Content processing method, apparatus, device, storage medium, and product

Abstract

The embodiment of the application provides a content processing method, device, equipment, storage medium and product, the method can be: predicting a first predicted content set about a target push service to be pushed to a target object at a first time; obtaining a first exposed content set about the target push service that has been exposed to the target object at the first time, and a first access content set of the target object; the content in the first access content set belongs to the first exposed content set; determining an evaluation index of an initial content push model according to the first exposed content set, the first access content set and the first predicted content set; adjusting the initial content push model according to the evaluation index of the initial content push model to obtain a target content push model, and the target content push model is used to push the content about the target push service to the target object. The application adopts the process, and the content prediction accuracy of the content push model can be improved.

Description

Technical Field

[0001] This invention relates to the field of computer technology, and in particular to a content processing method, apparatus, device, storage medium, and product. Background Technology

[0002] Push notification systems can recommend content to users, such as personalized content recommendations. Typically, the content recommendation model used in push notification systems is evaluated using recall. However, the inventors discovered that this evaluation metric is rather idealized and does not consider actual push scenarios. Therefore, optimizing the content recommendation model based on this metric resulted in a significant discrepancy between the model's content prediction results and the actual push situation, leading to low content prediction accuracy. Summary of the Invention

[0003] This application provides a content processing method, apparatus, device, storage medium, and product that can improve the content prediction accuracy of content push models.

[0004] In a first aspect, embodiments of this application provide a content processing method, including:

[0005] Obtain the first historical access data of the target object, wherein the first historical access data refers to the access data of the target object for the historical exposure content of the target push service within the first historical time period;

[0006] Invoke the initial content push model for the target push service to predict, based on the first historical access data, the first predicted set of content related to the target push service to be pushed to the target object at the first time.

[0007] Obtain the first set of exposure content related to the target push service that has been exposed to the target object in the first time, and the first set of access content of the target object; the content in the first set of access content belongs to the first set of exposure content.

[0008] The evaluation metrics for the initial content push model are determined based on the first exposed content set, the first accessed content set, and the first predicted content set.

[0009] Based on the evaluation metrics of the initial content push model, the initial content push model is adjusted to obtain a target content push model, which is used to push content about the target push service to the target object.

[0010] Secondly, embodiments of this application provide a content processing apparatus, including:

[0011] The acquisition module is used to acquire the first historical access data of the target object. The first historical access data refers to the access data of the target object for the historical exposure content of the target push service within the first historical time period.

[0012] The prediction module is used to invoke the initial content push model for the target push service, so as to predict the first set of predicted content about the target push service to be pushed to the target object at the first time based on the first historical access data.

[0013] The acquisition module is further configured to acquire a first set of exposure content about the target push service that has been exposed to the target object in the first time, and a first set of access content of the target object; the content in the first set of access content belongs to the first set of exposure content.

[0014] The determining module is used to determine the evaluation index of the initial content push model based on the first exposed content set, the first accessed content set, and the first predicted content set.

[0015] An adjustment module is used to adjust the initial content push model according to the evaluation indicators of the initial content push model to obtain a target content push model, which is used to push content about the target push service to the target object.

[0016] Thirdly, embodiments of this application provide a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps of the method described in the first aspect of claim.

[0017] Fourthly, embodiments of this application provide a computer-readable storage medium having a computer program stored thereon, characterized in that the computer program, when executed by a processor, implements the steps of the method described in the first aspect.

[0018] Fifthly, embodiments of this application provide a computer program product, including a computer program that, when executed by a processor, implements the steps of the method described in the first aspect.

[0019] In summary, the computer device can predict the first set of predicted content about the target push service to be pushed to the target object at the first moment, and can also obtain the first set of exposed content about the target push service that has already been exposed to the target object at the first moment, as well as the first set of accessed content of the target object. Therefore, the computer device can determine the evaluation index of the initial content push model based on the first set of exposed content, the first set of accessed content, and the first set of predicted content, so as to adjust the initial content push model and obtain the target content push model. This application introduces the calculation of the evaluation index of the content push model based on the actual exposure of the content to be pushed predicted by the model during the actual push process. This can obtain an evaluation index of the content prediction effect of the content push model that is more in line with the actual push scenario. Optimizing the content push model based on this evaluation index can effectively improve the content prediction accuracy of the target content push model. Attached Figure Description

[0020] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0021] Figure 1 This is a schematic diagram of the structure of a recommendation system provided in an embodiment of the present invention;

[0022] Figure 2 This is a flowchart illustrating a content processing method provided in an embodiment of the present invention;

[0023] Figure 3 This is a flowchart illustrating another content processing method provided in an embodiment of the present invention;

[0024] Figure 4 This is a schematic diagram of the structure of a content processing device provided in an embodiment of the present invention;

[0025] Figure 5 This is a schematic diagram of the structure of a computer device provided in an embodiment of the present invention. Detailed Implementation

[0026] The technical solutions of the present invention will now be described with reference to the accompanying drawings in the embodiments of the present invention.

[0027] Artificial intelligence (AI) is the theory, methods, technology, and application systems that use digital computers or machines controlled by digital computers to simulate, extend, and expand human intelligence, perceive the environment, acquire knowledge, and use that knowledge to achieve optimal results. In other words, AI is a comprehensive technology within computer science that attempts to understand the essence of intelligence and produce a new kind of intelligent machine that can react in a way similar to human intelligence. AI studies the design principles and implementation methods of various intelligent machines, enabling them to possess the functions of perception, reasoning, and decision-making.

[0028] Artificial intelligence (AI) is a comprehensive discipline encompassing a wide range of fields, including both hardware and software technologies. Fundamental AI technologies generally include sensors, dedicated AI chips, cloud computing, distributed storage, big data processing, operating / interactive systems, and mechatronics. AI software technologies primarily include computer vision, speech processing, natural language processing, and machine learning / deep learning.

[0029] Natural Language Processing (NLP) is an important field within computer science and artificial intelligence. It studies the theories and methods for enabling effective communication between humans and computers using natural language. NLP is a science that integrates linguistics, computer science, and mathematics. Therefore, research in this field involves natural language—the language people use in daily life—and thus it has a close relationship with linguistic research. NLP techniques typically include text processing, semantic understanding, machine translation, question answering, and knowledge graphs.

[0030] Machine Learning (ML) is a multidisciplinary field involving probability theory, statistics, approximation theory, convex analysis, and algorithm complexity theory. It specifically studies how computers can simulate or implement human learning behavior to acquire new knowledge or skills and reorganize existing knowledge structures to continuously improve their performance. Machine learning is the core of artificial intelligence and the fundamental way to endow computers with intelligence; its applications span all areas of artificial intelligence. Machine learning and deep learning typically include techniques such as artificial neural networks, belief networks, reinforcement learning, transfer learning, inductive learning, and instructional learning.

[0031] With the research and advancement of artificial intelligence (AI) technology, AI is being studied and applied in various fields, such as smart homes, smart wearable devices, virtual assistants, smart speakers, smart marketing, autonomous driving, drones, robots, smart healthcare, and smart customer service. It is believed that with the development of technology, AI will be applied in more fields and play an increasingly important role.

[0032] This application provides a content processing method involving artificial intelligence technologies such as natural language processing, machine learning, and deep learning. Through a recommendation model, a set of content to be pushed to an object can be generated, and the exposure set and access set of the object can be obtained. Then, by combining the set of content to be pushed, an evaluation index for the recommendation model can be determined.

[0033] See Figure 1 The recommended system shown, Figure 1 The recommendation system shown includes multiple terminals (including terminal 1, terminal 2, etc.) and server 10. Each terminal can establish communication with server 10. The content processing method described can be applied to... Figure 1 The recommendation system shown. Specifically, the content processing method can be applied to server 10.

[0034] The terminals include, but are not limited to, smartphones, tablets, laptops, desktop computers, laptops, wearable devices such as wearable watches, vehicle terminals, and aircraft. The server 10 can be a single server or a server cluster. A server can be a single physical server, a server cluster consisting of at least two physical servers, or a distributed system. It can also be a cloud server providing basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, Content Delivery Network (CDN), and big data and artificial intelligence platforms.

[0035] exist Figure 1In the aforementioned recommendation system, server 10 can generate a set of content to be pushed to an object, such as the user corresponding to terminal 1, through a content push model. It then obtains a set of content to be exposed to the object based on this set and pushes it to the object. Server 10 can also obtain the set of accessed content by the object in relation to the exposed content set, and combine this with the exposed content set and the set of content to be pushed to determine the evaluation metrics for the content push model. Based on these metrics, the content push model can be optimized, thereby improving its content prediction accuracy. Typically, the results analyzed by the recall module in a recommendation system differ somewhat from the final results presented to the user. This is because, in addition to the recall module's processing, modules such as sorting, reordering, and filtering are involved in data processing, leading to significant differences between the offline evaluation results and the actual online performance of the recommendation system's recall module. This solution primarily provides an evaluation metric that more closely approximates the online evaluation, helping businesses better assess the performance of different content push models when deployed online, even during the offline phase.

[0036] The content processing method will be specifically described below through the following embodiments.

[0037] Please see Figure 2 This is a flowchart illustrating a content processing method provided in an embodiment of this application. The method can be applied to computer devices. Specifically, the method includes the following steps:

[0038] S201. Obtain the first historical access data of the target object, wherein the first historical access data refers to the access data of the target object for the historical exposure content of the target push service within the first historical time period.

[0039] The target object can be any one or more objects that use the target push service, such as any one or more users who use the target push service. Alternatively, the target object can be any one or more objects that the target push service targets, such as any one or more users that the target push service targets.

[0040] The target push service can be any type of push service or a specified push service. Here, the target push service includes push services such as product push service, book push service, music push service, video push service, information push service, and advertising push service, etc., and this application embodiment does not limit this.

[0041] Here, the first historical time period represents a past period of time, which can be the most recent month, the most recent week, the most recent three days, the most recent day, etc., and this application embodiment does not impose any restrictions on this. For example, the first historical time period can be day tn to day t. Here, t is an integer greater than or equal to 1, and n is an integer less than or equal to t.

[0042] Historical exposure content refers to the exposure content related to the target push service that has been shown to the target audience. For example, if the target push service is a product push service, then historical exposure content refers to each product that has been shown to the target audience within a historical time period. If the target push service is a book push service, then historical exposure content refers to each book that has been shown to the target audience within a historical time period. If the target push service is a music push service, then historical exposure content refers to each piece of music that has been shown to the target audience within a historical time period. If the target push service is a video push service, then historical exposure content refers to each video that has been shown to the target audience within a historical time period. If the target push service is an information push service, then historical exposure content refers to each piece of information that has been shown to the target audience within a historical time period. If the target push service is an advertising push service, then historical exposure content refers to each advertisement that has been shown to the target audience within a historical time period. It should be noted that the above are merely examples of historical exposure content and should not constitute a limitation on this application.

[0043] The accessed data may include a set of accessed content, which may include one or more pieces of content. Here, the set of accessed content may refer to the set of content that has been clicked.

[0044] In one embodiment, the access data may further include at least one of the following: a set of content that has been liked, a set of content that has been saved, and a set of content that has been forwarded. In one embodiment, in addition to including the set of accessed content, the access data may also include the access duration for each piece of content in the set of accessed content.

[0045] S202. Invoke the initial content push model for the target push service to predict, based on the first historical access data, the first predicted content set for the target push service to be pushed to the target object at the first time.

[0046] In this embodiment, a computer device can input first historical access data into an initial content push model for a target push service, and generate a first predicted content set for the target push service to be pushed to the target object at a first time. The first time can be any time or a specified time later than the end time of the first historical time period. In one embodiment, the first time can be the day after the end time included in the first historical time period. For example, if the first historical time period is day tn - day t, then the first time can be day t+1. The first predicted content set can include one or more content items. The first predicted content set can be understood as the set of content mined by the initial content push model to be pushed to the target object at a first time.

[0047] In one embodiment, after obtaining the first predicted content set, the computer device can obtain the first exposure content set about the target push service to be exposed to the target object at a first time, and push the first exposure content set to the target object at a first time.

[0048] In one embodiment, the first exposure content set may include all or part of the content in the first predicted content set, or the first exposure content set may exclude any content from the first predicted content set. It should be noted that the specific content included in the first exposure content set depends on the specific processing strategy for the prediction results after obtaining them from the content push module.

[0049] In one embodiment, a computer device can determine a first sub-set of content to be exposed to a target object at a first time from a first set of predicted content, and define the first sub-set of content as the first set of exposed content. The first sub-set of content may include one or more pieces of content. The content exposed to the target object refers to the content actually pushed to the target object. Here, the first set of exposed sub-content refers to a set of actual pushed content pushed to the target object.

[0050] In one embodiment, the computer device determines the first sub-exposure content set to be exposed to the target object at a first time from the first predicted content set by: sorting the first predicted content set to obtain a sorted predicted content set; filtering the sorted predicted content set to obtain a filtered content set; and determining the filtered content set as the first sub-exposure content set related to the target push service to be exposed to the target object at a first time. In another embodiment, the computer device sorts the first predicted content set to obtain a sorted predicted content set by: performing an initial sort on the first predicted content set to obtain a first sorted content set; then performing a fine sort on the first sorted content set to obtain a second sorted content set; and finally obtaining the sorted predicted content set based on the second sorted content set. In yet another embodiment, the computer device can perform the initial sorting of the first predicted content set using a pre-trained first sorting model. Specifically, the computer device scores each item in the first predicted content set using a first ranking model, obtains a score for each item, sorts the items in the first predicted content set according to their scores from highest to lowest, obtains a first ranking result, and selects the top M items from the first predicted content set based on the first ranking result to construct a first ranked content set. In one embodiment, the first ranking model can be a machine learning model. In another embodiment, the computer device can perform a fine-tuning process on the first ranked content set using a pre-trained second ranking model. Specifically, the computer device scores each item in the first ranked content set using a second ranking model, obtains a score for each item, sorts the items in the second ranked content set according to their scores from highest to lowest, obtains a second ranking result, and selects the top N items from the first ranked content set based on the second ranking result to construct a second ranked content set, where N is a positive integer less than M. In one embodiment, the second ranking model can be a deep learning model. In one embodiment, the computer device obtains the sorted predicted content set based on the second sorted content set in the following ways: the computer device can determine the second sorted content set as the sorted predicted content set; or, the computer device can rearrange the second sorted content set to obtain the sorted predicted content set. This rearrangement process includes, but is not limited to, using random sampling. In one embodiment, the filtering process performed by the computer device on the sorted predicted content set includes, but is not limited to, filtering out content already accessed by the target object or filtering out content accessed by the target object within a second specified time period, etc., and this application embodiment does not limit this method.The second specified time period can be the past day, the past three days, the past week, the past month, etc., and this application embodiment does not limit this.

[0051] In one embodiment, the computer device may further acquire a second set of sub-content to be exposed to the target object at a first moment, and determine the first set of sub-content to be exposed to the target object as the first set of content. The second set of sub-content may include one or more pieces of content. The second set of sub-content points to another set of content actually pushed to the target object. The content included in the second set of sub-content is different from the content included in the first set of content. In one embodiment, the second set of sub-content may be content that the target object has not accessed, or popular content that the target object has not accessed. In one embodiment, popular content that the target object has not accessed can be determined as follows: The content published to the computer device is ordered from high to low or low to high popularity index at a first moment; the top K items in the popularity index ranking are selected; the content among the top K items that the target object has not accessed is determined as popular content that the target object has not accessed, where K is a positive integer.

[0052] In this embodiment, after the computer device exposes a first set of exposed content to the target object at a first moment, the target object can view the various contents in the first set of exposed content. The target object can select content of interest from the first set of exposed content for access. The computer device can record the target object's access data for the first set of exposed content.

[0053] In one application scenario, a target user can click on any piece of content (of interest) in the first set of exposed content and send a viewing command to the computer device. Upon receiving the viewing command, the computer device returns detailed information about the target content to the target user, who can then view this information. During this process, the computer device can identify the target content as the content accessed by the target user.

[0054] S203. Obtain the first set of exposure content about the target push service that has been exposed to the target object in the first time, and the first set of access content of the target object.

[0055] S204. Determine the evaluation metrics of the initial content push model based on the first exposed content set, the first accessed content set, and the first predicted content set.

[0056] In this embodiment, the computer device can acquire a first set of exposed content and a first set of accessed content. The first set of accessed content may include all content accessed by the target object in response to the first set of exposed content at a given time. The content in the first set of accessed content belongs to the first set of exposed content. After acquiring the first set of exposed content and the first set of accessed content, the computer device can obtain an evaluation metric for the initial content push model based on the first set of exposed content, the first set of accessed content, and the first set of predicted content. In the process of evaluating the content push model, this application incorporates the actual online exposure of the predicted content to be pushed to calculate the evaluation metric for the content push model. This evaluation metric better reflects the actual content prediction effect when the content push model is actually run for content prediction.

[0057] In one embodiment, the computer device determines the evaluation metric of the initial content push model based on the first exposed content set, the first accessed content set, and the first predicted content set in the following manner: determining the actual exposure volume of the target object with respect to the target push service based on the first exposed content set and the first predicted content set; determining the actual access volume of the target object with respect to the target push service based on the first accessed content set and the first predicted content set; and determining the evaluation metric of the initial content push model based on the actual access volume and the actual exposure volume.

[0058] In one embodiment, the computer device determines the actual exposure of the target object to the target push service based on the first exposed content set and the first predicted content set by: determining a first content intersection between the first exposed content set and the first predicted content set; counting the number of contents included in the first content intersection; and determining the number of contents included in the first content intersection as the actual exposure of the target object to the target push service. Here, the first content intersection refers to the content intersection between the first exposed content set and the first predicted content set, meaning content that appears in both the first exposed content set and the first predicted content set.

[0059] For example, if the first set of exposed content includes content 1, content 2, content 3, content 4, and content 5, and the first set of predicted content includes content 2, content 3, content 4, content 5, and content 6, the computer device can determine that the first content intersection between the first set of exposed content and the first set of predicted content includes content 2, content 3, content 4, and content 5. The computer device can also calculate that the first content intersection includes 4 items and determine 4 as the actual exposure volume of the target object regarding the target push service.

[0060] In one embodiment, the computer device determines the actual access volume of the target object to the target push service based on the first access content set and the first predicted content set by: determining a second content intersection between the first access content set and the first predicted content set; counting the number of contents included in the second content intersection; and determining the number of contents included in the second content intersection as the actual access volume of the target object to the target push service. Here, the second content intersection refers to the content intersection between the first access content set and the first predicted content set, meaning content that appears in both the first access content set and the first predicted content set.

[0061] For example, if the first accessed content set includes content 2 and content 3, and the first predicted content set includes content 2, content 3, content 4, content 5, and content 6, the computer device can determine that the second content intersection between the first accessed content set and the first predicted content set includes content 2 and content 3. The computer device can then calculate that the second content intersection includes two items, and can identify these two items as the actual access volume of the target object regarding the target push service.

[0062] In one embodiment, the computer device determines the evaluation metric of the initial content push model based on the actual number of visits and the actual number of exposures by: calculating the ratio between the actual number of visits and the actual number of exposures; determining the ratio between the actual number of visits and the actual number of exposures as a visit conversion metric; and determining the visit conversion metric as the evaluation metric of the initial content push model. In other words, the computer device can divide the actual number of visits by the actual number of exposures to obtain the ratio between the actual number of visits and the actual number of exposures, and then determine the ratio between the actual number of visits and the actual number of exposures as the evaluation metric for the initial content push model.

[0063] For example, if the actual number of visits is 2 and the actual number of impressions is 4, the computer device can determine 0.5 as the evaluation metric for the initial content push model.

[0064] In one embodiment, the computer device determines the evaluation metric of the content push model based on the first exposed content set, the first accessed content set, and the first predicted content set as follows: The computer device determines a first content intersection between the first exposed content set and the first predicted content set; counts the number of contents included in the first content intersection; and determines the number of contents included in the first content intersection as the actual exposure volume of the target object regarding the target push service. The computer device determines a second content intersection between the first accessed content set and the first predicted content set; counts the number of contents included in the second content intersection; and determines the number of contents included in the second content intersection as the actual access volume of the target object regarding the target push service. The computer device calculates the ratio between the actual access volume and the actual exposure volume; determines the ratio between the actual access volume and the actual exposure volume as an access conversion metric, and determines the access conversion metric as the evaluation metric of the initial content push model.

[0065] In one embodiment, considering that the first access content set contains content that the target object mistakenly accessed, the computer device determines the actual access volume of the target object regarding the target push service based on the first access content set and the first predicted content set by: deleting each piece of content that the target object mistakenly accessed from the first access content set to obtain a deleted access content set; determining a second content intersection between the deleted access content set and the first predicted content set; counting the number of contents included in the second content intersection; and determining the number of contents included in the second content intersection as the actual access volume of the target object regarding the target push service. Here, the second content intersection refers to the content intersection between the deleted access content set and the first predicted content set, meaning content that appears in both the deleted access content set and the first predicted content set. Since the actual access volume is usually a parameter reflecting the number of contents that an object is interested in, and the content that the target object mistakenly accessed indicates that the content is not of interest to the target object, including the mistakenly accessed content in the actual access volume statistics would result in an inaccurate actual access volume. However, this embodiment of the application performs the actual access volume statistics only after deleting the mistakenly accessed content, thus obtaining a more accurate actual access volume.

[0066] S205. Based on the evaluation metrics of the initial content push model, the initial content push model is adjusted to obtain a target content push model, which is used to push content about the target push service to the target object.

[0067] In this embodiment of the application, the computer device can adjust the initial content push model according to the evaluation index of the initial content push model until the evaluation index of the adjusted initial content push model is greater than or equal to the evaluation index threshold, or the adjusted initial content push model meets the stopping iteration condition (such as the number of iterations being greater than or equal to the preset number of iterations), thereby obtaining the target content push model and improving the content prediction accuracy based on the target content push model.

[0068] In one embodiment, assuming the initial push model is a first initial push model, the computer device can adjust the model parameters of the first initial content push model based on its evaluation index if the evaluation index of the initial content push model is less than the evaluation index threshold, thus obtaining an adjusted first initial content push model. The computer device can also obtain second historical access data of the target object, which refers to the historical exposure data of the target object for the target push service within a second historical time period. Then, it calls the adjusted first initial content push model to predict, based on the second historical access data, a second predicted content set about the target push service to be pushed to the target object at a second time. It also obtains the second exposure content set about the target push service that has already been exposed to the target object at a second time, and the target object's second access content set. The content in the second access content set belongs to the second exposure content set. Based on the second exposure content set, the second access content set, and the second predicted content set, it determines the evaluation index of the adjusted first initial content push model. If the evaluation index of the adjusted first initial content push model is greater than or equal to the evaluation index threshold, then the adjusted first initial content push model is determined as the target content push model. The above process is equivalent to continuously adjusting the model parameters of the initial content recommendation model until the evaluation index of the adjusted initial content push model is greater than or equal to the evaluation index threshold, thus obtaining the target content push model.

[0069] The second historical time period also represents a past time period. The second historical time period differs from the first historical time period. For example, the first historical time period could be day tn to day t. The second historical time period could be day t-n+1 to day t+1. The second time can be any time later than the end time of the second historical time period, or a specified time. In one embodiment, the second time can be the day after the end time included in the second historical time period. For example, if the second historical time period is day t-n+1 to day t+1, then the second time could be day t+2. The second predicted content set can include one or more content items. The second predicted content set can be understood as the set of content to be pushed to the target object at the second time, mined by the adjusted first initial content push model. In one embodiment, the second exposure content set can include all or part of the content in the second predicted content set, or the second exposure content set may not include any content in the second predicted content set. It should be noted that the specific content included in the second exposure content set depends on the specific processing strategy of the prediction results after obtaining the prediction results from the content push module. It should be noted that the method by which the computer device obtains the second set of exposed content about the target push service that has been exposed to the target object in the second time, and the second set of accessed content of the target object, can refer to the aforementioned method by which the computer device obtains the first set of exposed content about the target push service that has been exposed to the target object in the first time, and the first set of accessed content of the target object. This embodiment of the application will not elaborate further here. Correspondingly, the method by which the computer device determines the evaluation index of the adjusted first initial content push model based on the second set of exposed content, the second set of accessed content, and the second set of predicted content can refer to the aforementioned method by which the evaluation index of the adjusted initial content push model is determined based on the first set of exposed content, the first set of accessed content, and the first set of predicted content. This embodiment of the application will not elaborate further here.

[0070] visible, Figure 2In the illustrated embodiment, the computer device can obtain the first historical access data of the target object and invoke the initial content push model for the target push service to predict the first predicted content set of the target push service to be pushed to the target object at the first time based on the first historical access data. The computer device can also obtain the first exposed content set of the target push service that has been exposed to the target object at the first time, as well as the first access content set of the target object. The content in the first access content set belongs to the first exposed content set. Thus, the computer device can determine the evaluation index of the initial content push model based on the first exposed content set, the first access content set, and the first predicted content set. The initial content push model can be adjusted using the evaluation index of the initial content push model to obtain the target content push model. This process can effectively improve the content prediction accuracy of the target content push model.

[0071] The content processing method described below will be introduced with reference to the following examples.

[0072] The computer device pulls the set of content items exposed by each user on the (t+2)th day, item E(u), and the set of content items clicked by the user, item C(u), from the database.

[0073] The computer device invokes the initial content push model to predict the set of predicted content to be pushed to the user on day t+2, which is item P(u), based on the access data from day t-n+1 to day t+1.

[0074] Computer devices can determine the evaluation metrics for the initial content push model based on E(u), C(u), and P(u).

[0075] Assuming the evaluation metric is represented as ctr_offline, the computer device can substitute E(u), C(u), and P(u) into Formula 1 below to calculate the ctr_offline of the initial content push model.

[0076] ctr_offline=∑u∈U∣C(u)∩P(u)∣ / ∑u∈U∣E(u)∩P(u)∣Formula 1;

[0077] Where |C(u)∩P(u)∣ represents calculating the intersection of the predicted content set and the clicked content set, and the number of contents included in the intersection is determined as the actual number of clicks, which is the actual number of visits. |E(u)∩P(u)∣ represents calculating the intersection of the predicted content set and the online exposed content set, and the number of contents included in the intersection is taken as the actual number of exposures, which is the actual number of exposures. Clicks / exposures equals ctr_offline.

[0078] It's important to note that `ctr_offline` essentially simulates a portion of the online performance in advance. This prevents some predicted high-performing content from being removed by subsequent modules, resulting in that content ultimately going unexposed. This avoids a situation where the offline metrics calculated before running the content delivery model are too high, while the online metrics calculated during content prediction are too low. Furthermore, using the intersection of the predicted content set and the exposed content set as the actual exposure count prevents some poorly predicted content from being filtered out by subsequent modules but still counted as exposed content when calculating offline metrics before running the content delivery model. This avoids overestimating the offline metrics calculated before running the content delivery model and thus underestimating its performance.

[0079] The following examples further illustrate the difference between recall and the evaluation metrics determined in the embodiments of this application.

[0080] Suppose that the predicted content set of content recommendation model 1 is {id1,id2,id3,id4,id5,id6,id7,id8,id9,id10}, and the predicted content set of content recommendation model 2 is {id8,id9,id10,id11,id12,id13,id14,id15,id16,id17}. The content set clicked by the user is {id7,id8,id9,id10}. If we calculate the recall, we get recall = 4 / 10 = 0.4 for content recommendation model 1, while the result for content recommendation model 2 is recall = 3 / 10 = 0.3. In this case, content recommendation model 1 is considered superior to content recommendation model 2.

[0081] If the user's online exposure content set is {id1,id2,id3,id4,id5,id6,id7,id8,id9,id10,id11}, then for content push model 1, ctr_offline = |id7,id8,id9,id10| / |id1,id2,id3,id4,id5,id6,id7,id8,id9,id10| = 0.4. For content push model 2, ctr_offline = |id8,id9,id10| / |id8,id9,id10,id11| = 0.75. Clearly, content push model 2 is better in this case. Furthermore, content push model 1 has 6 bad cases, while content push model 2 only has one bad case.

[0082] Therefore, recall is a purely offline metric. Its calculation assumes that all content delivery models will actually expose users to their content, and it compares the performance of these models under this premise. However, this differs significantly from actual delivery scenarios, i.e., the content delivery behavior of the models running online. Especially when the predicted content set obtained from the content delivery model undergoes sorting and blending, it's almost impossible for all predicted content to be exposed. This is because recommendation systems require diversity, and the number of predicted content items in the set is limited. CTR_offline, on the other hand, considers actual online exposure and clicks during calculation, approximating the recall results as having undergone subsequent sorting and filtering, thus making the calculated evaluation metric closer to the online results.

[0083] Please see Figure 3 This is a flowchart illustrating another content processing method provided in an embodiment of this application. This method can be applied to the aforementioned computer devices. Specifically, the method may include the following steps:

[0084] S301. Obtain the first historical access data of the target object, wherein the first historical access data refers to the access data of the target object for the historical exposure content of the target push service within the first historical time period.

[0085] S302. Invoke the initial content push model for the target push service to predict, based on the first historical access data, the first predicted content set for the target push service to be pushed to the target object at the first time.

[0086] S303. Obtain the first set of exposed content about the target push service that has been exposed to the target object in the first time, and the first set of accessed content of the target object; the content in the first set of accessed content belongs to the first set of exposed content.

[0087] S304. Determine the evaluation metrics of the initial content push model based on the first exposed content set, the first accessed content set, and the first predicted content set.

[0088] Steps S301-S304 can be found in [reference needed]. Figure 2 Examples of embodiments are not described in detail here.

[0089] S305. When the number of initial content push models is at least two and the at least two initial content push models include a first initial content push model and a second initial content push model, the first initial content push model is adjusted according to the evaluation index of the first initial content push model to obtain a first candidate content push model.

[0090] S306. Based on the evaluation metrics of the second initial content push model, adjust the second initial content push model to obtain the second candidate content push model.

[0091] In this embodiment, when the number of initial content push models is at least two, and the at least two initial content push models include a first initial content push model and a second initial content push model, the computer device can adjust the model parameters of the first initial content push model according to the evaluation index of the first initial content push model to obtain a first candidate content push model; and adjust the model parameters of the second initial content push model according to the evaluation index of the second initial content push model to obtain a second candidate content push model. Here, the first candidate content push model refers to the candidate content push model obtained after adjusting the first initial content push model. The second candidate content push model refers to the candidate content push model obtained after adjusting the second initial content push model.

[0092] In one embodiment, the computer device may adjust the model parameters of the first initial content push model to obtain a first candidate content push model when the evaluation indicators of the first initial content push model do not meet the requirements; the computer device may also adjust the model parameters of the second initial content push model to obtain a second candidate content push model when the evaluation indicators of the second initial content push model do not meet the requirements.

[0093] In one embodiment, the computer device adjusts the first initial content push model based on the evaluation index of the first initial content push model to obtain a first candidate content push model. This adjustment can be achieved as follows: when the evaluation index of the first initial content push model is less than the evaluation index threshold, the computer device adjusts the model parameters of the first initial content push model based on the evaluation index, resulting in an adjusted first initial content push model; it then acquires the second historical access data of the target object, where the second historical access data refers to the access data of the target object for historical exposure content of the target push service within a second historical time period; and it invokes the adjusted first initial content push model to adjust the first candidate content push model based on the second historical data. Access data to predict a second set of predicted content about the target push service to be pushed to the target object at a second time; obtain a second set of exposed content about the target push service that has already been exposed to the target object at a second time, and a second set of accessed content for the target object; the content in the second set of accessed content belongs to the second set of exposed content; determine the evaluation index of the adjusted first initial content push model based on the second set of exposed content, the second set of accessed content, and the second set of predicted content; if the evaluation index of the adjusted first initial content push model is greater than or equal to the evaluation index threshold, then the adjusted first initial content push model is determined as the first candidate content push model.

[0094] In one embodiment, the computer device adjusts the second initial content push model according to the evaluation index of the second initial content push model to obtain a second candidate content push model. This can be achieved by: when the evaluation index of the second initial content push model is less than the evaluation index threshold, the computer device adjusts the model parameters of the second initial content push model according to the evaluation index of the second initial content push model to obtain an adjusted second initial content push model; acquiring the second historical access data of the target object, where the second historical access data refers to the access data of the target object for historical exposure content of the target push service within a second historical time period; and calling the adjusted second initial content push model to adjust the second candidate content push model according to the second historical data. Historical access data is used to predict a second set of predicted content about the target push service to be pushed to the target object at a second time; a second set of exposed content about the target push service that has already been exposed to the target object at the second time, and a second set of access content of the target object are obtained; the content in the second set of access content belongs to the second set of exposed content; an evaluation index for the adjusted second initial content push model is determined based on the second set of exposed content, the second set of access content, and the second set of predicted content; if the evaluation index of the adjusted second initial content push model is greater than or equal to the evaluation index threshold, then the adjusted second initial content push model is determined as a second candidate content push model. The method by which the computer device determines the evaluation index of the adjusted second initial content push model based on the second set of exposed content, the second set of access content, and the second set of predicted content can be found in the aforementioned method of determining the evaluation index of the adjusted initial content push model based on the first set of exposed content, the first set of access content, and the first set of predicted content, which will not be elaborated upon in this embodiment.

[0095] In one embodiment, the data type of the historical access data input to the first initial content push model is different from the data type of the historical access data input to the second initial content push model. Specifically, the data type of the first historical access data input to the first initial content push model is different from the data type of the first historical access data input to the second initial content push model. Correspondingly, the data type of the second historical access data input to the first initial content push model is different from the data type of the second historical access data input to the second initial content push model. In other words, the push strategy used by the first initial content push model is different from the push strategy used by the second initial content push model. In one embodiment, in addition to inputting historical access data into the corresponding content push model for content prediction, the object data of the target object can also be input into the corresponding content push model for content prediction.

[0096] S307. Obtain the evaluation metrics of the first candidate content push model and the evaluation metrics of the second candidate content push model.

[0097] The specific method for obtaining the evaluation metrics of the first candidate content push model can be found in the method for obtaining the evaluation metrics of the first initial content push model. Similarly, the specific method for obtaining the evaluation metrics of the second candidate content push model can be found in the method for obtaining the evaluation metrics of the first initial content. Further details are omitted here.

[0098] S308. If the evaluation index of the first candidate content push model is greater than or equal to the evaluation index of the second candidate content push model, then the first candidate content push model is determined as the target content push model.

[0099] S309. If the evaluation index of the first candidate content push model is less than the evaluation index of the second candidate content push model, then the second candidate content push model is determined as the target content push model.

[0100] In other words, the computer device can select a candidate content push model from the first and second candidate content push models based on their respective evaluation metrics, and use this selection as the target content push model. Specifically, the computer device can determine the first candidate content push model as the target content push model when its evaluation metric is greater than or equal to that of the second candidate content push model. Alternatively, the computer device can determine the second candidate content push model as the target content push model when its evaluation metric is less than that of the second candidate content push model. This application employs this process, allowing the computer device to select the candidate content push model with the better evaluation metric as the target content push model, thereby obtaining a superior target content push model and improving the accuracy of content push to the target audience.

[0101] In one application scenario, if the target push service is to push e-books, the computer device can obtain the third historical access data of the target object. This third historical access data refers to the access data of the target object for historically exposed e-books within a third historical time period. The device then invokes the target content push model to predict the set of e-books to be pushed to the target object at the third time based on the access data. Finally, it pushes the e-books to the target object based on the set of e-books to be pushed. Here, the third historical time period also refers to a past time period. The third time can be any time later than the end time of the third historical time period, or a specified time. In one embodiment, the third time can be the day after the end time included in the third historical time period.

[0102] For example, a computer device might expose multiple ebooks to a user between October 1st and October 7th. The user can click on specific ebooks within these ebooks to access them. The computer device can record the user's access data from October 1st to October 7th. Assume the access data indicates that the user accessed ebooks 1, 2, 3, 4, and 5 during this period. The computer device can then invoke a targeted content recommendation model to predict the set of ebooks to be recommended to the user on October 8th based on this access data. The computer device can then recommend ebooks to the target user based on this recommended set of ebooks.

[0103] When the above embodiments of this application are applied to specific products or technologies, user permission or consent is required, and the collection, use, and processing of data such as the target object's access data and access content sets must comply with the relevant laws, regulations, and standards of the relevant countries and regions. In other words, a computer device can only acquire this information when it obtains authorization from the user regarding the aforementioned information.

[0104] For example, a computer device displays a permission prompt interface on the terminal corresponding to the target object. This permission prompt interface is used to prompt the user that the access data and access content set of the target object are being collected. The collection of this information will begin after the user confirms the permission prompt interface; otherwise, it will end.

[0105] Please see Figure 4 This is a schematic diagram of the structure of a content processing apparatus provided in an embodiment of this application. The apparatus can be applied to the aforementioned computer devices. Specifically, the apparatus may include an acquisition module 401, a prediction module 402, a determination module 403, an adjustment module 404, and a push module 405.

[0106] The acquisition module is used to acquire the first historical access data of the target object. The first historical access data refers to the access data of the target object for the historical exposure content of the target push service within a first historical time period.

[0107] The prediction module is used to invoke the initial content push model for the target push service, so as to predict the first set of predicted content about the target push service to be pushed to the target object at the first time based on the first historical access data.

[0108] The acquisition module is further configured to acquire a first set of exposure content about the target push service that has been exposed to the target object in the first time, and a first set of access content for the target object; the content in the first set of access content belongs to the first set of exposure content.

[0109] The determination module is used to determine the evaluation index of the initial content push model based on the first exposed content set, the first accessed content set, and the first predicted content set.

[0110] An adjustment module is used to adjust the initial content push model according to the evaluation indicators of the initial content push model to obtain a target content push model, which is used to push content about the target push service to the target object.

[0111] In one embodiment, the determining module determines the evaluation metrics of the initial content push model based on the first exposed content set, the first accessed content set, and the first predicted content set, specifically as follows:

[0112] Based on the first set of exposed content and the first set of predicted content, determine the actual exposure of the target object with respect to the target push service;

[0113] Based on the first set of accessed content and the first set of predicted content, determine the actual number of accesses of the target object to the target push service;

[0114] The evaluation metrics for the initial content push model are determined based on the actual number of visits and the actual number of exposures.

[0115] In one embodiment, the determining module determines the actual exposure volume of the target object with respect to the target push service based on the first exposure content set and the first predicted content set, specifically as follows:

[0116] Determine the first content intersection between the first exposed content set and the first predicted content set;

[0117] Count the number of contents included in the first content intersection;

[0118] The number of contents included in the first content intersection is determined as the actual exposure of the target object with respect to the target push service.

[0119] In one embodiment, the determining module determines the actual access volume of the target object regarding the target push service based on the first access content set and the first predicted content set, specifically as follows:

[0120] Determine the second content intersection between the first accessed content set and the first predicted content set;

[0121] Count the number of contents included in the second content intersection;

[0122] The number of contents included in the second content intersection is determined as the actual number of visits by the target object to the target push service.

[0123] In one embodiment, the determining module determines the evaluation metrics of the initial content push model based on the actual number of visits and the actual number of exposures, specifically:

[0124] Calculate the ratio between the actual number of visits and the actual number of impressions;

[0125] The ratio between the actual number of visits and the actual number of exposures is determined as the visit conversion index, and the visit conversion index is determined as the evaluation index of the initial content push model.

[0126] In one embodiment, the number of initial content push models is at least two, including a first initial content push model and a second initial content push model. The adjustment module adjusts the initial content push models according to the evaluation metrics of the initial content push models to obtain the target content push model, specifically as follows:

[0127] Based on the evaluation metrics of the first initial content push model, the first initial content push model is adjusted to obtain the first candidate content push model;

[0128] Based on the evaluation metrics of the second initial content push model, the second initial content push model is adjusted to obtain the second candidate content push model;

[0129] Obtain the evaluation metrics of the first candidate content push model and the evaluation metrics of the second candidate content push model;

[0130] If the evaluation index of the first candidate content push model is greater than or equal to the evaluation index of the second candidate content push model, then the first candidate content push model is determined as the target content push model.

[0131] If the evaluation index of the first candidate content push model is less than the evaluation index of the second candidate content push model, then the second candidate content push model will be determined as the target content push model.

[0132] In one embodiment, the adjustment module adjusts the first initial content push model based on the evaluation metrics of the first initial content push model to obtain a first candidate content push model, specifically as follows:

[0133] If the evaluation index of the first initial content push model is less than the evaluation index threshold, the model parameters of the first initial content push model are adjusted according to the evaluation index of the first initial content push model to obtain the adjusted first initial content push model.

[0134] Obtain the second historical access data of the target object, wherein the second historical access data refers to the access data of the target object for the historical exposure content of the target push service within a second historical time period;

[0135] The adjusted first initial content push model is invoked to predict, based on the second historical access data, a second set of predicted content about the target push service to be pushed to the target object at a second time.

[0136] Obtain a second set of exposure content related to the target push service that has been exposed to the target object at a second time, and a second set of access content for the target object; the content in the second set of access content belongs to the second set of exposure content.

[0137] The evaluation metrics of the adjusted first initial content push model are determined based on the second exposed content set, the second accessed content set, and the second predicted content set.

[0138] If the evaluation index of the adjusted first initial content push model is greater than or equal to the evaluation index threshold, then the adjusted first initial content push model is determined as the first candidate content push model.

[0139] In one embodiment, the device further includes a push module, wherein the target push service is pushing e-books, and the push module is configured to:

[0140] Obtain the third historical access data of the target object, wherein the third historical access data refers to the access data of the target object for historically exposed e-books within a third historical time period;

[0141] The target content push model is invoked to predict the set of ebooks to be pushed to the target object at a third time based on the access data;

[0142] Based on the set of ebooks to be pushed, ebooks are pushed to the target object.

[0143] visible, Figure 4 In the illustrated embodiment, the content processing device can acquire first historical access data of the target object and invoke an initial content push model for the target push service to predict a first predicted content set of the target push service to be pushed to the target object at a first time based on the first historical access data. The content processing device can also acquire a first exposed content set of the target push service that has already been exposed to the target object at a first time, and a first access content set of the target object, wherein the content in the first access content set belongs to the first exposed content set. The content processing device can further determine the evaluation index of the initial content push model based on the first exposed content set, the first access content set, and the first predicted content set, so as to adjust the initial content push model using the evaluation index of the initial content push model to obtain the target content push model. This process can effectively improve the content prediction accuracy of the target content push model.

[0144] Please see Figure 5 This is a schematic diagram of the structure of a computer device provided in an embodiment of this application. Figure 5 As shown, the aforementioned computer device 1000 may include: a processor 1001, a network interface 1004, and a memory 1005. Furthermore, the computer device 1000 may also include: at least one communication bus 1002 and a user interface 1003. The communication bus 1002 is used to implement communication between these components. The user interface 1003 may include a display screen and a keyboard; the user interface 1003 may also include a standard wired interface and a wireless interface. The network interface 1004 may optionally include a standard wired interface and a wireless interface (such as a Wi-Fi interface). The memory 1005 may be high-speed RAM or non-volatile memory, such as at least one disk storage device. The memory 1005 may also optionally be at least one storage device located remotely from the aforementioned processor 1001. Figure 5 As shown, the memory 1005, which is a computer-readable storage medium, may include an operating system, a network communication module, a user interface module, and a device control application.

[0145] exist Figure 5 In the computer device 1000 shown, the network interface 1004 can send and receive data, such as receiving requests or pushing content; while the processor 1001 can be used to call the device control application stored in the memory 1005 to achieve:

[0146] Obtain the first historical access data of the target object, wherein the first historical access data refers to the access data of the target object for the historical exposure content of the target push service within the first historical time period;

[0147] Invoke the initial content push model for the target push service to predict, based on the first historical access data, the first predicted set of content related to the target push service to be pushed to the target object at the first time.

[0148] Obtain the first set of exposure content related to the target push service that has been exposed to the target object in the first time, and the first set of access content of the target object; the content in the first set of access content belongs to the first set of exposure content.

[0149] The evaluation metrics for the initial content push model are determined based on the first exposed content set, the first accessed content set, and the first predicted content set.

[0150] Based on the evaluation metrics of the initial content push model, the initial content push model is adjusted to obtain a target content push model, which is used to push content about the target push service to the target object.

[0151] In one embodiment, when determining the evaluation metrics of the initial content push model based on the first exposed content set, the first accessed content set, and the first predicted content set, the processor 1001 can be used to call the device control application stored in the memory 1005, specifically implemented as follows:

[0152] Based on the first set of exposed content and the first set of predicted content, determine the actual exposure of the target object with respect to the target push service;

[0153] Based on the first set of accessed content and the first set of predicted content, determine the actual number of accesses of the target object to the target push service;

[0154] The evaluation metrics for the initial content push model are determined based on the actual number of visits and the actual number of exposures.

[0155] In one embodiment, when determining the actual exposure volume of the target object with respect to the target push service based on the first exposure content set and the first predicted content set, the processor 1001 can be used to call the device control application stored in the memory 1005, specifically implemented as follows:

[0156] Determine the first content intersection between the first exposed content set and the first predicted content set;

[0157] Count the number of contents included in the first content intersection;

[0158] The number of contents included in the first content intersection is determined as the actual exposure of the target object with respect to the target push service.

[0159] In one embodiment, when determining the actual access volume of the target object regarding the target push service based on the first access content set and the first predicted content set, the processor 1001 can be used to call the device control application stored in the memory 1005, specifically implemented as follows:

[0160] Determine the second content intersection between the first accessed content set and the first predicted content set;

[0161] Count the number of contents included in the second content intersection;

[0162] The number of contents included in the second content intersection is determined as the actual number of visits by the target object to the target push service.

[0163] In one embodiment, when determining the evaluation metrics of the initial content push model based on the actual number of visits and the actual number of exposures, the processor 1001 can be used to call the device control application stored in the memory 1005, specifically implemented as follows:

[0164] Calculate the ratio between the actual number of visits and the actual number of impressions;

[0165] The ratio between the actual number of visits and the actual number of exposures is determined as the visit conversion index, and the visit conversion index is determined as the evaluation index of the initial content push model.

[0166] In one embodiment, the number of initial content push models is at least two, including a first initial content push model and a second initial content push model. When the initial content push models are adjusted according to the evaluation indicators of the initial content push models to obtain the target content push model, the processor 1001 can be used to call the device control application stored in the memory 1005, specifically implemented as follows:

[0167] Based on the evaluation metrics of the first initial content push model, the first initial content push model is adjusted to obtain the first candidate content push model;

[0168] Based on the evaluation metrics of the second initial content push model, the second initial content push model is adjusted to obtain the second candidate content push model;

[0169] Obtain the evaluation metrics of the first candidate content push model and the evaluation metrics of the second candidate content push model;

[0170] If the evaluation index of the first candidate content push model is greater than or equal to the evaluation index of the second candidate content push model, then the first candidate content push model is determined as the target content push model.

[0171] If the evaluation index of the first candidate content push model is less than the evaluation index of the second candidate content push model, then the second candidate content push model will be determined as the target content push model.

[0172] In one embodiment, when adjusting the first initial content push model based on the evaluation metrics of the first initial content push model to obtain a first candidate content push model, the processor 1001 can be used to call the device control application stored in the memory 1005, specifically implemented as follows:

[0173] If the evaluation index of the first initial content push model is less than the evaluation index threshold, the model parameters of the first initial content push model are adjusted according to the evaluation index of the first initial content push model to obtain the adjusted first initial content push model.

[0174] Obtain the second historical access data of the target object, wherein the second historical access data refers to the access data of the target object for the historical exposure content of the target push service within a second historical time period;

[0175] The adjusted first initial content push model is invoked to predict, based on the second historical access data, a second set of predicted content about the target push service to be pushed to the target object at a second time.

[0176] Obtain a second set of exposure content related to the target push service that has been exposed to the target object at a second time, and a second set of access content for the target object; the content in the second set of access content belongs to the second set of exposure content.

[0177] The evaluation metrics of the adjusted first initial content push model are determined based on the second exposed content set, the second accessed content set, and the second predicted content set.

[0178] If the evaluation index of the adjusted first initial content push model is greater than or equal to the evaluation index threshold, then the adjusted first initial content push model is determined as the first candidate content push model.

[0179] In one embodiment, the target push service is pushing e-books, and the processor 1001 can be used to call the device control application stored in the memory 1005, specifically implemented as follows:

[0180] Obtain the third historical access data of the target object, wherein the third historical access data refers to the access data of the target object for historically exposed e-books within a third historical time period;

[0181] The target content push model is invoked to predict the set of ebooks to be pushed to the target object at a third time based on the access data;

[0182] Based on the set of ebooks to be pushed, ebooks are pushed to the target object.

[0183] visible, Figure 5In the illustrated embodiment, the computer device can acquire first historical access data of the target object and invoke an initial content push model for the target push service to predict a first predicted content set of the target push service to be pushed to the target object at a first time based on the first historical access data. The computer device can also acquire a first exposed content set of the target push service that has already been exposed to the target object at a first time, as well as a first access content set of the target object, wherein the content in the first access content set belongs to the first exposed content set. The computer device can further determine the evaluation index of the initial content push model based on the first exposed content set, the first access content set, and the first predicted content set, so as to adjust the initial content push model using the evaluation index of the initial content push model to obtain the target content push model. This process can effectively improve the content prediction accuracy of the target content push model.

[0184] It should be understood that the computer device 1000 described in the embodiments of this application can execute the foregoing text. Figure 2 or Figure 3 The description of the content processing method in the embodiments can also be performed as described above. Figure 4 The description of the content processing apparatus in the corresponding embodiments will not be repeated here. Furthermore, the beneficial effects of using the same method will also not be repeated.

[0185] This application also provides a computer-readable storage medium storing a computer program executed by the aforementioned content processing apparatus. The computer program includes program instructions, which, when executed by the processor, enable the execution of the aforementioned content processing apparatus. Figure 2 Implementation examples or Figure 3 The description of the above-described content processing method in the embodiments will not be repeated here. Furthermore, the beneficial effects of using the same method will also not be repeated. For technical details not disclosed in the embodiments of the computer-readable storage medium involved in this application, please refer to the description of the method embodiments of this application.

[0186] As an example, the above program instructions can be deployed and executed on a computer device, or deployed and executed on at least two computer devices in one location, or executed on at least two computer devices distributed in at least two locations and interconnected by a communication network. At least two computer devices distributed in at least two locations and interconnected by a communication network can form a blockchain network.

[0187] The aforementioned computer-readable storage medium can be a content processing apparatus provided in any of the foregoing embodiments or a central storage unit of the aforementioned computer device, such as a hard disk or central storage of the computer device. The computer-readable storage medium can also be an external storage device of the computer device, such as a plug-in hard disk, smart media card (SMC), secure digital (SD) card, flash card, etc., provided on the computer device. Furthermore, the computer-readable storage medium may include both the central storage unit and external storage devices of the computer device. The computer-readable storage medium is used to store the computer program and other programs and data required by the computer device. The computer-readable storage medium can also be used to temporarily store data that has been output or will be output.

[0188] The terms "first," "second," etc., in the specification, claims, and drawings of this application are used to distinguish content in different media, rather than to describe a specific order. Furthermore, the term "comprising," and any variations thereof, are intended to cover non-exclusive inclusion. For example, a process, method, apparatus, product, or device that includes a series of steps or units is not limited to the listed steps or modules, but may optionally include steps or modules not listed, or may optionally include other step units inherent to these processes, methods, apparatuses, products, or devices.

[0189] This application also provides a computer program product, including a computer program / instructions, which, when executed by a processor, implement the foregoing description. Figure 2 or Figure 3 The description of the above-described content processing method in the embodiments will not be repeated here. Furthermore, the beneficial effects of using the same method will also not be repeated. For technical details not disclosed in the embodiments of the computer program product involved in this application, please refer to the description of the method embodiments of this application.

[0190] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this application.

[0191] The methods and related apparatus provided in this application are described with reference to the method flowcharts and / or structural diagrams provided in this application. Specifically, each block of the method flowchart and / or structural diagram, as well as combinations of blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable network-connected device to create a machine, such that the instructions, which execute via the processor of the computer or other programmable network-connected device, generate instructions for implementing the process. Figure 1 A schematic diagram of one or more processes and / or structures. Figure 1 The computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable network-connected device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 A schematic diagram of one or more processes and / or structures. Figure 1 The functions specified in one or more boxes. These computer program instructions may also be loaded onto a computer or other programmable network-connected device to cause a series of operational steps to be performed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable device for implementing the process. Figure 1 A process or multiple processes and / or structures illustrate the steps of the functions specified in one or more boxes.

[0192] The above-disclosed embodiments are merely preferred embodiments of this application and should not be construed as limiting the scope of this application. Therefore, any equivalent variations made in accordance with the claims of this application shall still fall within the scope of this application.

Claims

1. A content processing method, characterized in that, include: Obtain the first historical access data of the target object, wherein the first historical access data refers to the access data of the target object for the historical exposure content of the target push service within the first historical time period; Invoke the initial content push model for the target push service to predict, based on the first historical access data, the first predicted set of content related to the target push service to be pushed to the target object at the first time. Obtain the first set of exposure content related to the target push service that has been exposed to the target object in the first time, and the first set of access content of the target object; the content in the first set of access content belongs to the first set of exposure content. Based on the first set of exposed content and the first set of predicted content, determine the actual exposure of the target object with respect to the target push service; Based on the first set of accessed content and the first set of predicted content, determine the actual number of accesses of the target object to the target push service; Calculate the ratio between the actual number of visits and the actual number of impressions; The ratio between the actual number of visits and the actual number of exposures is determined as the visit conversion index, and the visit conversion index is determined as the evaluation index of the initial content push model. Based on the evaluation metrics of the initial content push model, the initial content push model is adjusted to obtain a target content push model, which is used to push content about the target push service to the target object.

2. The method according to claim 1, characterized in that, Determining the actual exposure volume of the target object with respect to the target push service based on the first exposure content set and the first predicted content set includes: Determine the first content intersection between the first exposed content set and the first predicted content set; Count the number of contents included in the first content intersection; The number of contents included in the first content intersection is determined as the actual exposure of the target object with respect to the target push service.

3. The method according to claim 1 or 2, characterized in that, Determining the actual access volume of the target object regarding the target push service based on the first access content set and the first predicted content set includes: Determine the second content intersection between the first accessed content set and the first predicted content set; Count the number of contents included in the second content intersection; The number of contents included in the second content intersection is determined as the actual number of visits by the target object to the target push service.

4. The method according to claim 1, characterized in that, The number of initial content push models is at least two, including a first initial content push model and a second initial content push model. The step of adjusting the initial content push models according to their evaluation metrics to obtain the target content push model includes: Based on the evaluation metrics of the first initial content push model, the first initial content push model is adjusted to obtain the first candidate content push model; Based on the evaluation metrics of the second initial content push model, the second initial content push model is adjusted to obtain the second candidate content push model; Obtain the evaluation metrics of the first candidate content push model and the evaluation metrics of the second candidate content push model; If the evaluation index of the first candidate content push model is greater than or equal to the evaluation index of the second candidate content push model, then the first candidate content push model is determined as the target content push model. If the evaluation index of the first candidate content push model is less than the evaluation index of the second candidate content push model, then the second candidate content push model will be determined as the target content push model.

5. The method according to claim 4, characterized in that, The step of adjusting the first initial content push model according to the evaluation metrics of the first initial content push model to obtain the first candidate content push model includes: If the evaluation index of the first initial content push model is less than the evaluation index threshold, the model parameters of the first initial content push model are adjusted according to the evaluation index of the first initial content push model to obtain the adjusted first initial content push model. Obtain the second historical access data of the target object, wherein the second historical access data refers to the access data of the target object for the historical exposure content of the target push service within a second historical time period; The adjusted first initial content push model is invoked to predict, based on the second historical access data, a second set of predicted content about the target push service to be pushed to the target object at a second time. Obtain a second set of exposure content related to the target push service that has been exposed to the target object at a second time, and a second set of access content for the target object; the content in the second set of access content belongs to the second set of exposure content. The evaluation metrics of the adjusted first initial content push model are determined based on the second exposed content set, the second accessed content set, and the second predicted content set. If the evaluation index of the adjusted first initial content push model is greater than or equal to the evaluation index threshold, then the adjusted first initial content push model is determined as the first candidate content push model.

6. The method according to claim 1, characterized in that, The target push service is to push e-books, and the method further includes: Obtain the third historical access data of the target object, wherein the third historical access data refers to the access data of the target object for historically exposed e-books within a third historical time period; The target content push model is invoked to predict the set of ebooks to be pushed to the target object at a third time based on the access data; Based on the set of ebooks to be pushed, ebooks are pushed to the target object.

7. A content processing apparatus, characterized in that, include: The acquisition module is used to acquire the first historical access data of the target object. The first historical access data refers to the access data of the target object for the historical exposure content of the target push service within the first historical time period. The prediction module is used to invoke the initial content push model for the target push service, so as to predict the first set of predicted content about the target push service to be pushed to the target object at the first time based on the first historical access data. The acquisition module is further configured to acquire a first set of exposure content about the target push service that has been exposed to the target object in the first time, and a first set of access content of the target object; the content in the first set of access content belongs to the first set of exposure content. The determining module is configured to determine the actual exposure volume of the target object with respect to the target push service based on the first exposure content set and the first predicted content set; and to determine the actual access volume of the target object with respect to the target push service based on the first access content set and the first predicted content set. Calculate the ratio between the actual number of visits and the actual number of impressions; The ratio between the actual number of visits and the actual number of exposures is determined as the visit conversion index, and the visit conversion index is determined as the evaluation index of the initial content push model. An adjustment module is used to adjust the initial content push model according to the evaluation indicators of the initial content push model to obtain a target content push model, which is used to push content about the target push service to the target object.

8. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 6.

10. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 6.

Images