WO2023142520A1

WO2023142520A1 - Information recommendation method and apparatus

Info

Publication number: WO2023142520A1
Application number: PCT/CN2022/124280
Authority: WO
Inventors: 田明杨; 刘侃
Original assignee: 北京沃东天骏信息技术有限公司; 北京京东世纪贸易有限公司
Priority date: 2022-01-26
Filing date: 2022-10-10
Publication date: 2023-08-03
Also published as: CN114493786A

Abstract

Disclosed in the present disclosure are an information recommendation method and apparatus. The specific embodiment comprises: in response to a trigger by a user for a target area in a display page, acquiring a first user characteristic corresponding to the user, wherein the target area corresponds to a plurality of pieces of item information; taking the first user characteristic and the plurality of pieces of item information as an input of a ranking model, and determining a ranking result of the plurality of pieces of item information according to an output of the ranking model, wherein the ranking model is obtained by means of training based on a plurality of second user characteristics and historical behavior information respectively corresponding to the plurality of second user characteristics; according to the ranking result and item aggregation content, determining target information to be recommended, wherein the item aggregation content is obtained by means of aggregation of recommended content of the plurality of pieces of item information; and recommending the target information by means of the display page.

Description

Information recommendation method and device

Cross References to Related Applications

This application claims the priority of the Chinese patent application No. 202210092529.5 filed on January 26, 2022, entitled "An Information Recommendation Method and Device", and the content disclosed in the above-mentioned Chinese patent application is cited in its entirety as the content of this application. part or all.

technical field

The present disclosure relates to the field of computer technology, in particular to an information recommendation method and device.

Background technique

E-commerce platforms usually conduct marketing through coupons, flash sales, and pre-sales.

When a user browses a piece of marketing information, the user often sees a list of items corresponding to the marketing information. For example, take a user browsing the marketing information corresponding to a coupon as an example. What the user sees is a list of all item information associated with the coupon. The item information is randomly arranged in the list, and the user may need to turn pages multiple times, Or search again and other operations to obtain the content recommendation you are interested in, thus reducing the effectiveness of information recommendation, thereby reducing the user experience.

Contents of the invention

In view of this, the embodiments of the present disclosure provide an information recommendation method and device, which can recommend aggregated content of items corresponding to a target area to a user through a display page according to a ranking result of a plurality of item information corresponding to a target area by a ranking model. As a result, the number of operations of the user is reduced, and the content of interest to the user can be accurately recommended to the user, the effectiveness and efficiency of information recommendation are improved, and user experience is further improved.

Furthermore, multiple item information can be used as the input of the content generation model to quickly generate and aggregate recommended content for multiple item information, reducing the data volume of recommended information, so that users can quickly obtain recommended key points from recommended information , thus improving the accuracy of recommendation information. In addition, according to the user's feedback on information recommendation, the ranking model is optimized to improve the recommendation accuracy of the ranking model, thereby improving the click conversion rate of items.

To achieve the above purpose, according to the first aspect of the embodiments of the present disclosure, an information recommendation method is provided, including:

Responding to the triggering of the target area by the user on the display page, acquiring the first user feature corresponding to the user; the target area corresponds to a plurality of item information;

Using the first user characteristics and the plurality of item information as the input of a ranking model, and determining the ranking result of the plurality of item information according to the output of the ranking model; the ranking model is based on a plurality of second user characteristics And the historical behavior information respectively corresponding to the plurality of second user characteristics is obtained through training;

Determine the target information to be recommended according to the sorting result and the item aggregation content; the item aggregation content is obtained by aggregation according to the recommendation content of the plurality of item information;

The target information is recommended through the display page.

In one or more embodiments of the present application, before determining the target information to be recommended according to the ranking result and the aggregated content of the item, it further includes:

Using the plurality of item information as the input of the content generation model to obtain the recommended content respectively corresponding to the plurality of item information; the recommended content includes one or more of the following: recommended short title, recommended short text, recommended Copywriting information, comment information and recommended short video information;

The recommended content is aggregated to obtain the item aggregated content.

In one or more embodiments of the present application, before receiving the user's trigger on the target area in the display page, it further includes:

Acquiring recommended target parameters, the recommended target parameters include any one or more of the following: return on investment, inventory consumption parameters, total value and number of people to be recommended;

Determine a plurality of item information corresponding to the target area according to the recommended target parameters, the type of the target area, and the item information that can be displayed on the display page.

In one or more embodiments of the present application, the item information includes item category;

Determine the target item information from the item information corresponding to the item category; wherein the target item information is the historical order quantity, and/or order completion amount, and/or item information whose order value is greater than a preset threshold;

The target item information is used as an input of the content generation model.

In one or more embodiments of the present application, the method also includes:

Obtain user feedback on the target information through the display page;

According to the feedback information, the ranking model is optimized according to a preset time period.

In one or more embodiments of the present application, the ranking model is trained based on LR algorithm, GBDT algorithm, Xgboost algorithm, LightGBM algorithm, xDeepFM algorithm, DeepFM algorithm and AutoInt algorithm.

According to the second aspect of the embodiments of the present disclosure, an information recommendation device is provided, including: an acquisition module, a sorting module, and a recommendation module; wherein,

The acquisition module is configured to acquire the first user feature corresponding to the user in response to the user's trigger on the target area in the display page; the target area corresponds to a plurality of item information;

The sorting module is configured to use the first user characteristics and the plurality of item information as input to a sorting model, and determine the sorting result of the multiple item information according to the output of the sorting model; the ranking model is Obtained by training according to multiple second user features and historical behavior information respectively corresponding to the multiple second user features;

The recommendation module is configured to determine the target information to be recommended according to the sorting result and the item aggregation content; the item aggregation content is obtained according to the recommendation content aggregation of the plurality of item information; the target information is passed through the Display pages for recommendations.

According to a third aspect of the embodiments of the present disclosure, an electronic device is provided, including:

one or more processors;

storage means for storing one or more programs,

When the one or more programs are executed by the one or more processors, the one or more processors are made to implement any one of the information recommendation methods provided in the first aspect above.

According to a fourth aspect of the embodiments of the present disclosure, there is provided a computer-readable medium on which a computer program is stored, and when the program is executed by a processor, any one of the information recommendation methods provided in the first aspect above is implemented. the method described.

The further effects of the above-mentioned non-conventional alternatives will be described below in conjunction with specific embodiments.

Description of drawings

The accompanying drawings are for better understanding of the present disclosure, and do not constitute an improper limitation of the present disclosure. in:

Fig. 1 is a schematic flowchart of an information recommendation method provided in one or more embodiments of the present disclosure;

FIG. 2 is a schematic structural diagram of an information recommendation device for an e-commerce platform provided in one or more embodiments of the present disclosure;

Fig. 3 is a schematic structural diagram of a sorting model provided in one or more embodiments of the present disclosure;

Fig. 4 is a schematic flowchart of an information recommendation method for an e-commerce platform provided in one or more embodiments of the present disclosure;

Fig. 5 is a schematic structural diagram of an information recommendation device provided in one or more embodiments of the present disclosure;

FIG. 6 is an exemplary system architecture diagram that may be applied in one or more embodiments of the present disclosure;

Fig. 7 is a schematic structural diagram of a computer system suitable for implementing a terminal device or a server in one or more embodiments of the present disclosure.

Detailed ways

Exemplary embodiments of the present disclosure are described below in conjunction with the accompanying drawings, which include various details of the embodiments of the present disclosure to facilitate understanding, and they should be regarded as exemplary only. Accordingly, those of ordinary skill in the art will recognize that various changes and modifications of the embodiments described herein can be made without departing from the scope and spirit of the disclosure. Also, descriptions of well-known functions and constructions are omitted in the following description for clarity and conciseness.

It should be noted that, in the case of no conflict, the embodiments of the present disclosure and the technical features in the embodiments can be combined with each other.

As shown in FIG. 1 , one or more embodiments of the present disclosure provide an information recommendation method, which may include the following steps S101 to S104:

Step S101: Responding to a user's trigger on a target area on a presentation page, acquire a first user characteristic corresponding to the user; the target area corresponds to a plurality of item information.

Step S102: Using the first user characteristics and the plurality of item information as input to a ranking model, and determining the ranking result of the plurality of item information according to the output of the ranking model; the ranking model is based on a plurality of first The two user features and the historical behavior information respectively corresponding to the plurality of second user features are obtained through training.

Step S103: Determine the target information to be recommended according to the sorting result and the item aggregation content; the item aggregation content is obtained by aggregation of the recommended content of the plurality of item information.

Step S104: Recommend the target information through the display page.

It is understandable that the display page may be a service system page that the user can browse. The target area can be an entry that can be triggered by the user. For example, in an e-commerce platform, the target area of the display page can be the content recommendation entry provided to the user such as the discovery page, planting grass page, and shopping, or it can be a coupon link , seckill page, pre-sale page and other marketing information entrances that can be viewed by users. It can be understood that the user can browse multiple item information by entering the corresponding page from these entrances or pages. The browsable item information is a plurality of item information corresponding to the target area corresponding to these entries.

The information recommendation method provided by one or more embodiments of the present disclosure can be applied to a display page of an e-commerce platform to recommend product information. When applied to an e-commerce platform, the information recommendation method provided by the embodiments of the present disclosure may be implemented by the information recommendation device shown in FIG. 2 . As shown in Figure 2, the device may include an application display module, a recommendation module, a data storage module, a marketing content generation module and a marketing algorithm module.

In the above step S101, the user can trigger the display page through the content entry or marketing entry of the application display module. Correspondingly, the content entry or marketing entry is the corresponding target area, and the browsable pages corresponding to the content entry or marketing entry The item information is a plurality of item information corresponding to the target area. In the above step S102, the ranking model can be set in the recommendation module, so as to realize the recommendation of thousands of people and thousands of faces according to the ranking results of the ranking model. In addition, the recommended content and item aggregation content in step S103 can be generated based on the marketing content generation module. When generating the recommended content, the marketing content generation module can generate recommended content corresponding to the marketing product based on the marketing product information, and combine the recommended content with the marketing Commodity aggregation is commodity aggregation content. The specific implementation manner of generating recommended content and item aggregation content will be further described in detail in the following embodiments.

Before receiving the user’s trigger on the target area in the display page, the item information corresponding to the target area can be determined first, and the following method provided by an embodiment of the present disclosure can be adopted: obtain recommended target parameters, and the recommended target parameters include the following Any one or more items: return on investment, inventory consumption parameters, total value, and number of people to be recommended; determine the target according to the recommended target parameters, the type of the target area, and the item information that can be displayed on the display page Multiple item information corresponding to the area.

The plurality of item information corresponding to the target area is a subset obtained from all displayable item information associated with the display page according to the recommended target parameters and the type of the target area.

It is understandable that the recommended target parameter may be determined by a combination of one or more parameters. The type of the target area may be a tab page entry, a link entry, an image entry, and the like.

When creating a marketing campaign on an e-commerce platform, a marketing goal is usually set for the marketing, which can be the return on investment of the items participating in the marketing, the expected quantity of inventory items to be consumed, the total value to be realized, or the user groups targeted for the marketing, and The number of recommended users, etc. Among them, the total value can be determined according to GMV (Gross Merchandise Volume, total merchandise transaction). When creating a marketing plan, these recommended target parameters can be input into a marketing algorithm module, so that the marketing algorithm module can output corresponding item information. The marketing algorithm module provided by the embodiment of the present disclosure may be shown in FIG. 2 .

After the multiple item information corresponding to the target area is determined, the aggregated content of the item corresponding to the multiple item information can be determined, and the following method provided by an embodiment of the present disclosure can be adopted: the multiple item information is used as the content generation model input to obtain the recommended content corresponding to the plurality of item information; the recommended content includes one or more of the following: recommended short title, recommended short text, recommended copy information, comment information and recommended short video information; The recommended content is aggregated to obtain the item aggregated content.

When applied to an e-commerce platform, the content generation model provided in one or more embodiments of the present disclosure may be the marketing content generation module shown in FIG. 2 . The marketing content generation module can use natural language processing technology and image processing technology to generate one or more recommended content corresponding to multiple item information from the content data of the business system according to multiple input item information, that is, recommend short titles , Recommended short text, recommended copy information, comment information and recommended short video information. Wherein, the comment information may include text comment information and picture comment information, and the picture comment information may be a posting comment or the like.

It can be understood that, when the item information is an item category, the following method provided by one or more embodiments of the present disclosure can be used to determine the target item information that can be used as input to the content generation model: from the items corresponding to the item category The target item information is determined in the information; wherein, the target item information is historical order quantity, and/or order completion quantity, and/or item information whose order value is greater than a preset threshold.

For example, the total quantity of historical orders of this item category is M, the total quantity of completed orders is N, and the total value of orders is S. The preset threshold can be set to M'=80%*M. In this item category, the item information whose historical order quantity is greater than M' is used as the target item information; the preset threshold can also be set to N'=80%*N , in this item category, the item information whose historical order completion amount is greater than N' is used as the target item information; the preset threshold can also be set as S'=80%*S, and in this item category, the historical order value Item information greater than S' is used as target item information. The final target item information can also be obtained by intersecting the three kinds of target item information that are greater than the preset threshold, that is, the historical order quantity is greater than M', the historical order completion amount is greater than N', and the historical order value is greater than S' The item information serves as the target item information.

After obtaining the recommended content corresponding to the plurality of item information, the multiple item information and the recommended content corresponding to the target area are aggregated to obtain item aggregation content. Taking the e-commerce platform as an example, its aggregation process can be expressed as:

in

Represents the m-dimensional item information contained in marketing k.

When applied to an e-commerce platform, the embodiment of the present disclosure can store the item aggregation content in a Redis database or a MYSQL database, or in an HBASE database or an Elasticsearch database through the data storage module in FIG. 2 .

After a plurality of item information in the target area is determined, and the items corresponding to the multiple item information are aggregated, generated and stored, when a trigger from the user for the target area is received, a possible implementation of step S102 is to target the multiple item information to User to sort. The first user characteristic corresponding to the user may be obtained first, and the historical behavior information corresponding to the first user characteristic may also be obtained at the same time. The first user feature can be the user's personal information, order information, etc., and the corresponding historical behavior information can be the content features browsed by the user, the item features of multiple items corresponding to the target area, the cross feature, the request scene feature and session characteristics, etc.

Wherein, the item feature is detailed information of the item, such as item identification, item category, and the like. The cross feature is a feature with richer dimensions obtained by crossing different features. The feature of the request scene may be a feature of the geographic location where the user triggering the target area is located. The session feature is the session feature saved when the user jumps between multiple sessions.

The multiple item information corresponding to the target area, the first user feature, and the user behavior information corresponding to the first user feature are input into the ranking model for sorting to obtain the order in which multiple item information is recommended to the user.

In a possible implementation manner of step S102, the ranking model is trained based on the LR algorithm, GBDT algorithm, Xgboost algorithm, LightGBM algorithm, xDeepFM algorithm, DeepFM algorithm and AutoInt algorithm. The ranking model is the model used in the recommendation module in Figure 2.

In one or more embodiments of the present disclosure, a schematic structural diagram of the sorting model may be as shown in FIG. 3 . The feature input may be user features and user behavior information, that is, user features, content features, item features, intersection features, request scene features, and session features. The training process of the sorting model can be obtained by using multiple user features and historical behavior information corresponding to the multiple user features.

Multiple sub-models are used in the ranking model in Figure 3 to make the ranking results match the content that the user is interested in. Among them, sub-models such as GBDT, LR, FM and Wide&Deep are used to preprocess the feature input. The cross network in the middle layer, that is, the deep learning model can include: xDeepFM, DeepFM, DCN, NFM, AFM, and AutoInt; and the feature representation can use DSIN, DIEN, and DIN, etc.; and MMOE and ESSM can be used as multi-task learning. Model. The sorting model can be combined with LR algorithm, GBDT algorithm, Xgboost algorithm and LightGBM to build a tree model for model training.

The deep learning models xDeepFM, DeepFM, and AutoInt used in it can combine depth and breadth to mine high-order and low-order cross information from user behavior information, improving the efficiency of model learning.

Understandably, machine learning models degrade rapidly when the data changes. In one embodiment of the present disclosure, the combination of LR algorithm and GBDT algorithm is used, so that the LR model can supplement the insufficiency of the online learning of the GBDT model, ensuring that the performance of the model will not drop too much when the online data changes. For example, in an e-commerce platform, the input characteristic data is usually large, and the input data changes rapidly. The strong online learning ability of the ranking model can ensure that the output results of the ranking model are more accurate.

It is understandable that usually the business system is constantly acquiring a large amount of user characteristics and user behavior information, and the ranking model usually needs to be updated and calibrated in real time, that is, it needs to use real-time data to continuously retrain the model. How quickly the model can be trained and how quickly the data can be processed is critical. Therefore, in an embodiment of the present disclosure, LightGBM is used in the sorting model, which can support efficient parallel training, thereby obtaining faster training speed, lower memory consumption, and the effect of quickly processing massive data, and improving information efficiency. Recommended efficiency.

In addition, using Xgboost can reduce the possibility of overfitting of the sorting model, making the loss of the model more accurate. When the training data is sparse, you can specify the default direction of the branch for the missing value or the specified value, which can greatly improve the model. training efficiency.

In one or more embodiments of the present disclosure, the target (target) of the output layer of the ranking model in FIG. 3 is usually composed of CTR (Click-Through-Rate, click through rate), CVR (Conversion Rate, conversion rate), GMV (Gross Merchandise Volume, total merchandise volume) is fused. The fusion of these targets can usually be fused using an exponential method or a linear weight method.

On top of the output layer of the ranking model, a rearrangement layer (Re-Rank layer) can also be set, which can re-rank the results of the output layer according to the regional characteristics, diversity characteristics, and user interest exploration ratio of multiple items. Sorting, so that the final result of the ranking model is more in line with the content that the user is interested in. The sorting result of the sorting model indicates the order in which multiple items corresponding to the target area are recommended on the display page.

According to the sorting result, the item aggregation content corresponding to the multiple item information is displayed on the display page for information recommendation to the user.

For example, on an e-commerce platform, the ranking results show items A, B, and C in descending order of recommendation. Item A ranks first, corresponding to the aggregated content of item A (recommended copy, recommended short video, comment information and other content) is displayed at the top of the display page, the aggregated content related to item B is ranked second, and so on.

During the use of the business system, the accuracy of the ranking model can be determined according to the user's feedback on the target information, and the ranking model can be updated and calibrated in real time. The following method provided by an embodiment of the present disclosure may be adopted: obtain user feedback information on the target information through the display page; optimize the ranking model according to a preset time period according to the feedback information.

It is understandable that the feedback information includes positive feedback information, for example, the user has a higher click-through rate, a longer browsing time, more sharing times, and a higher order volume for the top-ranked items in the target information. In other words, the recommendation degree Higher items generate higher attention and conversion rate; feedback information also includes negative feedback information, such as users not clicking on the top-ranked items, or directly closing the display page and other operations, in other words, higher recommendation The item does not match what the user is interested in.

Based on these positive and negative feedback information, the model is incrementally trained and calibrated. Among them, the preset duration can be in units of days, that is, incremental training is performed on the ranking model every day using the historical data of the previous day to calibrate the update; the preset duration can also be in hours, that is, the ranking model is used every hour before One hour of historical data for incremental training to calibrate updates. Continuous calibration and updating of the model can improve the accuracy of the ranking model, increase the possibility that the ranking results match the content of interest to the user, and thus improve the accuracy of information recommendation.

The information recommendation method provided by the embodiments of the present disclosure will be specifically described below with reference to FIG. 2 and FIG. 4 , taking an information recommendation method applied to an e-commerce platform as an example.

Figure 4 is an information recommendation method applied to an e-commerce platform, and its steps are as follows:

Step S401: According to the recommended target parameters, use the marketing algorithm module to create a marketing plan.

The marketing algorithm module can be shown in Figure 2, and the recommended target parameters can be return on investment, inventory consumption parameters, total value and number of people to be recommended. The marketing algorithm module can output product selection recommendations, quota recommendations and crowd recommendations. Among them, product selection recommendation refers to product information; quota recommendation refers to the specific discount amount of product information; crowd recommendation refers to the group of people recommended by product information.

Step S402: Utilize the marketing content generation module to generate product recommendation content, and obtain the aggregated content of the marketing product through aggregation.

As shown in Figure 2, while creating a marketing plan, the product information output by the marketing algorithm module can be used as the input of the marketing content generation module, and through NLP (Natural Language Processing, natural language processing) and image processing technology, product related information can be obtained. Recommended content, including recommended short titles, recommended short texts, recommended copywriting information, recommended short videos, and comment lists, etc.

When the product information is a product category, the product whose historical order quantity in the product category is greater than the preset threshold can be used as the input of the marketing content generation module; the product whose sales volume in the historical order is greater than the preset threshold can also be used as the marketing content to generate The input of the module; it is also possible to use the commodities with the total transaction amount of the historical order greater than the preset threshold as the input of the marketing content generation module.

Aggregate the generated recommended content and the marketing product list to obtain the aggregated marketing product content. The aggregated content of marketing products can include multi-dimensional information such as marketing product information, preferential information, listings, comments, and marketing copywriting.

Step S403: Store the aggregated content of the marketing commodity in the database.

As shown in Figure 2, the marketing aggregation content can be stored in the database through the data storage module, which can be MYSQL, Redis database, or HBASE, Elasticsearch database.

Step S404: Receive the user's trigger for the target area, and use the ranking model in the recommendation module to sort the multiple products to be recommended corresponding to the target area.

The target area can be a content entrance, such as discovery of good products-recommendation, discovery of good products-word-of-mouth, shopping, and posting orders with prizes. It can also be a marketing portal, such as product details discount coupons, shopping cart coupons, advertisement placement and seckill entrances.

The sorting result of the sorting module is also the order in which multiple products to be recommended are recommended for the current user.

Step S405: According to the result of the sorting model, recommend a plurality of products to be recommended and the aggregated content of corresponding marketing products on the display page.

As shown in Figure 2, on the e-commerce platform, through the RPC or HTTP protocol, the aggregated content of the marketing products stored in the database is displayed and recommended on the display page according to the sorting results of the products to be recommended.

Through the information recommendation method, the triggering of the marketing portal by the user can enable the user to browse the recommended content of the product that the user is interested in, which can increase the click conversion rate of the product and further improve the efficiency of marketing.

Similarly, the triggering of the content portal by the user can enable the user to browse the aggregated marketing information. For example, when a user triggers a discovery portal, the product to be recommended corresponding to the discovery portal and multiple marketing aggregation information corresponding to the product can be recommended to the user, so that the user can quickly find preferential information corresponding to multiple marketing related to the product of interest, Reduce the operation of users searching for preferential information and improve user experience.

According to the information recommendation method provided by the embodiment of the present disclosure, according to the sorting results of multiple item information corresponding to the target area by the sorting model, the aggregated content of the items corresponding to the target area can be recommended to the user through the display page, so that the user does not need additional operations That is, the content recommendation of interest can be obtained, the effectiveness and efficiency of the information recommendation are improved, and the user experience is further improved.

Furthermore, multiple item information can be used as the input of the content generation model to quickly generate and aggregate recommended content for multiple item information, reducing the workload of information recommendation; Optimized to improve the recommendation accuracy of the ranking model, thereby increasing the click conversion rate of items.

As shown in FIG. 5 , one or more embodiments of the present disclosure provide an information recommendation device 500, including: an acquisition module 501, a sorting module 502, and a recommendation module 503; wherein,

The obtaining module 501 is configured to obtain the first user feature corresponding to the user in response to the user's triggering of the target area in the display page; the target area corresponds to a plurality of item information;

The sorting module 502 is configured to use the first user characteristics and the plurality of item information as input of a sorting model, and determine the sorting result of the multiple item information according to the output of the sorting model; the sorting model It is obtained by training according to multiple second user features and historical behavior information respectively corresponding to the multiple second user features;

The recommendation module 503 is configured to determine the target information to be recommended according to the sorting result and the item aggregation content; the item aggregation content is obtained according to the recommendation content aggregation of the plurality of item information; the target information is passed through the The display page mentioned above is recommended.

In one or more embodiments of the present disclosure, with continued reference to FIG. 5 , the information recommendation device may further include: an aggregation module 504; wherein,

The aggregation module 504 is configured to use the plurality of item information as the input of the content generation model to obtain the recommended content respectively corresponding to the plurality of item information; the recommended content includes one or more of the following: recommended Short title, recommended short text, recommended copy information, comment information and recommended short video information; the recommended content is aggregated to obtain the aggregated content of the item.

In one or more embodiments of the present disclosure, the obtaining module 501 is configured to obtain recommended target parameters before receiving a user's trigger on the target area in the presentation page, and the recommended target parameters include any of the following or Multiple items: return on investment, inventory consumption parameters, total value, and number of people to be recommended; according to the recommended target parameters, the type of the target area, and the item information that can be displayed on the display page, determine the number of items corresponding to the target area item information.

In one or more embodiments of the present disclosure, the acquisition module 501 is configured to determine that the item information includes an item category; determine target item information from item information corresponding to the item category; wherein, the target The item information is historical order quantity, and/or order completion amount, and/or item information whose order value is greater than a preset threshold; the target item information is used as the input of the content generation model.

In one or more embodiments of the present disclosure, the sorting module 502 is configured to obtain user feedback on the target information through the display page; optimize the sorting model according to a preset duration according to the feedback information .

In one or more embodiments of the present disclosure, the sorting module 502 is configured to determine that the sorting model is trained based on LR algorithm, GBDT algorithm, Xgboost algorithm, LightGBM algorithm, xDeepFM algorithm, DeepFM algorithm and AutoInt algorithm.

In one or more embodiments of the present disclosure, the recommendation module 503 is configured to use the plurality of item information as an input of the content generation model before determining the target information to be recommended according to the sorting result and item aggregation content, Obtain the recommended content respectively corresponding to the plurality of item information; the recommended content includes one or more of the following: recommended short title, recommended short text, recommended copywriting information, comment information and recommended short video information; The recommended content is aggregated to obtain the aggregated content of the item.

According to the information recommendation device provided by one or more embodiments of the present disclosure, according to the sorting results of multiple item information corresponding to the target area by the sorting model, the item aggregation content corresponding to the target area can be recommended to the user through the display page. As a result, the number of operations of the user is reduced, and the content of interest to the user can be accurately recommended to the user, the effectiveness and efficiency of information recommendation are improved, and user experience is further improved.

Fig. 6 shows an exemplary system architecture 600 to which the information recommendation method or information recommendation device of one or more embodiments of the present disclosure can be applied.

As shown in FIG. 6 , a system architecture 600 may include

terminal devices

601 , 602 , and 603 , a network 604 and a server 605 . The network 604 is used as a medium for providing communication links between the

terminal devices

601 , 602 , 603 and the server 605 . Network 604 may include various connection types, such as wires, wireless communication links, or fiber optic cables, among others.

Users can use

terminal devices

601 , 602 , 603 to interact with server 605 via network 604 to receive or send messages and the like.

The

terminal devices

601, 602, and 603 may be various electronic devices that have a display screen and support information browsing, including but not limited to smart phones, tablet computers, laptop computers and desktop computers, and the like.

The server 605 may be a server that provides various services, for example, a server that provides information recommendation for users to use the

terminal devices

601, 602, 603 to trigger the display of the target area of the page. The server for information recommendation can sort multiple item information corresponding to the target area, generate item aggregate content, determine the target information to be recommended, and recommend the target information on the display page for display on the

terminal devices

601, 602, and 603 .

It should be understood that the numbers of terminal devices, networks and servers in FIG. 6 are only illustrative. According to the implementation needs, there can be any number of terminal devices, networks and servers.

Referring now to FIG. 7 , it shows a schematic structural diagram of a computer system 700 suitable for implementing a terminal device according to an embodiment of the present disclosure. The terminal device shown in FIG. 7 is only an example, and should not limit the functions and scope of use of this embodiment of the present disclosure.

As shown in FIG. 7 , a computer system 700 includes a central processing unit (CPU) 701 that can operate according to a program stored in a read-only memory (ROM) 702 or a program loaded from a storage section 708 into a random-access memory (RAM) 703 Instead, various appropriate actions and processes are performed. In the RAM 703, various programs and data required for the operation of the system 700 are also stored. The CPU 701, ROM 702, and RAM 703 are connected to each other via a bus 704. An input/output (I/O) interface 705 is also connected to the bus 704 .

The following components are connected to the I/O interface 705: an input section 706 including a keyboard, a mouse, etc.; an output section 707 including a cathode ray tube (CRT), a liquid crystal display (LCD), etc., and a speaker; a storage section 708 including a hard disk, etc. and a communication section 709 including a network interface card such as a LAN card, a modem, or the like. The communication section 709 performs communication processing via a network such as the Internet. A drive 710 is also connected to the I/O interface 705 as needed. A removable medium 711, such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, etc., is mounted on the drive 710 as necessary so that a computer program read therefrom is installed into the storage section 708 as necessary.

In particular, according to the disclosed embodiments of the present disclosure, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, the disclosed embodiments of the present disclosure include a computer program product, which includes a computer program carried on a computer-readable medium, where the computer program includes program codes for executing the methods shown in the flowcharts. In such an embodiment, the computer program may be downloaded and installed from a network via communication portion 709 and/or installed from removable media 711 . When this computer program is executed by a central processing unit (CPU) 701, the above-described functions defined in the system of the present disclosure are performed.

It should be noted that the computer-readable medium shown in the present disclosure may be a computer-readable signal medium or a computer-readable storage medium or any combination of the above two. A computer readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device, or device, or any combination thereof. More specific examples of computer-readable storage media may include, but are not limited to, electrical connections with one or more wires, portable computer diskettes, hard disks, random access memory (RAM), read-only memory (ROM), erasable Programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination of the above. In the present disclosure, a computer-readable storage medium may be any tangible medium that contains or stores a program that can be used by or in conjunction with an instruction execution system, apparatus, or device. In the present disclosure, however, a computer-readable signal medium may include a data signal propagated in baseband or as part of a carrier wave, carrying computer-readable program code therein. Such propagated data signals may take many forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the foregoing. A computer-readable signal medium may also be any computer-readable medium other than a computer-readable storage medium, which can send, propagate, or transmit a program for use by or in conjunction with an instruction execution system, apparatus, or device. . Program code embodied on a computer readable medium may be transmitted using any appropriate medium, including but not limited to wireless, wireline, optical fiber cable, RF, etc., or any suitable combination of the foregoing.

The flowchart and block diagrams in the Figures illustrate the architecture, functionality, and operation of possible implementations of systems, methods and computer program products according to various embodiments of the present disclosure. In this regard, each block in a flowchart or block diagram may represent a module, program segment, or portion of code that includes one or more logical functions for implementing specified executable instructions. It should also be noted that, in some alternative implementations, the functions noted in the block may occur out of the order noted in the figures. For example, two blocks shown in succession may, in fact, be executed substantially concurrently, or they may sometimes be executed in the reverse order, depending upon the functionality involved. It should also be noted that each block in the block diagrams or flowchart illustrations, and combinations of blocks in the block diagrams or flowchart illustrations, can be implemented by a dedicated hardware-based system that performs the specified function or operation, or can be implemented by a A combination of dedicated hardware and computer instructions.

The modules involved in the embodiments described in the present disclosure may be implemented by software or by hardware. The described modules can also be set in a processor, for example, it can be described as: a processor includes an acquisition module, a sorting module, and a recommendation module. Wherein, the names of these modules do not constitute a limitation of the module itself under certain circumstances, for example, the obtaining module may also be described as a "module for obtaining item information".

As another aspect, the present disclosure also provides a computer-readable medium, which may be included in the device described in the above embodiments, or may exist independently without being assembled into the device. The above-mentioned computer-readable medium carries one or more programs, and when the above-mentioned one or more programs are executed by the device, the device includes: responding to the user's trigger on the target area in the display page, acquiring the user's corresponding The first user feature; the target area corresponds to a plurality of item information; the first user feature and the plurality of item information are used as the input of the ranking model, and the plurality of item information is determined according to the output of the ranking model The sorting result; the sorting model is obtained by training according to multiple second user characteristics and the historical behavior information corresponding to the multiple second user features; according to the sorting result and the aggregated content of the item, determine the target information to be recommended ; The aggregated content of the item is obtained through aggregation of the recommended content of the plurality of item information; and the target information is recommended through the display page.

According to the technical solutions of the embodiments of the present disclosure, according to the sorting results of multiple item information corresponding to the target area by the sorting model, the aggregated content of the items corresponding to the target area can be recommended to the user through the display page. As a result, the number of operations of the user is reduced, and the content of interest to the user can be accurately recommended to the user, the effectiveness and efficiency of information recommendation are improved, and user experience is further improved.

The specific implementation manners described above do not limit the protection scope of the present disclosure. It should be apparent to those skilled in the art that various modifications, combinations, sub-combinations and substitutions may occur depending on design requirements and other factors. Any modifications, equivalent replacements and improvements made within the spirit and principles of the present disclosure shall be included within the protection scope of the present disclosure.

Claims

A method for recommending information, comprising:

Responding to the triggering of the target area by the user on the display page, acquiring the first user feature corresponding to the user; the target area corresponds to a plurality of item information;

Using the first user characteristics and the plurality of item information as the input of a ranking model, and determining the ranking result of the plurality of item information according to the output of the ranking model; the ranking model is based on a plurality of second user characteristics And the historical behavior information respectively corresponding to the plurality of second user characteristics is obtained through training;

Determine the target information to be recommended according to the sorting result and the item aggregation content; the item aggregation content is obtained by aggregation according to the recommendation content of the plurality of item information;

The target information is recommended through the display page.
The method according to claim 1, wherein, before determining the target information to be recommended according to the sorting result and the item aggregation content, further comprising:

Using the plurality of item information as the input of the content generation model to obtain the recommended content respectively corresponding to the plurality of item information; the recommended content includes one or more of the following: recommended short title, recommended short text, recommended Copywriting information, comment information and recommended short video information;

The recommended content is aggregated to obtain the item aggregated content.
The method according to claim 2, wherein, before receiving the user's trigger on the target area in the display page, further comprising:

Acquiring recommended target parameters, the recommended target parameters include any one or more of the following: return on investment, inventory consumption parameters, total value and number of people to be recommended;

A plurality of item information corresponding to the target area is determined according to the recommended target parameters, the type of the target area, and the item information that can be displayed on the display page.
The method according to claim 3, wherein the item information includes an item category;

Determine the target item information from the item information corresponding to the item category; wherein the target item information is the historical order quantity, and/or order completion amount, and/or item information whose order value is greater than a preset threshold;

The target item information is used as an input of the content generation model.
The method according to claim 1, further comprising:

Obtain user feedback on the target information through the display page;

According to the feedback information, the ranking model is optimized according to a preset time period.
The method according to claim 1, wherein,

The ranking model is trained based on LR algorithm, GBDT algorithm, Xgboost algorithm, LightGBM algorithm, xDeepFM algorithm, DeepFM algorithm and AutoInt algorithm.
An information recommendation device, including: an acquisition module, a sorting module, and a recommendation module; wherein,

The acquisition module is configured to acquire the first user feature corresponding to the user in response to the user's trigger on the target area in the display page; the target area corresponds to a plurality of item information;

The sorting module is configured to use the first user characteristics and the plurality of item information as input to a sorting model, and determine the sorting result of the multiple item information according to the output of the sorting model; the ranking model is Obtained by training according to multiple second user features and historical behavior information respectively corresponding to the multiple second user features;

The recommendation module is configured to determine the target information to be recommended according to the sorting result and the item aggregation content; the item aggregation content is obtained according to the recommendation content aggregation of the plurality of item information; the target information is passed through the Display pages for recommendations.
The device according to claim 7, further comprising: an aggregation module; wherein,

The aggregation module is configured to use the plurality of item information as the input of the content generation model to obtain the recommended content respectively corresponding to the plurality of item information; the recommended content includes one or more of the following: recommended short Title, recommended short text, recommended copy information, comment information and recommended short video information; the recommended content is aggregated to obtain the aggregated content of the item.
An electronic device comprising:

one or more processors;

storage means for storing one or more programs,

When the one or more programs are executed by the one or more processors, the one or more processors are made to implement the method according to any one of claims 1-6.
A computer-readable medium, on which a computer program is stored, and when the program is executed by a processor, the method according to any one of claims 1-6 is implemented.