Advertisement recommendation method, apparatus, device, medium, and program product

Abstract

The application provides an advertisement recommendation method and device, equipment, medium and program product. It can be applied to the fields of artificial intelligence, Internet and information recommendation. The advertisement recommendation method comprises the following steps: predicting a product sequence of interest of a target user in response to an interactive operation of the target user; screening a plurality of advertisements to determine a candidate advertisement sequence matched with the target user; taking each product in the product sequence as a reference object, determining a target advertisement matched with the product from the candidate advertisement sequence to obtain a plurality of target advertisements; and rearranging the plurality of target advertisements based on the arrangement order of each product in the product sequence to obtain a target advertisement sequence for recommending to the target user.

Description

Technical Field

[0001] This application relates to the fields of artificial intelligence, the Internet and information recommendation technology, and specifically to an advertising recommendation method, apparatus, device, medium and program product. Background Technology

[0002] With the rapid development of internet and artificial intelligence technologies, ad recommendation has become a crucial component of internet platforms, directly impacting user experience, advertiser conversion rates, and platform revenue. Currently, most mainstream ad recommendation systems employ a cascaded pipeline model. This model progressively filters ads through independent stages such as candidate recall, coarse ranking, fine ranking, and re-ranking, pushing out a sequence of candidate ads. While this reduces computational complexity and enables rapid delivery, it suffers from information decay and local optima issues, affecting recommendation accuracy. Summary of the Invention

[0003] In view of this, this application provides an advertising recommendation method, apparatus, device, medium, and program product.

[0004] One aspect of this application provides an advertising recommendation method, comprising: predicting a product sequence that the target user is interested in in response to an interactive operation of the target user; filtering multiple advertisements to determine a candidate advertisement sequence that matches the target user; determining target advertisements that match the products from the candidate advertisement sequence, using each product in the product sequence as a reference, to obtain multiple target advertisements; and rearranging the multiple target advertisements based on the order of the products in the product sequence to obtain a target advertisement sequence for recommendation to the target user.

[0005] According to embodiments of this application, predicting a product sequence that a target user is interested in includes: identifying historical products in the target user's historical behavior information to obtain multiple historical products; determining prompt information to represent the multiple historical products based on their respective attribute information; and using a large model to predict the products that the target user is interested in based on the prompt information to obtain a product sequence.

[0006] According to embodiments of this application, predicting a product sequence that a target user is interested in includes: determining the target user's favorite products from the operation object targeted by the interaction operation; determining prompt information to represent the favorite products based on the attribute information of the favorite products; and using a large model to predict the products that the target user is interested in based on the prompt information to obtain a product sequence.

[0007] According to an embodiment of this application, a large model is used to predict products of interest to a target user based on prompt information to obtain a product sequence. This includes: inputting prompt information into the large model to determine candidate products of interest to the target user, wherein candidate products include historical products from the target user's historical behavior information or products of interest involved in the target user's interactive operations; determining associated products that can be combined with the candidate products to obtain a product group including the candidate products and associated products; and ranking the product group based on the target user's level of interest in the candidate products and associated products to obtain a product sequence.

[0008] According to an embodiment of this application, taking each product in the product sequence as a reference, a target advertisement matching the product is determined from the candidate advertisement sequence to obtain multiple target advertisements, including: for each product in the product sequence, the following matching process is performed: based on the effect prediction index of each candidate advertisement in the candidate advertisement sequence and the semantic similarity between the product and each candidate advertisement, a target advertisement matching the product is determined, and the effect prediction index is used to characterize the importance of the candidate advertisement to the target user.

[0009] According to embodiments of this application, when performing the matching process for each product in the product sequence, the method further includes: removing the target advertisement that matches the product from the candidate advertisement sequence.

[0010] According to an embodiment of this application, when there are multiple product sequences and multiple target ad sequences, the ad recommendation method further includes: inputting multiple target ad sequences into an evaluation model and outputting an evaluation value for each target ad sequence, wherein the evaluation model is trained based on the historical interaction information of sample users on each of the multiple sample ad sequences; determining a recommended ad sequence from the multiple target ad sequences based on the evaluation values ​​of each target ad sequence; and recommending the recommended ad sequence to the target user.

[0011] According to an embodiment of this application, the advertising recommendation method further includes: when a candidate advertising sequence is determined, timing is performed; when a product sequence is not determined within a predetermined time period, multiple candidate advertisements are rearranged based on the effect prediction indicators of each candidate advertisement to obtain a rearrangement result; and based on the rearrangement result, a predetermined number of candidate advertisements are selected to obtain a target advertising sequence.

[0012] Another aspect of this application provides an advertising recommendation device, comprising: a prediction module for predicting a product sequence that the target user is interested in in response to an interactive operation of the target user; a first determination module for filtering multiple advertisements to determine a candidate advertisement sequence that matches the target user; a second determination module for determining target advertisements that match the products in the candidate advertisement sequence, using each product in the product sequence as a reference, to obtain multiple target advertisements; and a rearrangement module for rearranging the multiple target advertisements based on the order of the products in the product sequence to obtain a target advertisement sequence for recommendation to the target user.

[0013] Another aspect of this application provides an electronic device comprising: one or more processors; and a memory for storing one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors cause the one or more processors to implement the advertising recommendation method described above.

[0014] Another aspect of this application provides a computer-readable storage medium storing computer-executable instructions, which, when executed, are used to implement the above-described advertising recommendation method.

[0015] Another aspect of this application provides a computer program product including computer-executable instructions that, when executed, implement the aforementioned advertising recommendation method.

[0016] According to embodiments of this application, by responding to the interactive operations of target users in real time, the system predicts the product sequence that the target users are interested in, and uses each product in the product sequence as a reference to determine the target advertisements that match the products from the candidate advertisement sequence. This preserves information about user interests and reduces information loss during the filtering process. Furthermore, based on the arrangement order of the products in the product sequence, multiple target advertisements are rearranged, achieving global optimization of advertisement recommendation and at least partially overcoming the limitations of independent optimization at each stage of traditional cascading pipelines. This not only provides target users with advertisements tailored to their specific needs, improving user experience, but also increases ad click-through rates and conversion rates, optimizing advertiser performance and platform advertising revenue. Attached Figure Description

[0017] The above and other objects, features and advantages of this application will become clearer from the following description of embodiments with reference to the accompanying drawings, in which:

[0018] Figure 1 The illustrations depict application scenarios of advertising recommendation methods, apparatuses, devices, media, and program products according to embodiments of this application.

[0019] Figure 2 A flowchart illustrating an advertising recommendation method according to an embodiment of this application is shown schematically;

[0020] Figure 3 This illustration schematically shows a sequence of products that a target user is interested in, according to an embodiment of this application.

[0021] Figure 4 This illustration schematically shows a sequence of products that a target user is interested in, according to another embodiment of this application.

[0022] Figure 5 This illustration schematically depicts an advertising recommendation diagram according to an embodiment of this application;

[0023] Figure 6 A flowchart illustrating an advertising recommendation method according to another embodiment of this application is shown schematically;

[0024] Figure 7 A block diagram schematically illustrates an advertising recommendation device according to an embodiment of this application; and

[0025] Figure 8 A block diagram of an electronic device suitable for implementing an advertising recommendation method according to an embodiment of this application is illustrated schematically. Detailed Implementation

[0026] The embodiments of this application will now be described with reference to the accompanying drawings. However, it should be understood that these descriptions are exemplary only and are not intended to limit the scope of this application. In the following detailed description, numerous specific details are set forth to provide a thorough understanding of the embodiments of this application for ease of explanation. However, it will be apparent that one or more embodiments may be implemented without these specific details. Furthermore, descriptions of well-known structures and technologies are omitted in the following description to avoid unnecessarily obscuring the concepts of this application.

[0027] The terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the scope of this application. The terms “comprising,” “including,” etc., as used herein indicate the presence of features, steps, operations, and / or components, but do not exclude the presence or addition of one or more other features, steps, operations, or components.

[0028] All terms used herein (including technical and scientific terms) have the meanings commonly understood by those skilled in the art, unless otherwise defined. It should be noted that the terms used herein are to be interpreted in a manner consistent with the context of this specification, and not in an idealized or overly rigid way.

[0029] When using expressions such as "at least one of A, B and C", they should generally be interpreted in accordance with the meaning that is commonly understood by those skilled in the art (e.g., "a system having at least one of A, B and C" should include, but is not limited to, a system having A alone, a system having B alone, a system having C alone, a system having A and B, a system having A and C, a system having B and C, and / or a system having A, B and C, etc.).

[0030] In the embodiments of this application, the collection, updating, analysis, processing, use, transmission, provision, disclosure, and storage of data (e.g., including but not limited to user personal information) comply with relevant laws and regulations, are used for legitimate purposes, and do not violate public order and good morals. In particular, necessary measures have been taken to prevent unauthorized access to user personal information data and to safeguard user personal information security and network security.

[0031] In the embodiments of this application, the user's authorization or consent was obtained before obtaining or collecting the user's personal information.

[0032] In realizing the concept of this application, it was found that related technologies for online advertising systems typically employ a cascading pipeline model, sequentially performing four stages: recall, coarse ranking, fine ranking, and re-ranking. Each stage must complete before the next can begin, forming an information transmission chain. While this information transmission chain ensures system efficiency and stability to a certain extent, its shortcomings gradually become apparent as user needs become increasingly complex and diverse. For example, during the information transmission process, the understanding of user intent becomes increasingly one-sided as the stages progress. In pursuit of efficiency, the coarse ranking stage often uses simplified models and features, leading to the premature filtering of many potentially relevant candidate ads. Although the fine ranking stage uses more complex models, it only focuses on the value prediction of individual ads, ignoring the correlation and combined value between ads. By the time the re-ranking stage is reached, the system can only make minor adjustments within a severely limited candidate set, failing to achieve true global optimization. Furthermore, while the fine ranking stage scores each ad independently, the re-ranking stage can only make simple rule adjustments to the Top-K results from the fine ranking stage. This design prevents the system from jointly optimizing the entire presentation sequence, such as failing to recognize the functional complementarity between products, failing to understand the user's complete scenario needs, and failing to generate solutions that meet the user's overall intent.

[0033] In view of this, embodiments of this application provide an advertising recommendation method, including: predicting a product sequence that the target user is interested in in response to an interactive operation of the target user; filtering multiple advertisements to determine a candidate advertisement sequence that matches the target user; using each product in the product sequence as a reference object, determining target advertisements that match the products from the candidate advertisement sequence to obtain multiple target advertisements; and rearranging the multiple target advertisements based on the order of the products in the product sequence to obtain a target advertisement sequence for recommendation to the target user.

[0034] Figure 1 The illustrations depict application scenarios of advertising recommendation methods, apparatuses, devices, media, and program products according to embodiments of this application.

[0035] like Figure 1 As shown, the application scenario 100 according to this embodiment may include a first terminal device 101, a second terminal device 102, a third terminal device 103, a network 104, and a server 105.

[0036] Network 104 serves as a medium for providing a communication link between the first terminal device 101, the second terminal device 102, the third terminal device 103, and the server 105. Network 104 may include various connection types, such as wired or wireless communication links or fiber optic cables, etc.

[0037] Users can use the first terminal device 101, the second terminal device 102, and the third terminal device 103 to interact with the server 105 via the network 104 to receive or send messages, etc. Various communication client applications can be installed on the first terminal device 101, the second terminal device 102, and the third terminal device 103, such as shopping applications, web browser applications, search applications, instant messaging tools, email clients, social media platform software, etc. (for example only).

[0038] The first terminal device 101, the second terminal device 102, and the third terminal device 103 can be various electronic devices with displays and support web browsing, including but not limited to smartphones, tablets, laptops, and desktop computers.

[0039] Server 105 can be a server that provides various services, such as a backend management server that supports websites browsed by users using the first terminal device 101, the second terminal device 102, and the third terminal device 103 (this is just an example). The backend management server can analyze and process data such as received user requests, and feed back the processing results (such as web pages, information, or data obtained or generated according to user requests) to the terminal devices.

[0040] It should be noted that the advertising recommendation method provided in this application embodiment can generally be executed by server 105. Correspondingly, the advertising recommendation device provided in this application embodiment can generally be located in server 105. The advertising recommendation method provided in this application embodiment can also be executed by a server or server cluster that is different from server 105 and capable of communicating with the first terminal device 101, the second terminal device 102, the third terminal device 103, and / or server 105. Correspondingly, the advertising recommendation device provided in this application embodiment can also be located in a server or server cluster that is different from server 105 and capable of communicating with the first terminal device 101, the second terminal device 102, the third terminal device 103, and / or server 105.

[0041] It should be understood that Figure 1 The number of terminal devices, servers, and networks shown is merely illustrative. Depending on implementation needs, any number of terminal devices, servers, and networks can be included.

[0042] Figure 2 A flowchart illustrating an advertising recommendation method according to an embodiment of this application is shown.

[0043] like Figure 2 As shown, the advertising recommendation method includes operations S210 to S240.

[0044] In operation S210, in response to the interactive actions of the target user, the product series that the target user is interested in are predicted.

[0045] In operation S220, multiple advertisements are filtered to determine a sequence of candidate advertisements that match the target user.

[0046] In operation S230, each product in the product sequence is used as a reference object to determine the target advertisement that matches the product from the candidate advertisement sequence, resulting in multiple target advertisements.

[0047] In operation S240, based on the order of products in the product sequence, multiple target advertisements are rearranged to obtain a target advertisement sequence for recommendation to target users.

[0048] In this embodiment, the advertising recommendation method can be applied to internet advertising recommendation scenarios.

[0049] For example, when a target user interacts with an internet platform, such as an e-commerce platform or a content platform, the advertising recommendation system responds to the interaction by calling a prediction model to predict the product series that the target user may be interested in.

[0050] Interactive actions can include, but are not limited to, actions that reflect user behavior, such as logging into the platform, browsing platform content, searching for products, clicking on advertisements, adding items to the shopping cart, and placing orders.

[0051] Predictive models can include, but are not limited to, large models and deep learning models.

[0052] For example, deep learning models, such as collaborative filtering and neural network models, can be used to analyze the target user's historical behavior information, identify the user's interests and preferences, determine various products the user may be interested in, and arrange them into a product sequence according to the user's level of interest from highest to lowest. The products in the product sequence can be items the user has actually purchased or browsed, or other items that are functionally related to the purchased or browsed items. For example, if the target user has browsed tents multiple times in the past, the predicted product sequence could include tents, sleeping mats, sleeping bags, portable stoves, camping lights, folding chairs, etc.

[0053] This application does not specify a particular method for filtering multiple advertisements. For example, it can employ existing techniques such as filtering candidate advertisement sequences during the recall phase, or it can perform a further coarse-ranking and filtering of candidate advertisement sequences based on the initial screening during the recall phase, or it can perform a further fine-ranking and filtering of candidate advertisement sequences based on the initial screening during the coarse-ranking phase. For instance, recall algorithms such as collaborative recall, content recall, and trending topic recall can be used, using the target user's user information as the filtering basis, to recall advertisements related to the user's interests from the platform's advertisement library. These recalled advertisements are then integrated to form a candidate advertisement sequence. The advertisement library can store multiple advertisements of different types and for different products.

[0054] The number of ads in the candidate ad sequence can be set according to actual recommendation needs.

[0055] In one example, the matching degree between each candidate ad in the candidate ad sequence and each product can be calculated, thereby filtering out the target ads that match the product. For example, the relevant information of the candidate ads and the semantic information of the products can be vectorized to obtain their respective vectors, and then the matching degree can be obtained based on the similarity between the vectors. The candidate ad with the highest matching degree is determined as the target ad, thus obtaining multiple target ads.

[0056] For example, if the product in the product sequence is a tent, then ads in the candidate ad sequence that are related to or consistent with the tent function will be selected.

[0057] In one example, since the product sequence can be arranged from high to low according to the target user's interest level, multiple target ads can be rearranged based on the order of the product sequence. This will place the target ads that match the products that are ranked higher in the product sequence, i.e. the products that the user is more interested in, at the top, and place the target ads that match the products that are ranked lower in the product sequence, i.e. the products that the user is less interested in, at the bottom. This ensures that the user sees the ads that are most interesting to them first, and further improves the accuracy of ad recommendations.

[0058] According to embodiments of this application, by responding to the interactive operations of target users in real time, the system predicts the product sequence that the target users are interested in, and uses each product in the product sequence as a reference to determine the target advertisements that match the products from the candidate advertisement sequence. This preserves information about user interests and reduces information loss during the filtering process. Furthermore, based on the arrangement order of the products in the product sequence, multiple target advertisements are rearranged, achieving global optimization of advertisement recommendation and at least partially overcoming the limitations of independent optimization at each stage of traditional cascading pipelines. This not only provides target users with advertisements tailored to their specific needs, improving user experience, but also increases ad click-through rates and conversion rates, optimizing advertiser performance and platform advertising revenue.

[0059] The following is for reference. Figures 3-6 In conjunction with specific embodiments, Figure 2 The advertising recommendation method shown will be further explained.

[0060] Figure 3 The illustration shows a schematic diagram of a predicted sequence of products of interest to a target user according to an embodiment of this application.

[0061] According to one embodiment of this application, regarding the above... Figure 2 Operation S210, as shown, predicting the product sequence of interest to the target user, may include the following steps: identifying historical products from the target user's historical behavior information to obtain multiple historical products; determining prompt information to represent these historical products based on their respective attribute information; and using a large model based on the prompt information to predict the products of interest to the target user, thus obtaining the product sequence.

[0062] For example, such as Figure 3As shown, taking the interaction of a target user on an e-commerce platform as an example, when the target user clicks to enter the homepage of the e-commerce platform, this click-through interaction can trigger the advertising recommendation system to respond to the interaction and obtain historical behavior information based on the target user's user identifier. Historical behavior information may include, but is not limited to: browsing history, click history, purchase history, search keywords, etc. Multiple historical products can be identified from the historical behavior information, thereby overcoming the limitation of users' interest in a single product, comprehensively capturing users' multi-dimensional interest tendencies, and avoiding the one-sidedness of judging user intent based solely on a single product.

[0063] For example, product identifiers (SKUs) for historical products can be identified from historical behavioral information. The attribute information of each historical product can serve as semantic information for that product, mapping a unique SKU to the historical product. For instance, the attribute information of each historical product may include, but is not limited to: product shape, product function, product color, product price, product name, product category, product application scenario, product core parameters, and target user group.

[0064] For example, semantic information can be determined as prompt information. Figure 3 As shown, the prompt information is input into the large model, which then predicts the products that the target user is interested in based on this prompt information and outputs a product sequence. Because the prompt information integrates semantic information from multiple historical products, it can convey user interests relatively comprehensively, allowing the large model to fully understand the user's true intentions and reducing intention interpretation bias.

[0065] It should be noted that the acquisition of historical behavior information in this example was performed with the authorization of the target user, which complies with relevant data security and privacy protection regulations.

[0066] In another example, the attribute information of multiple historical products can be transformed into vectors of a preset vector dimension, resulting in multiple vectors. These vectors are then combined to generate the prompt information. The preset vector dimension can be determined based on the predictive performance of the large model and its understanding of the product. For an advertising recommendation scenario, this preset vector dimension can be preferably 3, illustratively. Thus, a product indicated by one SKU can be transformed into three tokens, such as "...".<a_1651><b_1682><c_273> ".

[0067] For example, each historical behavior in the historical behavior information can correspond to a product, thus products can be identified using historical behavior identifiers. For instance, if clicking on product A and placing an order for product B in the historical behavior information could result in a notification message like "..."<bt_0><a_1651><b_1682><c_273><bt_1><a_1902><b_1088><c_943> "<bt_0> "Represents a click,"<a_1651><b_1682><c_273> "Represents product A, "<bt_1> "Represents placing an order,"<a_1902><b_1088><c_943> "Represents product B."

[0068] In another example, user basic information can be concatenated with historical behavior information to obtain a prompt message. For example, user basic information may include, but is not limited to, user ID, membership status, membership level, gender, and age, and can be described in natural language. The resulting prompt message could be, for example:

[0069] "user_pin, Gold Member, Male, Yes, 26-35 years old"<bt_0><a_1651><b_1682><c_273><bt_1><a_1902><b_1088><c_943> "etc. user_pin represents the user identifier."

[0070] It should be noted that the user's basic information in this example can be obtained with the target user's authorization, which complies with relevant data security and privacy protection regulations.

[0071] In summary, by combining large-scale models with prompts for prediction, we can accurately uncover the potential user interests behind multiple historical products, generating an orderly product sequence that aligns with the overall needs of users. This effectively addresses the problem that the understanding of user intent becomes increasingly fragmented as the process of traditional information delivery progresses, further enhancing the accuracy of advertising recommendations and improving user experience and platform recommendation effectiveness.

[0072] In response to the above Figure 2 The operation S210 shown, predicting the product sequence that the target user is interested in, can be implemented not only by the method provided in the above embodiments, but also by the following operations: Based on the operation object targeted by the interaction operation, determine the products of interest to the target object from the operation object. Based on the attribute information of the products of interest, determine the prompt information used to represent the products of interest. Using a large model based on the prompt information, predict the products of interest to the target user to obtain the product sequence.

[0073] In this example, the object of operation can include the specific physical object directly pointed to by the interactive operation, such as the specific product page clicked by the target user, the set of products corresponding to the search keywords, or the product category browsing. The products to be followed can include all or some of the products in the specific product page, the set of products corresponding to the search keywords, or the product category browsing; this application does not impose specific limitations on these.

[0074] Figure 4 A schematic diagram illustrating a predicted sequence of products of interest to a target user according to another embodiment of this application is shown.

[0075] For example, such as Figure 4 As shown, taking a target user clicking to view a specific product page on an e-commerce platform as an example, this click-to-view interaction can trigger the advertising recommendation system to respond to the interaction. Based on the specific product page viewed, the system identifies the products of interest from that specific product page. For example, if a user clicks to view "portable folding camping tent," the specific product page will display products associated with the tent, thus allowing the system to identify these associated products as products of interest.

[0076] This application does not specify the specific implementation method for determining the target product of interest from the operation objects. For example, the user interaction frequency and duration of the interaction can be screened from the operation objects to select the interactive objects corresponding to the interaction behavior in the past few days as the product of interest. This can exclude the product of interest that is clicked accidentally or browsed briefly without a clear interest.

[0077] Product attribute information can be retrieved directly from the interactive platform's database, ensuring the accuracy of the information.

[0078] The prompt message may not include all the attribute details of the products of interest, but only the key information that reflects the interests of the target users. Its format can be preset according to the understanding capabilities of the large-scale model. Determining the prompt message based on the attribute information of the products of interest provides a clear interest direction for subsequent large-scale model predictions, avoiding prediction bias caused by ambiguous information.

[0079] In implementing the above embodiments, it was found that, taking a user browsing a "family camping tent" product details page as an example, traditional methods can usually only identify the core product "tent," while ignoring the user's potential needs in related scenarios such as "family," "overnight stay," "dining," and "leisure." Ultimately, the advertisements displayed in the recommendation section may be limited to single-category, similar-function products such as sleeping mats, sleeping bags, and camping lights, failing to proactively provide users with a complete camping solution encompassing sleep systems (such as sleeping mats and sleeping bags), cooking systems (such as portable stoves), lighting systems (such as camping lights), and leisure systems (such as folding chairs).

[0080] Based on this, in this example, a large model is used to predict products of interest to the target user based on the prompt information, resulting in a product sequence. This can include the following steps: inputting the prompt information into the large model to identify candidate products of interest to the target user; identifying related products that can be combined with the candidate products to obtain product groups including the candidate products and related products; and ranking the product groups based on the target user's level of interest in the candidate and related products to obtain the product sequence.

[0081] Candidate products may include historical products from the target user's historical behavior information or products of interest involved in the target user's interactive operations.

[0082] For example, when the candidate product is a "tent", related products that can be used in combination with the candidate product may include, but are not limited to, moisture-proof mats, sleeping bags, portable stoves, camping lights, folding chairs, etc.

[0083] Large models can include, but are not limited to, pre-trained language models with semantic understanding, association mining, and interest prediction capabilities. For example, they can be fine-tuned and optimized in advance using massive amounts of product data and user interest association data, enabling them to accurately identify user interest directions in prompts, mine the association relationships between candidate products and other products, and thus predict products that users may be potentially interested in. This application does not specifically limit the fine-tuning and optimization; for example, candidate products and related products can be used as training samples to allow the model to learn product association rules and user interest preference patterns, mine association relationships that are related to the functions of the products of interest, adapt to the scenario, and meet the potential needs of users, and ensure that the prediction results are consistent with the actual interests of users.

[0084] The identified prompts can be input into a fine-tuned large model. The large model performs semantic parsing on the prompts and extracts candidate products. Combining the relationships learned during model training, it predicts related products. Finally, the products are sorted according to the target user's level of interest in the related and candidate products, forming an ordered product sequence.

[0085] For example, a large-scale model can predict "baking oven" as a candidate product, and based on "baking oven," it can identify related products such as egg beaters, cake flour, oven mitts, and baking mold storage boxes. These products are then sorted from highest to lowest user interest, such as: egg beaters > cake flour > oven mitts > baking mold storage boxes, etc., and the final output is this sorted product sequence as the product sequence that the target user is interested in. This satisfies the target user's potential needs in related scenarios such as home baking, parent-child activities, simple desserts, and convenient operation.

[0086] According to embodiments of this application, based on identifying associated products that can be combined with candidate products, the application scenarios that users are interested in are discovered behind the candidate products, thereby constructing a product sequence that covers scenario requirements, which can solve the problem of insufficient scenario understanding in traditional methods.

[0087] According to the embodiments of this application, regarding the above... Figure 2 The operation S230 shown uses each product in the product sequence as a reference object to determine the target advertisement that matches the product from the candidate advertisement sequence, resulting in multiple target advertisements. It may include the following operation: for each product in the product sequence, perform the following matching process: determine the target advertisement that matches the product based on the effect prediction index of each candidate advertisement in the candidate advertisement sequence and the semantic similarity between the product and each candidate advertisement.

[0088] In this embodiment, performance prediction metrics can be used to characterize the importance of candidate ads to target users. For example, performance prediction metrics may include, but are not limited to, the predicted click-through rate (CTR), the value per thousand effective impressions (VPI), the predicted conversion rate, the predicted transaction volume, and the predicted creative quality. The predicted CTR characterizes the probability that a user will click on the ad. The VPI characterizes the revenue value that the ad can generate for the platform for every thousand effective impressions.

[0089] This application does not specify a particular method for determining the performance prediction metrics. For example, after performing regular recall, coarse ranking, and fine ranking sequentially, a pre-defined performance prediction model can be used to calculate the estimated click-through rate and the value per thousand effective impressions for each candidate ad, thereby obtaining the corresponding performance prediction metrics for each candidate ad. Alternatively, performance prediction metrics can be determined through statistical analysis by combining historical impression data, historical click data, and advertiser bidding information of the candidate ads.

[0090] This application does not specify a particular method for determining the semantic similarity between a product and each candidate advertisement. For example, it can be obtained by calculating the inner product of the word vectors of each product in the product sequence and each advertisement in the candidate advertisement sequence.

[0091] For example, the performance prediction metric and semantic similarity can be weighted and summed to obtain a matching score, and the target ad with the highest matching score to the product can be selected from the candidate ad sequence.

[0092] Figure 5 The illustration shows a schematic diagram of an advertising recommendation according to an embodiment of this application.

[0093] like Figure 5As shown, when target user 510 interacts through an e-commerce platform, the playback system 520 installed on the platform can sense the user's interaction in real time. Based on this interaction, it predicts a product sequence 540 and determines a candidate ad sequence 530 using a large model M. Then, based on the matching degree between the products in the product sequence and each candidate ad in the candidate ad sequence, it determines target ads that match each product in the product sequence, resulting in multiple target ads 550. These multiple target ads 550 are then rearranged based on the order of the products in the product sequence to obtain a target ad sequence 560, which is then recommended to target user 510. Through this generative architecture of predicting the product sequence 540 based on a large model M and the discriminative architecture of filtering and rearranging target ads, collaborative reasoning enables the transformation from isolated products to product sequences that can be combined and applied within a limited time and space.

[0094] This application does not specify the form of recommendation for the target ad sequence. For example, ads in the target ad sequence can be displayed to target users in a form that can be viewed intuitively, such as pop-ups, news feeds, or recommendation lists, so as to achieve accurate ad recommendation.

[0095] According to the embodiments of this application, target advertisements are determined based on a multi-dimensional approach using performance prediction metrics and semantic similarity. This approach ensures that target advertisements are highly compatible with products and scenarios that users are interested in through semantic similarity, avoiding a disconnect between advertisements and user interests and improving user experience. Furthermore, the performance prediction metrics take into account both ad click performance and platform revenue, achieving a balance between user interests, ad performance, and platform revenue. This further optimizes the accuracy of ad recommendations and solves problems such as information decay, local optima, and insufficient scenario understanding in traditional ad recommendations.

[0096] According to one embodiment of this application, when performing the matching process for each product in the product sequence, the method further includes: removing the target advertisement that matches the product from the candidate advertisement sequence.

[0097] For example, a product sequence might contain product 1, product 2, product 3, and product 4. By calculating the match between product 1 and each candidate ad in the candidate ad sequence, target ad 1 matching product 1 is determined. At this point, target ad 1 can be removed from the candidate ad sequence. When determining target ad 2 matching product 2, target ad 1 is removed from the candidate ad sequence. This ensures that duplicate target ads do not appear in the same target ad sequence, solving the resource waste problem caused by duplicate recommendations.

[0098] Figure 6 A flowchart illustrating an advertising recommendation method according to another embodiment of this application is shown.

[0099] According to one embodiment of this application, the advertising recommendation method may include, for example: Figure 6 Operations S610 to S660 are shown.

[0100] In operation S610, timing is performed when a candidate ad sequence is determined.

[0101] In operation S620, determine whether the product sequence has been determined within the predetermined time period. If yes, execute operations S650 to S660. If no, execute operations S630 to S640.

[0102] In operation S630, based on the performance prediction metrics of each candidate advertisement, multiple candidate advertisements are rearranged to obtain the rearrangement result.

[0103] In operation S640, based on the rearrangement results, a predetermined number of candidate advertisements are selected to obtain the target advertisement sequence.

[0104] In operation S650, for each product in the product sequence, the following matching process is performed: based on the performance prediction metrics of each candidate advertisement in the candidate advertisement sequence and the semantic similarity between the product and each candidate advertisement, the target advertisement that matches the product is determined.

[0105] In operation S660, based on the order of products in the product sequence, multiple target advertisements are rearranged to obtain a target advertisement sequence for recommendation to target users.

[0106] The scheduled time period can be set according to the actual need for timely delivery of recommended advertisements. For example, it can be flexibly adjusted based on the performance of the e-commerce platform and the prediction speed of the large model.

[0107] For the effect prediction indicators, please refer to the explanation of the effect prediction indicators in the above examples, which will not be repeated here.

[0108] The pre-ordered quantity can be flexibly adjusted based on the terminal display interface and user experience requirements.

[0109] According to the embodiments of this application, it is possible to ensure that even in extreme scenarios, a reasonable target ad sequence can still be output quickly, avoiding ad recommendation delays caused by waiting for product sequences and affecting user experience.

[0110] According to one embodiment of this application, for a large model, the embedding of products can support multi-beam search, and the number of beams is configurable. By configuring the number of beams, multiple product lists can be generated.

[0111] For example, when generating each product list, the large model can employ a fixed-position restricted generation scheme to ensure the output format meets expectations. The large model can generate tokens step-by-step, allowing only specific types of tokens to be output at each step, avoiding invalid output. The large model accurately determines the current generation position by continuously counting the number of generated tokens; initially, when the token length is 0, a placeholder is generated.<p_0> When the generated token is 3 characters long, the three codes identifying the first SKU have been completed, and a placeholder needs to be generated.<p_1> When the generated token is 7 characters long, it indicates that the second SKU is complete and needs to be generated.<p_2> When the generated token length is 11, it indicates that the third SKU is complete and needs to be generated.<p_3> .

[0112] The output format of a large model could be, for example, for each product sequence, containing 4 SKUs, each SKU consisting of something like "<a_xxx><b_xxx><c_xxx> "Represented by 3 vectors, using placeholders"<p_0> arrive<p_3> To indicate the position of the SKU in the product sequence, the following example shows four product sequences with a bundle size of four:

[0113] Product Series 1:

[0114] <p_0><a_1001><b_2001><c_3001><p_1><a_1002><b_2002><c_3002><p_2><a_1003><b_2003><c_3003><p_3><a_1004><b_2004><c_3004>

[0115] Product Series 2:

[0116] <p_0><a_1005><b_2005><c_3005><p_1><a_1006><b_2006><c_3006><p_2><a_1007><b_2007><c_3007><p_3><a_1008><b_2008><c_3008>

[0117] Product Series 3:

[0118] <p_0><a_1009><b_2009><c_3009><p_1><a_1010><b_2010><c_3010><p_2><a_1011><b_2011><c_3011><p_3><a_1012><b_2012><c_3012>

[0119] Product Series 4:

[0120] <p_0><a_1013><b_2013><c_3013><p_1><a_1014><b_2014><c_3014><p_2><a_1015><b_2015><c_3015><p_3><a_1016><b_2016><c_3016> .

[0121] For the output format of a large model, when predicting multiple product combinations, the numerical matrix output by the large model before normalization is processed by setting all values ​​in the entire numerical matrix to negative infinity. Based on the current product, bundle, and target placeholder, the corresponding target position in the numerical matrix is ​​precisely located, and the value at the target position is set to 0, while other positions remain negative infinity. Then, a Softmax transformation is performed. Due to mathematical properties, positions with a value of 0 are set to exp(0) = 1, and positions with negative infinity are set to exp(-∞) = 0. This determines that the probability of the target placeholder in the final probability distribution is 1, and the probability of other tokens is 0. This achieves the generation of product sequences at fixed positions.

[0122] According to one embodiment of this application, when there are multiple product sequences and multiple target advertising sequences, the advertising recommendation method may include, in addition to, [the following methods may be used]. Figure 2 In addition to operations S210 to S240, the following operations may also be included: inputting multiple target ad sequences into an evaluation model and outputting an evaluation value for each target ad sequence; determining a recommended ad sequence from the multiple target ad sequences based on the evaluation values ​​of each target ad sequence; and recommending the recommended ad sequence to the target user.

[0123] In this example, the evaluation model can be trained based on the historical interaction information of sample users with each of the multiple sample ad sequences. Each of the multiple sample ad sequences can be generated using the ad recommendation method described above.

[0124] For example, after obtaining multiple sample ad sequences, each sample ad sequence can be displayed to sample users. Then, the interaction status of sample users with each sample ad sequence in the historical interaction information can be counted to determine the target sample ad sequence that the sample users are most interested in. The target sample ad sequence is used as a positive sample, and the other sample ad sequences among the multiple sample ad sequences are used as negative samples. The positive and negative samples are used to train the neural network model to obtain the trained evaluation model.

[0125] According to the embodiments of this application, by predicting multiple product sequences that the target user is interested in, the user's potential interests can be captured more comprehensively. Multiple product sequences are matched to obtain corresponding multiple advertising sequences. Each advertising sequence is then comprehensively evaluated, and the final target advertising sequence is selected based on the evaluation results. This achieves a global consideration of advertising recommendation and avoids the drawbacks of getting stuck in local optima and failing to take into account the overall effect when only based on single sequence matching or single-dimensional selection.

[0126] Figure 7 A block diagram of an advertising recommendation device according to an embodiment of this application is shown schematically.

[0127] like Figure 7 As shown, the advertising recommendation device 700 includes a prediction module 710, a first determination module 720, a second determination module 730, and a rearrangement module 740.

[0128] The prediction module 710 is used to predict the product series that the target user is interested in in response to the target user's interactive operations.

[0129] The first determining module 720 is used to filter multiple advertisements and determine a sequence of candidate advertisements that match the target user.

[0130] The second determining module 730 is used to determine the target advertisement that matches the product from the candidate advertisement sequence, taking each product in the product sequence as a reference, and obtain multiple target advertisements.

[0131] The rearrangement module 740 is used to rearrange multiple target advertisements based on the order of products in the product sequence to obtain a target advertisement sequence for recommendation to target users.

[0132] According to an embodiment of this application, the prediction module 710 includes: an identification unit, a first sub-determination unit, and a first sub-prediction unit. The identification unit is used to identify historical products in the target user's historical behavior information to obtain multiple historical products. The first sub-determination unit is used to determine prompt information representing the multiple historical products based on their respective attribute information. The first sub-prediction unit is used to use a large model to predict products of interest to the target user based on the prompt information, thereby obtaining a product sequence.

[0133] According to an embodiment of this application, the prediction module 710 includes: a second sub-determination unit, a third sub-determination unit, and a second sub-prediction unit. The second sub-determination unit is used to determine the target user's interested products from the operation object based on the operation object targeted by the interactive operation. The third sub-determination unit is used to determine prompt information representing the interested products based on the attribute information of the interested products. The second sub-prediction unit is used to predict products of interest to the target user based on the prompt information using a large model, thereby obtaining a product sequence.

[0134] According to an embodiment of this application, a large model is used to predict products of interest to a target user based on prompt information to obtain a product sequence. This includes: inputting prompt information into the large model to determine candidate products of interest to the target user, wherein candidate products include historical products from the target user's historical behavior information or products of interest involved in the target user's interactive operations; determining associated products that can be combined with the candidate products to obtain a product group including the candidate products and associated products; and ranking the product group based on the target user's level of interest in the candidate products and associated products to obtain a product sequence.

[0135] According to an embodiment of this application, the second determining module 730 is further configured to perform the following matching process for each product in the product sequence: based on the effect prediction index of each candidate advertisement in the candidate advertisement sequence and the semantic similarity between the product and each candidate advertisement, determine the target advertisement that matches the product, wherein the effect prediction index is used to characterize the importance of the candidate advertisement to the target user.

[0136] According to embodiments of this application, when performing the matching process for each product in the product sequence, the method further includes: removing the target advertisement that matches the product from the candidate advertisement sequence.

[0137] According to embodiments of this application, when there are multiple product sequences and multiple target advertising sequences, the advertising recommendation device further includes: an evaluation module, a third determination module, and a recommendation module. The evaluation module is used to input multiple target advertising sequences into an evaluation model and output an evaluation value for each target advertising sequence. The evaluation model is trained based on the historical interaction information of sample users with each of the multiple sample advertising sequences. The third determination module is used to determine a recommended advertising sequence from the multiple target advertising sequences based on the evaluation values ​​of each target advertising sequence. The recommendation module is used to recommend the recommended advertising sequence to the target user.

[0138] According to embodiments of this application, the advertising recommendation device further includes: a timing module, a candidate advertising reordering module, and a filtering module. The timing module is used to time the process when a candidate advertising sequence is determined. The candidate advertising reordering module is used to reorder multiple candidate advertisements based on the performance prediction indicators of each candidate advertisement when a product sequence is not determined within a predetermined time period, thereby obtaining a reordering result. The filtering module is used to filter a predetermined number of candidate advertisements based on the reordering result to obtain a target advertising sequence.

[0139] Any one or more of the modules, submodules, units, and subunits according to the embodiments of this application, or at least part of the functions of any one or more of them, can be implemented in one module. Any one or more of the modules, submodules, units, and subunits according to the embodiments of this application can be implemented by dividing them into multiple modules. Any one or more of the modules, submodules, units, and subunits according to the embodiments of this application can be at least partially implemented as hardware circuits, such as field-programmable gate arrays (FPGAs), programmable logic arrays (PLAs), systems-on-a-chip, systems-on-a-substrate, systems-on-package, application-specific integrated circuits (ASICs), or implemented by hardware or firmware in any other reasonable manner by integrating or packaging circuits, or implemented in any one of software, hardware, and firmware, or in a suitable combination of any of these. Alternatively, one or more of the modules, submodules, units, and subunits according to the embodiments of this application can be at least partially implemented as computer program modules, which, when run, can perform corresponding functions.

[0140] For example, any plurality of the prediction module 710, the first determination module 720, the second determination module 730, and the rearrangement module 740 can be combined into one module / unit / subunit, or any one of these modules / units / subunits can be split into multiple modules / units / subunits. Alternatively, at least part of the functionality of one or more of these modules / units / subunits can be combined with at least part of the functionality of other modules / units / subunits and implemented in one module / unit / subunit. According to embodiments of this application, at least one of the prediction module 710, the first determination module 720, the second determination module 730, and the rearrangement module 740 can be at least partially implemented as hardware circuitry, such as a field-programmable gate array (FPGA), a programmable logic array (PLA), a system-on-a-chip, a system-on-a-substrate, a system-on-package, an application-specific integrated circuit (ASIC), or any other reasonable means of integrating or packaging the circuitry, or implemented in software, hardware, or firmware, or in any suitable combination of any of these three implementation methods. Alternatively, at least one of the prediction module 710, the first determination module 720, the second determination module 730, and the rearrangement module 740 may be implemented at least partially as a computer program module that can perform corresponding functions when the computer program module is run.

[0141] It should be noted that the advertising recommendation device part in the embodiments of this application corresponds to the advertising recommendation method part in the embodiments of this application. The description of the advertising recommendation device part is specifically referred to in the advertising recommendation method part, and will not be repeated here.

[0142] Figure 8A block diagram of an electronic device suitable for implementing an advertising recommendation method according to an embodiment of this application is illustrated schematically. Figure 8 The electronic device shown is merely an example and should not impose any limitation on the functionality and scope of use of the embodiments of this application.

[0143] like Figure 8 As shown, an electronic device 800 according to an embodiment of this application includes a processor 801, which can perform various appropriate actions and processes according to a program stored in a read-only memory ROM 802 or a program loaded from a storage portion 808 into a random access memory RAM 803. The processor 801 may include, for example, a general-purpose microprocessor (e.g., a CPU), an instruction set processor and / or an associated chipset and / or a special-purpose microprocessor (e.g., an application-specific integrated circuit (ASIC)), etc. The processor 801 may also include onboard memory for caching purposes. The processor 801 may include a single processing unit or multiple processing units for performing different actions of the method flow according to an embodiment of this application.

[0144] RAM 803 stores various programs and data required for the operation of electronic device 800. Processor 801, ROM 802, and RAM 803 are interconnected via bus 804. Processor 801 executes various operations of the method flow according to embodiments of this application by executing programs in ROM 802 and / or RAM 803. It should be noted that programs may also be stored in one or more memories other than ROM 802 and RAM 803. Processor 801 may also execute various operations of the method flow according to embodiments of this application by executing programs stored in one or more memories.

[0145] According to embodiments of this application, the electronic device 800 may further include an input / output (I / O) interface 805, which is also connected to a bus 804. The electronic device 800 may also include one or more of the following components connected to the input / output (I / O) interface 805: an input section 806 including a keyboard, mouse, etc.; an output section 807 including a cathode ray tube (CRT), liquid crystal display (LCD), etc., and a speaker, etc.; a storage section 808 including a hard disk, etc.; and a communication section 809 including a network interface card such as a LAN card, modem, etc. The communication section 809 performs communication processing via a network such as the Internet. A drive 810 is also connected to the input / output (I / O) interface 805 as needed. A removable medium 811, such as a disk, optical disk, magneto-optical disk, semiconductor memory, etc., is installed on the drive 810 as needed so that computer programs read from it can be installed into the storage section 808 as needed.

[0146] According to embodiments of this application, the method flow according to embodiments of this application can be implemented as a computer software program. For example, embodiments of this application include a computer program product comprising a computer program carried on a computer-readable storage medium, the computer program containing program code for performing the methods shown in the flowchart. In such embodiments, the computer program can be downloaded and installed from a network via communication section 809, and / or installed from removable medium 811. When the computer program is executed by processor 801, it performs the functions defined in the system of embodiments of this application. According to embodiments of this application, the systems, devices, apparatuses, modules, units, etc., described above can be implemented by computer program modules.

[0147] This application also provides a computer-readable storage medium, which may be included in the device / apparatus / system described in the above embodiments; or it may exist independently and not assembled into the device / apparatus / system. The computer-readable storage medium carries one or more programs, which, when executed, implement the method according to the embodiments of this application.

[0148] According to embodiments of this application, the computer-readable storage medium can be a non-volatile computer-readable storage medium. Examples include, but are not limited to: portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof. In this application, the computer-readable storage medium can be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device.

[0149] For example, according to embodiments of this application, a computer-readable storage medium may include the ROM 802 and / or RAM 803 described above and / or one or more memories other than ROM 802 and RAM 803.

[0150] Embodiments of this application also include a computer program product comprising a computer program containing program code for performing the methods provided in the embodiments of this application. When the computer program product is run on an electronic device, the program code is used to enable the electronic device to implement the advertising recommendation method provided in the embodiments of this application.

[0151] When the computer program is executed by the processor 801, it performs the functions defined in the system / apparatus of this application embodiment. According to the embodiments of this application, the systems, apparatuses, modules, units, etc., described above can be implemented by computer program modules.

[0152] In one embodiment, the computer program may rely on a tangible storage medium such as an optical storage device or a magnetic storage device. In another embodiment, the computer program may also be transmitted and distributed in the form of signals over a network medium, and may be downloaded and installed via the communication section 809, and / or installed from a removable medium 811. The program code contained in the computer program can be transmitted using any suitable network medium, including but not limited to: wireless, wired, etc., or any suitable combination thereof.

[0153] According to embodiments of this application, program code for executing the computer programs provided in the embodiments of this application can be written in any combination of one or more programming languages. Specifically, these computational programs can be implemented using high-level procedural and / or object-oriented programming languages, and / or assembly / machine languages. Programming languages ​​include, but are not limited to, languages ​​such as Java, C++, Python, "C", or similar programming languages. The program code can be executed entirely on the user's computing device, partially on the user's device, partially on a remote computing device, or entirely on a remote computing device or server. In cases involving remote computing devices, the remote computing device can be connected to the user's computing device via any type of network, including a local area network (LAN) or a wide area network (WAN), or it can be connected to an external computing device (e.g., via the Internet using an Internet service provider).

[0154] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of this application. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in a block diagram or flowchart, and combinations of blocks in a block diagram or flowchart, may be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions. Those skilled in the art will understand that the features described in the various embodiments of this application can be combined and / or combined in various ways, even if such combinations are not explicitly described in this application. In particular, without departing from the spirit and teachings of this application, the features described in the various embodiments of this application can be combined and / or combined in various ways. All such combinations and / or combinations fall within the scope of this application.

[0155] The embodiments of this application have been described above. However, these embodiments are merely illustrative and not intended to limit the scope of this application. Although various embodiments have been described above, this does not mean that the measures in the various embodiments cannot be used advantageously in combination. Without departing from the scope of this application, those skilled in the art can make various substitutions and modifications, all of which should fall within the scope of this application.

Claims

1. An advertising recommendation method, comprising: In response to the interactive actions of the target user, predict the product series that the target user is interested in; Multiple advertisements are filtered to determine a sequence of candidate advertisements that match the target user; Using each product in the product sequence as a reference, a target advertisement matching the product is determined from the candidate advertisement sequence, resulting in multiple target advertisements; Based on the order of the products in the product sequence, the multiple target advertisements are rearranged to obtain a target advertisement sequence for recommendation to the target user.

2. The method according to claim 1, wherein: The prediction of the product series that the target user is interested in includes: Identify historical products from the target user's historical behavior information to obtain multiple historical products; Based on the attribute information of each of the historical products, determine the prompt information used to represent the historical products. Using a large model based on the prompt information, the products that the target user is interested in are predicted, thus obtaining the product sequence.

3. The method according to claim 1, wherein: The prediction of the product series that the target user is interested in includes: Based on the target object of the interactive operation, determine the target object's favorite products from the target object; Based on the attribute information of the product of interest, determine the prompt information used to represent the product of interest; Using a large model based on the prompt information, the products that the target user is interested in are predicted, thus obtaining the product sequence.

4. The method according to claim 2 or 3, wherein, The process of using a large model to predict products of interest to the target user based on the prompt information, thereby obtaining the product sequence, includes: The prompt information is input into the large model to call the large model to determine the candidate products that the target user is interested in, wherein the candidate products include historical products in the target user's historical behavior information or products that the target user is interested in through interactive operations; Identify associated products that can be combined with the candidate products to obtain a product group that includes the candidate products and the associated products. Based on the target user's level of interest in the candidate products and related products, the product groups are sorted to obtain the product sequence.

5. The method according to claim 1, wherein: The step of determining target advertisements matching the products in the candidate advertisement sequence, using each product in the product sequence as a reference, yields multiple target advertisements, including: For each product in the product sequence, the following matching process is performed: Based on the performance prediction metrics of each candidate ad in the candidate ad sequence and the semantic similarity between the product and each candidate ad, the target ad that matches the product is determined, and the performance prediction metrics are used to characterize the importance of the candidate ad to the target user.

6. The method according to claim 5, wherein: When performing the matching process for each product in the product sequence, the method further includes: Remove the target ad that matches the product from the candidate ad sequence.

7. The method according to claim 5, wherein: When the product sequence includes multiple items and the target advertising sequence includes multiple items, the method further includes: Multiple target ad sequences are input into an evaluation model, which outputs an evaluation value for each target ad sequence. The evaluation model is trained based on the historical interaction information of sample users with each of the multiple sample ad sequences. Based on the evaluation values ​​of each of the target ad sequences, a recommended ad sequence is determined from the plurality of target ad sequences; The recommended ad sequence is recommended to the target user.

8. The method according to claim 1, wherein: The method further includes: If the candidate ad sequence is determined, timing begins; If the product sequence is not determined within the predetermined time period, the candidate advertisements are rearranged based on the effect prediction index of each candidate advertisement to obtain the rearrangement result. Based on the rearrangement results, a predetermined number of candidate advertisements are selected to obtain the target advertisement sequence.

9. An advertising recommendation device, comprising: The prediction module is used to predict the product series that the target user is interested in in response to the target user's interactive operations; The first determining module is used to filter multiple advertisements and determine a sequence of candidate advertisements that match the target user; The second determining module is used to determine the target advertisement that matches the product from the candidate advertisement sequence, taking each product in the product sequence as a reference, and obtain multiple target advertisements; The rearrangement module is used to rearrange multiple target advertisements based on the order of the products in the product sequence to obtain a target advertisement sequence for recommendation to the target user.

10. An electronic device, comprising: One or more processors; Memory, used to store one or more programs. Wherein, when the one or more programs are executed by the one or more processors, the one or more processors implement the method of any one of claims 1 to 8.

11. A computer-readable storage medium having stored thereon executable instructions that, when executed by a processor, cause the processor to perform the method of any one of claims 1 to 8.

12. A computer program product comprising a computer program that, when executed by a processor, implements the method according to any one of claims 1 to 8.

Images