Information recommendation method and device, electronic equipment and storage medium

Abstract

The present disclosure provides an information recommendation method and device, electronic equipment and storage medium, relates to the technical field of data processing, and particularly relates to the technical field of artificial intelligence, large models, deep learning, intelligent driving, intelligent recommendation and the like. The specific implementation scheme is as follows: in response to a scene demand of a target object for a target vehicle, a constraint condition required to meet the scene demand is obtained; prompt information of a large model is constructed based on the constraint condition; the prompt information is input to the large model to obtain at least one execution item recommended by the large model for the target vehicle; and the scene demand of the target object is responded to based on the at least one execution item. The embodiments of the present disclosure can automatically recommend an execution item meeting a scene demand for assisting a user in controlling a vehicle for any scene demand.

Description

TECHNICAL FIELD

[0001] The present disclosure relates to the technical field of data processing, and particularly relates to the technical fields of artificial intelligence, large models, deep learning, intelligent driving, intelligent recommendation, etc. BACKGROUND

[0002] With the continuous development of technology, more and more execution items are supported by vehicles to meet the different needs of users. However, due to the increasing number of execution items, many users do not know how to use these execution items, or even are not aware of the existence of the corresponding execution items.

[0003] In order to improve the user experience, in the use scene of the vehicle by the vehicle driver, the related technology will artificially set a combination rule to help the user use the vehicle. In the combination rule, the artificial will set the corresponding execution item for different situations. When the trigger condition defined in the combination rule is met, the user will be recommended the pre-set execution item to assist the user to better use the vehicle. SUMMARY

[0004] The present disclosure provides an information recommendation method and device, an electronic device, and a storage medium.

[0005] According to an aspect of the present disclosure, an information recommendation method is provided, comprising:

[0006] In response to a scene demand of a target object for a target vehicle, obtaining a constraint condition required to meet the scene demand;

[0007] Constructing prompt information of a large model based on the constraint condition;

[0008] Inputting the prompt information to the large model to obtain at least one execution item recommended by the large model for the target vehicle;

[0009] Responding to the scene demand of the target object based on the at least one execution item.

[0010] According to another aspect of the present disclosure, an information recommendation device is provided, comprising:

[0011] An obtaining module, configured to obtain a constraint condition required to meet a scene demand of a target object for a target vehicle in response to the scene demand;

[0012] A constructing module, configured to construct prompt information of a large model based on the constraint condition;

[0013] A generating module, configured to input the prompt information to the large model to obtain at least one execution item recommended by the large model for the target vehicle;

[0014] A responding module, configured to respond to the scene demand of the target object based on the at least one execution item.

[0015] According to another aspect of the present disclosure, an electronic device is provided, comprising:

[0016] at least one processor; and

[0017] a memory connected with the at least one processor in communication; wherein

[0018] the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the method described above.

[0019] According to another aspect of the present disclosure, a non-transitory computer readable storage medium storing computer instructions is provided, wherein the computer instructions are used to enable the computer to perform the method described above.

[0020] According to another aspect of the present disclosure, a computer program product is provided, comprising a computer program which, when executed by a processor, implements the method described above.

[0021] The embodiments of the present disclosure can automatically recommend an execution item meeting the scene requirement to assist the user in controlling the vehicle for any scene requirement.

[0022] It should be understood that the content described in this part is not intended to identify key or important features of the embodiments of the present disclosure, nor to limit the scope of the present disclosure. Other features of the present disclosure will become apparent from the following description. BRIEF DESCRIPTION OF DRAWINGS

[0023] The accompanying drawings are used to better understand the present scheme and do not limit the present disclosure. Among them:

[0024] Figure 1 is a flowchart of an information recommendation method according to an embodiment of the present disclosure;

[0025] Figure 2 is a flowchart of the interaction between the server, the target vehicle and the large model according to an embodiment of the present disclosure;

[0026] Figure 3 is a schematic diagram of the specific process of information recommendation according to an embodiment of the present disclosure;

[0027] Figure 4 is a structural schematic diagram of an information recommendation device according to an embodiment of the present disclosure;

[0028] Figure 5 is a block diagram of an electronic device for implementing the information recommendation method according to an embodiment of the present disclosure. DETAILED DESCRIPTION

[0029] Exemplary embodiments of the present disclosure are described herein with reference to the accompanying drawings, in which various details are set forth to facilitate an understanding of the present disclosure. It should be appreciated that various embodiments of the present disclosure can be practiced with variation of details and / or without using all of the available details. Accordingly, it should be understood that various embodiments of the present disclosure can be practiced without resorting to detail(s) not specifically set forth in the following description. For the purpose of clarity, technical concepts will be briefly described and descriptions of well-known functions and constructions will be omitted so as not to unnecessarily obscure the subject matter of the present disclosure.

[0030] In addition, in order to better illustrate the scheme provided by the embodiments of the present disclosure, a large number of specific details are given in the specific embodiments below. Those skilled in the art should understand that the embodiments of the present disclosure can also be implemented without some specific details. In some examples, methods, means, elements and circuits that are well known to those skilled in the art are not described in detail in order to highlight the main idea of the embodiments of the present disclosure.

[0031] As described above, the scene requirements of the vehicle are met by setting a combination rule. That is, when one or more conditions are met, the user is recommended a plurality of recommended contents set in advance by the user to assist the user in using the vehicle. For example, the user says "navigate to an address", and the vehicle will automatically plan a route to the address.

[0032] In the related art, the combination rule needs to be based on conditions for exhaustive enumeration, and when the preset trigger condition is met, the recommended item is given. However, this approach has the following problems:

[0033] First, when there is a slight difference from the trigger condition set by the rule, the recommended item cannot be hit. For example, the user's instruction has one more condition item, one less condition item, or similar but not identical condition items than the trigger condition, and then the recommended execution item is not obtained.

[0034] Second, because it needs to be manually enumerated, the configuration of all scenes depends on the understanding of the scene by the operation designer, and scenes that are not set will not hit the appropriate execution item.

[0035] Third, the level of the operation designer of the scene is uneven, which will affect the final result.

[0036] In addition, on the one hand, the user does not know how to use the vehicle components or the application software functions in the center screen to achieve their own needs or solve existing problems; on the other hand, it is not clear which components or software functions can achieve their own needs or solve existing problems, resulting in that the user's scene requirements are not well met.

[0037] Therefore, in the embodiments of the present disclosure, an information recommendation method is proposed, which meets the user's scene requirements through the processing capacity of the large model for information.

[0038] Large model generally refers to a model with a large number of parameters or complex structures in computer science. These models can be used for various tasks such as image recognition, natural language processing, etc. For example, a deep neural network is a common large model.

[0039] The large model used in the embodiments of the present disclosure can be a large language model, which refers to a deep learning model trained using a large amount of text data and can generate natural language text or understand the meaning of language text. The large language model can handle a variety of natural language tasks such as text classification, question answering, dialogue, etc.

[0040] In addition, the large model in the embodiments of the present disclosure can also be a large model for processing images, so that the user can use body language to describe the scene requirements.

[0041] In another possible implementation, the large language model and the large model for processing images can be combined in the embodiments of the present disclosure to implement information recommendation, so that the user's scene requirements can be analyzed comprehensively by combining the user's body language and written language, and the recommended execution items can be given.

[0042] As Figure 1 shown, the information recommendation method provided by the embodiments of the present disclosure includes the following contents:

[0043] S101, in response to the scene requirements of the target object for the target vehicle, obtaining the constraint conditions required to meet the scene requirements.

[0044] In some embodiments, the user can express his own requirements, questions or appeals, which will be referred to as scene requirements in the embodiments of the present disclosure. For example, the user can say "what should I do if I am tired", "how can I feel better if I am car-sick", "I want to be more romantic", etc. Therefore, the scene requirements in the embodiments of the present disclosure can be vague requirements for the target vehicle.

[0045] The constraint conditions in the embodiments of the present disclosure can be some boundary restrictions required to meet the user's scene requirements, so as to constrain the large model to reasonably plan the execution items. For example, it can be the temperature outside the vehicle, so as to reasonably control the temperature inside the vehicle to improve the user experience.

[0046] S102, constructing prompt information of the large model based on the constraint conditions.

[0047] The prompt information can help the large model better understand the task to be completed and make a reasonable response to meet the user's scene requirements.

[0048] S103, inputting the prompt information to the large model to obtain at least one execution item recommended by the large model for the target vehicle.

[0049] S104, constructing a prompt information based on the at least one execution item.

[0050] In the embodiments of the present disclosure, by obtaining the constraint condition required to meet the scene demand and using the constraint condition to construct the prompt information, the large model can be helped to reasonably understand and process the scene demand, so as to plan a reasonable recommended execution item for execution. The data analysis and inference ability of the large model can be used to process the open scene demand, without the need for artificial setting of combined rules, and the large model can feed back to various scene demands proposed by the user and reasonably give recommended execution items to meet the scene demand of the user. In addition, relying on the extensive knowledge reserve and processing speed of the large model, the large model can also infer and recommend more creative, highlight, unexpected and more specific recommended execution items.

[0051] In implementation, the large model can give the recommended reason and the corresponding scene title when giving the execution item, so as to better understand the recommended result by the target object and improve the user experience.

[0052] In the following, the embodiments of the present disclosure will be described from three aspects of constructing a prompt information, generating an execution item and optimizing a prompt information to explain the key content of completing the vehicle execution item recommendation based on the large model.

[0053] 1) Constructing a prompt information

[0054] In order to enable the large model to accurately understand and process the scene demand of the target object and improve the user experience, the constraint condition required to be obtained in the embodiments of the present disclosure can include at least one of the following:

[0055] a. The description information of the target object is a boundary condition set for the large model based on the characteristics of the target object itself. The description information of the target object is used for the large model to understand the preferences and personalized needs of the target object, so as to reasonably recommend suitable execution items for the target object, so that the vehicle can better provide services for the target object and improve the comfort of the target object.

[0056] The description information of the target object can include personalized information of the target object, such as user portrait. It should be noted that all the obtained user self information is obtained after the user is informed and agrees on the basis of strictly complying with relevant laws and regulations. At the same time of protecting the personal data security of the user, it will not be used for any unauthorized purpose.

[0057] b. Vehicle configuration information of the target vehicle.

[0058] In the embodiments of the present disclosure, the configuration information of the vehicle can be high, medium, and low configuration information of the target vehicle. The vehicle configuration information is a boundary condition set by the capability of the vehicle itself for the large model. Based on the vehicle configuration information of the target vehicle, the large model can understand the capability range of the target vehicle itself, so as to make reasonable recommendations and improve the accuracy of the recommendations.

[0059] c. Position information of the target vehicle.

[0060] As the name implies, the position information of the target vehicle is the position of the target vehicle. With the position information, the large model can understand the location of the target vehicle, so as to make recommendations for the target object in combination with the situation of the location, so as to improve the user experience. For example, the local customs and climate characteristics can be used to provide the user with an execution item. This can help users who are traveling to experience the local characteristics even in the car, so as to improve the user experience. For another example, according to the geographical and topographical conditions of the location, the driving mode can be reasonably set, so as to improve the comfort and safety of driving.

[0061] d. Environment information of the environment in which the target vehicle is located.

[0062] For example, the environment information can be condition information based on active perception, such as the weather condition of the day, the temperature of the day, the congestion condition of the road, etc.

[0063] Based on this information, the large model can understand the environment around the target object, so as to help the large model recommend reasonable execution items for the target object based on its own knowledge, so as to improve the user experience.

[0064] e. Position information of the target object in the target vehicle.

[0065] For example, the target object is located in the front right or the back left in the vehicle. By identifying the position of the target object in the vehicle, more personalized and customized services can be provided to users in different positions, so as to optimize the service experience. For example, different seats can have different capabilities, and by determining the position of the target object in the target vehicle, the execution item can be better recommended for the target object.

[0066] In summary, in the embodiments of the present disclosure, by obtaining the description information of the target object, the vehicle configuration information, the position information of the target vehicle, and the environment information of the environment in which the target vehicle is located, the situation of the target object can be described comprehensively and accurately, so that the large model can more accurately understand and meet the needs of the user, thereby further improving the user experience.

[0067] In some embodiments, based on the constraint conditions, the large model can better understand the needs of the user. In order to further improve the recommendation ability of the large model, in the embodiments of the present disclosure, the prompt information can additionally include at least one of the following information:

[0068] (1) Inference method for large model reference;

[0069] In the inference method, exemplary inference steps are included, and matters needing attention for at least one inference step; the matters needing attention can include at least one of the following information:

[0070] f. Analysis method of scene requirement.

[0071] The analysis method specifies the type of information obtained from the scene requirement, and can also include the format (such as Jason format) of the information type. For example, when the scene requirement is expressed in text, entity recognition can be performed on the scene requirement to obtain key information of the expected type, to help the large model filter out secondary information and retain the core content. In the case of expressing scene requirements through body movements, the action names corresponding to the body movements can be extracted from the body movement images, and the action names can be arranged in time sequence according to the order of generating the actions, and a text expression of the series of actions can also be generated to help the large model understand the scene requirement.

[0072] In short, here the analysis method can define how the large model parses the desired target content from the scene requirement. In practice, it can be set according to actual conditions, and the embodiments of the present disclosure do not limit this.

[0073] g. Examples of mutually exclusive execution items.

[0074] Mutually exclusive execution items refer to the case where two or more operations cannot be performed at the same time, that is, mutually exclusive execution items can be understood as execution items that violate common sense when executed simultaneously. For example, when driving in summer, simultaneously opening the car window completely and turning on the air conditioner to cool is against common sense. In specific implementation, it can be set according to actual conditions.

[0075] According to the examples of mutually exclusive execution items, the large model can help reduce confusion and misunderstanding, so that the generated execution items are more accurate and consistent, which helps to improve user experience.

[0076] h. Care knowledge of specified object.

[0077] Among them, the care knowledge is the knowledge required for caring for the specified object according to nursing science. These knowledge in the embodiments of the present disclosure will affect how to control the vehicle, so that the control of the vehicle can better adapt to the physical needs of the specified object.

[0078] Different target object groups should include different relevant knowledge. For example, the elderly, infants, and pregnant women all need additional care. Taking pregnant women as an example, the care knowledge can include the need to keep the air in the car fresh: try not to smoke or use perfume and other irritating odor items. Accordingly, when recommending execution items, the large model can consider the needs of pregnant women and reasonably set the following execution items, including: opening the car window for ventilation, and controlling the temperature in the car within an appropriate range, generally recommended between 20-25 degrees Celsius.

[0079] Based on this care knowledge, when facing different object groups, the large model can help better understand the knowledge related to it, thereby further improving the recommendation ability of the large model and enhancing the user experience.

[0080] i. Function explanation of execution item.

[0081] The function explanation of the execution item, as the name implies, is to explain the operation that the execution item can perform, and the direct or indirect benefits brought by performing these operations. These function explanations help the large model understand the functional characteristics of the execution item, so as to recommend the execution item that fits the user's needs.

[0082] Reasonable function explanations are given for execution items. For example, for the pregnant woman group, the generated execution item is to set the car air conditioner temperature to 23 degrees Celsius, and the function explanation corresponding to this execution item is that the increase in the body temperature of the pregnant woman can cause fetal hypoxia. For another example, the seat massage function can provide different massage methods for different parts. To help the large model better understand the capabilities of the execution item.

[0083] By giving reasonable function explanation, the large model can continuously learn and accumulate knowledge, improve its understanding of the execution item, and make the execution item generated by the large model more reasonable.

[0084] In summary, in the embodiments of the present disclosure, by explicitly focusing on the content of the matter of concern, the large model can better understand the scene requirements, exclude incorrect logic, and have more knowledge systems, thereby having more scene recommendation capabilities and better meeting the needs of users.

[0085] (2) Recommendation library supported by target vehicle.

[0086] The recommendation library includes a set of execution items supported by the target vehicle, or the recommendation library can include a set of execution items and media resources.

[0087] For example, for vehicles with different configurations, the vehicle user manual information of the vehicle configuration is obtained to build the recommendation library.

[0088] The execution item set supported by the target vehicle is constructed as a recommendation library to help the large model generate recommended execution items with information support, so as to ensure that the generated execution items are accurate and reliable.

[0089] In addition, some music resources, video resources and the like supported by the vehicle can be constructed as a recommendation library based on user needs. Based on the constructed media resources, the large model can understand the capabilities of the target vehicle in terms of media resources, so that the generated execution items are more personalized to improve user experience.

[0090] To sum up, in the embodiments of the present disclosure, a standardized or scalable recommendation library is provided to help the large model more accurately understand the functions and characteristics of the target vehicle and recommend more suitable information for different user needs.

[0091] In the embodiments of the present disclosure, by providing exemplary inference steps and matters needing attention, more information and guidance are provided for the large model, which can help the large model more clearly understand the key points and problems needing attention in the inference process, thereby improving the accuracy and efficiency of the inference.

[0092] 2) Responding to user needs based on execution items

[0093] The large model has extensive knowledge and can perform inference and deduction. However, the large model does not recommend in a closed manner with combinational rules, so initially, the large model may give inappropriate execution items after inference.

[0094] Therefore, based on at least one execution item responding to the scene needs of the target object, the following can be implemented:

[0095] Step A1, filtering execution items that meet the preset conditions from at least one execution item.

[0096] Specifically, execution items that meet the filtering conditions can be filtered from at least one execution item to obtain execution items that meet the preset conditions:

[0097] Among them, the execution items that meet the filtering conditions include at least one of the following:

[0098] (1) Execution items that violate vehicle usage common sense.

[0099] For example, during the driving of the target vehicle, the target object proposes the demand "a little stuffy, what should I do", in order to alleviate the discomfort of the user, the sunroof can be opened for ventilation, but at the same time, if it is raining, the sunroof cannot be opened. At this time, it can be more appropriate to open a small gap in the window.

[0100] The execution item that violates the common sense of using the vehicle is taken as a filtering condition, and a more humanized execution item is selected for recommendation, so as to improve the user experience.

[0101] (2) The execution item with an abnormal output format.

[0102] The standard output format of the execution item can use the JSON format, and any format that does not conform to the format is abnormal.

[0103] In addition, in the following, the output content will include "action (execution item)", "reason (recommended reason)", and "title (scene title)" as examples to introduce the output result of the large model. Therefore, each parameter of the output has corresponding format requirements, such as naming method, parameter value, and parameter quantity.

[0104] By setting the standard output format, the format of the execution item generated by the large model can be constrained within the desired range, improving the readability and subsequent operability of the execution item, and also improving the rationality of the recommended result, so as to improve the user experience.

[0105] The specific format requirements can be set according to the needs, and the embodiments of the present disclosure do not limit this.

[0106] (3) The execution item outside the capability range of the target vehicle.

[0107] The execution item that exceeds the support range cannot be generated. In this way, it is ensured that the execution item generated by the large model is executable by the target vehicle. For example, the execution item generated in a low-config target vehicle is a configuration that only high-config target vehicles have, which will reduce the user experience of using the vehicle.

[0108] In summary, in the embodiments of the present disclosure, by setting the preset condition, the execution item that meets the filtering condition is filtered out, and the execution item that meets the demand of the target object is automatically selected, and the inappropriate option is excluded. In this way, the accuracy and efficiency of the recommendation are improved, and the trouble caused by the inappropriate or invalid execution item is avoided.

[0109] After obtaining the execution item that meets the preset condition, in step A2, the execution item that meets the preset condition is recommended to the target object.

[0110] In some embodiments, the target object can issue a scene demand to the vehicle machine of the target vehicle through voice, and then the finally selected execution item is displayed on the main control screen corresponding to the vehicle machine.

[0111] Similarly, the target object can also identify the position of the target object in the target vehicle based on the target recognition device in the target vehicle. For example, the position of the target object in the target vehicle can be identified through a microphone array, and the recommended execution item can be displayed in the corresponding position area. If there are screens in front of multiple seats in the vehicle, after identifying the target object in a specific seat position, the execution item can be displayed on the screen in front of the seat where the target object is located, so as to facilitate the subsequent operation of the user.

[0112] Alternatively, in the case of supporting the control of the vehicle by the mobile terminal, the recommended execution item can also be displayed in the mobile terminal of the target object, so as to facilitate the control of the target object.

[0113] In the embodiments of the present disclosure, the execution item meeting the demand is automatically screened according to the preset condition and recommended to the target object. In this way, the rationality of the execution item is improved, the accuracy of the recommended result is improved, and a more user-demand-conforming recommendation service is provided.

[0114] In addition to the touch mode, the target object can control the recommended execution item through voice interaction or body language. The embodiments of the present disclosure can also modify the execution item meeting the preset condition in response to the modification demand of the target object. For example, in the initial stage, the recommendation result of the large model may not be able to very accurately meet the user's preference. Among the generated execution items, there may be a case that the recommended execution item does not completely meet the demand or expectation of the target object, and the target object can make a modification request, and then the corresponding execution item will be modified to meet the new demand.

[0115] In the embodiments of the present disclosure, the recommended execution item is flexibly adjusted and optimized according to the modification demand of the target object. In this way, a more personalized, user-demand-conforming recommendation service is provided, and the user experience and satisfaction are enhanced.

[0116] In some embodiments, based on professional knowledge, the large model can also output the recommendation reason of at least one execution item at the same time, and recommend the execution item meeting the preset condition and the recommendation reason to the target object.

[0117] For example, the scenario demand proposed by the target object is "I am very tired", and the generated execution item can be opening the window, playing energetic music, and the recommendation reason corresponding to opening the window is that fresh air and oxygen can stimulate the brain, making people feel more awake and alert, thereby reducing the feeling of fatigue. The recommendation reason for playing energetic music is that energetic music usually has fast rhythm and strong rhythm, and such rhythm can stimulate the brain of people, making people feel excited and energetic.

[0118] In the embodiments of the present disclosure, by providing the recommendation reason, the target object can understand why a specific execution item is recommended, so that the decision-making process is more explicit. In addition, the target object can better understand the recommendation result through the recommendation reason, so that the target object may be more willing to adopt the recommendation, thereby improving the effectiveness and satisfaction of the recommendation.

[0119] To help the target object quickly understand the output execution item, in the embodiments of the present disclosure, the large model also outputs a scene title for the recommended execution item; so as to store the scene title and the setting result of the target object for at least one execution item in the target address, wherein the setting result is the execution item finally adopted by the user; then, when the scene requirement corresponding to the scene title is obtained again, the setting result saved in the target address is used to respond to the scene requirement.

[0120] For example, the scene requirement proposed by the target object is "I am very sleepy, what should I do", and the scene title generated for it is "sleepy scene". The scene title and the corresponding execution item are saved in the recommendation library. When the target object proposes the scene requirement again, "I am very sleepy, what should I do", the large model can directly execute the execution item corresponding to the scene title saved in the recommendation library.

[0121] Among them, the scene title and the corresponding execution item saved in the recommendation library can be modified according to the user's requirement, and the modified scene title and the corresponding execution item will overwrite the original content. When the user proposes the modified scene requirement next time, the execution item called by the large model is the modified content.

[0122] In the embodiments of the present disclosure, by outputting the scene title for at least one execution item, the target object can more intuitively understand the generated execution item. Based on the saved setting response to the scene requirement, it is helpful to improve the intelligent degree of the large model, so that it can make more suitable recommendation information for the user when encountering the same scene requirement next time, and improve the user experience.

[0123] 3) Optimize the prompt information

[0124] Based on the foregoing description, it can be known that the prompt information can guide the large model to make reasonable reasoning and give appropriate recommendation results. Therefore, the accuracy of the recommendation results of the large model can be increased by continuously optimizing the prompt information.

[0125] In order to more effectively optimize the output content of the large model to meet the user's requirement, based on the generated execution item, the recommendation requirement in the prompt information can include the following steps:

[0126] Step B1, based on at least one requirement in the prompt information for the output content of the large model, determine the matching degree of at least one execution item and at least one requirement.

[0127] Step B2, determining a recommendation score of the current recommendation task based on the matching degree of the at least one execution item and the at least one requirement.

[0128] For example, in constructing the prompt information, the requirement output format can be a requirement. For example, the JSON format includes a scene title, an execution subject, and a recommendation reason. If the generated execution item is missing the scene title, the matching degree of the generated execution item and the requirement is low.

[0129] For another example, the generated execution items cannot be mutually exclusive, which can also be a requirement. The generated execution items cannot violate the car usage common sense, which is another requirement. Therefore, the information in the prompt information that limits the generated result can be a requirement for the execution item.

[0130] In the embodiments of the present disclosure, the recommendation score is calculated by weighting and summing the matching degrees of the generated execution items and the requirements. For example, there are three execution items A, B, and C, which can be used to determine the matching degrees with the requirements, and then the recommendation score is obtained by weighting and summing based on the weight factors of the requirements.

[0131] Of course, the recommendation score of each execution item can also be calculated, and then the average value is obtained to get the final recommendation score of the entire recommendation result. The specific calculation method is not limited in the embodiments of the present disclosure.

[0132] Step B3, the recommendation score of the current recommendation task is summarized into the recommendation task set;

[0133] Step B4, based on the recommendation task set, the recommendation task to be optimized whose recommendation score is lower than the preset threshold is screened out to obtain a recommendation task set;

[0134] Step B5, based on the recommendation task set to be optimized, the recommendation requirement in the prompt information is constructed.

[0135] In the embodiments of the present disclosure, by determining the matching degree of the execution item and the corresponding requirement, the recommendation score is summarized into the recommendation task set, the recommendation task to be optimized whose recommendation score is lower than the preset threshold is screened out, and the new recommendation requirement is constructed based on the recommendation task set to be optimized. The value of each recommendation task can be more accurately evaluated, which helps to comprehensively understand the overall performance of all recommendation tasks. The recommendation requirement formed thereby can enable the large model to focus on the parts that need to be improved, and can help the large model to better reason when facing similar problems, thereby improving the accuracy of the recommendation and the user satisfaction.

[0136] In the process of constructing the prompt information of the large model, correct examples can be given to make the large model generate execution items that meet the requirements. Correspondingly, in the embodiments of the present disclosure, the large model can also learn from the reverse error examples to optimize the recommendation results. Specifically, based on the set of recommendation tasks to be optimized, the recommendation requirements in the prompt information can be implemented as:

[0137] Step C1, filtering out recommendation examples with recommendation effects lower than the target effect from the set of recommendation tasks to be optimized as recommendation counterexamples;

[0138] In some embodiments, a large number of known scene requirements can be classified to obtain a set of recommendation tasks to be optimized under each scene requirement. For each set of recommendation tasks to be optimized under a scene requirement, a typical counterexample under the scene requirement is obtained. In order to enable the large model to further learn and infer based on the constructed counterexamples, in step C2, the recommendation requirements for the large model to learn error examples are constructed based on the recommendation counterexamples and added to the prompt information, so as to achieve the purpose of optimizing the recommendation results.

[0139] In the embodiments of the present disclosure, by constructing counterexamples, the large model can more accurately understand which recommendations do not meet the requirements, which helps the model to more comprehensively understand the scene requirements, optimize the reasoning and derivation process, and generate more accurate and reasonable execution items.

[0140] In the embodiments of the present disclosure, the information recommendation method provided can be executed by the target vehicle or completed by the target vehicle interacting with the server.

[0141] In the case of interaction, as shown in Figure 2 The target vehicle can collect the scene requirements of the target object and provide the constraint conditions required for constructing the prompt information, including user portrait, vehicle configuration information, vehicle location information, and surrounding environment information.

[0142] Based on the scene requirements and the constraint conditions, a recommendation request is constructed and sent to the server. The server can respond to the recommendation request, construct the prompt information, and then call the large model for processing to obtain the recommendation results of the large model. After reasonable filtering and screening, the final recommendation results are obtained. The recommendation results include execution items, reasons for recommending the execution items, and scene titles. The recommendation request of the target vehicle can be responded based on the recommendation results. The target vehicle displays the recommendation results and interacts with the target object after obtaining the recommendation results.

[0143] In order to facilitate the systematic understanding of the information recommendation method provided by the embodiments of the present disclosure, the method will be described below. Figure 3

[0144] As shown in Figure 3 ​As shown, the information input of the large language model includes user demand, such as user voice: "I am very sleepy, what should I do"; and multi-dimensional perception of the vehicle, such as vehicle location information, in-vehicle environment, user portrait, etc. In addition, the information of the vehicle is also part of the input of the large model, such as obtaining the vehicle configuration from the knowledge base according to the vehicle model, and the voice control items supported by the vehicle.

[0145] Based on the above information, the Prompt of the large language model is constructed, the Prompt information is brought into the large language model, and the large language model performs reasoning to generate recommended titles, recommended execution items, and recommended reasons through demand management and scene management. After post-processing operations (such as format processing, exception handling, and excluding execution items that are mutually exclusive and vehicle capabilities), the vehicle configuration that is not supported by the target vehicle or the voice control item that is not supported by the vehicle is filtered out, and the final recommended execution item is generated. The recommended execution item can be displayed in categories, for example, including:

[0146] Voice execution item, that is, an execution item that supports voice interaction;

[0147] Large model generation item, such as pictures and videos generated by the large model, which can guide the user to understand the capabilities or operation modes of the execution item, or guide the user to understand the operation to be performed;

[0148] Voice broadcast item, that is, content that needs to be broadcast by voice;

[0149] Manual operation item, that is, an item that needs to be manually operated by the user;

[0150] Information service item, that is, some service information, such as explanations of the title, execution item, etc.

[0151] The recommended execution item is displayed to the target object while being saved to the recommendation library. When the same user demand is received again, the saved recommended display item can be directly called. In this process, the execution item that has been opened does not need to be recommended again.

[0152] In the embodiments of the present disclosure, some general training samples can be pre-calibrated, and some training samples with user preference characteristics and some expert set training samples can also be included. These training samples include scene demand and its corresponding reasonable execution item, recommended reason, and scene title. Vehicle information and corresponding professional knowledge are provided to fine-tune the pre-trained large model to obtain a large model that can reasonably recommend execution items.

[0153] For example, Figure 3As shown, when the artificial expert has a preference requirement for scene execution, the scene data can also be provided, combined with the constraint conditions obtained to meet the scene requirements, and used for model training or model fine-tuning of the large language model to generate the final execution item, so that the large model can generate an execution item that meets a certain preference.

[0154] Based on the same technical concept, the present embodiment also provides an information recommendation device 400, which comprises Figure 4 As shown, comprising:

[0155] The acquisition module 401 is configured to acquire constraint conditions required to meet the scene requirements of the target object for the target vehicle;

[0156] The construction module 402 is configured to construct prompt information of the large model based on the constraint conditions;

[0157] The generation module 403 is configured to input the prompt information into the large model to obtain at least one execution item recommended by the large model for the target vehicle;

[0158] The response module 404 is configured to respond to the scene requirements of the target object based on the at least one execution item.

[0159] In some embodiments, the constraint conditions include at least one of the following:

[0160] The description information of the target object;

[0161] The vehicle configuration information of the target vehicle;

[0162] The position information of the target vehicle;

[0163] The environmental information of the environment in which the target vehicle is located;

[0164] The position information of the target object in the target vehicle.

[0165] In some embodiments, the prompt information includes at least one of the following information:

[0166] The inference method for the large model to refer to; the inference method includes exemplary inference steps and matters needing attention for at least one inference step;

[0167] The recommendation library supported by the target vehicle.

[0168] In some embodiments, the matters needing attention include at least one of the following information:

[0169] The analysis method of the scene requirement;

[0170] Examples of mutually exclusive execution items;

[0171] The care knowledge of the designated object;

[0172] The function explanation of the execution item.

[0173] In some embodiments, the recommendation library includes a set of execution items supported by the target vehicle; or,

[0174] The set of execution items and the media resource are included in the recommendation library.

[0175] In some embodiments, the response module includes:

[0176] The screening unit is configured to screen the execution item that meets the preset condition from the at least one execution item;

[0177] The recommendation unit is configured to recommend the execution item that meets the preset condition to the target object.

[0178] In some embodiments, the screening unit is specifically configured to:

[0179] Filter out the execution item that meets the filtering condition from the at least one execution item to obtain the execution item that meets the preset condition;

[0180] The execution item that meets the filtering condition includes at least one of:

[0181] The execution item that violates the car usage common sense;

[0182] The execution item with an abnormal output format;

[0183] The execution item beyond the capability range of the target vehicle.

[0184] In some embodiments, the device further includes:

[0185] The modification module is configured to modify the execution item that meets the preset condition in response to a modification demand of the target object.

[0186] In some embodiments, the large model further outputs a recommendation reason of the at least one execution item;

[0187] The recommendation unit is specifically configured to recommend the execution item that meets the preset condition to the target object, including:

[0188] The execution item that meets the preset condition and the recommendation reason are recommended to the target object.

[0189] In some embodiments, the large model further outputs a scene title for the at least one execution item; and the device further includes:

[0190] The processing module is configured to store the scene title and a setting result of the target object on the at least one execution item in a target address; and in the case that the scene title corresponding scene demand is acquired again, the setting result saved in the target address is used to respond to the scene demand.

[0191] In some embodiments, further comprising an optimization module, configured to:

[0192] determine a matching degree between the at least one execution item and the at least one requirement based on the at least one requirement for the large model output content in the prompt information;

[0193] determine a recommendation score of the current recommendation task based on the matching degree between the at least one execution item and the at least one requirement;

[0194] aggregate the recommendation score of the current recommendation task into a recommendation task set;

[0195] based on the recommendation task set, filter out a recommendation task with a recommendation score lower than a preset threshold to obtain a to-be-optimized recommendation task set;

[0196] based on the to-be-optimized recommendation task set, construct a recommendation requirement in the prompt information.

[0197] In some embodiments, the optimization module is specifically configured to:

[0198] filter out a recommendation example with a recommendation effect lower than a target effect from the to-be-optimized recommendation task set as a recommendation counterexample;

[0199] construct a recommendation requirement for the large model learning error example based on the recommendation counterexample and add it to the prompt information.

[0200] The specific functions and examples of the modules, sub-modules, and units of the apparatuses of the embodiments of the present disclosure are described in the related description of the corresponding steps in the above method embodiments, which will not be described here.

[0201] According to the embodiments of the present disclosure, the present disclosure further provides an electronic device, a readable storage medium and a computer program product.

[0202] Figure 5 A schematic block diagram of an example electronic device 500 that can be used to implement embodiments of the present disclosure is shown. The electronic device is intended to represent various forms of digital computers, such as laptops, desktops, workstations, personal digital assistants, servers, blade servers, mainframes, and other appropriate computers. The electronic device can also represent various forms of mobile devices, such as personal digital assistants, cellular telephones, smartphones, wearable devices, and other similar computing devices. The components shown here, their connections and relationships, and their functions, are meant to be examples only, and are not intended to limit implementations of the present disclosure described and / or claimed in this document.

[0203] As Figure 5As shown, the device 500 includes a computing unit 501 that can perform various appropriate actions and processes in accordance with a computer program stored in a read-only memory (ROM) 502 or a computer program loaded from a storage unit 508 into a random access memory (RAM) 503. In the RAM 503, various programs and data required for the operation of the device 500 can also be stored. The computing unit 501, the ROM 502, and the RAM 503 are connected to each other through a bus 504. An input / output (I / O) interface 505 is also connected to the bus 504.

[0204] A plurality of components in the device 500 are connected to the I / O interface 505, including: an input unit 506, such as a keyboard, a mouse, etc.; an output unit 507, such as various types of displays, speakers, etc.; a storage unit 508, such as a magnetic disk, an optical disk, etc.; and a communication unit 509, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 509 allows the device 500 to exchange information / data with other devices through a computer network, such as the Internet, and / or various telecommunication networks.

[0205] The computing unit 501 can be various general and / or special-purpose processing components with processing and computing capabilities. Some examples of the computing unit 501 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various special-purpose artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, a digital signal processor (DSP), and any appropriate processor, controller, microcontroller, etc. The computing unit 501 performs various methods and processes described above, such as the information recommendation method. For example, in some embodiments, the information recommendation method can be implemented as a computer software program that is tangibly embodied in a machine-readable medium, such as the storage unit 508. In some embodiments, part or all of the computer program can be loaded and / or installed on the device 500 via the ROM 502 and / or the communication unit 509. When the computer program is loaded into the RAM 503 and executed by the computing unit 501, one or more steps of the information recommendation method described above can be performed. Alternatively, in other embodiments, the computing unit 501 can be configured to perform the information recommendation method by any other appropriate means, such as by means of firmware.

[0206] The various embodiments of the systems and techniques described above can be implemented in digital electronic circuitry, integrated circuitry, a field programmable gate array (FPGA), an application specific integrated circuit (ASIC), a system on a chip (SOC), a complex programmable logic device (CPLD), computer hardware, firmware, software, and / or combinations thereof. These various embodiments can include implementation in one or more computer programs that are executable and / or interpretable on a programmable system including at least one programmable processor, which can be special or general purpose, coupled to receive data and instructions from, and to transmit data and instructions to, a storage system, at least one input device, and at least one output device.

[0207] Program code for carrying out methods of the present disclosure can be written in any combination of one or more programming languages. The program code can be provided to a processor or controller of a general purpose computer, special purpose computer, or other programmable data processing apparatus to produce a machine, such that the program code, when executed by the processor or controller, produces the functions / operations specified in the flowcharts and / or the block diagrams. The program code can be executed entirely on a machine, partially on a machine, partially on a machine as a stand-alone software package, or entirely on a remote machine or server.

[0208] In the context of the present disclosure, a machine-readable medium can be a tangible medium that contains or stores a program for use by or in connection with an instruction execution system, apparatus, or device. The machine-readable medium can be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium can include but is not limited to an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any suitable combination of the foregoing. More specific examples of the machine-readable storage medium will include one or more lines of electrical connections, portable computer disks, hard disk drives, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or Flash memory), optical fibers, portable compact disc read-only memories (CD-ROMs), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing.

[0209] To provide for interaction with a user, the systems and techniques described here can be implemented on a computer having a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user and a keyboard and a pointing device (e.g., a mouse or a trackball) by which the user can provide input to the computer. Other kinds of devices can be used to provide for interaction with a user as well; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form, including acoustic, speech, or tactile input.

[0210] The systems and techniques described here can be implemented in a computing system that includes a back end component (e.g., as a data server), or that includes a middleware component (e.g., an application server), or that includes a front end component (e.g., a user computer having a graphical user interface or a Web browser through which a user can interact with an implementation of the systems and techniques described here), or any combination of such back end, middleware, or front end components. The components of the system can be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include a local area network (LAN), a wide area network (WAN), and the Internet.

[0211] The computer system can include clients and servers. A client and server are generally remote from each other and typically interact through a communication network. The relationship of client and server arises by virtue of computer programs running on the respective computers and having a client-server relationship to each other. The server can be a cloud server, a server of a distributed system, or a server combined with a blockchain.

[0212] It should be understood that various forms of flow shown above can be used, with steps reordered, added, or removed. For example, the steps recited in the present disclosure can be performed in parallel, in series, or in a different order, without limitation herein, so long as the desired results of the technology disclosed in the present disclosure are achieved.

[0213] The specific embodiments described above are not intended to be limiting, and persons skilled in the art will appreciate that various modifications, combinations, sub-combinations and alternatives can be made to the specific embodiments without departing from the principles of the disclosure. Accordingly, modifications, equivalent alternatives and improvements should be included within the scope of the disclosure.

Claims

1. An information recommendation method, comprising: obtaining a constraint condition required to meet a scene demand of a target object for a target vehicle in response to the scene demand; constructing prompt information of a large model based on the constraint condition; inputting the prompt information into the large model to obtain at least one execution item recommended by the large model for the target vehicle; responding to the scene demand of the target object based on the at least one execution item; wherein the execution item is used to control a whole vehicle component or an application software function in a center control screen of the vehicle; wherein the prompt information further comprises: an inference method for reference by the large model, wherein the inference method comprises an exemplary inference step and a matter of attention for at least one inference step; a recommendation library supported by the target vehicle; the matter of attention comprises at least two of the following information: an analysis method of the scene demand; an example of mutually exclusive execution items; care knowledge of a specified object; a function explanation of the execution item; the recommendation library comprises an execution item set supported by the target vehicle; or the recommendation library comprises the execution item set and media resources.

2. The method of claim 1, wherein, the constraint condition comprises at least one of the following: description information of the target object; vehicle configuration information of the target vehicle; location information of the target vehicle; environment information of an environment in which the target vehicle is located; location information of the target object in the target vehicle.

3. The method of claim 1, wherein, the responding to the scene demand of the target object based on the at least one execution item comprises: screening an execution item satisfying a preset condition from the at least one execution item; and recommending the execution item satisfying the preset condition to the target object.

4. The method of claim 3, wherein, the screening the execution item satisfying the preset condition from the at least one execution item comprises: filtering out an execution item satisfying a filtering condition from the at least one execution item to obtain the execution item satisfying the preset condition; wherein the execution item satisfying the filtering condition comprises at least one of the following: an execution item contrary to a vehicle usage common sense; an execution item with an abnormal output format; an execution item beyond a capability range of the target vehicle.

5. The method of claim 3, further comprising: in response to a modification demand of the target object, modifying the execution item satisfying the preset condition.

6. The method of claim 3, wherein, the large model further outputs a recommendation reason of the at least one execution item; the recommending the execution item satisfying the preset condition to the target object comprises: recommending the execution item satisfying the preset condition and the recommendation reason to the target object.

7. The method of claim 3, wherein, the large model further outputs a scene title for the at least one execution item; and the method further comprises: storing the scene title and a setting result of the target object for the at least one execution item in a target address in association; and in a case of again obtaining the scene demand corresponding to the scene title, responding to the scene demand based on the setting result saved in the target address.

8. The method of any one of claims 1-7, further comprising: determining a matching degree between the at least one execution item and at least one requirement for output content of the large model based on the at least one requirement in the prompt information. determine a recommendation score of the current recommendation task based on the matching degree between the at least one execution item and the at least one requirement; aggregate the recommendation score of the current recommendation task into a recommendation task set; screen out a recommendation task set with a recommendation score lower than a preset threshold based on the recommendation task set, to obtain a to-be-optimized recommendation task set; construct a recommendation requirement in the prompt information based on the to-be-optimized recommendation task set.

9. The method of claim 8, wherein, The method of constructing the recommendation requirement in the prompt information based on the to-be-optimized recommendation task set comprises: screening out a recommendation example with a recommendation effect lower than a target effect from the to-be-optimized recommendation task set as a recommendation counterexample; constructing a recommendation requirement of the large model learning error example based on the recommendation counterexample and adding it to the prompt information.

10. An information recommendation apparatus, comprising: an acquisition module configured to acquire constraint conditions required to meet a scene demand of a target object for a target vehicle in response to the scene demand; a construction module configured to construct prompt information of a large model based on the constraint conditions; a generation module configured to input the prompt information into the large model to obtain at least one execution item recommended by the large model for the target vehicle; a response module configured to respond to the scene demand of the target object based on the at least one execution item; wherein the execution item is used to control a whole vehicle component or an application software function in a center control screen of the vehicle; wherein the prompt information further comprises: an inference method for reference by the large model, wherein the inference method comprises exemplary inference steps and attention items for at least one inference step; a recommendation library supported by the target vehicle; the attention items comprise at least two of the following information: an analysis method of the scene demand; an example of mutually exclusive execution items; care knowledge of a specified object; a function explanation of the execution item; the recommendation library comprises an execution item set supported by the target vehicle; or the recommendation library comprises the execution item set and media resources.

11. The apparatus of claim 10, wherein, The constraint conditions comprise at least one of the following: description information of the target object; vehicle configuration information of the target vehicle; location information of the target vehicle; environment information of an environment where the target vehicle is located; location information of the target object in the target vehicle.

12. The apparatus of claim 10, wherein, The response module comprises: a screening unit configured to screen execution items meeting a preset condition from the at least one execution item; a recommendation unit configured to recommend the execution items meeting the preset condition to the target object.

13. The apparatus of claim 12, wherein, The screening unit is specifically configured to: filter out execution items meeting a filtering condition from the at least one execution item to obtain the execution items meeting the preset condition; wherein the execution items meeting the filtering condition comprise at least one of the following: an execution item contrary to a vehicle usage common sense; an execution item with an abnormal output format; an execution item beyond a capability range of the target vehicle.

14. The apparatus of claim 12, further comprising: a modification module configured to modify the execution items meeting the preset condition in response to a modification demand of the target object.

15. The apparatus of claim 12, wherein, The large model further outputs a recommendation reason of the at least one execution item; The recommendation unit is specifically configured to recommend the execution item meeting the preset condition and the recommendation reason to the target object.

16. The apparatus of claim 12, wherein, The large model also outputs a scene title for the at least one execution item; and the device further includes: A processing module is configured to store the scene title and a setting result of the target object on the at least one execution item in a target address; and in a case where the scene title corresponds to the scene requirement is acquired again, the setting result saved in the target address is used to respond to the scene requirement.

17. The apparatus of any one of claims 10-16, further comprising an optimization module configured to: determine a matching degree of the at least one execution item and at least one requirement for the large model output content based on the at least one requirement in the prompt information; determine a recommendation score of the current recommendation task based on the matching degree of the at least one execution item and the at least one requirement; aggregate the recommendation score of the current recommendation task into a recommendation task set; based on the recommendation task set, filter out a recommendation task with a recommendation score lower than a preset threshold to obtain a to-be-optimized recommendation task set; based on the to-be-optimized recommendation task set, construct a recommendation requirement in the prompt information.

18. The apparatus of claim 17, wherein, The optimization module is specifically configured to: filter out a recommendation example with a recommendation effect lower than a target effect from the to-be-optimized recommendation task set as a recommendation counterexample; based on the recommendation counterexample, construct a recommendation requirement of a learning error example of the large model and add it to the prompt information.

19. An electronic device, comprising: at least one processor; and a memory connected in communication with the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the method of any one of claims 1-9.

20. A non-transitory computer readable storage medium having stored thereon computer instructions, wherein, The computer instructions are used to enable the computer to perform the method of any one of claims 1-9.

21. A computer program product comprising a computer program which, when executed by a processor, implements the method of any one of claims 1-9.

Images