Guiding statement determination method and device, storage medium and electronic device

Abstract

The invention discloses a guidance statement determination method and device, a storage medium and an electronic device, and relates to the technical field of smart home, the guidance statement determination method comprises the following steps: obtaining an interaction statement of a target object; a target intention and an explicit-implicit degree corresponding to the interaction statement are determined through a target interaction model, the explicit-implicit degree is used for indicating the reflection degree of the target intention in the interaction statement, and the target intention is the intention expressed by the target object through the interaction statement; and through the target interaction model, determining a target guide statement according to the target intention and the explicit-implicit degree. By adopting the technical scheme, the problem that the guide statement cannot be well determined is solved.

Description

Technical Field

[0001] The present application relates to the field of smart home technology, and more specifically, to a method and device for determining a guide sentence, a storage medium, and an electronic device. Background Art

[0002] In the traditional field of smart home appliances, the interaction between users and devices mostly relies on a rule-based matching system, which requires a large amount of corpus to be defined in advance in order to match intent based on semantic similarity. This method has a good response to clear and direct user instructions, but when faced with complex or ambiguous user expressions, the accuracy of recognition and response is low, and it is unable to guide customer intent. It is also difficult to maintain and update, and has poor scalability.

[0003] With regard to the problem in the related art that the guide statement cannot be determined well, no effective solution has been proposed yet.

[0004] Therefore, it is necessary to improve the related technology to overcome the above-mentioned defects in the related technology. Summary of the invention

[0005] The embodiments of the present invention provide a method and device for determining a guide sentence, a storage medium, and an electronic device, so as to at least solve the problem that the guide sentence cannot be determined well.

[0006] According to one aspect of an embodiment of the present invention, a method for determining a guide sentence is provided, comprising: obtaining an interaction sentence of a target object; determining a target intention and a degree of explicitness and invisibility corresponding to the interaction sentence through a target interaction model, wherein the degree of explicitness and invisibility is used to indicate the degree to which the target intention is reflected in the interaction sentence, and the target intention is the intention expressed by the target object through the interaction sentence; determining a target guide sentence according to the target intention and the degree of explicitness and invisibility through the target interaction model.

[0007] In an exemplary embodiment, the target interaction model is used to determine a target guiding sentence according to the target intent and the degree of explicitness and invisibility, including: determining a plurality of guiding sentences corresponding to the target intent through the target interaction model; and determining the target guiding sentence from the plurality of guiding sentences according to the degree of explicitness and invisibility through the target interaction model.

[0008] In an exemplary embodiment, the target guided sentence is determined from the multiple guided sentences according to the degree of visibility, including: when the degree of visibility is used to indicate that the degree of embodiment of the target intention in the interactive sentence is greater than or equal to a first preset threshold, the guided sentence whose guiding ability is greater than or equal to a second preset threshold among the multiple guided sentences is determined as the target guided sentence; when the degree of visibility is used to indicate that the degree of embodiment of the target intention in the interactive sentence is less than the first preset threshold, the guided sentence whose guiding ability is less than the second preset threshold among the multiple guided sentences is determined as the target guided sentence.

[0009] In an exemplary embodiment, the method further includes: obtaining a training data set, wherein the training data set includes multiple groups of data, each group of the multiple groups of data includes: sample interaction sentences, the intention corresponding to the sample interaction sentences, the degree of explicitness or implicitness of the intention corresponding to the sample interaction sentences, and the guiding sentences corresponding to the sample interaction sentences; using the training data set to train a general interaction model, and when the general interaction model meets a convergence condition, determining the general interaction model that meets the convergence condition as the target interaction model, wherein when the general interaction model does not meet the convergence condition, the model parameters of the general interaction model are adjusted.

[0010] In an exemplary embodiment, the general interaction model is trained using the training data set, and when the general interaction model meets the convergence condition, the general interaction model that meets the convergence condition is determined as the target interaction model, including: using the training data set to train the first network in the general interaction model, and when the first network meets the first convergence condition, determining the first network that meets the first convergence condition as an intention recognition network, wherein, when the first network does not meet the first convergence condition, the model parameters of the first network are adjusted, and the intention recognition network is used to recognize the intention of a sentence and the degree of explicitness of the intention; using the training data set to train the second network in the general interaction model, and when the second network meets the second convergence condition, determining the second network that meets the second convergence condition as a guide sentence generation network, wherein, when the second network does not meet the second convergence condition, the model parameters of the second network are adjusted, and the guide sentence generation network is used to generate a guide sentence according to the intention of the sentence and the degree of explicitness of the intention; wherein the target interaction model includes the intention recognition network and the guide sentence generation network, and the convergence condition includes the first convergence condition and the second convergence condition.

[0011] In an exemplary embodiment, after determining the target guiding sentence according to the target intention and the degree of explicitness and invisibility through the target interaction model, the method also includes: obtaining feedback information of the target object; saving the feedback information, the interaction sentence, the target intention, the degree of explicitness and the target guiding sentence as a piece of data into a target database; when the number of data items in the target database is greater than a third preset threshold, using the data in the target database to train the target interaction model to update the target interaction module.

[0012] In an exemplary embodiment, after determining the target guidance statement according to the target intention and the degree of visibility through the target interaction model, the method also includes: voice playing the target guidance statement; listening to the response statement of the target object; or broadcasting a prompt statement and listening to the response statement of the target object, wherein the prompt statement is used to prompt the target object to feedback a response statement.

[0013] According to another aspect of an embodiment of the present invention, a device for determining a guide sentence is also provided, including: an acquisition module, used to acquire an interaction sentence of a target object; a first determination module, which determines a target intention and a degree of explicitness and invisibility corresponding to the interaction sentence through a target interaction model, wherein the degree of explicitness and invisibility is used to indicate the degree to which the target intention is reflected in the interaction sentence, and the target intention is the intention of the target object to execute through the interaction sentence; and a second determination module, which determines a target guide sentence according to the target intention and the degree of explicitness and invisibility through the target interaction model.

[0014] According to another aspect of the embodiments of the present invention, a computer-readable storage medium is provided, in which a computer program is stored, wherein the computer program is configured to execute the above-mentioned method for determining the boot statement when running.

[0015] According to another aspect of an embodiment of the present invention, there is provided an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the method for determining the boot statement through the computer program.

[0016] According to another aspect of the embodiments of the present application, a computer program product is provided, including a computer program, and the method for determining the above position when the computer program is executed by a processor.

[0017] Through the present invention, the intention corresponding to the acquired interaction sentence and the degree of visibility of the intention are determined through the target interaction model, and then the target guiding sentence is determined according to the intention and the degree of visibility through the target interaction model. Since the intention and the degree of visibility of the interaction sentence sent by the target object are taken into consideration when determining the guiding sentence, the accuracy of the interaction sentence in response to the target object is improved, thereby solving the problem of not being able to determine the guiding sentence well, and thus improving the user experience. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present application and, together with the description, serve to explain the principles of the present application.

[0019] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, for ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative labor.

[0020] Figure 1 is a hardware environment schematic diagram of a method for determining a guide statement according to an embodiment of the present application;

[0021] Figure 2 is a flow chart of a method for determining a guide sentence according to an embodiment of the present invention;

[0022] Figure 3 It is a schematic diagram of the first part of the flow chart of a system for determining a guide sentence according to an embodiment of the present invention;

[0023] Figure 4 is a schematic diagram of the second part of the flow chart of a system for determining a guide sentence according to an embodiment of the present invention;

[0024] Figure 5 is a schematic diagram of the third part of the flow chart of a system for determining a guide sentence according to an embodiment of the present invention;

[0025] Figure 6 is a schematic diagram of the fourth part of the flow chart of a system for determining a guide sentence according to an embodiment of the present invention;

[0026] Figure 7 is a schematic diagram of generating a guide sentence by a training interaction model according to an embodiment of the present invention;

[0027] Figure 8 is a schematic diagram of generating a guide sentence by a test interaction model according to an embodiment of the present invention;

[0028] Fig. 9It is a structural block diagram of a device for determining a guide sentence according to an embodiment of the present invention. DETAILED DESCRIPTION

[0029] In order to enable those skilled in the art to better understand the solution of the present application, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present application.

[0030] It should be noted that the terms "first", "second", etc. in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the present application described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device comprising a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0031] According to one aspect of an embodiment of the present application, a method for determining a guide sentence is provided. The method for determining a guide sentence is widely used in smart home (Smart Home), smart home, smart home device ecology, smart house (IntelligenceHouse) ecology and other whole-house intelligent digital control application scenarios. Optionally, in this embodiment, the above-mentioned method for determining a guide sentence can be applied to Figure 1 In the hardware environment composed of the terminal device 102 and the server 104 shown in FIG. Figure 1 As shown, the server 104 is connected to the terminal device 102 via a network, and can be used to provide services (such as application services, etc.) for the terminal or a client installed on the terminal. A database can be set on the server or independently of the server to provide data storage services for the server 104. Cloud computing and / or edge computing services can be configured on the server or independently of the server to provide data computing services for the server 104.

[0032] The above network may include but is not limited to: wired network, wireless network. The above wired network may include but is not limited to: wide area network, metropolitan area network, local area network, and the above wireless network may include but is not limited to: WIFI (Wireless Fidelity), Bluetooth. The terminal device 102 may be but is not limited to a PC, a mobile phone, a tablet computer, a smart air conditioner, a smart range hood, a smart refrigerator, a smart oven, a smart stove, a smart washing machine, a smart water heater, a smart washing equipment, a smart dishwasher, a smart projection equipment, a smart TV, a smart clothes drying rack, a smart curtain, a smart audio and video, a smart socket, a smart speaker, a smart speaker, a smart fresh air equipment, a smart kitchen and bathroom equipment, a smart bathroom equipment, a smart sweeping robot, a smart window cleaning robot, a smart mopping robot, a smart air purification equipment, a smart steamer, a smart microwave oven, a smart kitchen treasure, a smart purifier, a smart water dispenser, a smart door lock, etc.

[0033] In order to solve the above problem, a method for determining a guide sentence is provided in this embodiment. Figure 2 1 is a flow chart of a method for determining a guide sentence according to an embodiment of the present invention, including but not limited to being applied in a voice interaction system, and the process includes the following steps S202-S206:

[0034] Step S202: Obtaining the interaction statement of the target object;

[0035] Optionally, in this step S202, the system first monitors and receives the interactive statements between the user (target object) and the smart home appliance. These statements can be instructions or inquiries issued by the user in various ways such as voice, text or gestures. For example, the user may say, "The food in the refrigerator is not fresh", or "How to adjust the water temperature of the washing machine". The system needs to be able to capture and understand these statements, which is the basis for subsequent processing.

[0036] Step S204: determining the target intention and degree of explicitness and implicitness corresponding to the interaction sentence through the target interaction model, wherein the degree of explicitness and implicitness is used to indicate the degree to which the target intention is reflected in the interaction sentence, and the target intention is the intention expressed by the target object through the interaction sentence;

[0037] Optionally, the powerful natural language processing capabilities of the target interaction model can be used to deeply analyze the user's input content and identify the user's explicit and implicit intentions. The target interaction model has learned a large amount of language patterns and user behavior data through training, and can capture the subtle differences in user expressions, thereby more accurately understanding the user's intentions. For example, in the field of home appliances, users may raise clear requirements such as "the refrigerator's cooling effect is not good", or they may just mention potential problems such as "the food in the refrigerator is not as cold as before". The target interaction model can accurately extract the user's real needs.

[0038] Optionally, after the system obtains the user's sentences, it will then use a pre-trained target interaction model (which can be a large language model) to analyze and understand these sentences to determine the user's true intention and the degree of explicitness of the intention. For example, for the sentence "The food in the refrigerator is not fresh", the model can recognize that the user may need to check or adjust the temperature setting of the refrigerator, which is a potential and relatively obscure demand with a low degree of explicitness. In contrast, "adjust the water temperature of the washing machine" is a clear and direct instruction with a high degree of explicitness.

[0039] It should be noted that the degree of explicitness is determined by analyzing the contextual information, tone, keywords, etc. contained in the sentence through the model. For explicit intent, the model can directly read clear operation instructions from the sentence; while for implicit intent, the model needs to interpret it in combination with a wider context and background knowledge. This distinction helps the system respond to user needs more flexibly and provide solutions that are more in line with the situation.

[0040] Step S206: Determine the target guidance sentence according to the target intention and the degree of explicitness and implicitness through the target interaction model.

[0041] Optionally, the identified intents are divided into two categories: strong association and weak association. For obvious user intents, the system will provide strong guidance, such as directly providing a solution to the refrigerator's refrigeration function for "the refrigerator's refrigeration effect is not good"; for potential user intents, the system will provide weak guidance, such as recommending users to check the refrigerator's settings or surrounding environment for "the food in the refrigerator is not as cold as before".

[0042] Optionally, after clarifying the user's intention and its degree of explicitness, the system will generate corresponding guiding sentences according to different types of intentions through the target interaction model. For explicit intentions, direct solutions or operation instructions will be generated, such as "Please lower the temperature setting of the refrigerator to 2 degrees Celsius." For implicit intentions, the system will generate more inspiring and guiding sentences to help users self-diagnose or understand possible solutions, such as "Considering that the food is not fresh, it may be because the temperature of the refrigerator is not set properly. You can check whether the temperature of the refrigerator is within the appropriate range."

[0043] It should be noted that by distinguishing the degree of explicitness of user intentions, the system can respond to user needs more intelligently and provide services that are both accurate and personalized, thereby improving the experience and efficiency of user interaction with smart home appliances. In addition, the system can adapt to the expression habits and contexts of different users, and can accurately understand even vague or implicit expressions, which greatly enhances the adaptability and intelligence of the system. By determining the degree of explicitness, the system can adjust its recommendation and guidance strategies according to the clarity of user intentions, avoiding excessive intervention or invalid recommendations, and allowing users to obtain a more natural and comfortable service experience.

[0044] In the above steps, the intention corresponding to the obtained interaction sentence and the degree of explicitness of the intention are determined through the target interaction model, and then the target guiding sentence is determined according to the intention and the degree of explicitness through the target interaction model. Since the intention and degree of explicitness of the interaction sentence sent by the target object are taken into consideration when determining the guiding sentence, the accuracy of the interaction sentence in response to the target object is improved, thereby solving the problem of not being able to determine the guiding sentence well, and improving the user experience.

[0045] In an exemplary embodiment, the target interaction model is used to determine the target guiding sentence according to the target intention and the degree of explicitness and implicitness, which can be achieved by the following steps S11-S12:

[0046] Step S11: determining a plurality of guiding sentences corresponding to the target intention through the target interaction model;

[0047] Optionally, when the target interaction model recognizes the user's target intent, it will not immediately generate a single guide sentence, but will generate multiple guide sentences related to the intent based on the model's training and knowledge base data. For example, if the user mentions "the washing machine cannot start normally", the model may generate multiple guide sentences such as "check whether the power is on", "check whether there is an error code displayed", "check whether the washing machine door is closed", etc. This step reflects the flexibility and comprehensiveness of the system, which can provide users with multiple possible solutions, thereby improving the efficiency of problem solving and user satisfaction.

[0048] Step S12: Determine the target guiding sentence from the multiple guiding sentences according to the degree of explicitness and invisibility through the target interaction model.

[0049] Optionally, after generating multiple guiding sentences, the system will select the most appropriate guiding sentence for feedback through the target interaction model according to the degree of explicitness of the target intent. For explicit intent, the system usually selects direct and clear guiding sentences, such as "Check whether the washing machine door is closed" in the above example, because such guiding sentences can directly solve the problem. For implicit intent, the system tends to choose more inspiring and gentler guiding sentences to avoid excessive intervention or misleading users. For example, if the user says "The clothes are always not washed clean when using the washing machine recently", this may be an intention that implies concern about the cleaning ability of the washing machine. The system will select more guiding sentences such as "Try to use the deep cleaning mode of the washing machine, or check whether the filter needs to be cleaned" from multiple guiding sentences.

[0050] It should be noted that generating multiple guidance sentences and filtering them according to the explicitness of the intent can help quickly identify and solve the problems faced by users. Especially when dealing with implicit intent, gentle guidance can often inspire users to self-diagnose and find the problem more quickly. In addition, by selecting guidance sentences according to the explicitness of the intent, the system can provide more personalized services, ensuring that the guidance is not only accurate but also fits the user's needs and context, improving the user experience.

[0051] In an exemplary embodiment, determining the target guiding sentence from the plurality of guiding sentences according to the degree of exposure can be implemented by the following steps S21 to S22, wherein step S21 and step S22 are performed in different situations:

[0052] Step S21: when the degree of visibility is used to indicate that the degree of embodiment of the target intention in the interactive sentence is greater than or equal to a first preset threshold, determining a guide sentence whose guiding ability is greater than or equal to a second preset threshold among the multiple guide sentences as the target guide sentence;

[0053] Optionally, when the intention expressed by the user is relatively obvious and its degree of explicitness reaches the first preset threshold set by the system, the system considers that the user's needs are relatively direct and can be clearly guided. At this time, the system will filter out those sentences whose guiding capabilities (i.e., the efficiency and effectiveness of solving problems or meeting needs) exceed the second preset threshold from the multiple guiding sentences related to the target intention as target guiding sentences. For example, if the user says "how to adjust the air conditioner temperature", which is an explicit intention, the system will select the most direct and operational one from the guiding sentences such as "adjust the temperature button" and "use the temperature button on the remote control", such as "use the temperature + / - button on the remote control to adjust the air conditioner temperature" as the target guiding sentence.

[0054] Optionally, a historical interaction record table and / or a target database stored in the system is obtained, wherein the historical interaction record table and the target database record guide sentences with user feedback weights, and the user feedback weights are used to reflect the degree of user feedback on the guide sentences with user feedback weights; when there are guide sentences with user feedback weights among the multiple guide sentences, the guide sentences with user feedback weights are preferentially determined as the target guide sentences.

[0055] It should be noted that through this strategy, the system can quickly respond to users' direct needs and provide guidance that is highly operational and efficient, thereby improving user experience and satisfaction.

[0056] Step S22: When the degree of visibility is used to indicate that the degree to which the target intention is reflected in the interactive sentence is less than the first preset threshold, the guiding sentence whose guiding ability is less than the second preset threshold among the multiple guiding sentences is determined as the target guiding sentence.

[0057] Optionally, when the user's target intention is relatively obscure and its degree of visibility does not reach the first preset threshold, the system believes that the user's needs may require more detailed exploration and guidance. In this case, the system will select those sentences with relatively low guidance ability but more inspiring and gentle from multiple guidance sentences as target guidance sentences. For example, the user may simply express "recently, the sleep quality has not been good", and the system recognizes that this may be related to the indoor temperature, but the user's intention is not direct, so the system chooses "You may need to check the indoor temperature and humidity, which will affect the sleep quality". This is a more gentle guidance sentence, instead of directly giving specific temperature adjustment suggestions.

[0058] It should be noted that this strategy is suitable for dealing with implicit needs. Through inspiring and gentle guidance, it avoids excessive intervention in users. At the same time, it can also gradually guide users to discover and solve problems, enhancing the intelligence of the system and the natural and smooth user experience.

[0059] It should be noted that steps S21 and S22 implement differentiated guiding statement selection strategies by distinguishing the degree of explicitness or implicitness of user intentions, which can ensure that the system can respond quickly to explicit needs while exploring and satisfying implicit needs in a detailed and gentle manner when processing user interactions, effectively improving the intelligence and personalization level of interaction between smart home appliances and users, and greatly enhancing the comfort and satisfaction of user experience by avoiding excessive intervention and providing reasonable guidance. This differentiated strategy based on explicitness or implicitness can significantly improve the adaptability and intelligence of the system.

[0060] In an exemplary embodiment, the method further comprises the following steps S31-S32:

[0061] Step S31: obtaining a training data set, wherein the training data set includes multiple groups of data, each group of the multiple groups of data includes: a sample interaction sentence, an intention corresponding to the sample interaction sentence, a degree of explicitness or implicitness of the intention corresponding to the sample interaction sentence, and a guide sentence corresponding to the sample interaction sentence;

[0062] Optionally, each of the multiple groups of data further includes: a response statement corresponding to the sample interaction statement, and a response method, wherein the response statement is used to directly answer the question asked in the sample interaction statement.

[0063] Optionally, each sample data includes at least: sample interaction sentences: original sentences issued by users when interacting with the intelligent system, such as "the food in the refrigerator is no longer fresh"; the intention corresponding to the sample interaction sentence: the user intention determined by manual annotation or preliminary data processing, such as adjusting the temperature of the refrigerator; the degree of explicitness of the intention: a quantitative assessment of the clarity of the intention in the sentence, a high degree of explicitness means that the intention in the sentence is obvious, while a low degree of explicitness means that the intention is more obscure; the guiding sentence corresponding to the sample interaction sentence: for a given intention and degree of explicitness, a preset or verified sentence that can effectively guide the user, such as "check whether the refrigerator temperature is set between 2-4 degrees Celsius".

[0064] Optionally, prepare Figure 7 There are 200 to 300 question-answer pairs in total, which may include knowledge questions and answers about household appliances, marking whether there is an intention to control the device and feasible standard instructions, and both positive and negative samples.

[0065] Step S32: Use the training data set to train a universal interaction model, and when the universal interaction model meets the convergence condition, determine the universal interaction model that meets the convergence condition as the target interaction model, wherein when the universal interaction model does not meet the convergence condition, the model parameters of the universal interaction model are adjusted.

[0066] Optionally, after obtaining the training data set, the next step is to use this data to train an initial general interaction model. The training process involves adjusting the model parameters so that the model can accurately predict user intent and its degree of explicitness and generate effective guiding sentences. Training will continue until the model's performance indicators (such as accuracy, loss function value, etc.) reach a pre-set convergence condition, which usually means that the model's performance on the training data has stabilized and can handle various user inputs well.

[0067] Optionally, by using a training data set labeled with explicitness and guide sentences, the model can learn how to generate guide sentences of corresponding levels for intentions with different explicitness, thereby improving the personalization and intelligence level of the service.

[0068] It should be noted that the above steps are key steps in building and optimizing the target interaction model. By using training data containing the degree of explicitness and guiding sentences, the system can not only accurately identify the user's intentions, but also provide personalized guidance services based on the degree of explicitness of the intentions. At the same time, it ensures the maturity and continuous optimization capabilities of the model, providing users with a more natural, smooth and personalized interaction experience, which is an important guarantee for realizing intelligent recommendation and guidance in the technical solution of the present invention.

[0069] In an exemplary embodiment, the general interaction model is trained using the training data set, and when the general interaction model satisfies a convergence condition, the general interaction model satisfying the convergence condition is determined as the target interaction model, which can be achieved by the following steps S41-S42:

[0070] Step S41: using the training data set to train the first network in the general interaction model, and when the first network satisfies a first convergence condition, determining the first network that satisfies the first convergence condition as an intention recognition network, wherein when the first network does not satisfy the first convergence condition, the model parameters of the first network are adjusted, and the intention recognition network is used to recognize the intention of a sentence and the degree of explicitness and implicitness of the intention;

[0071] Optionally, in step S41, the system focuses on training the first network in the general interaction model - the intent recognition network. Using the sample interaction sentences in the training data set and the corresponding intent and degree of explicitness, the intent recognition network is trained to recognize the potential intent in the user's sentences and the clarity of these intents. During the training process, the parameters of the network will be continuously adjusted to optimize its recognition performance. When the recognition error rate is reduced to a certain level, or the loss function value reaches a preset threshold, that is, when the first convergence condition is met, the network is deemed to have completed training and is formally determined as an intent recognition network for actual user intent recognition.

[0072] Optionally, conduct offline intent extraction testing, run prompts in batches to traverse all the knowledge of current home appliances, find qualified knowledge, test the effect of prompts, and continue to optimize the effect of prompts, and finally form a stable, standardized output and prompts that meet the requirements of the home appliance industry, so as to achieve effect reference Figure 7 .

[0073] It should be noted that through special training, the intent recognition network can accurately understand user needs, whether it is explicit needs expressed directly or implicit needs that require reasoning. This lays a solid foundation for the next step of guiding sentence generation and ensures the accuracy and depth of the system's understanding of user intent.

[0074] Step S42: using the training data set to train the second network in the general interaction model, and when the second network satisfies a second convergence condition, determining the second network that satisfies the second convergence condition as a guide sentence generation network, wherein when the second network does not satisfy the second convergence condition, the model parameters of the second network are adjusted, and the guide sentence generation network is used to generate a guide sentence according to the intention of the sentence and the degree of explicitness of the intention;

[0075] Among them, the target interaction model includes the intention recognition network and the guide sentence generation network, and the convergence condition includes the first convergence condition and the second convergence condition.

[0076] Optionally, in step S42, the system trains a second network, the guide sentence generation network. Based on the intention and degree of explicitness of the user's sentence, and the guide sentence samples in the training data set, the network learns how to generate guide sentences suitable for specific intentions and degrees of explicitness. Similar to step S41, when the error between the generated guide sentence and the expected result is minimized and the second convergence condition is met, the second network is determined as the guide sentence generation network. If the condition is not met, the network parameters continue to be adjusted until the optimization target is reached.

[0077] It should be noted that the training of the guidance sentence generation network enables the system to generate guidance sentences that match the user's intention and degree of explicitness, which not only improves the naturalness of the user's interaction with the intelligent system, but also enhances the system's responsiveness and user satisfaction. The generated guidance sentences can provide a variety of services from direct instructions to heuristic suggestions according to the clarity of the intention, meeting the needs of users in different situations.

[0078] It should be noted that by decomposing the general interaction model and training it separately into an intent recognition network and a guide sentence generation network, the high-precision recognition ability of deep learning and the generation ability of natural language processing are effectively integrated to form an efficient and intelligent target interaction model. Such a model can accurately understand the depth and breadth of user intentions, and at the same time generate guide sentences that are both in line with the needs and natural and smooth, greatly improving the user's experience of interacting with smart home appliances and achieving more intelligent, humane and efficient services. In addition, by setting the convergence conditions, the training process of the model is ensured to reach the optimal state, avoiding the problem of overtraining or undertraining, and ensuring the stability and accuracy of the system. This strategy of separate training and compound use is the key to achieving high-quality user interaction in the field of smart home appliances.

[0079] In an exemplary embodiment, after determining the target guiding sentence according to the target intention and the degree of explicitness and implicitness through the target interaction model, the method further includes the following steps S51-S53:

[0080] Step S51: Obtaining feedback information of the target object;

[0081] Optionally, the feedback information of the target object is used to annotate a weight value to the target guiding sentence, so that the target guiding sentence is determined as a guiding sentence with a user feedback weight.

[0082] Optionally, after the model generates the target guidance statement and provides it to the user, the system will collect the user's feedback information on the guidance statement. This may include whether the user follows the guidance statement, the effect after the operation, the user's satisfaction evaluation of the guidance statement, etc. Feedback information is a key data source for model optimization, which can reveal the degree of match between the model output and the actual needs of the user.

[0083] Optionally, a corresponding weight or score is added to the guide sentence according to the actual situation of whether the user has used the intent recommended in the target guide sentence in multiple rounds of conversations or within 24 hours thereafter.

[0084] It should be noted that collecting user feedback information can help the system understand the actual effect of the guidance statement, provide a basis for subsequent model optimization, and enhance user participation in the system, making the service more in line with user needs.

[0085] Step S52: saving the feedback information, the interactive sentence, the target intention, the degree of visibility and the target guiding sentence as one piece of data into a target database;

[0086] Optionally, the system saves the user's feedback information and the context information when generating the guide sentence (including the original interaction sentence, the identified target intent, the degree of explicitness of the intent, and the generated target guide sentence) into the target database. This database is the data pool for model training and optimization, and its content richness and quality directly affect the performance of the model.

[0087] Optionally, the generated new guiding sentence is manually reviewed and then entered into the target database, and in actual use, the system marks the weight value of "user feedback" according to the user's feedback.

[0088] It should be noted that by saving the complete interaction context and feedback results, the target database can provide the model with highly relevant and specific learning cases, promote the model to learn user preferences and actual feedback, and thus improve the accuracy and applicability of its generated guidance sentences.

[0089] Step S53: When the number of data items in the target database is greater than a third preset threshold, the target interaction model is trained using the data in the target database to update the target interaction module.

[0090] Optionally, when the target database accumulates a certain amount of data (i.e., the number of data items exceeds a third preset threshold), the system will use this data to retrain the target interaction model to update model parameters and optimize model performance. This process is cyclical, and as data accumulates, the model can continue to learn and adapt, improving its intelligent decision-making capabilities and user satisfaction.

[0091] It should be noted that through continuous model updates, the system can better understand and meet user needs, reduce misleading or inappropriate suggestions, and improve the intelligence level and user experience of the entire service process. In addition, this mechanism helps the system capture changes in user behavior and needs, ensuring the timeliness and adaptability of the model.

[0092] It should be noted that steps S51 to S53 form a closed-loop optimization process by collecting user feedback, saving complete interaction cases, and regularly using these cases to train and update the model. This process not only ensures that the model can self-learn and adjust according to actual user feedback, but also promotes a more natural, efficient and satisfactory interaction between the smart home appliance system and the user, and is a key technical strategy for realizing user personalized services and improving interaction quality.

[0093] In an exemplary embodiment, after determining the target guiding sentence according to the target intention and the degree of explicitness and implicitness through the target interaction model, the method further includes the following steps S61 to S63, wherein steps S62 and S63 are not executed in any order:

[0094] Step S61: voice playing the target guiding sentence;

[0095] Optionally, after generating the target guidance sentence, the system will play the guidance sentence to the user through voice. This interactive method is more humane and can adapt to the usage habits of different users, especially for those users who may not be convenient to read text when operating smart home appliances.

[0096] It should be noted that voice playback guidance sentences can improve the efficiency of information transmission, reduce the user's operation steps, and provide a more natural interactive experience. Especially for users with visual impairments or in busy scenarios, voice interaction can significantly improve their convenience and experience in using smart home appliances.

[0097] Step S62: monitoring the response statement of the target object;

[0098] Optionally, the system will continue to monitor the response statements that the user may issue after receiving the guidance statement. This may be the user's confirmation of the guidance statement, feedback on the operation results, or further questions or requirements. Monitoring response statements is an important way for the system to obtain user feedback, which can help the system understand the effect of the guidance statement and whether further guidance or adjustment is needed.

[0099] It should be noted that monitoring response statements can enable the system to adjust its subsequent operations in a timely manner and provide more personalized services. It can also collect user behavior data to provide a basis for subsequent model optimization and personalized recommendations.

[0100] Step S63: broadcasting a prompt statement, and monitoring a response statement of the target object, wherein the prompt statement is used to prompt the target object to feedback a response statement.

[0101] Optionally, in some cases, the system will actively prompt and encourage users to provide feedback. For example, after playing the guidance sentence, the system may broadcast a sentence "Please tell me whether you have followed the guidance or have other questions". Such a prompt sentence can stimulate the user's willingness to provide feedback and ensure that the system can obtain the user's feedback information in a timely manner.

[0102] It should be noted that active prompts enhance the interaction between users and the system and increase the probability of user feedback, which helps the system to understand user needs more comprehensively and improve its service quality and user satisfaction. In this way, the system can better adapt to the specific situation of users and provide more timely and effective help.

[0103] It should be noted that by broadcasting guidance statements by voice, monitoring the user's response statements, and actively prompting the user for feedback, the smart home appliance system can provide a more humane and efficient user interaction experience. Voice interaction is naturally integrated into the user's life scenes, reducing the user's cognitive burden and improving the convenience of their interaction with smart home appliances. At the same time, by monitoring response statements and actively prompting feedback, the system can continuously obtain real-time information from users, which not only helps to adjust services in real time, but also collects a large amount of data for subsequent model training and optimization, ensuring continuous improvement and personalization of services. This interactive method combines the intelligence of technology with the humanization of user experience. It is a key technical strategy for achieving efficient communication between smart home appliances and users, and can significantly improve user satisfaction and usage experience with smart home appliances.

[0104] Obviously, the above-described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. In order to better understand the above method, the above process is described below in conjunction with the embodiments, but it is not intended to limit the technical solutions of the embodiments of the present invention. Figure 3 , Figure 4 , Figure 5 and Figure 6 This is a schematic diagram of an optional complete system for this application, specifically:

[0105] 1. Intent recognition module: Utilize the powerful natural language processing capabilities of the big model (i.e. the target interaction model mentioned above) to conduct in-depth analysis of the user's input content and identify the user's explicit and potential intentions. The big model has learned a large amount of language patterns and user behavior data through training, and can capture the subtle differences in user expressions, thereby more accurately understanding the user's intentions. For example, in the field of home appliances, users may raise clear requirements such as "the refrigerator's cooling effect is not good", or they may only mention potential problems such as "the food in the refrigerator is not as cold as before". The big model can accurately extract the user's real needs;

[0106] 2. Intent classification mechanism: The identified intents are divided into two categories: strong correlation and weak correlation. For obvious user intents, the system will provide strong guidance, such as directly providing a solution for the refrigerator's refrigeration function; for potential user intents, the system will provide weak guidance, such as recommending users to check the refrigerator's settings or surrounding environment;

[0107] 3. Personalized recommendation system: Based on the user's intent classification, the system will provide personalized guidance and recommendations. For strong-related intents, the system will directly provide solutions or answers; for weak-related intents, the system will recommend related content or services based on the user's preferences and historical behavior;

[0108] 4. User feedback mechanism: In order to continuously optimize the accuracy of recommendations and guidance, the system has established a user feedback mechanism. Users can evaluate the recommendation results, and the system will adjust the recommendation algorithm based on user feedback to achieve self-learning and optimization.

[0109] 1. Household appliances knowledge training data set (for equipment control guidance):

[0110] The product team, together with the big data team and the home appliance industry, established a training data set for the Natural Language Processing (NLP) team, which consisted of approximately 200 to 300 question-answer pairs, including knowledge questions and answers about home appliances. The product team marked whether there was an intention to control the device and whether feasible standard instructions were spoken, with both positive and negative samples.

[0111] 2. Household appliances knowledge test data set (for equipment control guidance):

[0112] In addition, prepare a test data set. The test method refers to Figure 8 It is used for NLP testing during the test, with about 200 to 300 question and answer pairs, including knowledge questions and answers about household appliances. The product team marks whether there is an intention to control the device and whether there are feasible standard instructions. There are both positive and negative samples.

[0113] 3. Intent extraction prompt and test:

[0114] Conduct offline intent extraction tests, run prompts in batches to traverse all the knowledge of current home appliances, find out the knowledge that meets the conditions, and test the prompt effect. NLP continues to optimize the prompt effect, and finally forms a stable, standardized output and prompt that meets the requirements of the home appliance industry. The effect is referenced Figure 7 .

[0115] 4. Prompts and tests generated from knowledge-based guidance:

[0116] Conduct an experiment on generating device control guidance scripts for household appliance knowledge question and answer pairs that meet the requirements, run prompts in batches, generate scripts with device control operation guidance, and test the prompt effect. The effect is referenced Figure 8 .

[0117] 5. Add cloud-based business processes to implement the logic of knowledge-based device control guidance:

[0118] (1) Knowledge response: Design a new cloud-based business process. In the knowledge database, if knowledge with the above labels is hit, the status of a series of current household devices will be checked to determine which speech to respond to the user. Among multiple speeches with the same intent, add a weight value of "user feedback". In the speech selection logic that cannot be excluded by the status check, use the speech with a high "user feedback" weight; if there is no hit in the knowledge base but it is indeed classified in the knowledge field, perform a fallback response and add the corpus to the "corpus pool to be optimized" to wait for a new round of HomeGPT parsing and generation;

[0119] (2) Response method: The speech presented to the user uses two presentation schemes. One scheme is to have a multi-round dialogue switch. When giving an answer, it adds the question "Do you want to execute" and turns on the pickup waiting. The other scheme is to have no multi-round dialogue. At the end of the speech, the suggestion is completed in the form of "You can say to me..." The switch between the two schemes depends on whether the subsequent user actually executes the cloud-based suggestions.

[0120] 6. Added a scoring mechanism for knowledge response + guidance of speech through user feedback:

[0121] (1) User feedback: Based on the actual situation of whether the user intends to use the recommendation in multiple rounds of conversation or within the next 24 hours, the corresponding weight or score is added to this knowledge reply + control guidance speech;

[0122] (2) Automatic operation: After the first five steps have been run for three to four periods of time and have achieved good results, the background will automatically run the process from b to e in accordance with the needs of knowledge classification. The new words generated will be manually reviewed and entered into the knowledge base. In actual use, the system will mark the weight value of "user feedback" based on user feedback.

[0123] It should be noted that this application also has the following advantages:

[0124] First, by using the intention extraction capability of the large model, we can more accurately obtain the user's explicit or implicit intentions. Whether the user's needs are direct or implicit, this model can be used to accurately extract them. In this way, we can more accurately understand the user's needs and provide users with more accurate services;

[0125] Secondly, according to the obviousness of the user's intention, different degrees of guidance and recommendation are carried out. For obvious user intentions, strong guidance will be carried out, such as directly providing solutions; for less obvious user intentions, weak guidance will be carried out, such as giving relevant recommendations. In this way, the personalized needs of users can be better met;

[0126] Finally, a user feedback mechanism was established. By collecting user feedback, we can understand user satisfaction with the service and suggestions for service improvement. This will enable us to continuously optimize services and improve user experience.

[0127] In other words, the application can more accurately extract and understand user needs, provide personalized services, and can continuously optimize based on user feedback, thereby greatly improving user satisfaction and usage experience.

[0128] Through the description of the above implementation methods, those skilled in the art can clearly understand that the method according to the above embodiment can be implemented by means of software plus a necessary general hardware platform, and of course can also be implemented by hardware, but in many cases the former is a better implementation method. Based on such an understanding, the technical solution of the present invention is essentially or the part that contributes to the prior art can be embodied in the form of a software product, which is stored in a storage medium (such as ROM / RAM, a magnetic disk, or an optical disk), and includes a number of instructions for a terminal device (which can be a mobile phone, a computer, a server, or a network device, etc.) to execute the methods of various embodiments of the present invention.

[0129] In the present embodiment, a device for determining a guide statement is also provided, and the device is used to implement the above-mentioned embodiment and preferred implementation mode, and the description has been made no further. As used below, the term "module" can implement a combination of software and / or hardware of a predetermined function. Although the equipment described in the following embodiments is preferably implemented in software, the implementation of hardware, or a combination of software and hardware is also possible and conceived.

[0130] Fig. 9 is a structural block diagram of a device for determining a guide sentence according to an embodiment of the present invention, the device comprising:

[0131] An acquisition module 92 is used to acquire the interaction statement of the target object;

[0132] A first determination module 94 determines the target intention corresponding to the interaction sentence and the degree of explicitness or implicitness of the target intention through a target interaction model, wherein the degree of explicitness or implicitness of the target intention is used to indicate the degree to which the target intention is reflected in the interaction sentence;

[0133] The second determination module 96 is used to determine the target guidance sentence according to the target intention and the degree of explicitness or implicitness of the target intention through the target interaction model.

[0134] The above-mentioned device determines the intention corresponding to the acquired interactive sentence and the degree of visibility of the intention through the target interaction model, and then determines the target guiding sentence according to the intention and the degree of visibility through the target interaction model. Since the intention and degree of visibility of the interactive sentence sent by the target object are taken into consideration when determining the guiding sentence, the accuracy of the interactive sentence in response to the target object is improved, thereby solving the problem of not being able to determine the guiding sentence well, and thus improving the user experience.

[0135] In an exemplary embodiment, the second determination module 96 is also used to determine multiple guiding sentences corresponding to the target intent through the target interaction model; through the target interaction model, determine the target guiding sentence from the multiple guiding sentences according to the degree of explicitness or implicitness of the target intent.

[0136] In an exemplary embodiment, the second determination module 96 is also used to determine, as the target guided sentence, a guided sentence having a guiding ability greater than or equal to a second preset threshold among the multiple guided sentences when the degree of visibility is used to indicate that the target intention is reflected in the interactive sentence to a degree greater than or equal to a first preset threshold; and to determine, as the target guided sentence, a guided sentence having a guiding ability less than the second preset threshold among the multiple guided sentences when the degree of visibility is used to indicate that the target intention is reflected in the interactive sentence to a degree less than the first preset threshold.

[0137] In an exemplary embodiment, the above-mentioned device also includes: a training module, used to obtain a training data set, wherein the training data set includes multiple groups of data, each group of data in the multiple groups of data includes: sample interaction sentences, the intention corresponding to the sample interaction sentences, the degree of explicitness or implicitness of the intention corresponding to the sample interaction sentences, and the guiding sentences corresponding to the sample interaction sentences; using the training data set to train a general interaction model, and when the general interaction model meets the convergence condition, determining the general interaction model that meets the convergence condition as the target interaction model, wherein when the general interaction model does not meet the convergence condition, the model parameters of the general interaction model are adjusted.

[0138] In an exemplary embodiment, the training module is also used to train the first network in the general interaction model using the training data set, and when the first network satisfies a first convergence condition, determine the first network that satisfies the first convergence condition as an intent recognition network, wherein, when the first network does not satisfy the first convergence condition, the model parameters of the first network are adjusted, and the intent recognition network is used to recognize the intent of a sentence and the degree of explicitness of the intent; use the training data set to train the second network in the general interaction model, and when the second network satisfies a second convergence condition, determine the second network that satisfies the second convergence condition as a guide sentence generation network, wherein, when the second network does not satisfy the second convergence condition, the model parameters of the second network are adjusted, and the guide sentence generation network is used to generate a guide sentence according to the intent of the sentence and the degree of explicitness of the intent; wherein the target interaction model includes the intent recognition network and the guide sentence generation network, and the convergence conditions include the first convergence condition and the second convergence condition.

[0139] In an exemplary embodiment, the above-mentioned device also includes: a feedback module, which is used to obtain feedback information of the target object after determining the target guidance sentence according to the target intention and the degree of explicitness of the target intention through the target interaction model; saving the feedback information, the interaction sentence, the target intention, the degree of explicitness of the target intention and the target guidance sentence as a piece of data into a target database; a training module, which is also used to use the data in the target database to train the target interaction model to update the target interaction module when the number of data items in the target database is greater than a third preset threshold.

[0140] In an exemplary embodiment, the above-mentioned device also includes: a playback module, which is used to voice play the target guidance sentence after determining the target guidance sentence through the target interaction model according to the target intention and the degree of explicitness of the target intention; monitor the response sentence of the target object; or broadcast a prompt sentence and monitor the response sentence of the target object, wherein the prompt sentence is used to prompt the target object to feedback a response sentence.

[0141] An embodiment of the present invention further provides a computer-readable storage medium, in which a computer program is stored, wherein the computer program is configured to execute the steps of any of the above method embodiments when running.

[0142] Optionally, in this embodiment, the storage medium may be configured to store a computer program for performing the following steps:

[0143] S1, obtain the target object’s interactive statements;

[0144] S2, determining the target intention and degree of explicitness and implicitness corresponding to the interactive sentence through a target interaction model, wherein the degree of explicitness and implicitness is used to indicate the degree to which the target intention is reflected in the interactive sentence, and the target intention is the intention expressed by the target object through the interactive sentence;

[0145] S3, determining a target guiding sentence according to the target intention and the degree of explicitness and implicitness through the target interaction model.

[0146] In an exemplary embodiment, the computer-readable storage medium may include, but is not limited to, various media that can store computer programs, such as a USB flash drive, a read-only memory (ROM), a random access memory (RAM), a mobile hard disk, a magnetic disk or an optical disk.

[0147] For specific examples in this embodiment, reference may be made to the examples described in the above embodiments and exemplary implementation modes, and this embodiment will not be described in detail herein.

[0148] An embodiment of the present invention further provides an electronic device, including a memory and a processor, wherein a computer program is stored in the memory, and the processor is configured to run the computer program to execute the steps in any one of the above method embodiments.

[0149] Optionally, in this embodiment, the processor may be configured to perform the following steps through a computer program:

[0150] S1, obtain the target object’s interactive statements;

[0151] S2, determining the target intention and degree of explicitness and implicitness corresponding to the interactive sentence through a target interaction model, wherein the degree of explicitness and implicitness is used to indicate the degree to which the target intention is reflected in the interactive sentence, and the target intention is the intention expressed by the target object through the interactive sentence;

[0152] S3, determining a target guiding sentence according to the target intention and the degree of explicitness and implicitness through the target interaction model.

[0153] In an exemplary embodiment, the electronic device may further include a transmission device and an input / output device, wherein the transmission device is connected to the processor, and the input / output device is connected to the processor.

[0154] For specific examples in this embodiment, reference may be made to the examples described in the above embodiments and exemplary implementation modes, and this embodiment will not be described in detail herein.

[0155] An embodiment of the present application further provides a computer program product, which includes a computer program. When the computer program is executed by a processor, the steps in any one of the above method embodiments are implemented.

[0156] An embodiment of the present application further provides another computer program product, including a non-volatile computer-readable storage medium, wherein the non-volatile computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps in any of the above method embodiments are implemented.

[0157] The embodiments of the present application also provide a computer program, which includes computer instructions stored in a computer-readable storage medium; a processor of a computer device reads the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions, so that the computer device performs the steps of any one of the above method embodiments.

[0158] Obviously, those skilled in the art should understand that the above modules or steps of the present invention can be implemented by a general computing device, they can be concentrated on a single computing device, or distributed on a network composed of multiple computing devices, they can be implemented by a program code executable by a computing device, so that they can be stored in a storage device and executed by the computing device, and in some cases, the steps shown or described can be executed in a different order than here, or they can be made into individual integrated circuit modules, or multiple modules or steps therein can be made into a single integrated circuit module for implementation. Thus, the present invention is not limited to any specific combination of hardware and software.

[0159] The above is only a preferred embodiment of the present application. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present application. These improvements and modifications should also be regarded as the scope of protection of the present application.

Claims

1. A method for determining a guide sentence, characterized in that: include: Get the interaction statement of the target object; Determining the target intention and degree of explicitness and implicitness corresponding to the interaction sentence through the target interaction model, wherein the degree of explicitness and implicitness is used to indicate the degree to which the target intention is reflected in the interaction sentence, and the target intention is the intention expressed by the target object through the interaction sentence; Through the target interaction model, the target guiding sentence is determined according to the target intention and the degree of explicitness and implicitness.

2. The method for determining a guiding sentence according to claim 1, characterized in that: Determining a target guiding sentence according to the target intention and the degree of explicitness and implicitness through the target interaction model includes: Determining a plurality of guiding sentences corresponding to the target intention through the target interaction model; The target guiding sentence is determined from the plurality of guiding sentences according to the degree of explicitness and invisibility through the target interaction model.

3. The method for determining a guiding sentence according to claim 2, characterized in that: Determining the target guiding sentence from the plurality of guiding sentences according to the degree of visibility includes: In the case where the degree of visibility is used to indicate that the degree to which the target intention is reflected in the interactive sentence is greater than or equal to a first preset threshold, determining a guide sentence whose guiding ability is greater than or equal to a second preset threshold among the multiple guide sentences as the target guide sentence; When the degree of visibility is used to indicate that the degree to which the target intention is reflected in the interactive sentence is less than the first preset threshold, the guiding sentence whose guiding ability among the multiple guiding sentences is less than the second preset threshold is determined as the target guiding sentence.

4. The method for determining a guiding sentence according to claim 1, characterized in that: The method further includes: acquiring a training data set, wherein the training data set includes multiple groups of data, each group of the multiple groups of data includes: a sample interaction sentence, an intention corresponding to the sample interaction sentence, a degree of explicitness or implicitness of the intention corresponding to the sample interaction sentence, and a guide sentence corresponding to the sample interaction sentence; The general interaction model is trained using the training data set, and when the general interaction model satisfies a convergence condition, the general interaction model that satisfies the convergence condition is determined as the target interaction model, wherein when the general interaction model does not satisfy the convergence condition, the model parameters of the general interaction model are adjusted.

5. The method for determining a guiding sentence according to claim 4, characterized in that: The method further comprises: training a universal interaction model using the training data set, and determining the universal interaction model satisfying the convergence condition as the target interaction model when the universal interaction model satisfies the convergence condition, comprising: Using the training data set to train the first network in the general interaction model, and when the first network satisfies a first convergence condition, determining the first network that satisfies the first convergence condition as an intent recognition network, wherein when the first network does not satisfy the first convergence condition, the model parameters of the first network are adjusted, and the intent recognition network is used to recognize the intent of a sentence and the degree of explicitness of the intent; The second network in the general interaction model is trained using the training data set, and when the second network satisfies a second convergence condition, the second network satisfying the second convergence condition is determined as a guide sentence generation network, wherein when the second network does not satisfy the second convergence condition, the model parameters of the second network are adjusted, and the guide sentence generation network is used to generate a guide sentence according to the intention of the sentence and the degree of explicitness of the intention; Among them, the target interaction model includes the intention recognition network and the guide sentence generation network, and the convergence condition includes the first convergence condition and the second convergence condition.

6. The method for determining a guiding sentence according to claim 1, characterized in that: After determining the target guiding sentence according to the target intention and the degree of explicitness and implicitness through the target interaction model, the method further includes: Obtaining feedback information from the target object; The feedback information, the interactive sentence, the target intention, the degree of explicitness and the target guiding sentence are saved as one piece of data in a target database; When the number of data items in the target database is greater than a third preset threshold, the target interaction model is trained using the data in the target database to update the target interaction module.

7. The method for determining a guiding sentence according to claim 1, characterized in that: After determining the target guiding sentence according to the target intention and the degree of explicitness and implicitness through the target interaction model, the method further includes: Voice playing of the target guiding sentence; Monitor the response statement of the target object; or A prompt statement is broadcasted, and a response statement of the target object is monitored, wherein the prompt statement is used to prompt the target object to feedback a response statement.

8. A device for determining a guide sentence, characterized in that: include: An acquisition module is used to obtain the interaction statements of the target object; A first determination module determines the target intention and the degree of explicitness and implicitness corresponding to the interactive sentence through a target interaction model, wherein the degree of explicitness and implicitness is used to indicate the degree to which the target intention is reflected in the interactive sentence, and the target intention is the intention of the target object to execute through the interactive sentence; The second determination module determines the target guiding sentence according to the target intention and the degree of explicitness and implicitness through the target interaction model.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium includes a stored program, wherein the program executes the method according to any one of claims 1 to 7 when executed.

10. An electronic device comprising a memory and a processor, characterized in that: A computer program is stored in the memory, and the processor is configured to execute the method according to any one of claims 1 to 7 through the computer program.

Images