Information pushing method and device based on large model, intelligent agent and electronic equipment

Abstract

The invention provides an information pushing method and device based on a large model, an intelligent agent, electronic equipment, a computer readable storage medium and a computer program product, and relates to the technical field of artificial intelligence such as large models. The method comprises the following steps: determining event information to be pushed based on a push indication message; pushing information of the to-be-pushed event is generated based on the to-be-pushed event information and the pushing preference of the target object by adopting a large model, and the pushing information meets the pushing preference; and pushing the push information to the target object, and determining a new push preference of the target object from the candidate preferences by adopting a large model based on the first behavior information of the target object to the push information. According to the method, the individuation of the pushed information can be improved, so that the browsing rate is improved.

Description

Technical Field

[0001] The present disclosure relates to the field of computer technologies, specifically to the field of artificial intelligence technologies such as large models, and particularly to an information push method, device, electronic device, and computer-readable storage medium based on a large model. Background Art

[0002] Application push is one of the main ways to attract users to use applications. However, the information pushed by applications often has a low browsing rate because it does not meet user interests and has low personalization. Summary of the Invention

[0003] Embodiments of the present disclosure propose an information push method, device, intelligent agent, electronic device, computer-readable storage medium, and computer program product based on a large model, which can improve the personalization of pushed information and thus improve the browsing rate.

[0004] In a first aspect, embodiments of the present disclosure propose an information push method based on a large model, including: determining event information to be pushed based on a push indication message; using a large model to generate push information for the event to be pushed based on the event information to be pushed and the push preference of a target object, where the push information meets the push preference; pushing the push information to the target object, and using a large model to determine a new push preference of the target object from candidate preferences based on the first behavior information of the target object for the push information.

[0005] In a second aspect, embodiments of the present disclosure propose an information push device, including: a first determination module, a first generation module, a push module, and a second determination module. The first determination module is configured to determine event information to be pushed based on a push indication message. The first generation module is configured to use a large model to generate push information for the event to be pushed based on the event information to be pushed and the push preference of a target object, where the push information meets the push preference. The push module is configured to push the push information to the target object. The second determination module is configured to use a large model to determine a new push preference of the target object from candidate preferences based on the first behavior information of the target object for the push information.

[0006] In a third aspect, embodiments of the present disclosure provide an intelligent agent configured to execute the method described in any implementation manner of the first aspect.

[0007] In a fourth aspect, embodiments of the present disclosure provide an electronic device, which includes: at least one processor; and a memory communicatively connected to the at least one processor; where the memory stores instructions executable by the at least one processor, and when the instructions are executed by the at least one processor, the at least one processor is enabled to execute the method described in any implementation manner of the first aspect.

[0008] Fifth aspect, embodiments of the present disclosure provide a non-transitory computer-readable storage medium storing computer instructions, which when executed by a computer can implement the method described in any implementation manner of the first aspect.

[0009] Sixth aspect, embodiments of the present disclosure provide a computer program product including a computer program, which when executed by a processor can implement the method described in any implementation manner of the first aspect.

[0010] It should be understood that the content described in this part is not intended to identify the key or important features of the embodiments of the present disclosure, nor is it used to limit the scope of the present disclosure. Other features of the present disclosure will become easily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] By reading the detailed description of the non-limiting embodiments with reference to the following drawings, other features, objectives, and advantages of the present disclosure will become more apparent:

[0012] Figure 1 is an exemplary system architecture to which the present disclosure can be applied;

[0013] Figure 2 is a flowchart of a method for information push based on a large model provided by an embodiment of the present disclosure;

[0014] Figure 3 is a flowchart of another method for information push based on a large model provided by an embodiment of the present disclosure;

[0015] Figure 4 is a flowchart of yet another method for information push based on a large model provided by an embodiment of the present disclosure;

[0016] Figure 5 is a schematic flowchart of a method for information push based on a large model in an application scenario provided by an embodiment of the present disclosure;

[0017] Figure 6 is a structural block diagram of an information push device provided by an embodiment of the present disclosure;

[0018] Figure 7 is a schematic structural diagram of an electronic device suitable for executing the method for information push based on a large model provided by an embodiment of the present disclosure. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0019] The following describes exemplary embodiments of the present disclosure with reference to the accompanying drawings. Various details of the embodiments of the present disclosure are included to facilitate understanding, and they should be considered merely exemplary. Therefore, those of ordinary skill in the art should recognize that various changes and modifications can be made to the embodiments described herein without departing from the scope and spirit of the present disclosure. Similarly, for the sake of clarity and conciseness, descriptions of well-known functions and structures are omitted below. It should be noted that, without conflict, the embodiments in the present disclosure and the features in the embodiments can be combined with each other.

[0020] In the technical solution of the present disclosure, the processing of the collection, storage, use, processing, transmission, provision, and disclosure of the user's personal information complies with the provisions of relevant laws and regulations and does not violate public order and good customs.

[0021] Figure 1 Exemplary system architecture 100 is shown, which can apply the embodiments of the information push method, device, intelligent agent, electronic device, and computer-readable storage medium based on the large model of the present disclosure.

[0022] As Figure 1 shown, system architecture 100 may include terminal devices 101, 102, 103, network 104, and server 105. Network 104 is used to provide a medium for communication links between terminal devices 101, 102, 103 and server 105. Network 104 may include various connection types, such as wired, wireless communication links, or fiber optic cables, etc.

[0023] Users can use terminal devices 101, 102, 103 to interact with server 105 through network 104 to receive or send messages, etc. Various applications for realizing information communication between the two can be installed on terminal devices 101, 102, 103 and server 105, such as e-commerce applications, news media applications, social media applications, financial service applications, health and fitness applications, instant messaging applications, etc.

[0024] The terminal devices 101, 102, 103 and the server 105 can be either hardware or software. When the terminal devices 101, 102, 103 are hardware, they can be various electronic devices with a display screen, including but not limited to smartphones, tablets, laptop computers, desktop computers, and so on; when the terminal devices 101, 102, 103 are software, they can be installed in the above-listed electronic devices, and can be implemented as multiple software or software modules, or can be implemented as a single software or software module, and no specific limitation is made here. When the server 105 is hardware, it can be implemented as a distributed server cluster composed of multiple servers, or can be implemented as a single server; when the server is software, it can be implemented as multiple software or software modules, or can be implemented as a single software or software module, and no specific limitation is made here.

[0025] The server 105 can provide various services through various built-in applications. When the server 105 runs an application, the following effects can be achieved: determining the event information to be pushed based on the push indication message; using a large model to generate the push information of the event to be pushed based on the event information to be pushed and the push preference of the target object, where the push information meets the push preference; pushing the push information to the target object, and using a large model to determine a new push preference of the target object from the candidate preferences based on the first behavior information of the target object for the push information.

[0026] It should be noted that the push indication message can be obtained from the terminal devices 101, 102, 103 through the network 104, and can also be pre-stored locally in the server 105 in various ways. Therefore, when the server 105 detects that these data have been stored locally, it can choose to directly obtain these data from the local. In this case, the exemplary system architecture 100 may not include the terminal devices 101, 102, 103 and the network 104.

[0027] Since generating the push information for the events to be pushed requires a large amount of computing resources and strong computing power, the information push method based on large models provided in the subsequent embodiments of the present disclosure is generally executed by the server 105 with strong computing power and a large amount of computing resources. Correspondingly, the information push device is generally also set in the server 105. However, it should also be noted that when the terminal devices 101, 102, and 103 also have computing power and computing resources that meet the requirements, the terminal devices 101, 102, and 103 can also complete the above operations that were originally done by the server 105 through the applications installed on them, and then output the same results as the server 105. Especially in the case where there are multiple terminal devices with different computing capabilities at the same time, when the application determines that the terminal device where it is located has strong computing power and a large amount of remaining computing resources, the terminal device can be allowed to execute the above operations, thereby appropriately reducing the computing pressure on the server 105. Correspondingly, the information push device can also be set in the terminal devices 101, 102, and 103. In this case, the exemplary system architecture 100 may not include the server 105 and the network 104.

[0028] It should be understood that Figure 1 the numbers of terminal devices, networks, and servers in

[0029] Application push is one of the main ways to attract users to use applications. However, the information pushed by applications often has a low browsing rate because it does not meet the user's interests and has low personalization. For example, only generating push information based on event information without considering the user's preferences to customize personalized push information, resulting in low personalization of push information.

[0030] In some embodiments, the push information is manually written and designed by humans. Manually creating the content of push information is time-consuming and difficult to update in real time, and it requires a lot of mental effort, manpower, and time. And manually creating push information for users in different languages and regions is complex and error-prone. Improper push frequency and time may lead to user loss.

[0031] In some embodiments, pre-designed push templates are used. The push templates only include fixed frameworks or formats and do not include specific content. When generating push information based on the push templates, only key variables (such as specific content) need to be filled in or selected, resulting in homogenized content and being difficult to meet complex customization requirements.

[0032] In some embodiments, push information is automatically generated based on preset rules or conditions (such as triggering specific messages according to user behavior). The solution is complex, requires careful design of rules and logic, is not easy to implement, and has a high dependence on data quality. Inaccurate data may lead to improper pushing and may cause user loss.

[0033] The method provided by the embodiments of the present disclosure is applicable to various information push scenarios. When applied to e-commerce applications, it can automatically generate personalized product recommendations and promotional information push, increasing the sales conversion rate. When applied to news media applications, it can automatically generate personalized news push according to user preferences, improving user activity. When applied to social media applications, it can generate personalized notifications and updates to keep users engaged and interactive. When applied to financial service applications, it can generate real-time push of market dynamics and account activities to help users make timely financial decisions. When applied to health and fitness applications, it can automatically generate exercise tips and health advice based on users' health data and goals.

[0034] Please refer to Figure 2 , Figure 2 which is a flowchart of a large model-based information push method provided by the embodiments of the present disclosure. The process 200 includes the following steps:

[0035] Step 201: Determine the event information to be pushed based on the push indication message.

[0036] Optionally, the above execution entity can obtain the push indication message in various ways. For example, the above execution entity obtains the push indication message from other devices. Another example is that the above execution entity obtains the push indication message based on manual operations, such as manually inputting the event information to be pushed.

[0037] Optionally, the event to be pushed can be an event in various scenarios. For example, the event to be pushed can be an operation activity, consumption content, account activity, health exercise tip, news push, or product recommendation, etc.

[0038] Optionally, the event information to be pushed may include, but is not limited to, at least one of the following: the name of the event to be pushed, the content of the event to be pushed, and the reference copywriting of the event to be pushed. Among them, the reference copywriting is different from the template that only needs to fill in or select key variables. The reference copywriting includes the complete push information. The main difference between the two is that the reference copywriting includes specific content, while the push template does not include specific content. The push information generated based on the reference copywriting is more flexible and easier to meet the personalized customization requirements.

[0039] For example, the name of the event to be pushed is: Spring Festival Blessing Red Packet Activity. The content of the event to be pushed is: Scan the code to send blessings and participate in the red packet sharing activity. The reference copywriting of the event to be pushed is "Spring Festival Blessing, Scan the Code to Win Benefits". The above is only a possible example and does not constitute a limitation.

[0040] When the above-mentioned execution entity obtains a push instruction message based on manual operation, for example, it can manually specify the name, content, and reference copywriting of the event to be pushed, or it can manually specify the name and / or content of the event to be pushed. Based on the name and / or content of the event to be pushed, the above-mentioned execution entity obtains reference copywriting applicable to the name and content of the event to be pushed from the database. The database includes the name, content, and reference copywriting of the event, and there is a mapping relationship between the name, content, and reference copywriting of the event.

[0041] Step 202: Use a large model to generate push information for the event to be pushed based on the event information to be pushed and the push preferences of the target object.

[0042] Exemplarily, the large model can use a large language model. A large language model (LLM) refers to a type of deep learning model trained with a large amount of text data. This type of model has the ability to understand natural language and generate natural language. They are usually based on a neural network architecture and have hundreds of millions or even billions of parameters, which enables them to capture complex language patterns and context relationships.

[0043] Exemplarily, the push information meets the push preferences of the target object. The number of push preferences of the target object can be one or more.

[0044] Exemplarily, the push preferences of the target object can refer to the preferences of the target object for the push information. For example, the push preferences of the target object can include the categories of push information that the target object is interested in and / or the categories of push information that the target object is not interested in, such as the interested categories (or tags): shopping, discounts, and the non-interested categories (or tags): games. For example, by using a large model to combine the event information to be pushed and the categories of push information that the target object is interested in, generating push information that meets the categories that the target object is interested in can enhance the target object's interest in the push information and increase the probability of browsing the push information.

[0045] For example, the above-mentioned execution entity can obtain the push preferences of the target object from other devices, or the above-mentioned execution entity can obtain the push preferences of the target object based on the portrait data of the target object. Alternatively, the above-mentioned execution entity can obtain the push preferences and candidate preferences of the target object based on the portrait data of the target object. The present disclosure is not limited thereto. In addition, the above-mentioned execution entity can obtain the push preferences of the target object before determining the event information to be pushed based on the push instruction message in the above-mentioned step 201, or it can also obtain the push preferences of the target object after the above-mentioned step 201 and before the above-mentioned step 202 generates the push information for the event to be pushed. The present disclosure is not limited to the timing at which the above-mentioned execution entity obtains the push preferences of the target object.

[0046] Step 203: Push the push information to the target object.

[0047] Exemplarily, the above execution entity can push the push information to the target object in various forms, such as in the form of a pop-up window on the client desktop, in the form of a notification bar of the client, etc.

[0048] Step 204: Use the large model to determine the new push preference of the target object from the candidate preferences based on the first behavior information of the target object for the push information.

[0049] Optionally, the candidate preferences may include at least one push tag, and the push preference may include at least one push tag that the target object is interested in. The above execution entity uses the large model to update the push preference of the target object based on the first behavior information of the target object for the push information, in combination with one or more predefined candidate preferences, to obtain the new push preference of the target object. Among them, the number of the obtained new push preferences of the target object may be one or more.

[0050] The obtained new push preference of the target object may be the same as or different from the push preference before the update. For example, the push preference of the target object includes preference 1 and preference 2, and the new push preference of the target object includes preference 1 and preference 2. Another example, the push preference of the target object includes preference 1 and preference 2, and the new push preference of the target object includes preference 2 and preference 3. Still another example, the push preference of the target object includes preference 1 and preference 2, and the new push preference of the target object includes preference 3 and preference 4.

[0051] For example, the first behavior information may include but is not limited to at least one of the following: viewing behavior (such as click operation, etc.), sharing behavior, liking behavior, collecting behavior, browsing time, not interested operation, viewing times, sharing times, and viewing moment.

[0052] Optionally, the candidate preferences may be preset. For example, the candidate preferences may be set according to the information push requirements, or may be determined after using the large model to analyze the data related to the push information of a large number of samples. The push tags included in the candidate preferences are related to the push scenario. For example, for e-commerce applications, the push tags included in the candidate preferences may include but are not limited to: product categories and promotional activities. Among them, the product categories may include but are not limited to one or more of the following: clothing, electronic products, food, cosmetics, and sports-related products.

[0053] Exemplarily, assuming that the candidate preferences include push preference 1 to push preference 8, the first behavior information of the target object for the push information can be input to the large model, and the large model filters out the new push preferences of the target object from push preference 1 to push preference 8 based on the first behavior information. In this way, the range of push preferences (candidate preferences) is fixed, and determining at least one preference from the candidate preferences within the fixed range as the new push preference of the target object can make the obtained new push preference of the target object more accurate.

[0054] Optionally, using the large model to determine the new push preference of the target object from the candidate preferences based on the first behavior information of the target object for the push information may include: using the large model to determine the new push preference of the target object based on the first behavior information of the target object for the push information, the candidate preferences, and the push preference of the target object. In this way, the range of push preferences (candidate preferences) is fixed. As the number of push information pushed to the user increases, the push preferences are continuously updated and iterated, and the obtained new push preferences will be more and more accurate. Furthermore, the personalization of the push information pushed to the target object will also be higher and higher.

[0055] The method provided in the embodiments of the present disclosure uses a large model to combine the event information to be pushed and the push preference of the target object to generate push information that conforms to the push preference of the target object. The generated push information is personalized. Compared with directly generating push information based on event information, it is more helpful to improve the personalization of push information, can enhance the interest of the target object in the push information, and thus improve the browsing rate. The logical rules are simple and the solution is easy to implement. Using a large model to generate push information can reduce the required manpower and improve efficiency.

[0056] In addition, the range of push preferences (candidate preferences) is fixed, and determining at least one preference from the candidate preferences within the fixed range as the new push preference of the target object can make the obtained new push preference of the target object more accurate, and thus can improve the personalization of the new push information.

[0057] In some embodiments, the push preference of the target object may be obtained based on the portrait data of the target object and the candidate preferences, where the portrait data of the target object may include but is not limited to at least one of the following: the historical behavior information of the target object for the historical push information, object attribute information, content preference, tone preference, and time preference.

[0058] Optionally, the acquisition of the push preference of the target object and the acquisition of information related to obtaining the push preference (such as portrait data) require the authorization of the target object.

[0059] For example, the behavioral information may include, but is not limited to, at least one of the following: viewing behavior (such as click operations, etc.), sharing behavior, liking behavior, favoriting behavior, browsing time, disinterest operation, view count, share count, and viewing moment. The historical behavioral information of the target object regarding the historical push information may refer to the above-mentioned behavioral information.

[0060] For example, the object attribute information may be the attribute information of the target object. The category of the group to which the target object belongs can be determined based on the attribute information of the target object, and the push preference of the target object can be determined based on the push preferences and candidate preferences of the category of the group to which the target object belongs. For example, in the case where historical behavioral information is not obtained, the push preference that matches or is roughly consistent with the candidate preference among the push preferences of the category of the group to which the target object belongs can be used as the push preference of the target object.

[0061] For example, content preference may refer to the preference of an object for content. The content preference can be obtained based on the behavioral information of the object for the content. There is a difference between the content and the push information. Taking a social media application as an example, the push information is the notification pushed by the social media application, such as a system notification, and the content can be the works published by other objects.

[0062] For example, the tone preference may include the category of the interested tone and / or the category of the disinterested tone. For example, the tone categories include: affectionate tone, serious tone, declarative tone, exclamatory tone, interrogative tone, imperative tone, etc. For example, assuming that the tone preference of the target object is the exclamatory tone, push information with an exclamatory tone can be generated based on the tone preference of the target object, which can make the push information more in line with the interests of the target object. Optionally, the push preference of the target object may include the tone preference.

[0063] For example, the time preference can be obtained based on the viewing moment of the target object for the historical push information. For example, if the number of times the target user views the historical push information at the first moment reaches the preset threshold, it is determined that the time preference includes the first moment. Or, if the number of times the target object views the historical push information at the first moment is the largest, the first moment is taken as the time preference. The push moment of the push information can be determined based on this time preference. For example, the push moment of the push information is determined as the first moment, and the push information is pushed to the target object at the first moment when the target object is interested, which can customize the push information personalized and thus improve the browsing rate of the target object for the push information. Optionally, the push preference of the target object may include the time preference.

[0064] Optionally, based on the time preference, the first moment for pushing the push information is determined. The above step 203 of pushing the push information to the target object may include: at the first moment, pushing the push information to the target object.

[0065] In this way, by inferring the viewing time of the historical push information by the target object, it is inferred that the target object is more interested in clicking on the push information at a certain / some time points. Subsequently, pushing to the target object at this certain / some time points can increase the viewing rate.

[0066] Optionally, the viewing frequency preference of the target object for the push information can also be determined by the viewing time. Based on the frequency preference, the first frequency for pushing this push information is determined. The above step 203 of pushing the push information to the target object may include: based on the time preference and the frequency preference, pushing the push information to the target object. This can further increase the viewing rate.

[0067] Optionally, the viewing time of the target object for the historical push information can be obtained from the mapped information recorded between the target object identifier, the target object name, the viewing time, and the push information.

[0068] Exemplarily, based on the historical behavior information of the target object for the historical push information, the category of the push information that the target object is interested in and / or the category of the push information that the target object is not interested in can be obtained.

[0069] Optionally, the category range of the push information can be pre-divided. For example, the category range of the push information includes categories 1 to 8. It is possible to determine whether the target object is interested in the push information based on the behavior information of the target object for the push information, and determine which category in the category range of the push information this push information belongs to. For example, it belongs to category 1. Combining whether the target object is interested in this push information and the category 1 it belongs to, it is determined whether the target object is interested in the push information of category 1.

[0070] For example, the category range of the push information can be divided manually, or it can be divided by a large model.

[0071] Exemplarily, based on the object attribute information, the category to which the target object belongs can be obtained. The push preferences corresponding to each category can be obtained based on the sample information. The push preference of the category to which the target object belongs is obtained from the push preferences corresponding to each category. Based on the push preference of the category to which the target object belongs, the push preference of the target object is determined (for example, the category of the push information that the target object is interested in and / or the category of the push information that the target object is not interested in).

[0072] Exemplarily, based on the content preference of the target object, the content category that the target object is interested in can be obtained. Based on the content category that the target object is interested in, the category of the push information that the target object is interested in and / or the category of the push information that the target object is not interested in is determined.

[0073] By combining the historical behavior information, object attribute information, and content preferences of the target object for historical push messages, the categories of push messages that the generated target object is interested in and / or the categories of push messages that the target object is not interested in can be made more accurate and personalized. Combining candidate preferences can make the obtained push preferences both personalized and accurate. Based on this push preference, personalized push messages for each individual can be generated, which can improve the personalization of push messages, help enhance the target object's interest in push messages, and thus increase the browsing rate. The push messages generated by combining the categories, tone preferences, and / or time preferences of the push messages that the target object is interested in (and / or not interested in) can further conform to the target object's interests and tone preferences. Pushing this push message to the target object based on the time preference can further increase the browsing rate.

[0074] Optionally, the push preference of the target object can be obtained by using a large model based on the portrait data of the target object and candidate preferences. The large model can be used as a tool for classifying and merging push message categories, replacing the process of manually classifying and labeling push message categories, and can reduce the proliferation of push message categories.

[0075] Optionally, the above-mentioned execution entity can directly generate push messages for the event to be pushed based on the event information to be pushed and the portrait data of the target object. For example, the above-mentioned execution entity can generate push messages for the event to be pushed based on the event information to be pushed and at least one of the following: the categories of push messages that the target object is interested in, the categories of push messages that the target object is not interested in, the historical behavior information of the target object for historical push messages, object attribute information, content preferences, tone preferences, and time preferences.

[0076] In some embodiments, the push preference of the target object may include at least one of the following: the categories of push messages that the target object is interested in, the categories of push messages that the target object is not interested in, the historical behavior information of the target object for historical push messages, object attribute information, content preferences, tone preferences, and time preferences.

[0077] In some embodiments, in step 202 above, using a large model to generate push messages for the event to be pushed based on the event information to be pushed and the push preference of the target object may include: using the large model to generate push messages based on the event information to be pushed, the push preference of the target object, and preset rules.

[0078] Optionally, the preset rules may include, but are not limited to, at least one of the following: language type, word count requirement, and style requirement. For example, the language type includes Chinese, English, etc. The word count requirement refers to the limit on the number of words in the push message. The style requirement may refer to the requirements for the display style of the push message, such as text alignment, font, color, size, display background, display pattern, and whether emoticons can be present, etc.

[0079] For example, the preset rules may be generated based on the event to be pushed. For example, if the name of the event to be pushed is: Spring Festival Blessing Red Envelope Activity, the color of the display background may be red. In this way, the push message obtained by combining the event information to be pushed, the push preferences of the target object, and the preset rules can not only meet the personalized needs of the target object, but also enhance the personalization for the event to be pushed, thereby improving the browsing rate.

[0080] In some embodiments, in step 202 above, the large model is used to generate the push message of the event to be pushed based on the event information to be pushed and the push preferences of the target object, which may include: generating a first prompt based on the event information to be pushed and the push preferences; inputting the first prompt into the large model to obtain the push message.

[0081] Exemplarily, the first prompt can indicate constructing the push message of the event to be pushed that meets the push preferences of the target object. For example, the first prompt can indicate constructing the push message that meets the push preferences of the target object and is related to the event to be pushed.

[0082] Optionally, in step 202 above, the large model is used to generate the push message of the event to be pushed based on the event information to be pushed and the push preferences of the target object, which may include: generating a second prompt based on the event information to be pushed, the push preferences, and the preset rules; inputting the second prompt into the large model to obtain the push message. Among them, the second prompt can indicate constructing the push message of the event to be pushed that meets the push preferences of the target object and the preset rules.

[0083] For example, suppose that in the event information to be pushed, the name of the event to be pushed is: Spring Festival blessing and red envelope sharing activity. The content of the event to be pushed is: Scan the code to send blessings and participate in the red envelope sharing activity. The reference copy of the event to be pushed is "Spring Festival blessing, scan the code to win benefits". Recent content preferences include "Graduate Entrance Examination Teaching Video" and "Fantasy Novels", and have clicked on the push information "Fans Real Welfare Activities" and "Free Novel Collection", often click on the push information at "19:00" and "20:00", and the tone preference includes the "intimate" tone. Generate prompt words based on the above event information to be pushed and push preferences: You are an expert in writing push copy for an application. There is a Spring Festival blessing and red envelope activity. The content of the activity is to scan the code to send blessings and participate in the red envelope sharing activity. Recently, you prefer "Graduate Entrance Examination Teaching Video" and "Fantasy Novels". You have clicked on the push copy "Fans Real Welfare Activities" and "Free Novel Collection". You often click on the push copy at "19:00" and "20:00". You like the "intimate" tone. Please generate a recommended copy and a recommended push time for her. The recommended copy is limited to 12 to 25 words (optional). The reference copy is "Spring Festival blessing, scan the code to win benefits". Enter the above prompt words into the large model to obtain push information (or push copy): Dear, send blessings and welcome good luck, scan the code to win red envelopes and gifts for the postgraduate entrance examination! . Recommended time: 19:00 or 20:00.

[0084] Combining the event information to be pushed and the push preferences of the target object (and also combining preset rules) can generate more personalized prompt words, and inputting the prompt words into the big model can customize personalized push information, which helps to increase the user's interest in the push information, thereby increasing the browsing rate. In addition, compared with manually generating push information, the push information generated by the big model in this application is not only more accurate and more personalized, but also reduces the required labor, improves the efficiency of generating push information, and can achieve real-time update of push information.

[0085] Please refer to Figure 3 , Figure 3 A flowchart of another large model-based information push method provided in an embodiment of the present disclosure, wherein process 300 includes the following steps.

[0086] Step 301: Determine event information to be pushed based on the push indication message.

[0087] Step 302: Generate push information of the event to be pushed based on the event information to be pushed and the push preference of the target object using the big model.

[0088] Regarding the specific implementation of step 301-step 302, reference may be made to the description of the above-mentioned step 201-step 202, which will not be repeated here.

[0089] Step 303: Input the pushed information into the big model for review and obtain a first review result.

[0090] Optionally, the training data of the large model includes sample information that does not conform to the first rule, such as cases where the review fails or failure cases, and the training data can also include sample information that conforms to the first rule. For example, through the embedding engineering of the large model, the large model learns this training data. The push information is input into the large model for review, and the large model can directly give the review result and the reason for non-pass.

[0091] Exemplarily, the training data may include cases where the review fails or failure cases, such as risk control problem cases. For example, in the cases in the training data: the following words cannot appear in the push information: price. Then the failure cases may include: Surprise price 188 yuan or Surprise membership price 500 yuan.

[0092] For example, the above-mentioned execution entity can review the content of the push information through the large model. For example, it can review whether the content conforms to the rules, whether it contains sensitive words, whether it contains prohibited words, etc., which can reduce the risk of the push information.

[0093] For example, if it is determined that the push information conforms to the rules, does not contain sensitive words, and does not contain prohibited words or phrases, the first review result includes passing the review. If it is determined that the push information does not conform to the rules, contains sensitive words, and / or contains prohibited words or phrases, the first review result includes failing the review.

[0094] Step 304: In response to determining that the first review result includes passing the review, push the push information to the target object.

[0095] Optionally, the above-mentioned pushing the push information to the target object may include: In response to determining that the first review result includes passing the review, push the push information to the target object.

[0096] Optionally, the method provided by the embodiments of the present disclosure may further include: using the large model to determine the new push preference of the target object from the candidate preferences based on the first behavior information of the target object for the push information. The specific implementation can refer to the above step 204 and will not be elaborated here.

[0097] Step 305: In response to determining that the first review result includes failing the review, use the large model to generate new push information based on the event information to be pushed, the push preference, and the push information.

[0098] Optionally, the method provided by the embodiments of the present disclosure may further include: using the large model to determine the new push preference of the target object from the candidate preferences based on the behavior information of the target object for the new push information. The specific implementation can refer to the above step 204 and will not be elaborated here.

[0099] If the first review result includes passing the review, proceed to the next step, such as pushing the message to the target object or conducting the next review (see Figure 4 the method shown). If the first review result includes failing the review, regenerate the push message until the push message passes the review. For example, the push message can be regenerated based on the push message that failed the review, or the push message can be directly generated based on the event information to be pushed and the push preference without considering the push message that failed the review.

[0100] Optionally, the new push message conforms to the push preference of the target object and is different from the push message. For example, the new push message conforms to the push preference of the target object and is different from the push message that failed the review. Generating a new push message based on the push message that failed the review can increase the probability of the new push message passing the review.

[0101] Optionally, the first review result can also include the reason for failing the review. The above-mentioned execution entity can generate a new push message based on the event information to be pushed, the push preference, the push message, and the reason for failing the review. This can avoid problems similar to the reason for failing the review in the new push message and can further increase the probability of the new push message passing the review.

[0102] In some embodiments, step 303 above of inputting the push message into the large model for review to obtain the first review result may include: generating a third prompt word based on the push message; inputting the third prompt word into the large model to obtain the first review result.

[0103] Optionally, the third prompt word instructs to review the push message based on the first rule.

[0104] Optionally, the training data of the large model includes sample information that does not conform to the first rule, as specifically described in the corresponding part of step 303 above.

[0105] Exemplarily, the above-mentioned execution entity can generate a third prompt word for instructing to review the push message based on the event information, and input the third prompt word into the large model. The large model reviews the third prompt word to obtain the first review result. Reviewing the push message by combining the prompt word with the large model can improve the accuracy of the review result.

[0106] In some embodiments, generating the push message and the review can be completed based on one prompt word. For example, generating a prompt word based on the event information to be pushed and the push preference, the prompt word instructs to construct a push message for the event to be pushed that meets the push preference of the target object and to review the generated push message. Inputting the prompt word into the large model to obtain the push message and the review result, which can also include the reason for not passing.

[0107] Please refer to Figure 4 ,Figure 4 It is a flowchart of another information push method based on a large model provided by an embodiment of the present disclosure, where process 400 includes the following steps.

[0108] Step 401: Determine the event information to be pushed based on the push indication message.

[0109] Step 402: Use the large model to generate the push information of the event to be pushed based on the event information to be pushed and the push preference of the target object.

[0110] Regarding the specific implementation manners of steps 401 - 402, reference may be made to the description of the above steps 201 - 202, which will not be elaborated here.

[0111] Step 403: Input the push information into the large model for review to obtain the first review result.

[0112] Regarding the specific implementation manner of step 403, reference may be made to the description of the above step 303, which will not be elaborated here.

[0113] Step 404: In response to determining that the first review result includes passing the review, review the push information based on the second rule to obtain the second review result.

[0114] For example, the second rule may include but is not limited to at least one of the following: format requirements, no sensitive words, no prohibited words or phrases.

[0115] After reviewing the push information based on the large model, the push information can also be reviewed again based on the second rule. By reviewing the push information multiple times, the risk of the push information can be further reduced.

[0116] Optionally, in response to determining that the first review result includes passing the review, the above execution subject may output the push information for manual review of the push information based on the second rule, and the above execution subject obtains the second review result based on the operation of manual review. Humans can not only review whether it complies with the rules, sensitive words, prohibited words or phrases, etc., but also review the aesthetics of the push information, which can further improve the personalization of the push information and thus increase the browsing rate.

[0117] Optionally, the order of reviewing the push information based on the second rule and reviewing the push information based on the large model may not be limited. For example, the push information may be reviewed based on the second rule first, and after passing the review, the push information may be reviewed based on the large model. It is also possible to only perform manual review, and after passing the review, directly push the push information to the target object.

[0118] Step 405: In response to determining that the second review result includes that the push information complies with the second rule, push the push information to the target object.

[0119] Optionally, the push message is pushed to the target object, including: in response to determining that the second review result includes that the push message conforms to the second rule, pushing the push message to the target object.

[0120] Optionally, the method provided by the embodiments of the present disclosure may further include: using a large model to determine a new push preference of the target object from candidate preferences based on the first behavior information of the target object for the push message. The specific implementation may refer to step 204 above and will not be elaborated here.

[0121] Step 406: In response to determining that the second review result includes that the push message does not conform to the second rule, use a large model to generate a new push message based on the event information to be pushed, the push preference, and the push message.

[0122] Optionally, the method provided by the embodiments of the present disclosure may further include: using a large model to determine a new push preference of the target object from candidate preferences based on the behavior information of the target object for the new push message. The specific implementation may refer to step 204 above and will not be elaborated here.

[0123] If the second review result includes passing the review, proceed to the next step, such as pushing the push message to the target object. If the first review result includes not passing the review, regenerate the push message until the push message passes the review. For example, a new push message can be regenerated based on the push message that did not pass the review, or a new push message can be directly generated based on the event information to be pushed and the push preference without considering the push message that did not pass the review.

[0124] In some embodiments, in step 202 above, using a large model to generate a push message for the event to be pushed based on the event information to be pushed and the push preference of the target object may include: using a large model to generate a new push message based on the event information to be pushed, the push preference of the target object, and the push message that did not pass the review. Among them, the number of push messages that did not pass the review can be one or more, and the push message that did not pass the review is the push message corresponding to the event (or push indication message) to be pushed this time. For example, after determining the event information to be pushed in step 201 and before step 202, a push message for the event to be pushed is generated, but this push message did not pass the review by manual and / or large model, etc. The new push message meets the push preference and is different from the push message that did not pass the review.

[0125] In some embodiments, the above-mentioned using a large model to determine a new push preference of the target object from candidate preferences based on the first behavior information of the target object for the push message may include: using a large model to update the portrait data of the target object based on the first behavior information; using a large model to determine a new push preference of the target object from candidate preferences based on the updated portrait data of the target object.

[0126] Exemplarily, after pushing the push information to the target object, the above-mentioned execution entity can further obtain the first behavior information of the target object regarding the push information, and use a large model to update the portrait data of the target object based on the first behavior information and the push information. For example, add the first behavior information and the push information to the portrait data before the update to obtain new portrait data. If the portrait data before the update includes the same push information and the corresponding behavior information is consistent with the first behavior information, the portrait data may not be updated either.

[0127] Taking the view behavior included in the first behavior information as an example, assuming the first behavior information includes the view behavior, it is considered that the user is interested in the push information, convert the push information into an interested category, and combine this interested category with the set of interested categories in the push preference before the update (including at least one category of push information that the target object is interested in) to obtain a new set of interested categories. Additionally, if the set of uninterested categories includes this interested category, this interested category in the set of uninterested categories can be deleted. For example, add the obtained interested category to the set of interested categories before the update to obtain a new set of interested categories. If the set of interested categories before the update includes the newly obtained interested category, the set of interested categories may not be updated, or it can be updated. For example, increasing the weight of this interested category can make the personalization more accurate.

[0128] For example, the view time in the first behavior information can be updated to the portrait data. If the portrait data before the update includes this view time, it may not be updated, or it can be updated. For example, increasing the weight of this view time can push the push information to the target object at one or several view times with higher weights.

[0129] By continuously updating or supplementing the portrait data of the target object to increase the amount of reference data, the newly obtained push preference can better match the interests of the target object, and the personalization of subsequent push information can be improved.

[0130] In some embodiments, the above-mentioned method of using a large model to determine the new push preference of the target object from the candidate preferences based on the updated portrait data of the target object may include: generating a fourth prompt word based on the updated portrait data of the target object, the push preference of the target object, and the candidate preferences; inputting the fourth prompt word into the large model to obtain the new push preference of the target object.

[0131] Optionally, the fourth prompt word indicates updating the push preference of the target object based on the updated portrait data of the target object and the candidate preferences.

[0132] Exemplarily, the above-mentioned execution entity may generate a fourth prompt word based on the updated portrait data of the target object, the previous push preferences, and the candidate preferences, indicating to update the push preferences of the target object based on the updated portrait data of the target object, the push preferences of the target object, and the candidate preferences, and input the fourth prompt word into the large model. The large model generates new push preferences based on the fourth prompt word. By using the large model to update the push preferences through the prompt word method, the accuracy of the push preferences can be improved, thereby improving the personalization of subsequent push information.

[0133] The present disclosure also provides an agent, which is configured to be able to implement the methods described in any of the above embodiments. Exemplarily, the agent includes an input module, a processing module, and an output module. Among them, the input module is used to determine the event information to be pushed based on the push indication message. The processing module is used to use the large model to generate the push information of the event to be pushed based on the event information to be pushed and the push preferences of the target object, where the push information meets the push preferences. The output module is used to push the push information to the target object. The technical effects brought by the embodiments related to the agent provided in this embodiment can be respectively referred to the corresponding descriptions in the above method embodiments, and will not be elaborated here.

[0134] For better understanding, the present disclosure also gives a specific implementation solution in combination with a specific application scenario. Please refer to Figure 5 the process 500 as shown.

[0135] Step 501: Manually input the event information to be pushed.

[0136] The server or the agent obtains the manually input push event information. The server in the embodiments of the present disclosure may be replaced by any entity capable of executing the methods provided in the embodiments of the present disclosure, such as an agent, etc.

[0137] Step 502: The server requests the portrait data of the target object from the database.

[0138] Step 503: The database sends the portrait data of the target object to the server.

[0139] Step 504: The server generates a first prompt word based on the event information to be pushed and the portrait data of the target object.

[0140] Figure 5 Taking the example of directly generating the push information of the event to be pushed based on the event information to be pushed and the portrait data of the target object for elaboration.

[0141] Step 505: The server inputs the first prompt word into the large model.

[0142] Step 506: The large model obtains the push information and the push time based on the first prompt word, and sends the push information and the push time to the server.

[0143] Optionally, the server can generate a second prompt word based on the event information to be pushed, the portrait data of the target object, and the preset rules, input the second prompt word to the large model, and the large model obtains the push information and the push time based on the second prompt word.

[0144] Step 507: The server generates a third prompt word based on the push information.

[0145] Step 508: The server inputs the third prompt word to the large model.

[0146] Step 509: The large model reviews the push information based on the third prompt word, obtains the first review result, and sends the first review result to the server.

[0147] Step 510: The server displays the push information for manual review of the push information.

[0148] Optionally, the server can display the push information for manual review of the push information in response to determining that the first review result includes passing the review.

[0149] Step 511: Manually review the push information and return the second review result to the server.

[0150] Step 512: The server constructs the push information in response to determining that the second review result includes that the push information complies with the second rule.

[0151] Step 513: The server sends the push information to the application at the push time.

[0152] Step 514: The application registers the push information and displays it to the target object.

[0153] Step 515: The target object clicks on the push information.

[0154] Step 516: The application obtains the first behavior information of the target object for the push information and sends the first behavior information to the server.

[0155] Step 517: The server requests the portrait data of the target object from the database.

[0156] Step 518: The database returns the portrait data of the target object to the server.

[0157] Step 519: The server generates a fourth prompt word based on the portrait data of the target object, the push information, and the first behavior information.

[0158] Step 520: The server inputs the fourth prompt word to the large model.

[0159] Step 521: The large model obtains new portrait data of the target object based on the fourth prompt word and sends the new portrait data of the target object to the server.

[0160] Step 522: The server writes the new portrait data of the target object into the database.

[0161] Step 523: The database returns the write result to the server.

[0162] Further reference Figure 6 , as an implementation of the methods shown in the above figures, the present disclosure provides an embodiment of an information push device. This device embodiment corresponds to the above method embodiment, and this device can be specifically applied to various electronic devices.

[0163] As Figure 6 shown, the information push device 600 of this embodiment may include a first determination module 601, a first generation module 602, a push module 603, and a second determination module 604. Among them, the first determination module 601 is configured to determine the event information to be pushed based on the push indication message. The first generation module 602 is configured to use the large model to generate push information of the event to be pushed based on the event information to be pushed and the push preference of the target object. The second determination module 604 is configured to use the large model to determine a new push preference of the target object from the candidate preferences based on the first behavior information of the target object for the push information. Among them, the push information satisfies the push preference. The push module 603 is configured to push the push information to the target object.

[0164] In this embodiment, for the specific processing of the first determination module 601, the first generation module 602, the push module 603, and the second determination module 604 in the information push device 600 and the technical effects brought by them, reference can be made to the relevant descriptions of the steps in the above embodiments respectively, and details will not be repeated here.

[0165] In some optional implementation manners of this embodiment, the push preference of the target object is obtained based on the portrait data of the target object and the candidate preferences, where the portrait data of the target object includes at least one of the following: the historical behavior information of the target object for the historical push information, object attribute information, content preference, tone preference, and time preference.

[0166] In some optional implementation manners of this embodiment, the first generation module 602 is further configured to: generate a first prompt word based on the event information to be pushed and the push preference, where the first prompt word instructs to construct push information of the event to be pushed that satisfies the push preference of the target object; input the first prompt word into the large model to obtain the push information.

[0167] In some alternative implementation manners of this embodiment, the first generation module 602 is further configured to: use a large model to generate push information based on the event information to be pushed, the push preferences of the target object, and preset rules, where the preset rules include at least one of the following: language type, word count requirement, and style requirement.

[0168] In some alternative implementation manners of this embodiment, the first generation module 602 is further configured to: generate a second prompt based on the event information to be pushed, the push preferences of the target object, and preset rules, where the second prompt instructs to construct push information for the event to be pushed that meets the push preferences of the target object and the preset rules; input the second prompt into the large model to obtain the push information.

[0169] In some alternative implementation manners of this embodiment, the information push device 600 further includes: an acquisition module. The acquisition module is configured to input the push information into the large model for review to obtain a first review result. The push module 603 is further configured to, in response to determining that the first review result includes passing the review, push the push information to the target object.

[0170] In some alternative implementation manners of this embodiment, the acquisition module is further configured to: generate a third prompt based on the push information, where the third prompt instructs to review the push information based on the first rule; input the third prompt into the large model to obtain the first review result, where the training data of the large model includes sample information that does not conform to the first rule.

[0171] In some alternative implementation manners of this embodiment, the information push device 600 further includes: a second generation module. The second generation module is configured to, in response to determining that the first review result includes not passing the review, use the large model to generate new push information based on the event information to be pushed, the push preferences, and the push information, where the new push information conforms to the push preferences and is different from the push information.

[0172] In some alternative implementation manners of this embodiment, the information push device 600 further includes: a review module and a third generation module. The review module is configured to, in response to determining that the first review result includes passing the review, review the push information based on the second rule to obtain a second review result. The push module 603 is further configured to, in response to determining that the second review result includes that the push information conforms to the second rule, push the push information to the target object. The third generation module is configured to, in response to determining that the second review result includes that the push information does not conform to the second rule, use the large model to generate new push information based on the event information to be pushed, the push preferences, and the push information, where the new push information conforms to the push preferences and is different from the push information.

[0173] In some alternative implementation manners of this embodiment, the second determination module 604 is further configured to: update the portrait data of the target object by using a large model based on the first behavior information; determine a new push preference of the target object from the candidate preferences by using the large model based on the updated portrait data of the target object.

[0174] In some alternative implementation manners of this embodiment, the second determination module 604 is further configured to: generate a fourth prompt word based on the updated portrait data of the target object, the push preference of the target object, and the candidate preferences, where the fourth prompt word indicates updating the push preference of the target object based on the updated portrait data of the target object and the candidate preferences; input the fourth prompt word into the large model to obtain the new push preference of the target object.

[0175] In some alternative implementation manners of this embodiment, the information push device 600 further includes: a third determination module. The third determination module is configured to determine a first moment for pushing the above push information based on the time preference. The push module 603 is further configured to push the push information to the target object at the first moment.

[0176] In some alternative implementation manners of this embodiment, the event information to be pushed includes at least one of the following: the name of the event to be pushed, the content of the event to be pushed, and the reference copywriting of the event to be pushed.

[0177] This embodiment exists as a device embodiment corresponding to the above method embodiment. For the technical effects brought by the embodiments involved in the information push device provided in this embodiment, reference may be made to the corresponding relevant descriptions in the above method embodiment respectively, and details are not described herein again.

[0178] According to an embodiment of the present disclosure, the present disclosure further provides an electronic device, which includes: at least one processor; and a memory communicatively connected to the at least one processor; wherein, the memory stores instructions executable by the at least one processor, and when the instructions are executed by the at least one processor, the at least one processor is enabled to implement the method described in any of the above embodiments.

[0179] According to an embodiment of the present disclosure, the present disclosure further provides a readable storage medium, which stores computer instructions for enabling a computer to implement the method described in any of the above embodiments when executed.

[0180] According to an embodiment of the present disclosure, the present disclosure further provides a computer program product, which can implement the method described in any of the above embodiments when executed by a processor.

[0181] Figure 7FIG. shows a schematic block diagram of an exemplary electronic device 700 that can be used to implement embodiments of the present disclosure. The electronic device is intended to represent various forms of digital computers, such as, laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as, personal digital processors, cellular phones, smart phones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely exemplary and are not intended to limit the implementation of the present disclosure described and / or claimed herein.

[0182] As Figure 7 shown, the device 700 includes a computing unit 701 that can perform various appropriate actions and processes according to a computer program stored in a read-only memory (ROM) 702 or a computer program loaded from a storage unit 708 into a random access memory (RAM) 703. In the RAM 703, various programs and data required for the operation of the device 700 can also be stored. The computing unit 701, the ROM 702, and the RAM 703 are connected to each other via a bus 704. An input / output (I / O) interface 705 is also connected to the bus 704.

[0183] A plurality of components in the device 700 are connected to the I / O interface 705, including: an input unit 706, such as a keyboard, a mouse, etc.; an output unit 707, such as various types of displays, speakers, etc.; a storage unit 708, such as a magnetic disk, an optical disk, etc.; and a communication unit 709, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 709 allows the device 700 to exchange information / data with other devices via a computer network such as the Internet and / or various telecommunication networks.

[0184] The computing unit 701 can be various general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of the computing unit 701 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various dedicated artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, a digital signal processor (DSP), and any suitable processor, controller, microcontroller, etc. The computing unit 701 executes the various methods and processes described above, such as the information push method based on a large model. For example, in some embodiments, the information push method based on a large model can be implemented as a computer software program that is tangibly contained in a machine-readable medium, such as the storage unit 708. In some embodiments, part or all of the computer program can be loaded and / or installed onto the device 700 via the ROM 702 and / or the communication unit 709. When the computer program is loaded into the RAM 703 and executed by the computing unit 701, one or more steps of the information push method based on a large model described above can be executed. Alternatively, in other embodiments, the computing unit 701 can be configured to execute the information push method based on a large model in any other suitable way (e.g., by means of firmware).

[0185] Various embodiments of the systems and techniques described above in this document can be implemented in digital electronic circuit systems, integrated circuit systems, field-programmable gate arrays (FPGAs), application-specific integrated circuits (ASICs), application-specific standard products (ASSPs), system-on-a-chip systems (SOCs), complex programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various embodiments can include: being implemented in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which can be a special-purpose or general-purpose programmable processor, and can receive data and instructions from a storage system, at least one input device, and at least one output device, and transmit the data and instructions to the storage system, the at least one input device, and the at least one output device.

[0186] The program code for implementing the methods of the present disclosure can be written in any combination of one or more programming languages. These program codes can be provided to a processor or controller of a general-purpose computer, a special-purpose computer, or other programmable data processing device, such that when the program codes are executed by the processor or controller, the functions / operations specified in the flowcharts and / or block diagrams are implemented. The program codes can be executed entirely on the machine, partially on the machine, as an independent software package partially on the machine and partially on a remote machine, or entirely on a remote machine or server.

[0187] In the context of this disclosure, a machine-readable medium can be a tangible medium that can contain or store 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. The machine-readable medium can include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination of the foregoing. More specific examples of the machine-readable storage medium would include an electrical connection based on one or more wires, a portable computer diskette, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or Flash memory), an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0188] For providing interaction with a user, the systems and techniques described herein 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 also be used to provide interaction with the user; for example, the 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 input, voice input, or tactile input).

[0189] The systems and techniques described herein can be implemented in a computing system that includes backend components (e.g., as a data server), or a computing system that includes middleware components (e.g., an application server), or a computing system that includes frontend components (e.g., a user computer having a graphical user interface or a web browser through which the user can interact with an implementation of the systems and techniques described herein), or a computing system that includes any combination of such backend components, middleware components, or frontend 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.

[0190] A computer system may include a client and a server. The client and the server are generally far from each other and usually interact via a communication network. The relationship between the client and the server is generated by computer programs running on the respective computers and having a client-server relationship with each other. The server may be a cloud server, also known as a cloud computing server or a cloud host, which is a host product in the cloud computing service system to address the deficiencies of difficult management and weak business scalability existing in traditional physical hosts and virtual private server (VPS) services. The server may also be a server of a distributed system or a server combined with a blockchain.

[0191] According to the technical solution of the embodiment of the present disclosure, a large model is used to combine the event information to be pushed and the push preference of the target object to generate push information that conforms to the push preference of the target object. The generated push information is personalized. Compared with directly generating push information based on event information, it is more helpful to improve the personalization of push information, can enhance the interest of the target object in the push information, and thus improve the browsing rate. The logical rule is simple and the solution is easy to implement. In addition, using a large model to generate push information can reduce the required manual labor and improve efficiency.

[0192] It should be understood that various forms of the processes shown above can be used, with steps reordered, added, or deleted. For example, the steps recited in the present disclosure can be executed in parallel, sequentially, or in a different order, as long as the desired results of the technical solution disclosed in the present disclosure can be achieved. There is no limitation herein.

[0193] The above specific embodiments do not constitute a limitation on the protection scope of the present disclosure. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present disclosure shall be included within the protection scope of the present disclosure.

Claims

1. A method for pushing information based on a large model, comprising: Determine the event information to be pushed based on the push indication message; Generate push information of the event to be pushed based on the event information to be pushed and the push preference of the target object using the large model, wherein the push information satisfies the push preference; Pushing the push information to the target object; The large model is used to determine a new push preference of the target object from candidate preferences based on the first behavior information of the target object in response to the push information.

2. The method according to claim 1, wherein: The push preference of the target object is obtained based on the portrait data of the target object and the candidate preferences, wherein the portrait data of the target object includes at least one of the following: historical behavior information of the target object for historical push information, object attribute information, content preference, tone preference and time preference.

3. The method according to claim 1 or 2, wherein: The adopting of the large model to generate push information of the event to be pushed based on the event information to be pushed and the push preference of the target object includes: Based on the event information to be pushed and the push preference, generating a first prompt word, wherein the first prompt word indicates to construct push information of the event to be pushed that meets the push preference of the target object; The first prompt word is input into the large model to obtain the push information.

4. The method according to claim 1 or 2, wherein: The adopting of the large model to generate push information of the event to be pushed based on the event information to be pushed and the push preference of the target object includes: The push information is generated by using the big model based on the event information to be pushed, the push preference of the target object and preset rules, wherein the preset rules include at least one of the following: language type, word count requirement and style requirement.

5. The method according to claim 4, wherein: The adopting of the large model to generate the push information based on the event information to be pushed, the push preference of the target object and preset rules includes: The large model is used to generate a second prompt word based on the event information to be pushed, the push preference of the target object and the preset rule, wherein the second prompt word indicates to construct push information of the event to be pushed that meets the push preference of the target object and the preset rule; The second prompt word is input into the large model to obtain the push information.

6. The method according to claim 1 or 2, wherein: The method further comprises: Inputting the pushed information into the large model for review to obtain a first review result; The step of pushing the push information to the target object includes: In response to determining that the first review result includes passing the review, the push information is pushed to the target object.

7. The method according to claim 6, wherein: The step of inputting the pushed information into the large model for review and obtaining a first review result includes: generating a third prompt word based on the push information, wherein the third prompt word indicates to review the push information based on the first rule; The third prompt word is input into the large model to obtain the first audit result, wherein the training data of the large model includes sample information that does not conform to the first rule.

8. The method according to claim 6, wherein: The method further comprises: In response to determining that the first audit result includes failure to pass the audit, the big model is used to generate new push information based on the event information to be pushed, the push preferences and the push information, wherein the new push information complies with the push preferences and is different from the push information.

9. The method according to claim 6, wherein: The method further comprises: In response to determining that the first review result includes passing the review, reviewing the push information based on a second rule to obtain a second review result; The pushing of the push information to the target object includes: in response to determining that the second review result includes that the push information complies with the second rule, pushing the push information to the target object; In response to determining that the second review result includes that the push information does not comply with the second rule, the big model is used to generate new push information based on the event information to be pushed, the push preference and the push information, wherein the new push information complies with the push preference and is different from the push information.

10. The method according to claim 1 or 2, wherein: The adopting the large model to determine the new push preference of the target object from candidate preferences based on the first behavior information of the target object to the push information includes: Using the large model to update the portrait data of the target object based on the first behavior information; The large model is used to determine a new push preference of the target object from the candidate preferences based on the updated portrait data of the target object.

11. The method according to claim 10, wherein: The adopting the large model based on the updated portrait data of the target object to determine the new push preference of the target object from the candidate preferences includes: Generate a fourth prompt word based on the updated portrait data of the target object, the push preference of the target object and the candidate preference, wherein the fourth prompt word indicates updating the push preference of the target object based on the updated portrait data of the target object and the candidate preference; The fourth prompt word is input into the macro model to obtain a new push preference of the target object.

12. The method according to claim 2, wherein: The method further comprises: Determining a first time to push the push information based on the time preference; The step of pushing the push information to the target object includes: At the first moment, the push information is pushed to the target object.

13. The method according to claim 1 or 2, wherein: The event information to be pushed includes at least one of the following: the name of the event to be pushed, the content of the event to be pushed, and the reference text of the event to be pushed.

14. An information push device, comprising: A first determining module is configured to determine the event information to be pushed based on the push indication message; A first generating module is configured to generate push information of the event to be pushed based on the event information to be pushed and the push preference of the target object using a large model, wherein the push information satisfies the push preference; A push module, configured to push the push information to the target object; The second determination module is configured to determine a new push preference of the target object from candidate preferences based on the first behavior information of the target object to the push information by using the large model.

15. An agent configured to perform the method according to any one of claims 1-13.

16. An electronic device, comprising: at least one processor; as well as a memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed 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 according to any one of claims 1 to 13.

17. A non-transitory computer-readable storage medium storing computer instructions, the computer instructions being used to cause the computer to execute the method of any one of claims 1-13.

18. A computer program product, comprising a computer program, which, when executed by a processor, implements the steps of the method according to any one of claims 1 to 13.

Images