Article recommendation method and device

Abstract

The invention discloses an article recommendation method and device. A specific embodiment of the method comprises the following steps: according to a group portrait of a private domain group where a target user is located, vectorizing an article recommendation request of the target user to obtain a vectorized recommendation request; vectorizing article information of candidate articles recalled according to the article recommendation request to obtain vectorized article information; and according to the vectorization recommendation request and the vectorization article information, determining an article to be recommended of the target user from the candidate articles. The invention provides an article recommendation method applied to a private domain, and the method integrates private domain group portraits, and achieves the refined operation of thousands of groups and thousands of faces. On the basis of meeting the current recommendation demand of the private domain user, the richness and accuracy of article recommendation are increased.

Description

TECHNICAL FIELD

[0001] Embodiments of the present application relate to the technical field of computer, in particular to the technical field of item recommendation, and more particularly to an item recommendation method and device, a computer readable medium and an electronic device. BACKGROUND

[0002] In recent years, major Internet giants and brand merchants in China are all working hard to build and operate private domains, with the core purpose being to transfer public domain users (referring to users who have not been converted to a specific WeChat group or application private domain) to the private domain, and then continuously convert and monetize them through activities, promotions, private chats, and other means, opening up new growth space for the increasingly dwindling incremental market. Currently, in the field of private domain operation, artificial methods are mainly used to reply to consumer questions and respond to consumer product demand. SUMMARY

[0003] Embodiments of the present application provide an item recommendation method and device, a computer readable medium and an electronic device.

[0004] In a first aspect, embodiments of the present application provide an item recommendation method, comprising: vectorizing an item recommendation request of a target user according to a group portrait of a private domain group in which the target user is located, to obtain a vectorized recommendation request; vectorizing item information of candidate items recalled according to the item recommendation request, to obtain vectorized item information; and determining a to-be-recommended item of the target user from the candidate items according to the vectorized recommendation request and the vectorized item information.

[0005] In some examples, the above-mentioned vectorizing the item recommendation request of the target user according to the group portrait of the private domain group in which the target user is located, to obtain the vectorized recommendation request, comprises: vectorizing the item recommendation request according to the group portrait of the private domain group in which the target user is located, through a pre-trained item recommendation model, to obtain the vectorized recommendation request, wherein the item recommendation model is used to represent the corresponding relationship between the item recommendation request and the to-be-recommended item.

[0006] In some examples, the above-mentioned vectorizing the item information of the candidate items recalled according to the item recommendation request, to obtain the vectorized item information, comprises: recalling a plurality of candidate items according to the item recommendation request, through a pre-trained item recall model, wherein the item recall model is used to represent the corresponding relationship between the item recommendation request and the candidate items; and vectorizing the item information of the plurality of candidate items respectively, to obtain a plurality of vectorized item information.

[0007] In some examples, the recalling, by the pre-trained item recall model, the plurality of candidate items according to the item recommendation request comprises: performing word segmentation on the item recommendation request by using a pre-trained natural language understanding model to obtain a word segmentation result; and recalling the plurality of candidate items according to the word segmentation result by using the item recall model.

[0008] In some examples, the vectorizing, respectively, the item information of the plurality of candidate items to obtain the plurality of vectorized item information comprises: vectorizing, respectively, the item information of the plurality of candidate items by using a pre-trained item recommendation model to obtain the plurality of vectorized item information, wherein the item recommendation model is used to represent a corresponding relationship between the item recommendation request and the to-be-recommended item.

[0009] In some examples, the determining, from the candidate items, the to-be-recommended item of the target user according to the vectorized recommendation request and the vectorized item information comprises: determining a similarity between the vectorized recommendation request and the vectorized item information; and determining, from the candidate items, the to-be-recommended item of the target user according to the similarity.

[0010] In some examples, the vectorizing, according to the group portrait of the private domain group in which the target user is located, the item recommendation request of the target user to obtain the vectorized recommendation request comprises: determining, by using a pre-trained wake-up model, whether the target user has an item recommendation demand according to the item recommendation request; and in response to determining that the target user has the item recommendation demand, vectorizing, according to the group portrait of the private domain group in which the target user is located, the item recommendation request to obtain the vectorized recommendation request.

[0011] In a second aspect, an embodiment of the present application provides an item recommendation apparatus, comprising: a first vectorization unit configured to vectorize, according to a group portrait of a private domain group in which a target user is located, an item recommendation request of the target user to obtain a vectorized recommendation request; a second vectorization unit configured to vectorize item information of a candidate item recalled according to the item recommendation request to obtain vectorized item information; and a determination unit configured to determine, from the candidate items, a to-be-recommended item of the target user according to the vectorized recommendation request and the vectorized item information.

[0012] In some examples, the first vectorization unit is further configured to vectorize, according to the group portrait of the private domain group in which the target user is located, the item recommendation request by using a pre-trained item recommendation model to obtain the vectorized recommendation request, wherein the item recommendation model is used to represent a corresponding relationship between the item recommendation request and the to-be-recommended item.

[0013] In some examples, the second vectorization unit is further configured to: recall, by a pre-trained item recall model, the plurality of candidate items according to the item recommendation request, wherein the item recall model is used to represent a corresponding relationship between the item recommendation request and the candidate items; and vectorize, respectively, item information of the plurality of candidate items to obtain a plurality of vectorized item information.

[0014] In some examples, the second vectorization unit is further configured to: perform word segmentation processing on the item recommendation request by a pre-trained natural language understanding model to obtain a word segmentation result; and recall, by the item recall model, the plurality of candidate items according to the word segmentation result.

[0015] In some examples, the second vectorization unit is further configured to: vectorize, respectively, item information of the plurality of candidate items by a pre-trained item recommendation model to obtain a plurality of vectorized item information, wherein the item recommendation model is used to represent a corresponding relationship between the item recommendation request and the item to be recommended.

[0016] In some examples, the determination unit is further configured to: determine a similarity between the vectorized recommendation request and the vectorized item information; and determine, according to the similarity, the item to be recommended for the target user from the candidate items.

[0017] In some examples, the first vectorization unit is further configured to: determine, by a pre-trained wake-up model, whether the target user has an item recommendation demand according to the item recommendation request; and in response to determining that the target user has the item recommendation demand, vectorize the item recommendation request according to a group portrait of a private domain group in which the target user is located to obtain the vectorized recommendation request.

[0018] In a third aspect, an embodiment of the present application provides a computer readable medium having a computer program stored thereon, wherein the program, when executed by a processor, implements the method described in any implementation manner of the first aspect.

[0019] In a fourth aspect, an embodiment of the present application provides an electronic device, including: one or more processors; and a storage device having one or more programs stored thereon, wherein when the one or more programs are executed by the one or more processors, the one or more processors implement the method described in any implementation manner of the first aspect.

[0020] The article recommendation method and device provided by the embodiment of the application can vectorize an article recommendation request of a target user according to a group portrait of a private domain group in which the target user is located, obtain a vectorized recommendation request, vectorize article information of a candidate article recalled according to the article recommendation request, obtain vectorized article information, and determine a to-be-recommended article of the target user from the candidate article according to the vectorized recommendation request and the vectorized article information, thereby providing an article recommendation method applied to a private domain, fusing a private domain group portrait, and realizing fine operation of 'thousands of groups and thousands of faces'. On the basis of meeting the current recommendation demand of a private domain user, the richness and accuracy of article recommendation are increased. BRIEF DESCRIPTION OF DRAWINGS

[0021] Other features, objects, and advantages of the application will become more apparent from the following detailed description of non-limiting embodiments read in conjunction with the accompanying drawings:

[0022] Figure 1 is an exemplary system architecture diagram to which an embodiment of the application can be applied;

[0023] Figure 2 is a flowchart of one embodiment of an article recommendation method according to the application;

[0024] Figure 3 is a schematic diagram of an application scenario of the article recommendation method according to the embodiment;

[0025] Figure 4 is a flowchart of another embodiment of an article recommendation method according to the application;

[0026] Figure 5 is a structural diagram of one embodiment of an article recommendation device according to the application;

[0027] Figure 6 is a structural schematic diagram of a computer system suitable for implementing the embodiments of the application. DETAILED DESCRIPTION

[0028] The application will be further described in detail below with reference to the drawings and embodiments. It can be understood that the specific embodiments described herein are only used to explain the related application, and not to limit the application. In addition, it should be noted that, for the sake of description, only the parts related to the application are shown in the drawings.

[0029] It should be noted that the embodiments in the application and the features in the embodiments can be combined with each other without conflict. The application will be described in detail below with reference to the drawings and in combination with the embodiments.

[0030] It should be noted that in the technical solutions of the present disclosure, the collection, collection, updating, analysis, processing, use, transmission, storage and the like of user personal information are in line with the relevant legal regulations, are used for legal purposes, and do not violate public order and good customs. Necessary measures are taken for user personal information to prevent illegal access to user personal information data, and to maintain user personal information security, network security and national security.

[0031] Figure 1 An exemplary architecture 100 to which the item recommendation method and device of the present application can be applied is shown.

[0032] As shown in Figure 1 The system architecture 100 can include terminal devices 101, 102, 103, a network 104, and a server 105. The terminal devices 101, 102, 103 are communicatively connected to form a topological network, and the network 104 is used as a medium to provide communication links between the terminal devices 101, 102, 103 and the server 105. The network 104 can include various connection types, such as wired, wireless communication links, or optical fiber cables, etc.

[0033] The terminal devices 101, 102, 103 can interact with the server 105 through the network 104 to receive or send data, etc. The terminal devices 101, 102, 103 can be hardware devices or software that support network connection to interact and process data. When the terminal devices 101, 102, 103 are hardware, they can be various electronic devices that support network connection, information acquisition, interaction, display, processing, etc., including but not limited to smartphones, vehicle-mounted computers, tablet computers, e-book readers, laptop computers, and desktop computers, etc. When the terminal devices 101, 102, 103 are software, they can be installed in the above-mentioned electronic devices. They can be implemented as multiple software or software modules for providing distributed services, or as a single software or software module. No specific limitation is made here.

[0034] The server 105 can be a server that provides various services, such as a background processing server that acquires an item recommendation request sent by a private domain user through a terminal device 101, 102, 103, combines the group portrait of the private domain group in which the target user is located, and recommends items to the private domain user. As an example, the server 105 can be a cloud server.

[0035] It should be noted that the server can be hardware or software. When the server is hardware, it can be implemented as a distributed server cluster composed of multiple servers, or as a single server. When the server is software, it can be implemented as multiple software or software modules (for example, software or software modules used to provide distributed services), or as a single software or software module. This is not specifically limited here.

[0036] It should also be noted that the item recommendation method provided by the embodiments of the present application can be executed by a server, or by a terminal device, or by a server and a terminal device in cooperation with each other. Accordingly, the item recommendation apparatus includes various parts (for example, various units), which can all be arranged in the server, or all be arranged in the terminal device, or be arranged in the server and the terminal device respectively.

[0037] It should be understood that Figure 1 The number of terminal devices, networks and servers inmay be only illustrative. According to the needs of implementation, there can be any number of terminal devices, networks and servers. When the electronic device on which the item recommendation method runs does not need to transmit data with other electronic devices, the system architecture can only include the electronic device (for example, a server or a terminal device) on which the item recommendation method runs.

[0038] With reference to Figure 2 , a flow 200 of one embodiment of the item recommendation method is shown, which includes the following steps:

[0039] In step 201, the item recommendation request of a target user is vectorized according to the group portrait of the private domain group in which the target user is located, to obtain a vectorized recommendation request.

[0040] In this embodiment, the execution subject (for example, the terminal device or the server) of the item recommendation method can obtain the item recommendation request of the target user from a remote or from a local through a wired network connection mode or a wireless network connection mode, and vectorize the item recommendation request of the target user according to the group portrait of the private domain group in which the target user is located, to obtain a vectorized recommendation request. Figure 1

[0041] The target user can be any private domain user in the private domain group, and the private domain group can be a group in various private domains, including but not limited to a private domain group in the private domain owned and controlled by an enterprise or an individual, such as an official website, a program mall, a public account, a WeChat group, an application platform, etc.

[0042] A private domain user profile is the result of a detailed description and analysis of the private domain user group. It generally includes demographic data such as age, gender, marital status, occupation, and income; geographical distribution data such as place of residence and birthplace; education level data; consumer behavior data such as purchasing habits, consumption preferences, and purchasing power; and social media data such as interests and social relationships. The aforementioned entities, with the authorization of the private domain users, can acquire or collect this information, and then analyze and statistically process this data to obtain the user profile.

[0043] As an example, firstly, the aforementioned executing entity can vectorize the target user's item recommendation request to obtain an initial vectorized recommendation request; then, it can vectorize the group profile of the target user's private domain to obtain a vectorized group profile; finally, it can combine the initial vectorized recommendation request and the vectorized group profile to obtain a vectorized recommendation request. The combination method could be, for example, vector concatenation or vector addition.

[0044] As another example, after obtaining the initial vectorized recommendation request and the vectorized group profile, the aforementioned execution entity can use the initial vectorized recommendation request and the vectorized group profile as input to a self-attention mechanism or a Transformer model, and perform cross-encoding to obtain the vectorized recommendation request.

[0045] In some optional implementations of this embodiment, the execution entity can perform step 201 as follows:

[0046] By using a pre-trained item recommendation model, the item recommendation request is vectorized based on the group profile of the target user's private domain, resulting in a vectorized recommendation request.

[0047] Among them, the item recommendation model is used to represent the correspondence between item recommendation requests and items to be recommended.

[0048] Item recommendation models can be neural network models with strong natural language understanding capabilities, such as BERT (Bidirectional Encoder Representations from Transformers), GPT (Generative Pre-trained Transformer), ALBERT (A Lite BERT), and RoBERTa (Robustly Optimized BERT Pretraining Approach), which are improved versions of BERT.

[0049] Item recommendation models can be trained in the following way:

[0050] First, a first training sample set is obtained. This set includes item recommendation request samples, the group profile of the private domain group to which the user belongs, item information of candidate items, and tags for items to be recommended. The tags for items to be recommended are used to characterize whether a candidate item is suitable for recommendation. Then, a machine learning algorithm is used, with the item recommendation request samples, group profiles, and item information as input, and the tags for items to be recommended corresponding to the input items as the expected output, to train an item recommendation model.

[0051] During the initial training of the item recommendation model, the item recommendation request samples are vectorized based on the group profile of the user's private domain. This results in vectorized recommendation request samples, which are continuously learned during iterative training to obtain vectorized recommendation request samples that can represent comprehensive and accurate semantic information.

[0052] In this implementation, the aforementioned executing entity can input the group profile of the target user's private domain group and the item recommendation request into the item recommendation model, and obtain a vectorized recommendation request through the item recommendation model.

[0053] In this implementation, a pre-trained item recommendation model is used to vectorize the item recommendation request based on the group profile of the target user's private domain group, resulting in a vectorized recommendation request. With the help of the item recommendation model's powerful natural language understanding capabilities, the vectorized recommendation request possesses comprehensive and accurate semantic information of both the item recommendation request and the group profile.

[0054] In some optional implementations of this embodiment, the execution entity can perform step 201 as follows:

[0055] First, a pre-trained wake-up model is used to determine whether the target user has a need for item recommendations based on the item recommendation request.

[0056] The wake-up model can use a neural network model with classification capabilities to determine whether the target user has a need for item recommendations based on the item recommendation request.

[0057] As an example, a wake-up model can be trained in the following way:

[0058] First, a second training sample set is obtained. This set includes training samples of item recommendation requests and demand labels representing whether each request has an item recommendation need. Then, a machine learning approach is used, with the item recommendation request samples as output and the corresponding demand labels as the expected output, to train a wake-up model.

[0059] Second, in response to the determination that the target user has a need for item recommendations, the item recommendation request is vectorized based on the group profile of the target user's private domain group, resulting in a vectorized recommendation request.

[0060] Once it is determined that the target user has a need for item recommendations, the recommendation process for the item recommendation request for the target user is initiated. That is, based on the group profile of the target user's private domain group, the item recommendation request is vectorized to obtain a vectorized recommendation request, and the items to be recommended to the target user are determined through subsequent processes.

[0061] If it is determined that the target user does not have a need for item recommendations, the recommendation process for item recommendation requests targeting the target user will not be initiated.

[0062] In this implementation, the recommendation process for the target user's item recommendation request is only activated in response to the target user's explicit expression of the item recommendation request, which ensures the effectiveness of the recommendation process and helps save computational resources for the recommendation process.

[0063] Step 202: Vectorize the item information of the candidate items recalled based on the item recommendation request to obtain vectorized item information.

[0064] In this embodiment, the aforementioned execution entity can vectorize the item information of the candidate items recalled according to the item recommendation request to obtain vectorized item information.

[0065] First, the aforementioned executing entity recalls multiple candidate items from the item database based on the item recommendation request; then, for each of the multiple candidate items, the item information of the candidate item is vectorized to obtain the vectorized item information of the candidate item.

[0066] During the recall process, the aforementioned implementing entities can employ recall methods such as content-based recall, collaborative filtering, and deep neural network-based recall to recall multiple candidate items from the item database.

[0067] In content-based recall methods, the aforementioned implementing entity can use the similarity between items to recall items similar to the target items liked by users, identifying them as candidate items. In collaborative filtering methods, both the similarity between users and the similarity between items are determined simultaneously. Items similar to the target items liked by a second user are recalled to a first user and identified as candidate items. Here, the second user is a user similar to the first user. In deep neural network-based recall methods, a corresponding candidate item set is generated using a deep neural network, and then the items in the candidate item set are sorted to determine multiple candidate items for recall.

[0068] In some optional implementations of this embodiment, the execution entity can perform step 202 as follows:

[0069] First, a pre-trained item recall model is used to recall multiple candidate items based on item recommendation requests.

[0070] Among them, the item recall model is used to characterize the correspondence between item recommendation requests and candidate items.

[0071] Item recall models can employ neural network models with item recall capabilities, such as RNN (Recurrent Neural Network) models and DSSM (Deep Structured Semantic Models).

[0072] The item recall model can be trained in the following way:

[0073] First, a third training sample set is obtained. This set includes training samples of item recommendation requests and recall tags. The recall tags represent candidate items to be recalled. Then, a machine learning approach is used, with the item recommendation request samples as output and the corresponding recall tags as the expected output, to train an item recall model.

[0074] In this implementation, the aforementioned executing entity can input an item recommendation request into the item recall model, and determine multiple candidate items through the item recall model.

[0075] Second, the item information of multiple candidate items is vectorized to obtain multiple vectorized item information.

[0076] As an example, the aforementioned execution entity performs encoding operations on the item information of multiple candidate items to obtain multiple vectorized item information.

[0077] In this implementation, multiple candidate items are recalled through a pre-trained item recall model. The item information of the multiple candidate items is vectorized, which improves the efficiency and accuracy of candidate item determination, and thus improves the efficiency and accuracy of the determination of the obtained vectorized item information.

[0078] In some optional implementations of this embodiment, the execution entity can perform the first step as follows: First, it performs word segmentation on the item recommendation request using a pre-trained natural language understanding model to obtain the word segmentation result; then, it uses an item recall model to recall multiple candidate items based on the word segmentation result.

[0079] Specifically, the aforementioned execution entity uses a pre-trained natural language understanding model to segment the item recommendation request into words, obtaining segmentation results representing multiple keywords in the item recommendation request; then, through an item recall model, it recalls at least one candidate item based on each keyword in the segmentation results; and combines at least one candidate item corresponding to each keyword to obtain multiple candidate items.

[0080] In this implementation, the method of first performing word segmentation based on natural language understanding on the item recommendation request and then recalling the items helps to improve the matching degree between the recalled candidate items and the item recommendation request.

[0081] In some optional implementations of this embodiment, the execution entity can perform the second step as follows:

[0082] By using a pre-trained item recommendation model, the item information of multiple candidate items is vectorized to obtain multiple vectorized item information. The item recommendation model is used to represent the correspondence between item recommendation requests and the items to be recommended.

[0083] The item recommendation model used to vectorize item information and the item recommendation model used to vectorize item recommendation requests are the same model.

[0084] In this implementation, the powerful natural language understanding capabilities of the item recommendation model are leveraged to enable the vectorized item information to possess comprehensive and accurate semantic information.

[0085] Step 203: Based on the vectorized recommendation request and the vectorized item information, determine the items to be recommended to the target user from the candidate items.

[0086] In this embodiment, the aforementioned execution entity can determine the item to be recommended to the target user from the candidate items based on the vectorized recommendation request and the vectorized item information, and recommend the item to the target user.

[0087] As an example, for each vectorized recommendation request, the aforementioned executing entity or the electronic device communicatively connected to the executing entity is provided with target vectorized item information corresponding to the vectorized recommendation request. The target vectorized item information is the vectorized item information that matches the corresponding vectorized recommendation request; in other words, the item represented by the target vectorized item information can be an item that satisfies the item recommendation request. The aforementioned executing entity can match the vectorized item information with the target vectorized item information, and determine the candidate items represented by vectorized item information that is the same as or similar to the target vectorized item information as the items to be recommended to the target user.

[0088] In some optional implementations of this embodiment, the execution entity can perform step 203 as follows:

[0089] First, determine the similarity between the vectorized recommendation request and the vectorized item information.

[0090] As an example, the aforementioned execution entity can perform a dot product operation between the vectorized recommendation request and the vectorized item information, and the resulting dot product represents the similarity between the vectorized recommendation request and the vectorized item information.

[0091] Second, based on similarity, identify the items to be recommended to the target user from the candidate items.

[0092] As an example, the aforementioned execution entity can sort multiple candidate items in descending order of similarity scores, and then determine the top-ranked preset number of candidate items as the recommended items for the target user. The preset number can be set according to actual circumstances; for example, a preset number of 3.

[0093] In this implementation, the target user's recommended items are determined from the candidate items based on the similarity between the determined vectorized recommendation request and the vectorized item information, thereby improving the matching degree between the determined target user and the recommended items.

[0094] See also Figure 3 , Figure 3 This is a schematic diagram 300 illustrating an application scenario of the item recommendation method according to this embodiment. Figure 3In the application scenario, target user 3011 in consumer WeChat group 301 (private domain group 1) sends an item recommendation request 304 to server 303 via terminal device 302. After obtaining the item recommendation request 304, server 303 vectorizes the item recommendation request 304 of target user 3011 based on the group profile 305 of consumer WeChat group 301 to which target user 3011 belongs, to obtain a vectorized recommendation request 306; vectorizes the item information of candidate items 307 recalled according to the item recommendation request, to obtain vectorized item information 308; and determines the item to be recommended 309 of target user from candidate items 307 based on the vectorized recommendation request 306 and the vectorized item information 308.

[0095] The method provided in the above embodiments of this application obtains a QR code image by determining the QR code region in the image to be processed; determines a super-resolution strategy for the QR code image based on its resolution; processes the QR code image according to the super-resolution strategy to obtain a super-resolution QR code image; and decodes the QR code in the super-resolution QR code image to generate a decoding result. This provides a method for recommending items by dynamically determining a super-resolution strategy for processing a QR code image based on its resolution, improving the quality of the processed super-resolution QR code image and helping to improve the accuracy of the QR code decoding result.

[0096] Continue to refer to Figure 4 The illustration shows a schematic flow 400 of another embodiment of the item recommendation method according to this application, including the following steps:

[0097] Step 401: Using a pre-trained wake-up model, determine whether the target user has a need for item recommendations based on the target user's item recommendation request.

[0098] Step 402: In response to determining that the target user has a need for item recommendations, the item recommendation request is vectorized using a pre-trained item recommendation model based on the group profile of the target user's private domain group, resulting in a vectorized recommendation request.

[0099] Among them, the item recommendation model is used to represent the correspondence between item recommendation requests and items to be recommended.

[0100] Step 403: The item recommendation request is segmented using a pre-trained natural language understanding model to obtain the segmentation results.

[0101] Step 404: Recall multiple candidate items based on the word segmentation results using a pre-trained item recall model.

[0102] Among them, the item recall model is used to characterize the correspondence between item recommendation requests and candidate items.

[0103] Step 405: Using a pre-trained item recommendation model, the item information of multiple candidate items is vectorized to obtain multiple vectorized item information.

[0104] Step 406: Determine the similarity between the vectorized recommendation request and the vectorized item information.

[0105] Step 407: Based on similarity, identify the items to be recommended to the target user from the candidate items.

[0106] As can be seen from this embodiment, with Figure 2 Compared to the corresponding embodiments, the process 400 of the item recommendation method in this embodiment specifically illustrates the vectorization process of item recommendation requests based on the item recommendation model and incorporating the group profile of private domain groups, and the vectorization process of item recommendation requests based on the item recommendation model. It integrates the private domain group profile to achieve refined operation of "a thousand groups, a thousand faces"; while meeting the current recommendation needs of private domain users, it increases the richness and accuracy of item recommendations.

[0107] Continue to refer to Figure 5 As an implementation of the methods shown in the above figures, this application provides an embodiment of an item recommendation device, which is similar to... Figure 2 Corresponding to the method embodiments shown, this device can be specifically applied to various electronic devices.

[0108] like Figure 5 As shown, the item recommendation device 500 includes: a first vectorization unit 501, configured to vectorize the item recommendation request of the target user based on the group profile of the private domain group to which the target user belongs, to obtain a vectorized recommendation request; a second vectorization unit 502, configured to vectorize the item information of the candidate items recalled according to the item recommendation request, to obtain vectorized item information; and a determination unit 503, configured to determine the item to be recommended by the target user from the candidate items based on the vectorized recommendation request and the vectorized item information.

[0109] In some examples, the first vectorization unit 501 mentioned above is further configured to: vectorize the item recommendation request based on the group profile of the private domain group to which the target user belongs through a pre-trained item recommendation model to obtain a vectorized recommendation request, wherein the item recommendation model is used to represent the correspondence between the item recommendation request and the item to be recommended.

[0110] In some examples, the second vectorization unit 502 described above is further configured to: recall multiple candidate items according to the item recommendation request using a pre-trained item retrieval model, wherein the item retrieval model is used to characterize the correspondence between the item recommendation request and the candidate items; and vectorize the item information of the multiple candidate items to obtain multiple vectorized item information.

[0111] In some examples, the second vectorization unit 502 mentioned above is further configured to: perform word segmentation processing on the item recommendation request through a pre-trained natural language understanding model to obtain the word segmentation result; and recall multiple candidate items based on the word segmentation result through an item recall model.

[0112] In some examples, the second vectorization unit 502 is further configured to: vectorize the item information of multiple candidate items using a pre-trained item recommendation model to obtain multiple vectorized item information, wherein the item recommendation model is used to represent the correspondence between item recommendation requests and items to be recommended.

[0113] In some examples, the determining unit 503 is further configured to: determine the similarity between the vectorized recommendation request and the vectorized item information; and determine the item to be recommended to the target user from the candidate items based on the similarity.

[0114] In some examples, the first vectorization unit 501 described above is further configured to: determine whether the target user has a need for item recommendations based on the item recommendation request using a pre-trained wake-up model; and in response to determining that the target user has a need for item recommendations, vectorize the item recommendation request based on the group profile of the private domain group to which the target user belongs, to obtain a vectorized recommendation request.

[0115] In this embodiment, the first vectorization unit in the item recommendation device vectorizes the target user's item recommendation request based on the group profile of the private domain group to which the target user belongs, resulting in a vectorized recommendation request; the second vectorization unit vectorizes the item information of the candidate items recalled based on the item recommendation request, resulting in vectorized item information; the determination unit determines the item to be recommended for the target user from the candidate items based on the vectorized recommendation request and the vectorized item information, thereby providing an item recommendation device applied to the private domain, integrating the private domain group profile to achieve refined operation of "personalized recommendations for each group"; and increasing the richness and accuracy of item recommendations while meeting the current recommendation needs of private domain users.

[0116] The following is for reference. Figure 6 It illustrates a device suitable for implementing embodiments of this application (e.g., Figure 1 The diagram shows the structure of the computer system 600 of the devices 101, 102, 103, and 105.Figure 6 The device shown is merely an example and should not impose any limitations on the functionality and scope of use of the embodiments of this application.

[0117] like Figure 6 As shown, the computer system 600 includes a processor (e.g., CPU, Central Processing Unit) 601, which can perform various appropriate actions and processes based on programs stored in read-only memory (ROM) 602 or programs loaded from storage section 608 into random access memory (RAM) 603. The RAM 603 also stores various programs and data required for the operation of the system 600. The processor 601, ROM 602, and RAM 603 are interconnected via a bus 604. An input / output (I / O) interface 605 is also connected to the bus 604.

[0118] The following components are connected to I / O interface 605: an input section 606 including a keyboard, mouse, etc.; an output section 607 including a cathode ray tube (CRT), liquid crystal display (LCD), etc., and speakers, etc.; a storage section 608 including a hard disk, etc.; and a communication section 609 including a network interface card such as a LAN card, modem, etc. The communication section 609 performs communication processing via a network such as the Internet. A drive 610 is also connected to I / O interface 605 as needed. A removable medium 611, such as a disk, optical disk, magneto-optical disk, semiconductor memory, etc., is installed on drive 610 as needed so that computer programs read from it can be installed into storage section 608 as needed.

[0119] Specifically, according to embodiments of this application, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments of this application include a computer program product comprising a computer program carried on a computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via communication section 609, and / or installed from removable medium 611. When the computer program is executed by processor 601, it performs the functions defined in the methods of this application.

[0120] It should be noted that the computer-readable medium of this application can be a computer-readable signal medium or a computer-readable storage medium, or any combination thereof. A computer-readable storage medium can be, for example,—but not limited to—an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of a computer-readable storage medium may include, but are not limited to: an electrical connection having one or more wires, a portable computer disk, a hard disk, 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 thereof. In this application, a computer-readable storage medium can be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device. In this application, a computer-readable signal medium can include a data signal propagated in baseband or as part of a carrier wave, carrying computer-readable program code. Such propagated data signals can take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. Computer-readable signal media can also be any computer-readable medium other than computer-readable storage media, which can send, propagate, or transmit a program for use by or in connection with an instruction execution system, apparatus, or device. The program code contained on the computer-readable medium can be transmitted using any suitable medium, including but not limited to: wireless, wire, optical fiber, RF, etc., or any suitable combination thereof.

[0121] Computer program code for performing the operations of this application can be written in one or more programming languages ​​or a combination thereof. Programming languages ​​include object-oriented programming languages—such as Java, Smalltalk, and C++—as well as conventional procedural programming languages—such as the "C" language or similar programming languages. The program code can be executed entirely on the client computer, partially on the client computer, as a standalone software package, partially on the client computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving remote computers, the remote computer can be connected to the client computer via any type of network—including a local area network (LAN) or a wide area network (WAN)—or can be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0122] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of apparatus, methods, and computer program products according to various embodiments of this application. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0123] The units described in the embodiments of this application can be implemented in software or hardware. The described units can also be housed in a processor; for example, it can be described as: a processor including a first vectorization unit, a second vectorization unit, and a determination unit. The names of these units do not necessarily limit the unit itself; for example, the first vectorization unit can also be described as "a unit that vectorizes the target user's item recommendation request based on the group profile of the target user's private domain group to obtain a vectorized recommendation request."

[0124] In another aspect, this application also provides a computer-readable medium, which may be included in the device described in the above embodiments; or it may exist independently and not assembled into the device. The computer-readable medium carries one or more programs, which, when executed by the device, cause the computer device to: vectorize the target user's item recommendation request based on the group profile of the target user's private domain group, obtaining a vectorized recommendation request; vectorize the item information of candidate items recalled based on the item recommendation request, obtaining vectorized item information; and determine the target user's recommended item from the candidate items based on the vectorized recommendation request and the vectorized item information.

[0125] The above description is merely a preferred embodiment of this application and an explanation of the technical principles employed. Those skilled in the art should understand that the scope of the invention involved in this application is not limited to technical solutions formed by specific combinations of the above-described technical features, but should also cover other technical solutions formed by arbitrary combinations of the above-described technical features or their equivalents without departing from the above-described inventive concept. For example, technical solutions formed by substituting the above features with (but not limited to) technical features with similar functions disclosed in this application.

Claims

1. A method for recommending items, including: Based on the group profile of the target user's private domain, the target user's item recommendation request is vectorized to obtain a vectorized recommendation request; The item information of the candidate items recalled according to the item recommendation request is vectorized to obtain vectorized item information; Based on the vectorized recommendation request and the vectorized item information, the items to be recommended to the target user are determined from the candidate items.

2. The method according to claim 1, wherein, The step of vectorizing the target user's item recommendation request based on the group profile of the target user's private domain group to obtain a vectorized recommendation request includes: By using a pre-trained item recommendation model, the item recommendation request is vectorized based on the group profile of the private domain group to which the target user belongs, resulting in the vectorized recommendation request. The item recommendation model is used to represent the correspondence between the item recommendation request and the item to be recommended.

3. The method according to claim 1, wherein, The step of vectorizing the item information of candidate items recalled according to the item recommendation request to obtain vectorized item information includes: The item recall model is used to recall multiple candidate items based on the item recommendation request, wherein the item recall model is used to characterize the correspondence between the item recommendation request and the candidate items. The item information of the multiple candidate items is vectorized to obtain multiple vectorized item information.

4. The method according to claim 3, wherein, The pre-trained item recall model recalls multiple candidate items based on the item recommendation request, including: The item recommendation request is segmented using a pre-trained natural language understanding model to obtain the segmentation results. The item recall model is used to recall the multiple candidate items based on the word segmentation results.

5. The method according to claim 3, wherein, The step of vectorizing the item information of the multiple candidate items to obtain multiple vectorized item information includes: The item information of the multiple candidate items is vectorized using a pre-trained item recommendation model to obtain the multiple vectorized item information. The item recommendation model is used to represent the correspondence between item recommendation requests and items to be recommended.

6. The method according to claim 1, wherein, The step of determining the recommended items for the target user from the candidate items based on the vectorized recommendation request and the vectorized item information includes: Determine the similarity between the vectorized recommendation request and the vectorized item information; Based on the similarity, the items to be recommended to the target user are determined from the candidate items.

7. The method according to claim 1, wherein, The step of vectorizing the target user's item recommendation request based on the group profile of the target user's private domain group to obtain a vectorized recommendation request includes: The pre-trained wake-up model determines whether the target user has a need for item recommendations based on the item recommendation request. In response to determining that the target user has a need for item recommendations, the item recommendation request is vectorized based on the group profile of the private domain group to which the target user belongs, to obtain the vectorized recommendation request.

8. An item recommendation device, comprising: The first vectorization unit is configured to vectorize the target user's item recommendation request based on the group profile of the private domain group to which the target user belongs, so as to obtain a vectorized recommendation request; The second vectorization unit is configured to vectorize the item information of the candidate items recalled according to the item recommendation request to obtain vectorized item information. The determining unit is configured to determine the item to be recommended to the target user from the candidate items based on the vectorized recommendation request and the vectorized item information.

9. A computer-readable medium having a computer program stored thereon, wherein, When the program is executed by the processor, it implements the method as described in any one of claims 1-7.

10. An electronic device, comprising: One or more processors; Storage device, on which one or more programs are stored, When the one or more programs are executed by the one or more processors, the one or more processors implement the method as described in any one of claims 1-7.

Images