Information recommendation method and device, computer device, and storage medium

Abstract

The application relates to the fields of artificial intelligence and intelligent medical treatment, and provides an information recommendation method, device and equipment and a computer storage medium, the method comprising the following steps: obtaining static features of to-be-recommended information based on a preset multi-modal model, wherein the static features are determined according to visual features and text features of the to-be-recommended information; determining dynamic features of the to-be-recommended information according to a behavior sequence of the to-be-recommended information; performing clustering processing on the to-be-recommended information according to the static information and the dynamic information, and determining information types of the to-be-recommended information; and determining target recommended information corresponding to content information from the to-be-recommended information based on the information types corresponding to the content information according to a browsing operation of the content information. The information recommendation method provided by the application can be applied to a medical platform, and diversified information features are combined as the basis for information recommendation, so that the accuracy of information recommendation is improved.

Description

Technical Field

[0001] This application relates to the fields of artificial intelligence and smart healthcare, and in particular to an information recommendation method, apparatus, computer equipment, and storage medium. Background Technology

[0002] With the rise of intelligent recommendation algorithms, their application in content platforms is becoming increasingly common. These algorithms recommend content that a user might be interested in based on their browsing history. For example, a content platform could be a medical platform where a user searches for medical information, and the platform recommends other content based on the relevance of the user's browsing history to other content. However, existing intelligent recommendation algorithms often rely on content tags, some of which are even user-added. The criteria for determining recommended content are often relatively singular, making it difficult to accurately identify user intent and resulting in low accuracy in recommended content. Summary of the Invention

[0003] The main objective of this application is to provide an information recommendation method, apparatus, device, and computer storage medium, with the aim of improving the accuracy of information recommendation.

[0004] Firstly, this application provides an information recommendation method, which includes the following steps:

[0005] Based on a preset multimodal model, static features of the information to be recommended are obtained, including those determined based on the visual and textual features of the information to be recommended.

[0006] Based on the behavioral sequence of the information to be recommended, determine the dynamic characteristics of the information to be recommended;

[0007] Based on the static features and the dynamic features, the information to be recommended is clustered to determine the information type of the information to be recommended;

[0008] Based on the browsing operation of the content information, and according to the information type corresponding to the content information, target recommended information corresponding to the content information is determined from the information to be recommended.

[0009] Secondly, this application also provides an information recommendation device, the information recommendation device comprising:

[0010] The static information acquisition module is used to acquire static features of the information to be recommended based on a preset multimodal model. The static features include those determined based on the visual features and text features of the information to be recommended.

[0011] The dynamic information acquisition module is used to determine the dynamic features of the information to be recommended based on the behavioral sequence of the information to be recommended.

[0012] The information clustering module is used to perform clustering processing on the information to be recommended based on the static features and the dynamic features, and to determine the information type of the information to be recommended.

[0013] The information recommendation module is used to determine the target recommended information corresponding to the content information from the information to be recommended based on the browsing operation on the content information and the information type corresponding to the content information.

[0014] Thirdly, this application also provides a computer device, the computer device including a processor, a memory, and a computer program stored in the memory and executable by the processor, wherein when the computer program is executed by the processor, it implements the information recommendation method as described above.

[0015] Fourthly, this application also provides a computer-readable storage medium storing a computer program, wherein when the computer program is executed by a processor, it implements the information recommendation method described above.

[0016] This application provides an information recommendation method, apparatus, device, and computer storage medium. The application obtains static features of information to be recommended based on a preset multimodal model. These static features include those determined based on the visual and textual features of the information to be recommended. Dynamic features of the information to be recommended are determined based on a sequence of behaviors related to the information. The information to be recommended is then clustered based on the static and dynamic features to determine its information type. Finally, based on browsing operations on the content information and the corresponding information type, target recommended information is determined from the information to be recommended that matches the content information. By combining diverse information features as the basis for information recommendation, the target recommended information is more aligned with the user's actual intentions, thus improving the accuracy of information recommendation. Attached Figure Description

[0017] To more clearly illustrate the technical solutions of the embodiments of this application, the drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 A flowchart illustrating an information recommendation method provided in an embodiment of this application;

[0019] Figure 2 This application provides a usage scenario diagram of an information recommendation method according to an embodiment of the present application.

[0020] Figure 3 A schematic block diagram of an information recommendation device provided in an embodiment of this application;

[0021] Figure 4 This is a schematic block diagram of the structure of a computer device according to an embodiment of this application. Detailed Implementation

[0022] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0023] The flowchart shown in the attached diagram is for illustrative purposes only and does not necessarily include all content and operations / steps, nor does it necessarily have to be performed in the order described. For example, some operations / steps can be broken down, combined, or partially merged, so the actual execution order may change depending on the actual situation.

[0024] This application provides an information recommendation method, apparatus, computer device, and computer-readable storage medium.

[0025] The following detailed description of some embodiments of this application is provided in conjunction with the accompanying drawings. Unless otherwise specified, the following embodiments and features can be combined with each other.

[0026] Please refer to Figure 1 , Figure 1 This is a flowchart illustrating an information recommendation method provided in an embodiment of this application. This information recommendation method can be used in a terminal or server to recommend information to users browsing content on a content platform. The terminal can be an electronic device such as a mobile phone, tablet, laptop, desktop computer, personal digital assistant, or wearable device; the server can be a standalone server, a server cluster, or a cloud server providing basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, content delivery networks (CDNs), and big data and artificial intelligence platforms.

[0027] For example, the information recommendation method provided in this application embodiment can be applied to a medical platform to recommend content related to the medical information browsed by the user.

[0028] Please refer to Figure 2 , Figure 2 This is a usage scenario diagram provided by an embodiment of this application. For example... Figure 2 As shown, the static features of the information to be recommended are determined based on visual and textual features, while the dynamic features of the information to be recommended are determined based on the user sequence of the users who perform the operation on the information to be recommended. After determining the dynamic and static features of the information to be recommended, the information type of the information to be recommended is determined based on the dynamic and static information.

[0029] like Figure 1 As shown, the information recommendation method includes steps S101 to S104.

[0030] Step S101: Based on a preset multimodal model, obtain the static features of the information to be recommended. The static features include those determined based on the visual features and text features of the information to be recommended.

[0031] For example, the information to be recommended in the content platform may include various forms of information, such as image information and voice information; of course, it is not limited to these, and may also include voice information, video information, etc., without limitation.

[0032] For example, the multimodal model can process various forms of information and integrate the features of various forms of information to obtain static features that can summarize the overall features of the information to be recommended.

[0033] For example, the static features can be obtained based on the visual features of the image information and the text features of the text information in the information to be recommended. Of course, it is not limited to this. The voice features of the voice information in the information to be recommended can also be used to determine the static features, which is not limited here.

[0034] In some implementations, the step of obtaining static features of the information to be recommended based on a preset multimodal model includes obtaining the static features based on the visual and textual features of the information to be recommended, including: fusing the visual and textual features of the information to be recommended based on a preset self-attention model to obtain the static features.

[0035] For example, the multimodal model can be a self-attention model (Vision-and-Language Transformer, ViLT) that can fuse visual and textual features.

[0036] For example, by inputting the visual and textual features of the information to be recommended into ViLT, static features that can be used to describe the information to be recommended can be obtained. These static features can be represented by vectors, but are not limited to this, and are not restricted here.

[0037] In some embodiments, the method further includes: acquiring image information from the information to be recommended; splitting the image information into multiple sub-image information; inputting the sub-image information into a preset image feature extraction network to obtain sub-image features corresponding to the sub-image information; determining the image features of the image information based on the sub-image features to obtain the visual features of the information to be recommended.

[0038] For example, directly extracting features from the image information in the recommendation information requires a large amount of computation. Instead, the image information can be split into sub-images (patch) and features can be extracted from each sub-image. Then, the image features of the image information can be obtained by combining the features of each patch, which can be used as the visual features of the recommendation information.

[0039] For example, determining the image features of the image information based on the sub-image features can also be implemented using a transformer, but is not limited to this method. The sub-image features and the image features can be represented by vectors, but are not limited to this method either.

[0040] In some embodiments, the method further includes: acquiring text information from the information to be recommended; performing word segmentation on the text information to obtain entity text corresponding to the text information; and determining the text features of the entity text based on a preset bidirectional encoding model to obtain the text features of the information to be recommended.

[0041] For example, the text information in the recommendation information is segmented to obtain the entity text with actual meaning in the text information, and meaningless text such as auxiliary words and modal particles in the text information is filtered out to improve the accuracy of text features.

[0042] For example, features are extracted from the entity text using a Bidirectional Encoder Representation from Transformers (BERT) model to obtain the text features of the entity text corresponding to the information to be recommended. These text features can be represented by vectors, but are not limited to this, and are not restricted here.

[0043] Step S102: Determine the dynamic characteristics of the information to be recommended based on the behavioral sequence of the information to be recommended.

[0044] For example, the sequence of behaviors of the information to be recommended may be a sequence of users [user1, user2, ..., user] who perform browsing operations on the information to be recommended. n ].

[0045] For example, since dynamic features change over time, they can be updated based on a preset period. For instance, the preset period could be 7 days, with the dynamic features of the information to be recommended being redefined every 7 days.

[0046] In some implementations, determining the dynamic features of the information to be recommended based on the behavioral sequence of the information to be recommended includes: classifying the behavioral information according to the recording time of the behavioral information in the behavioral sequence to obtain a long-term behavioral sequence and a short-term behavioral sequence; establishing a dual-tower model based on the long-term behavioral sequence and the short-term behavioral sequence, the dual-tower model being used to determine the long-term dynamic features and short-term dynamic features of the information to be recommended; and determining the dynamic features of the recommended information based on the long-term dynamic features and the short-term dynamic features if the similarity between the long-term dynamic features and the short-term dynamic features is greater than a preset threshold.

[0047] For example, since there may be a large number of behavioral sequences for each piece of information to be recommended, and the dual-tower model has the characteristic of fast processing speed, the dynamic characteristics of the information to be recommended are determined by the dual-tower model, which improves the processing efficiency of processing behavioral sequences and reduces processing time.

[0048] For example, since long-term dynamic features and short-term dynamic features target the same information to be recommended, and the similarity between the two is high, the dynamic features of the recommended information are determined based on the long-term dynamic features and short-term dynamic features when the similarity between the long-term dynamic features and short-term dynamic features is greater than a preset threshold.

[0049] For example, the similarity can be determined based on the similarity between long-term dynamic features and short-term dynamic features, but it is not limited to this.

[0050] In some implementations, the behavior sequence includes: a user sequence that performs browsing operations on the information to be recommended; the step of classifying the behavior information according to the recording time of the behavior information in the behavior sequence to obtain a long-term behavior sequence and a short-term behavior sequence includes: classifying the user information according to the recording time of the user information in the user sequence to obtain a long-term user sequence and a short-term user sequence.

[0051] For example, users who browse the same information to be recommended usually have certain common characteristics. Therefore, the characteristics of the information to be recommended can be determined by the user who clicks on the information to be recommended and performs the browsing operation.

[0052] For example, user sequences can be divided into short-term user sequences and long-term user sequences. The short-term user sequence can be, for example, a sequence of users who viewed the information to be recommended within the last 7 days; the long-term user sequence can be a sequence of users who viewed the information to be recommended between 7 days ago and 30 days ago.

[0053] In some implementations, classifying the user information according to the recording time of the user information in the user sequence to obtain long-term user sequences and short-term user sequences includes at least one of the following: filtering the user information in the long-term user sequence based on the browsing duration of the user corresponding to the user information when performing a browsing operation on the information to be recommended; filtering the user information in the long-term user sequence based on the preference intensity of the user corresponding to the user information for the information to be recommended, wherein the preference intensity is determined based on whether the user performs a preset operation on the information to be recommended; and filtering the user information in the long-term user sequence based on the user attributes corresponding to the user information in the user sequence.

[0054] For example, since long-term user sequences are sequences of users who perform browsing operations on the recommended information over a relatively long period of time, their data volume may be large. Therefore, user information in long-term user sequences can be filtered to reduce the data volume of long-term user sequences and improve the efficiency of obtaining dynamic features.

[0055] For example, user information in a long-term user sequence can be filtered based on the browsing duration of the user corresponding to the user information when browsing the recommended information. It is understood that some users may have browsed the recommended information due to accidental clicks, and users who clicked accidentally have lower relevance to the characteristics of the recommended information. Since users who clicked accidentally usually exit the page and stop browsing the recommended information within a short time, they can be filtered based on their browsing duration. Specifically, if the browsing duration of the user corresponding to the user information is less than a preset duration, the user information is filtered.

[0056] For example, the strength of a user's preference for the recommended information can be determined based on the user's actions performed on the information to be recommended. The preset actions performed by the user on the recommended information may include: purchasing, adding to cart, etc. It is understood that if a user has performed a purchase or added to cart action on the recommended information, it indicates that the user has a strong preference for the recommended information, and the browsing actions performed by the user on the recommended information are more reliable. Specifically, if the user corresponding to the user information has not performed the preset action on the recommended information, the user information is filtered.

[0057] For example, users who browse the same recommended information typically have similar user attributes, including user age, user region, etc. Therefore, user information can be filtered based on user attributes. Specifically, the sparsity of user information in the user sequence is determined based on the user information; if the sparsity of the user information is greater than a preset sparsity, the user information is filtered.

[0058] Step S103: Based on the static features and the dynamic features, perform clustering processing on the information to be recommended to determine the information type of the information to be recommended.

[0059] For example, based on the K-means clustering algorithm, the information to be recommended is clustered according to the static features and the dynamic features.

[0060] For example, by clustering the information to be recommended based on static and dynamic features, information with similar features can be identified as the same information type, so that the content information to be recommended to the user can be determined according to the information type.

[0061] Step S104: Based on the browsing operation for the content information, and based on the information type corresponding to the content information, determine the target recommendation information corresponding to the content information from the information to be recommended.

[0062] For example, the content information is the information currently being viewed by the user. Based on the information type to which the content information belongs, the information to be recommended that belongs to the same information type as the content information is determined as the target recommended information, and the target recommended information is displayed to the user on the page of the content information.

[0063] For example, if there are many pieces of information to be recommended in the information type, the target recommended information can be determined based on the similarity between the information to be recommended in the information type and the content information. Specifically, the recommendation order of the information to be recommended is determined based on the similarity between the information to be recommended in the information type and the content information, and the information to be recommended within a preset number of recommendation orders is determined as the target recommended information.

[0064] For example, if the preset quantity is 5, meaning that the number of target recommended information items displayed on the content information page is 5, then the top 5 recommended information items with the highest similarity to the content information will be determined as target recommended information. The preset quantity can be determined according to actual needs and is not limited here.

[0065] For example, if the number of recommended information items in the information type is small, such as when the number is less than the preset number, recommended information items belonging to other information types can be identified as target recommended information. Specifically, the other information type can be the information type with the highest similarity to the information type corresponding to the content information.

[0066] For example, the similarity may be determined based on a pre-defined distance, but it is not limited thereto and is not restricted here.

[0067] The information recommendation method provided in the above embodiments obtains static features of the information to be recommended based on a preset multimodal model. These static features include those determined based on the visual and textual features of the information to be recommended. Dynamic features of the information to be recommended are determined based on a sequence of behaviors related to the information. The information to be recommended is then clustered based on the static and dynamic features to determine its information type. Finally, based on browsing operations on the content information and the corresponding information type, target recommended information corresponding to the content information is determined from the information to be recommended. By combining diverse information features as the basis for information recommendation, the target recommended information is more aligned with the user's actual intent, thus improving the accuracy of information recommendation.

[0068] Please see Figure 3 , Figure 3 This is a schematic diagram of an information recommendation device provided in an embodiment of this application. The information recommendation device can be configured in a server or terminal to execute the aforementioned information recommendation method.

[0069] like Figure 3 As shown, the information recommendation device includes: a static information acquisition module 110, a dynamic information acquisition module 120, an information clustering module 130, and an information recommendation module 140.

[0070] The static information acquisition module 110 is used to acquire static features of the information to be recommended based on a preset multimodal model. The static features include those determined based on the visual features and text features of the information to be recommended.

[0071] The dynamic information acquisition module 120 is used to determine the dynamic features of the information to be recommended based on the behavioral sequence of the information to be recommended.

[0072] The information clustering module 130 is used to perform clustering processing on the information to be recommended based on the static features and the dynamic features, and to determine the information type of the information to be recommended.

[0073] The information recommendation module 140 is used to determine target recommended information corresponding to the content information from the information to be recommended based on the browsing operation on the content information and the information type corresponding to the content information.

[0074] For example, the dynamic information acquisition module 120 also includes an information classification submodule, a dual-tower model submodule, and a dynamic feature determination submodule.

[0075] The classification submodule is used to classify the behavior information according to the recording time of the behavior information in the behavior sequence to obtain long-term behavior sequences and short-term behavior sequences;

[0076] The dual-tower model submodule is used to establish a dual-tower model based on the long-term behavior sequence and the short-term behavior sequence. The dual-tower model is used to determine the long-term dynamic features and short-term dynamic features of the information to be recommended.

[0077] The dynamic feature determination submodule is used to determine the dynamic features of the recommendation information based on the long-term dynamic features and the short-term dynamic features if the similarity between the long-term dynamic features and the short-term dynamic features is greater than a preset threshold.

[0078] For example, the classification submodule includes a user sequence classification submodule.

[0079] The user sequence classification submodule is used to classify the user information according to the recording time of the user information in the user sequence to obtain long-term user sequences and short-term user sequences.

[0080] For example, the user sequence classification submodule includes at least one of a browsing duration filtering submodule, a preference intensity filtering submodule, and an information attribute filtering submodule.

[0081] The browsing duration filtering submodule is used to filter user information in the long-term user sequence based on the browsing duration of the user corresponding to the user information when performing browsing operations on the information to be recommended;

[0082] The preference intensity filtering submodule is used to filter user information in the long-term user sequence based on the preference intensity of the user corresponding to the user information for the information to be recommended, wherein the preference intensity is determined according to whether the user performs a preset operation on the information to be recommended;

[0083] The information attribute filtering submodule is used to filter the user information in the long-term user sequence based on the user attributes corresponding to the user information in the user sequence.

[0084] For example, the static information acquisition module 110 includes a feature fusion submodule.

[0085] The feature fusion submodule is used to fuse the visual and textual features of the information to be recommended based on a preset self-attention model to obtain the static features.

[0086] For example, the information recommendation device further includes an image segmentation module, an image feature extraction module, and an image feature determination module.

[0087] An image segmentation module is used to obtain image information from the information to be recommended and to segment the image information into multiple sub-image information.

[0088] The image feature extraction module is used to input the sub-image information into a preset image feature extraction network to obtain the sub-image features corresponding to the sub-image information;

[0089] The image feature determination module is used to determine the image features of the image information based on the sub-image features, thereby obtaining the visual features of the information to be recommended.

[0090] For example, the information recommendation device further includes a text segmentation module and a text feature determination module.

[0091] The text segmentation module is used to obtain the text information in the information to be recommended, perform word segmentation on the text information, and obtain the entity text corresponding to the text information.

[0092] The text feature determination module is used to determine the text features of the entity text based on a preset bidirectional encoding model, thereby obtaining the text features of the information to be recommended.

[0093] It should be noted that those skilled in the art will understand that, for the sake of convenience and brevity, the specific working processes of the above-described apparatus and its modules and units can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0094] The methods and apparatus of this application can be used in a wide variety of general-purpose or special-purpose computing system environments or configurations. Examples include: personal computers, server computers, handheld or portable devices, tablet devices, multiprocessor systems, microprocessor-based systems, set-top boxes, programmable consumer electronics devices, network PCs, minicomputers, mainframe computers, and distributed computing environments including any of the above systems or devices. This application can be described in the general context of computer-executable instructions executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, etc., that perform specific tasks or implement specific abstract data types. This application can also be practiced in distributed computing environments where tasks are performed by remote processing devices connected via a communication network. In distributed computing environments, program modules can reside in local and remote computer storage media, including storage devices.

[0095] For example, the above-described method and apparatus can be implemented as a computer program, which can be used in, for example... Figure 4 It runs on the computer device shown.

[0096] Please see Figure 4 , Figure 4 This is a schematic block diagram illustrating the structure of a computer device provided in an embodiment of this application. The computer device may be a server or a terminal.

[0097] like Figure 4 As shown, the computer device includes a processor, a memory, and a network interface connected via a system bus, wherein the memory may include a storage medium and internal memory.

[0098] The storage medium may store the operating system and computer programs. These computer programs include program instructions that, when executed, cause the processor to perform any information recommendation method.

[0099] The processor provides computing and control capabilities, supporting the operation of the entire computer device.

[0100] Internal memory provides an environment for the execution of computer programs stored in storage media. When these computer programs are executed by a processor, the processor can perform any information recommendation method.

[0101] This network interface is used for network communication, such as sending assigned tasks. Those skilled in the art will understand that... Figure 4 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0102] It should be understood that the processor can be a Central Processing Unit (CPU), but it can also be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. Among these, a general-purpose processor can be a microprocessor or any conventional processor.

[0103] In one embodiment, the processor is configured to run a computer program stored in memory to perform the following steps:

[0104] Based on a preset multimodal model, static features of the information to be recommended are obtained, including those determined based on the visual and textual features of the information to be recommended.

[0105] Based on the behavioral sequence of the information to be recommended, determine the dynamic characteristics of the information to be recommended;

[0106] Based on the static features and the dynamic features, the information to be recommended is clustered to determine the information type of the information to be recommended;

[0107] Based on the browsing operation of the content information, and according to the information type corresponding to the content information, target recommended information corresponding to the content information is determined from the information to be recommended.

[0108] It should be noted that those skilled in the art will understand that, for the sake of convenience and brevity, the specific working process of the above-described information recommendation can be referred to the corresponding process in the aforementioned information recommendation control method embodiment, and will not be repeated here.

[0109] This application also provides a computer-readable storage medium storing a computer program, the computer program including program instructions, and the method implemented when the program instructions are executed can be referred to various embodiments of the method recommended in this application.

[0110] The computer-readable storage medium may be an internal storage unit of the computer device described in the foregoing embodiments, such as the hard disk or memory of the computer device. The computer-readable storage medium may also be an external storage device of the computer device, such as a plug-in hard disk, SmartMedia Card (SMC), Secure Digital (SD) card, or Flash Card equipped on the computer device.

[0111] It should be understood that the terminology used in this specification is for the purpose of describing particular embodiments only and is not intended to limit the scope of the application. As used in this specification and the appended claims, the singular forms “a,” “an,” and “the” are intended to include the plural forms unless the context clearly indicates otherwise.

[0112] It should also be understood that the term "and / or" as used in this specification and the appended claims refers to any combination and all possible combinations of one or more of the associated listed items, and includes such combinations. It should be noted that, herein, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or system that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or system. Without further limitation, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or system that includes that element.

[0113] The sequence numbers of the embodiments in this application are for descriptive purposes only and do not represent the superiority or inferiority of the embodiments. The above descriptions are merely specific implementations of this application, but the scope of protection of this application is not limited thereto. Any person skilled in the art can easily conceive of various equivalent modifications or substitutions within the technical scope disclosed in this application, and these modifications or substitutions should all be covered within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. An information recommendation method, characterized in that, The method includes: Based on a preset multimodal model, static features of the information to be recommended are obtained, including those determined based on the visual and textual features of the information to be recommended. The dynamic characteristics of the information to be recommended are determined based on the behavioral sequence of the information to be recommended, wherein the behavioral sequence includes a sequence of users performing browsing operations on the information to be recommended. Based on the static features and the dynamic features, the information to be recommended is clustered to determine the information type of the information to be recommended; Based on the browsing operation of the content information, and based on the information type corresponding to the content information, the information to be recommended that belongs to the same information type as the content information is determined as the target recommended information; The dynamic features include long-term dynamic features and short-term dynamic features. Determining the dynamic features of the information to be recommended based on the behavioral sequence targeting the information to be recommended includes: Based on the recording time of the behavioral information in the behavioral sequence, the behavioral information is classified to obtain long-term behavioral sequences and short-term behavioral sequences; A dual-tower model is established based on the long-term behavior sequence and the short-term behavior sequence. The dual-tower model is used to determine the long-term dynamic features and short-term dynamic features of the information to be recommended.

2. The information recommendation method according to claim 1, characterized in that, The step of determining the dynamic features of the information to be recommended based on the behavioral sequence of the information to be recommended further includes: If the similarity between the long-term dynamic features and the short-term dynamic features is greater than a preset threshold, the dynamic features of the recommendation information are determined based on the long-term dynamic features and the short-term dynamic features.

3. The information recommendation method according to claim 2, characterized in that, The behavior sequence includes: a user sequence performing browsing operations on the information to be recommended; the step of classifying the behavior information according to the recording time of the behavior information in the behavior sequence to obtain long-term behavior sequences and short-term behavior sequences includes: Based on the recording time of user information in the user sequence, the user information is classified to obtain long-term user sequences and short-term user sequences.

4. The information recommendation method according to claim 3, characterized in that, The step of classifying the user information according to the recording time of the user information in the user sequence to obtain long-term user sequences and short-term user sequences includes at least one of the following: Based on the browsing time of the user corresponding to the user information when performing browsing operations on the information to be recommended, the user information in the long-term user sequence is filtered; Based on the user's preference intensity for the information to be recommended, the user information in the long-term user sequence is filtered, wherein the preference intensity is determined according to whether the user performs a preset operation on the information to be recommended; The user information in the long-term user sequence is filtered based on the user attributes corresponding to the user information in the user sequence.

5. The information recommendation method according to claim 1, characterized in that, The static features of the information to be recommended are obtained based on a preset multimodal model. These static features are determined based on the visual and textual features of the information to be recommended, including: Based on a preset self-attention model, the visual and textual features of the information to be recommended are fused to obtain the static features.

6. The information recommendation method according to any one of claims 1-5, characterized in that, The method further includes: Obtain image information from the information to be recommended, and split the image information into multiple sub-image information; The sub-image information is input into a preset image feature extraction network to obtain the sub-image features corresponding to the sub-image information; Based on the sub-image features, the image features of the image information are determined to obtain the visual features of the information to be recommended.

7. The information recommendation method according to any one of claims 1-5, characterized in that, The method further includes: Obtain the text information from the information to be recommended, perform word segmentation on the text information, and obtain the entity text corresponding to the text information; Based on a preset bidirectional encoding model, the text features of the entity text are determined, and the text features of the information to be recommended are obtained.

8. An information recommendation device, characterized in that, The information recommendation device includes: The static information acquisition module is used to acquire static features of the information to be recommended based on a preset multimodal model. The static features include those determined based on the visual features and text features of the information to be recommended. The dynamic information acquisition module is used to determine the dynamic characteristics of the information to be recommended based on the behavioral sequence of the information to be recommended, wherein the behavioral sequence includes a user sequence of performing browsing operations on the information to be recommended; The information clustering module is used to perform clustering processing on the information to be recommended based on the static features and the dynamic features, and to determine the information type of the information to be recommended. The information recommendation module is used to determine the information to be recommended, which belongs to the same information type as the content information, as the target recommended information based on the browsing operation on the content information and the information type corresponding to the content information. The dynamic features include long-term dynamic features and short-term dynamic features. Determining the dynamic features of the information to be recommended based on the behavioral sequence targeting the information to be recommended includes: Based on the recording time of the behavioral information in the behavioral sequence, the behavioral information is classified to obtain long-term behavioral sequences and short-term behavioral sequences; A dual-tower model is established based on the long-term behavior sequence and the short-term behavior sequence. The dual-tower model is used to determine the long-term dynamic features and short-term dynamic features of the information to be recommended.

9. A computer device, characterized in that, The computer device includes a processor, a memory, and a computer program stored in the memory and executable by the processor, wherein when the computer program is executed by the processor, it implements the steps of the information recommendation method as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program, wherein when the computer program is executed by a processor, it implements the steps of the information recommendation method as described in any one of claims 1 to 7.

Images