information through voice. For example, smart toys use voice output devices to output voice to answer questions.

[0125] The above-mentioned action patterns can be understood as patterns of expressing information through physical actions. For example, intelligent toys can output actions through action output devices, such as controlling the eyes, limbs, head and other parts to make specific actions (such as moving the eyes, rotating the limbs, shaking the head, wagging the tail and other actions) to answer questions.

[0126] The aforementioned voice output can be understood as the smart toy converting answer data into playable voice information through voice.

[0127] The aforementioned action output can be understood as the intelligent toy performing corresponding physical actions based on the answer data, such as moving its eyes, rotating its arms, or moving other parts. Through action output, users can obtain answers in a direct or interactive way.

[0128] In one possible embodiment, when the smart toy's output mode is voice mode and action mode, the answer data is converted into audible voice information through voice, and the answer data is expressed through the smart toy's physical actions, such as moving the eyes, rotating the limbs, shaking the head, wagging the tail, etc.; when the smart toy's output mode is voice mode, the answer data is converted into audible voice information through voice; when the smart toy's output mode is action mode, the answer data is expressed through the smart toy's physical actions, such as moving the eyes, rotating the arms, or moving other parts.

[0129] Optionally, after the intent detection step when the smart toy is woken up by the user, if a functional intent is detected, search data and a function database are acquired; based on the search data, the target function data is searched in the function database, and the smart toy is controlled to output the target function data.

[0130] In this embodiment of the invention, the above-mentioned functional intent can be understood as the needs or purposes expressed by the user through specific function buttons on the smart toy, such as nursery rhymes, stories, Tang poems, etc.

[0131] The search data mentioned above can be understood as searching for data in the functional database that corresponds to the user's functional intent. For example, if the user's functional intent is nursery rhymes, the search data will be nursery rhymes; if the user's functional intent is stories, the search data will be stories, and so on.

[0132] The aforementioned functional database can be understood as a database used to store and manage various functions, such as nursery rhyme data, story data, Tang poetry data, etc.

[0133] The aforementioned target function data can be understood as the function data corresponding to the user's functional intent. For example, if the user's functional intent is nursery rhymes, then searching for nursery rhymes in the function database and finding nursery rhyme data in the function database is the target function data; if the user's functional intent is stories, then searching for stories in the function database and finding story data in the function database is the target function data.

[0134] Furthermore, after retrieving the target function data, the system controls the smart toy to output the target function data. This output is a pre-set output format, which can be voice output or / or action output.

[0135] In one possible embodiment, when the user's intended function is detected as nursery rhymes, the system searches for nursery rhymes in the function database. Once a nursery rhyme is found, the system controls the smart toy to output the nursery rhyme data. When the user's intended function is detected as stories, the system searches for stories in the function database. Once a story is found, the system controls the smart toy to output the story data. When the user's intended function is detected as Tang Dynasty poetry, the system searches for Tang Dynasty poetry in the function database. Once a Tang Dynasty poetry is found, the system controls the smart toy to output the Tang Dynasty poetry data.

[0136] like Figure 2 As shown, an embodiment of the present invention provides a control device for a smart toy, the control device for the smart toy comprising:

[0137] The intent detection module 201 is used to detect intent when the smart toy is woken up by the user;

[0138] The determination module 202 is used to acquire feeding data if a feeding intention is detected, and to determine the growth value increased by the smart toy based on the feeding data;

[0139] The first acquisition module 203 is used to acquire question and answer data and a target knowledge base if a question and answer intent is detected. The level of the target knowledge base corresponds to the current growth value of the smart toy, and different levels of the target knowledge base have knowledge data corresponding to different user ages.

[0140] The first control module 204 is used to perform question-and-answer processing on the question-and-answer data based on the target knowledge base to obtain answer data, and to control the smart toy to output the answer data.

[0141] Optionally, the intent detection module 201 includes:

[0142] The acquisition submodule is used to acquire the user's voice data when the smart toy is woken up by the user; the intent recognition submodule is used to perform intent recognition on the voice data to obtain the user's intent.

[0143] Optionally, the determining module 202 includes:

[0144] The acquisition submodule is used to control the smart toy to acquire the current image if the user's intention is to feed the user.

[0145] The first determining submodule is used as the image recognition submodule to perform image recognition on the current image and identify the type of feeding food in the current image;

[0146] The second determining submodule is used to determine the growth value corresponding to the feeding food based on the type of feeding food;

[0147] The third determining submodule is used to determine the growth value increased by the smart toy based on the growth value corresponding to the feeding food.

[0148] Optionally, the first acquisition module 203 includes:

[0149] The semantic recognition submodule is used to sequentially perform speech recognition and semantic recognition on the speech data to obtain the semantic recognition result corresponding to the speech data.

[0150] The fourth determining submodule is used to determine the user's question and answer data based on the semantic recognition results.

[0151] Optionally, the first acquisition module 203 includes:

[0152] The fifth determining submodule is used to determine the current growth value of the smart toy;

[0153] The sixth determination submodule is used to determine the target knowledge base corresponding to the current growth value based on the correspondence between the growth value stage and the knowledge base level. Different growth value stages correspond to different levels of the knowledge base.

[0154] Optionally, the first control module 204 includes:

[0155] A submodule is constructed to build prompt words for a large language model based on the question-and-answer data and the target knowledge base. The prompt words are used to instruct the large language model to call the target knowledge base to perform question-and-answer processing on the question-and-answer data.

[0156] The question-and-answer processing submodule is used to input the prompt words into the large language model for question-and-answer processing to obtain answer data.

[0157] Optionally, the first control module 204 includes:

[0158] The seventh determining submodule is used to determine, based on the semantics of the answer data, whether the output model of the smart toy is a voice mode and / or a motion mode;

[0159] The control submodule is used to control the smart toy to output the answer data via voice and / or action based on the voice mode and / or the action mode.

[0160] Optionally, the device further includes:

[0161] The second acquisition module is used to acquire search data and a function database if a function intent is detected.

[0162] The second control module is used to search for target function data in the function database based on the search data, and control the smart toy to output the target function data.

[0163] like Figure 3 As shown, this embodiment of the invention also provides an electronic device, including a processor, which can execute any of the above-described control methods for intelligent toys.

[0164] Specifically, it includes a processor 301 and a memory 302, as well as a computer program stored in the memory 302 and capable of running on the processor 301, which executes the control method for the intelligent toy, wherein:

[0165] The processor 301 executes the calculator program for the control method of the intelligent toy stored in the memory 302, and performs the following steps:

[0166] When the smart toy is activated by the user, intent detection is performed;

[0167] If a feeding intention is detected, feeding data is acquired, and the growth value increased by the smart toy is determined based on the feeding data.

[0168] If a question-and-answer intent is detected, question-and-answer data and a target knowledge base are obtained. The level of the target knowledge base corresponds to the current growth value of the smart toy. Different levels of the target knowledge base have knowledge data corresponding to different user ages.

[0169] Based on the target knowledge base, the question-and-answer data is processed to obtain answer data, and the smart toy is controlled to output the answer data.

[0170] Optionally, the process of processor 301 performing intent detection when the smart toy is awakened by the user includes:

[0171] When the smart toy is activated by the user, it acquires the user's voice data;

[0172] The user's intent is obtained by performing intent recognition on the voice data.

[0173] Optionally, the process of acquiring feeding data and determining the increased growth value of the smart toy based on the feeding data, performed by processor 301, includes:

[0174] If the user's intention is to feed the user, then control the smart toy to capture the current image;

[0175] The current image is subjected to image recognition to identify the type of feeding food in the current image; based on the type of feeding food, the growth value corresponding to the feeding food is determined;

[0176] Based on the growth value corresponding to the feeding food, the growth value increased by the smart toy is determined.

[0177] Optionally, the acquisition of question-and-answer data performed by processor 301 includes:

[0178] The speech data is sequentially subjected to speech recognition and semantic recognition to obtain the semantic recognition result corresponding to the speech data;

[0179] Based on the semantic recognition results, the user's question and answer data is determined.

[0180] Optionally, the acquisition of the target knowledge base executed by processor 301 includes:

[0181] Determine the current growth value of the intelligent toy;

[0182] Based on the correspondence between the growth value stage and the knowledge base level, the target knowledge base corresponding to the current growth value is determined, and different growth value stages correspond to different knowledge base levels.

[0183] Optionally, the process executed by processor 301, which involves performing question-and-answer processing on the question-and-answer data based on the target knowledge base to obtain answer data, includes:

[0184] Based on the question-and-answer data and the target knowledge base, prompt words for the large language model are constructed. The prompt words are used to instruct the large language model to call the target knowledge base to perform question-and-answer processing on the question-and-answer data.

[0185] The prompt words are input into the large language model for question-and-answer processing to obtain answer data.

[0186] Optionally, the processor 301's execution of controlling the smart toy to output the answer data includes:

[0187] Based on the semantics of the answer data, the output model of the smart toy is determined to be a voice mode and / or a motion mode;

[0188] Based on the voice pattern and / or the action pattern, the smart toy is controlled to output the answer data via voice and / or action.

[0189] Optionally, after intent detection is performed when the smart toy is awakened by the user, the method executed by the processor 301 further includes:

[0190] If a functional intent is detected, search data and the functional database are retrieved;

[0191] Based on the search data, the target function data is found in the function database, and the smart toy is controlled to output the target function data.

[0192] This invention also provides a computer-readable storage medium storing a computer program. When the computer program is executed by a processor, it implements the various processes of the intelligent toy control method provided in this invention and achieves the same technical effect. To avoid repetition, it will not be described again here.

[0193] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The program can be stored in a computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. The storage medium can be a magnetic disk, optical disk, read-only memory (ROM), or random access memory (RAM), etc.

[0194] The above description discloses only preferred embodiments of the present invention and should not be construed as limiting the scope of the present invention. Therefore, equivalent variations made in accordance with the claims of the present invention are still within the scope of the present invention.

Claims

1. A control method for an intelligent toy, characterized in that, The method includes the following steps: When the smart toy is in a wake-up state, intent detection is performed; If a feeding intention is detected, feeding data is acquired, and the growth value increased by the smart toy is determined based on the feeding data. If a question-and-answer intent is detected, question-and-answer data and a target knowledge base are obtained. The level of the target knowledge base corresponds to the current growth value of the smart toy. Different levels of the target knowledge base have knowledge data corresponding to different user ages. Based on the target knowledge base, the question-and-answer data is processed to obtain answer data, and the smart toy is controlled to output the answer data.

2. The control method for the intelligent toy as described in claim 1, characterized in that, When the smart toy is activated by the user, the intent detection process includes: When the smart toy is activated by the user, it acquires the user's voice data; The user's intent is obtained by performing intent recognition on the voice data.

3. The control method for the intelligent toy as described in claim 2, characterized in that, The process of acquiring feeding data and determining the increased growth value of the smart toy based on the feeding data includes: If the user's intention is to feed the user, then control the smart toy to capture the current image; Perform image recognition on the current image to identify the type of feeding food in the current image; Based on the type of feeding food, determine the corresponding growth value of the feeding food; Based on the growth value corresponding to the feeding food, the growth value increased by the smart toy is determined.

4. The control method for the intelligent toy as described in claim 2, characterized in that, The acquisition of question-and-answer data includes: The speech data is sequentially subjected to speech recognition and semantic recognition to obtain the semantic recognition result corresponding to the speech data; Based on the semantic recognition results, the user's question and answer data is determined.

5. The control method for the intelligent toy as described in claim 4, characterized in that, The acquisition of the target knowledge base includes: Determine the current growth value of the intelligent toy; Based on the correspondence between the growth value stage and the knowledge base level, the target knowledge base corresponding to the current growth value is determined, and different growth value stages correspond to different knowledge base levels.

6. The control method for the intelligent toy as described in claim 5, characterized in that, The step of performing question-and-answer processing on the question-and-answer data based on the target knowledge base to obtain answer data includes: Based on the question-and-answer data and the target knowledge base, prompt words for the large language model are constructed. The prompt words are used to instruct the large language model to call the target knowledge base to perform question-and-answer processing on the question-and-answer data. The prompt words are input into the large language model for question-and-answer processing to obtain answer data.

7. The control method for the intelligent toy as described in claim 6, characterized in that, The control of the smart toy to output the answer data includes: Based on the semantics of the answer data, the output model of the smart toy is determined to be a voice mode and / or a motion mode; Based on the voice pattern and / or the action pattern, the smart toy is controlled to output the answer data via voice and / or action.

8. The control method for the intelligent toy as described in claim 1, characterized in that, After performing intent detection when the smart toy is activated by the user, the method further includes: If a functional intent is detected, search data and the functional database are retrieved; Based on the search data, the target function data is found in the function database, and the smart toy is controlled to output the target function data.

9. A control device for an intelligent toy, characterized in that, The control device for the intelligent toy includes: An intent detection module is used to detect intent when the smart toy is woken up by the user; The determination module is used to acquire feeding data if a feeding intention is detected, and to determine the growth value increased by the smart toy based on the feeding data; The first acquisition module is used to acquire question and answer data and a target knowledge base if a question and answer intent is detected. The level of the target knowledge base corresponds to the current growth value of the smart toy, and different levels of the target knowledge base have knowledge data corresponding to different user ages. The first control module is used to perform question-and-answer processing on the question-and-answer data based on the target knowledge base to obtain answer data, and to control the smart toy to output the answer data.

10. An electronic device, characterized in that, include: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor, when executing the computer program, implements the steps of the control method for the intelligent toy as claimed in any one of claims 1 to 8.

11. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, implements the steps of the control method for the intelligent toy as described in any one of claims 1 to 8.

Images