Material recommendation method and device, and electronic device

Abstract

The application discloses a material recommendation method and device and electronic equipment. The method comprises the following steps: generating different user vectors and different material vectors; in response to a set trigger event, obtaining a recall strategy matched with a target scene; obtaining a user vector and a material vector matched with the target scene from the different user vectors and the different material vectors through the recall strategy; and obtaining a target material recommended for a target user according to the recall strategy, the user vector and the material vector matched with the target scene.

Description

Technical Field

[0001] This disclosure relates to the field of Internet technology, and more specifically, to a material recommendation method, a material recommendation device, an electronic device including at least one computing device and at least one storage device, and a computer-readable storage medium. Background Technology

[0002] With the widespread adoption of smart devices, purchasing goods through electronic devices such as mobile phones, tablets, and laptops has become an important way for users to buy things in their daily lives. As the number of products online continues to grow, how to recommend the products that users are most interested in from this vast array of goods is a key concern for users.

[0003] In product recommendation scenarios, the first step is the recall step, which involves retrieving items that the user might be interested in from a massive number of products. However, current recall methods only use a single user feature or a single product feature, resulting in limited feature usage and inaccurate recommendation results. Furthermore, existing technologies process user or product information online to extract the relevant user or product features when product recommendations are needed, wasting significant online resources. Summary of the Invention

[0004] One objective of this disclosure is to provide a new technical solution for material recommendation.

[0005] According to a first aspect of this disclosure, a material recommendation method is provided, comprising:

[0006] Generate different user vectors and different material vectors;

[0007] In response to a set trigger event, retrieve a recall strategy that matches the target scenario;

[0008] The recall strategy is used to obtain user vectors and material vectors that match the target scenario from the different user vectors and the different material vectors.

[0009] Based on the recall strategy, the user vector and the material vector matching the target scenario, the target material recommended to the target user is obtained.

[0010] Optionally, generating different user vectors and different material vectors includes:

[0011] Obtain a training dataset; wherein each piece of training data in the training dataset includes user features and corresponding material features;

[0012] The training dataset is processed to obtain user feature sets and material feature sets;

[0013] The user feature set is processed by a first preset model to generate different user vectors; and the material feature set is processed by a second preset model to generate different material vectors.

[0014] Optionally, obtaining the training dataset includes:

[0015] Provides a primary configuration interface;

[0016] Obtain the dataset input through the first configuration interface as the training dataset.

[0017] Optionally, after generating different user vectors and different material vectors, the method further includes:

[0018] The different user vectors and the different material vectors are stored in the vector management system;

[0019] The step of obtaining user vectors and material vectors matching the target scenario from the different user vectors and the different material vectors through the recall strategy includes:

[0020] The recall strategy is used to obtain user vectors and material vectors that match the target scenario from the vector management system.

[0021] Optionally, the recall strategy includes at least one of u2i, i2i, u2i2i, u2u2i, and u2tag2i.

[0022] Optionally, obtaining a recall strategy matching the target scenario includes:

[0023] Provide a second configuration interface for configuring the recall strategy;

[0024] The recall strategy input through the second configuration interface is obtained as the recall strategy that matches the target scenario.

[0025] Optionally, the target scenario includes any one of the following: information recommendation scenario, product recommendation scenario, and short video recommendation scenario.

[0026] Optionally, obtaining the target material recommended for the target user based on the recall strategy, the user vector and the material vector matching the target scenario includes:

[0027] Based on the recall strategy, candidate materials are obtained by matching the user vector and material vector with the target scenario;

[0028] Select recommended materials that match the target user from the candidate materials;

[0029] The recommended materials are then recommended to the target user.

[0030] Optionally, selecting recommended materials that match the target user from the candidate materials includes:

[0031] After performing coarse and fine sorting on the candidate materials, recommended materials matching the target user are obtained.

[0032] Optionally, selecting recommended materials that match the target user from the candidate materials includes:

[0033] After performing coarse sorting, fine sorting, and rearrangement on the candidate materials, recommended materials matching the target user are obtained.

[0034] Optionally, the set triggering event includes at least one of the following: when the set time is reached, or when a request from the target user is received.

[0035] According to a second aspect of this disclosure, a material recommendation device is also provided, comprising:

[0036] The generation module is used to generate different user vectors and different material vectors;

[0037] The first acquisition module is used to acquire a recall strategy that matches the target scenario in response to a set trigger event;

[0038] The second acquisition module is used to acquire user vectors and material vectors that match the target scenario from the different user vectors and the different material vectors through the recall strategy.

[0039] The recommendation module is used to obtain target materials recommended to target users based on the recall strategy, user vectors and material vectors that match the target scenario.

[0040] Optionally, the generation module is specifically used for:

[0041] Obtain a training dataset; wherein each piece of training data in the training dataset includes user features and corresponding material features;

[0042] The training dataset is processed to obtain user feature sets and material feature sets;

[0043] The user feature set is processed using a first preset model to generate different user vectors; and...

[0044] The material feature set is processed by a second preset model to generate different material vectors.

[0045] Optionally, the generation module is specifically used for:

[0046] Provides a primary configuration interface;

[0047] Obtain the dataset input through the first configuration interface as the training dataset.

[0048] Optionally, the device further includes a storage module.

[0049] The storage module is used to: store the different user vectors and the different material vectors into the vector management system;

[0050] The second acquisition module is specifically used to: acquire user vectors and material vectors that match the target scenario from the vector management system through the recall strategy.

[0051] Optionally, the recall strategy includes at least one of u2i, i2i, u2i2i, u2u2i, and u2tag2i.

[0052] Optionally, the second acquisition module is specifically used for:

[0053] Provide a second configuration interface for configuring the recall strategy;

[0054] The recall strategy input through the second configuration interface is obtained as the recall strategy that matches the target scenario.

[0055] Optionally, the target scenario includes any one of the following: information recommendation scenario, product recommendation scenario, and short video recommendation scenario.

[0056] Optionally, the recommendation module is specifically used for:

[0057] Based on the recall strategy, candidate materials are obtained by matching the user vector and material vector with the target scenario;

[0058] Select recommended materials that match the target user from the candidate materials;

[0059] The recommended materials are then recommended to the target user.

[0060] Optionally, the recommendation module is specifically used for:

[0061] After performing coarse and fine sorting on the candidate materials, recommended materials matching the target user are obtained.

[0062] Optionally, the recommendation module is specifically used for:

[0063] After performing coarse sorting, fine sorting, and rearrangement on the candidate materials, recommended materials matching the target user are obtained.

[0064] Optionally, the set triggering event includes at least one of the following: when the set time is reached, or when a request from the target user is received.

[0065] According to a third aspect of this disclosure, an electronic device includes at least one computing device and at least one storage device, wherein the at least one storage device is configured to store instructions for controlling the at least one computing device to perform the method described in accordance with the first aspect above; or, the device implements the device described in the second aspect above through the computing device and the storage device.

[0066] According to a fourth aspect of this disclosure, a computer-readable storage medium is also provided, wherein a computer program is stored thereon, the computer program implementing the method described in the first aspect above when executed by a processor.

[0067] One beneficial effect of this disclosure is that the methods, apparatus, and electronic devices according to embodiments of this disclosure generate different user vectors and different material vectors offline, pre-processing user features and material features into different user vectors and different material vectors to avoid resource waste caused by performing similar operations online. Simultaneously, when material recommendations are needed online, a recall strategy is first obtained. Then, the recall strategy is used to pull back user vectors and material vectors matching the target scenario from different user vectors and different material vectors. Based on the recall strategy, user vectors, and material vectors, target materials recommended to the target user are obtained. That is, for material recommendation recall, it combines user vectors and material vectors to quickly find materials matching the user's interests. This is equivalent to creating a recall model utilizing multi-feature fusion, thereby improving the accuracy of material recommendations. Attached Figure Description

[0068] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments of the invention and, together with their description, serve to explain the principles of the invention.

[0069] Figure 1 This is a schematic diagram of the hardware structure of an electronic device according to an embodiment of the present disclosure;

[0070] Figure 2 This is a schematic flowchart of a material recommendation method according to an embodiment of the present disclosure;

[0071] Figure 3 This is a schematic block diagram of a material recommendation device according to an embodiment of the present disclosure;

[0072] Figure 4 This is a schematic block diagram of an electronic device according to an embodiment of the present disclosure;

[0073] Figure 5 This is a schematic diagram of the hardware structure of an electronic device according to another embodiment of the present disclosure. Detailed Implementation

[0074] Various exemplary embodiments of the present invention will now be described in detail with reference to the accompanying drawings. It should be noted that, unless otherwise specifically stated, the relative arrangement, numerical expressions, and values ​​of the components and steps set forth in these embodiments do not limit the scope of the invention.

[0075] The following description of at least one exemplary embodiment is merely illustrative and is in no way intended to limit the invention or its application or use.

[0076] Techniques, methods, and equipment known to those skilled in the art may not be discussed in detail, but where appropriate, such techniques, methods, and equipment should be considered part of the specification.

[0077] In all the examples shown and discussed herein, any specific values ​​should be interpreted as merely exemplary and not as limitations. Therefore, other examples of exemplary embodiments may have different values.

[0078] It should be noted that similar labels and letters in the following figures indicate similar items; therefore, once an item is defined in one figure, it does not need to be discussed further in subsequent figures.

[0079] <Hardware Configuration>

[0080] The method of this disclosure embodiment can be implemented by at least one electronic device, that is, the apparatus 3000 for implementing the method can be arranged on the at least one electronic device. Figure 1 The hardware structure of any electronic device is shown. Figure 1 The electronic device shown can be a portable computer, desktop computer, workstation, server, etc., or any other device with a processor or other computing device and a memory or other storage device, which is not limited here.

[0081] like Figure 1As shown, the electronic device 1000 may include a processor 1100, a memory 1200, an interface device 1300, a communication device 1400, a display device 1500, an input device 1600, a speaker 1700, a microphone 1800, etc. The processor 1100 is used to execute computer programs. These computer programs can be written using instruction sets based on architectures such as x86, Arm, RISC, MIPS, and SSE. The memory 1200 may include, for example, ROM (Read-Only Memory), RAM (Random Access Memory), or non-volatile memory such as a hard disk. The interface device 1300 may include, for example, a USB interface or a headphone jack. The communication device 1400 may be capable of wired or wireless communication, specifically including Wi-Fi communication, Bluetooth communication, 2G / 3G / 4G / 5G communication, etc. The display device 1500 may be, for example, an LCD screen or a touch screen. The input device 1600 may include, for example, a touch screen, a keyboard, or motion input. Electronic device 1000 can output voice information through speaker 1700 and collect voice information through microphone 1800, etc.

[0082] Figure 1 The electronic devices shown are merely illustrative and are in no way intended to limit the invention, its application, or use. In embodiments applied to this disclosure, the memory 1200 of the electronic device 1000 is used to store instructions for controlling the processor 1100 to operate in order to perform the material recommendation method of embodiments of this disclosure. Those skilled in the art can design instructions based on the schemes disclosed herein. How the instructions control the processor to operate is well known in the art and will not be described in detail here.

[0083] In one embodiment, an electronic device is provided that includes at least one computing device and at least one storage device, the at least one storage device being used to store instructions for controlling the at least one computing device to perform a method according to any embodiment of the present disclosure.

[0084] The device may include at least one Figure 1 The electronic device 1000 shown is intended to provide at least one computing device, such as a processor, and at least one storage device, such as a memory, without limitation herein.

[0085] <Method Implementation>

[0086] Figure 2 This is a flowchart illustrating a material recommendation method according to an embodiment of the present disclosure, which is executed by an electronic device 1000, such as... Figure 2 As shown, the method may include the following steps S2100 to S2400:

[0087] Step S2100: Generate different user vectors and different material vectors.

[0088] User vectors, also known as user embeddings, integrate different user features. Material vectors, also known as material embeddings, integrate different material features, including but not limited to information, products, and short videos.

[0089] User characteristics are features related to users, including but not limited to user attributes such as user age, user gender, and user occupation. Material characteristics are features related to materials, including but not limited to material attributes such as material number, material name, material title, and material content.

[0090] Different user vectors, or different user embeddings, include user vectors corresponding to different scenarios. Different material vectors, or different material embeddings, include user vectors corresponding to different scenarios. For example, different user vectors include, but are not limited to, user vectors corresponding to news recommendation scenarios, product recommendation scenarios, and short video recommendation scenarios. Similarly, different material vectors include, but are not limited to, material vectors corresponding to news recommendation scenarios, product recommendation scenarios, and short video recommendation scenarios. Specifically, the material vector for news recommendation scenarios is a vector composed of different news features, the material vector for product recommendation scenarios is a vector composed of different product features, and the material vector for short video recommendation scenarios is a vector composed of different short video features.

[0091] In this embodiment, generating different user vectors and different material vectors in step S2100 may further include the following steps S2110 to S2140:

[0092] Step S2110: Obtain the training dataset.

[0093] Each training data point in the training dataset includes user features and corresponding material features. The more training data included in the training dataset, the more accurate the generated user and material vectors will be. However, after a certain amount of training data is collected, the increase in the accuracy of the training results will become increasingly slow until it stabilizes. Therefore, the required amount of training data can be determined by balancing the accuracy of the training results with the data processing cost.

[0094] In this step S2110, the training data in the training dataset involves data from different scenarios. For example, the training data in the training dataset includes user data and product data for the product recommendation scenario; for another example, the training data in the training dataset includes user data and information data for the information recommendation scenario; and for yet another example, the training data in the training dataset includes user data and short video data for the short video recommendation scenario.

[0095] In this embodiment, the step S2110 of obtaining the training dataset may further include: providing a first configuration interface; obtaining the dataset input through the first configuration interface as the training dataset.

[0096] The first configuration interface involves configuration items related to the dataset. These items offer different import methods for training data, such as importing locally stored training data into electronic device 1000, importing data from a database into electronic device 1000, and importing data periodically via HDFS. Users can use the specific import methods provided by this configuration item to import training data and form a training dataset.

[0097] Step S2120: Process the training dataset to obtain user feature set and material feature set.

[0098] In this step S2120, all training data in the obtained training dataset can be preprocessed to obtain a user feature set and a material feature set. The user feature set is a set composed of all user features, and the material feature set is a set composed of all material features.

[0099] Step S2130: Process the user feature set using the first preset model to generate different user vectors.

[0100] The first preset model can be a collaborative filtering (CF) model or a factor machine (FM) model; this embodiment does not limit it.

[0101] In this step S2130, different user embeddings, i.e. user vectors, can be generated through the first preset model. For example, user vectors corresponding to information recommendation scenarios, product recommendation scenarios, and short video recommendation scenarios can be generated.

[0102] Step S2140: The material feature set is processed by the second preset model to generate different material vectors.

[0103] The second preset model can be a collaborative filtering model or a factorization machine model; this embodiment does not limit it.

[0104] In this step S2140, different material embeddings, i.e. material vectors, can be generated through the second preset model. For example, material vectors corresponding to information recommendation scenarios, product recommendation scenarios, and short video recommendation scenarios can be generated.

[0105] Based on steps S2110 to S2140 above, the training dataset is processed offline, and different user vectors and different material vectors are generated through their respective models for online use, thus avoiding the waste of resources in online services.

[0106] After generating different user vectors and different material vectors, proceed to:

[0107] Step S2200: In response to the set trigger event, obtain the recall strategy that matches the target scenario.

[0108] The set trigger events include at least one of the following: reaching a set time or receiving a request from the target user. The set time can be a value set according to the actual application scenario and actual needs; the set time can be 8:00 AM every day.

[0109] The target users are those who need material recommendations. Requests from the target users include, but are not limited to, refresh requests and search requests.

[0110] The target scenario is the scenario where the material is located. The target scenario includes any one of the above-mentioned information recommendation scenario, product recommendation scenario and short video recommendation scenario. Of course, the target scenario may also include other scenarios, but this embodiment does not limit it.

[0111] The recall strategy includes at least one of u2i, i2i, u2i2i, u2u2i, and u2tag2i. That is, it can be a single recall strategy used as a recall strategy that matches the target scenario, or it can be a combination of multiple recall strategies used as a recall strategy that matches the target scenario.

[0112] The i2i above refers to the process of moving from one item to another, i.e., item to item. It requires calculating the similarity between the two items, as there is a high probability that the two items will be viewed at the same time. When one item is viewed, the other item is recommended to the user.

[0113] The above u2i refers to the process of a user moving to an item, i.e., user to item. It usually refers to the user's direct behavior, such as when a user views an item, the user will be recommended the same item again.

[0114] The u2i2i model above refers to a user moving from one item to another. It combines u2i and i2i, first obtaining u2i data, then expanding upon it using i2i data. This allows users to move from one node to another, reaching a third node, thus enabling recommendations. For example, if a user views headphones (u2i), similar or related products (i2i) are found and recommended to the user.

[0115] The above u2u2i refers to the process of going from one user to another, and then to a specific product. Typically, user attributes such as gender, age, and occupation are used to calculate similarity and obtain the u2u. Alternatively, data can be mined from behavioral data; for example, if the content and videos viewed are largely similar, users can be considered to belong to the same category. For instance, if User 1 and User 2 are similar, and User 2 views a certain product, that product will be recommended to User 1.

[0116] The above u2tag2i refers to the intermediate node being a tag, not u or i. It first calculates the user's tag preferences and then matches them with the item list. This strategy has good generalization performance and the recommended content is not so narrow. For example, if the user likes science fiction, then all science fiction content will be recommended.

[0117] In this embodiment, the electronic device 1000 is pre-set with a set trigger event. When the trigger event is triggered, the electronic device 1000 can obtain a recall strategy that matches the target scenario. Then, according to the recall strategy, it can obtain user vectors and material vectors that match the target scenario from different user vectors and different material vectors. Furthermore, according to the recall strategy, user vectors, and material vectors, it can obtain target materials recommended for the target user, thereby providing the user with efficient and accurate material recommendations.

[0118] In this embodiment, step S2200, obtaining the recall strategy matching the target scenario, may further include the following steps S2210 to S2220:

[0119] Step S2210: Provide a second configuration interface for configuring the recall strategy.

[0120] The second configuration interface contains configuration items for recall strategies. Through these configuration items, you can configure a recall strategy that matches the target scenario.

[0121] Step S2220: Obtain the recall strategy input through the second configuration interface as the recall strategy that matches the target scenario.

[0122] In this step S2220, the recall strategy configured through the configuration items of the second configuration interface can be at least one of u2i, i2i, u2i2i, u2u2i, and u2tag2i.

[0123] Understandably, the target scenarios involve different specific contexts. For example, in the product recommendation scenario, besides the homepage hot news recommendation scenario, other recommendation scenarios may also be involved. For the homepage hot recommendation scenario within the product recommendation scenario, different combinations of recall strategies can be used to recommend popular products and products that users like. However, the number of user-like products recommended in this scenario is relatively small.

[0124] Based on the above steps S2210 to S2220, the configuration interface allows users to configure recall strategies that match the target scenario, thus meeting user needs.

[0125] After executing the response to the set trigger event and obtaining the recall strategy matching the target scenario, proceed to:

[0126] Step S2300: Using the recall strategy, obtain user vectors and material vectors that match the target scenario from the different user vectors and the different material vectors.

[0127] In this embodiment, after generating different user vectors and different material vectors according to the above step S2100, the different user vectors and different material vectors are stored in the vector management system in the electronic device 1000 so that the electronic device 1000 can obtain user vectors and material vectors that match the target scenario from the different user vectors and different material vectors through a recall strategy.

[0128] In this embodiment, step S2300, which involves obtaining user vectors and material vectors matching the target scenario from the different user vectors and different material vectors through the recall strategy, may further include: obtaining user vectors and material vectors matching the target scenario from the vector management system through the recall strategy.

[0129] After executing the recall strategy to obtain user vectors and material vectors matching the target scenario from the different user vectors and the different material vectors, proceed to:

[0130] Step S2400: Based on the recall strategy, the user vector and the material vector matching the target scenario, obtain the target material recommended for the target user.

[0131] In this embodiment, after obtaining the recall strategy, user vector, and material vector that match the target scenario, the target material recommended for the target user can be obtained based on the recall strategy, user vector, and material vector that match the target scenario.

[0132] In this embodiment, step S2400, based on the recall strategy, the user vector and the material vector matching the target scenario, to obtain the target material recommended for the target user, may further include the following steps S2410 to S2430:

[0133] Step S2410: Based on the recall strategy, obtain candidate materials by matching the user vector and material vector with the target scenario.

[0134] In this step S2410, it is equivalent to recalling the candidate materials through the recall step, and the number of these candidate materials is usually very large.

[0135] Step S2420: Select recommended materials that match the target user from the candidate materials.

[0136] In one example, step S2420, which selects recommended materials that match the target user from the candidate materials, may further include: after coarse and fine sorting the candidate materials, obtaining recommended materials that match the target user.

[0137] Understandably, after the recall, due to the large number of recalled materials, the materials will be sorted in a coarse and fine manner, gradually reducing the scale of materials and increasing the personalization level. Finally, the final recommendation results will be returned in descending order of the probability of user clicks and other behaviors.

[0138] Coarse ranking sorts and truncates the recall results based on a small amount of feature data; it uses the fewest features. Fine ranking, on the other hand, combines feature data provided by a feature service with a complex AI model service to achieve precise ranking; it uses the most features.

[0139] In one example, step S2420, which selects recommended materials that match the target user from the candidate materials, may further include: performing coarse ranking, fine ranking, and rearrangement on the candidate materials to obtain recommended materials that match the target user.

[0140] The reordering calculation incorporates various business rules and adjusts the sorted list obtained from the fine-order calculation (e.g., pinning, mandatory recommendation, etc.) to obtain the final recommendation result.

[0141] The coarse-sorting and rearrangement calculations have been explained in the examples above. It is understandable that since the materials are sorted by fine-sorting calculation, they may not fully meet business needs and user experience. Therefore, it is necessary to consider indicators such as the diversity of materials to re-sort the materials that have already been sorted.

[0142] Step S2430: Recommend the recommended materials to the target user.

[0143] Based on the above steps S2410 to S2430, after sequentially going through recall, coarse sorting, fine sorting and / or rearrangement, the material scale is gradually reduced and the personalization precision is improved to provide users with personalized recommendation results.

[0144] According to the method of this disclosure, different user vectors and different material vectors are generated offline. User features and material features are pre-processed into different user vectors and different material vectors to avoid resource waste caused by performing similar operations online. Simultaneously, when material recommendation is needed online, a recall strategy is first obtained. Then, the recall strategy is used to retrieve user vectors and material vectors matching the target scenario from different user vectors and different material vectors. Based on the recall strategy, user vectors, and material vectors, target materials recommended to the target user are obtained. In other words, for material recommendation recall, it combines user vectors and material vectors to quickly find materials matching the user's interests. This is equivalent to creating a recall model utilizing multi-feature fusion, thereby improving the accuracy of material recommendation.

[0145] The following is an example of a material recommendation method, which may include the following steps S3100 to S3900:

[0146] Step S3100: Obtain the training dataset.

[0147] Step S3200: Process the training dataset to obtain the user feature set and the material feature set.

[0148] Step S3300: Process the user feature set using the first preset model to generate different user vectors.

[0149] Step S3400: Process the material feature set using the second preset model to generate different material vectors.

[0150] Step S3500: Store different user vectors and different material vectors into the vector management system.

[0151] Step S3600: In response to the set trigger event, obtain the recall strategy that matches the target scenario.

[0152] Step S3700: Using a recall strategy, retrieve user vectors and material vectors that match the target scenario from the vector management system.

[0153] Step S3800: Based on the recall strategy, obtain candidate materials by matching the user vector and material vector with the target scenario.

[0154] Step S3900: After coarse sorting, fine sorting and rearrangement of candidate materials, recommended materials that match the target user are obtained and recommended materials are recommended to the target user.

[0155] <Device Embodiment>

[0156] In this embodiment, a material recommendation device 3000 is also provided, such as... Figure 3 As shown, the material recommendation device 3000 includes a generation module 3100, a first acquisition module 3200, a second acquisition module 3300, and a recommendation module 3400, which are used to implement the material recommendation method provided in this embodiment. Each module of the material recommendation device 3000 can be implemented by software or by hardware, and no limitation is made here.

[0157] The generation module 3100 is used to generate different user vectors and different material vectors.

[0158] The first acquisition module 3200 is used to acquire a recall strategy that matches the target scenario in response to a set trigger event.

[0159] The second acquisition module 3300 is used to acquire user vectors and material vectors that match the target scenario from the different user vectors and the different material vectors through the recall strategy.

[0160] The recommendation module 3400 is used to obtain target materials recommended to target users based on the recall strategy, user vectors and material vectors that match the target scenario.

[0161] In one embodiment, the generation module 3100 is specifically configured to: acquire a training dataset; wherein each training data in the training dataset includes user features and corresponding material features; process the training dataset to obtain a user feature set and a material feature set; process the user feature set using a first preset model to generate different user vectors; and process the material feature set using a second preset model to generate different material vectors.

[0162] In one embodiment, the generation module 3100 is specifically used to: provide a first configuration interface; and obtain the dataset input through the first configuration interface as the training dataset.

[0163] In one embodiment, the device 3000 further includes a storage module (not shown in the figure).

[0164] The storage module is used to: store the different user vectors and the different material vectors into the vector management system;

[0165] The second acquisition module 3200 is specifically used to: acquire user vectors and material vectors that match the target scenario from the vector management system through the recall strategy.

[0166] In one embodiment, the recall strategy includes at least one of u2i, i2i, u2i2i, u2u2i, and u2tag2i.

[0167] In one embodiment, the second acquisition module 3200 is specifically used to: provide a second configuration interface for configuring the recall strategy; and acquire the recall strategy input through the second configuration interface as the recall strategy matching the target scenario.

[0168] In one embodiment, the target scenario includes any one of an information recommendation scenario, a product recommendation scenario, and a short video recommendation scenario.

[0169] In one embodiment, the recommendation module 3400 is specifically configured to: obtain candidate materials based on the recall strategy, user vectors and material vectors matching the target scenario; select recommended materials matching the target user from the candidate materials; and recommend the recommended materials to the target user.

[0170] In one embodiment, the recommendation module 3400 is specifically used to: after performing coarse and fine ranking on the candidate materials, obtain recommended materials that match the target user.

[0171] In one embodiment, the recommendation module 3400 is specifically used to: after performing coarse ranking, fine ranking and rearrangement on the candidate materials, obtain recommended materials that match the target user.

[0172] In one embodiment, the set triggering event includes at least one of the following: when a set time is reached, or when a request is received from the target user.

[0173] <Equipment Example>

[0174] Corresponding to the above method embodiments, this embodiment also provides an electronic device, such as... Figure 4 As shown, it may include a material recommendation device 3000 according to any embodiment of the present disclosure, for implementing the material recommendation method of any embodiment of the present disclosure.

[0175] like Figure 5 As shown, the electronic device 4000 may further include a processor 4200 and a memory 4100, the memory 4100 being used to store executable instructions; the processor 4200 being used to control the electronic device according to the instructions to perform a material recommendation method according to any embodiment of the present disclosure.

[0176] Each module of the above-described device 4000 can be implemented by the processor 4200 executing the instructions to perform the method according to any embodiment of the present disclosure.

[0177] This invention can be a system, method, and / or computer program product. A computer program product may include a computer-readable storage medium having computer-readable program instructions loaded thereon for causing a processor to implement various aspects of the invention.

[0178] Computer-readable storage media can be tangible devices capable of holding and storing instructions for use by an instruction execution device. Computer-readable storage media can be, for example—but not limited to—electrical storage devices, magnetic storage devices, optical storage devices, electromagnetic storage devices, semiconductor storage devices, or any suitable combination thereof. More specific examples (a non-exhaustive list) of computer-readable storage media include: portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), static random access memory (SRAM), portable compact disc read-only memory (CD-ROM), digital multifunction disc (DVD), memory sticks, floppy disks, mechanical encoding devices, such as punch cards or recessed protrusions storing instructions thereon, and any suitable combination thereof. The computer-readable storage media used herein are not to be construed as transient signals themselves, such as radio waves or other freely propagating electromagnetic waves, electromagnetic waves propagating through waveguides or other transmission media (e.g., light pulses through fiber optic cables), or electrical signals transmitted through wires.

[0179] The computer-readable program instructions described herein can be downloaded from computer-readable storage media to various computing / processing devices, or downloaded via a network, such as the Internet, local area network, wide area network, and / or wireless network, to an external computer or external storage device. The network may include copper transmission cables, fiber optic transmission, wireless transmission, routers, firewalls, switches, gateway computers, and / or edge servers. A network adapter card or network interface in each computing / processing device receives the computer-readable program instructions from the network and forwards them to the computer-readable storage media in the respective computing / processing device.

[0180] The computer program instructions used to perform the operations of this invention may be assembly instructions, instruction set architecture (ISA) instructions, machine instructions, machine-dependent instructions, microcode, firmware instructions, state setting data, or source code or object code written in any combination of one or more programming languages, including object-oriented programming languages ​​such as Smalltalk, C++, etc., and conventional procedural programming languages ​​such as the "C" language or similar programming languages. The computer-readable program instructions may be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving a remote computer, the remote computer may be connected to the user's computer via any type of network—including a local area network (LAN) or a wide area network (WAN)—or may be connected to an external computer (e.g., via the Internet using an Internet service provider). In some embodiments, electronic circuitry, such as programmable logic circuitry, field-programmable gate arrays (FPGAs), or programmable logic arrays (PLAs), is personalized by utilizing state information from the computer-readable program instructions. This electronic circuitry can execute the computer-readable program instructions to implement various aspects of the invention.

[0181] Various aspects of the present invention are described herein with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It should be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer-readable program instructions.

[0182] These computer-readable program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing apparatus to produce a machine such that, when executed by the processor of the computer or other programmable data processing apparatus, they create means for implementing the functions / actions specified in one or more blocks of the flowchart and / or block diagram. These computer-readable program instructions can also be stored in a computer-readable storage medium that causes a computer, programmable data processing apparatus, and / or other device to operate in a particular manner; thus, the computer-readable medium storing the instructions comprises an article of manufacture that includes instructions for implementing aspects of the functions / actions specified in one or more blocks of the flowchart and / or block diagram.

[0183] Computer-readable program instructions may also be loaded onto a computer, other programmable data processing apparatus, or other device to cause a series of operational steps to be performed on the computer, other programmable data processing apparatus, or other device to produce a computer-implemented process, thereby causing the instructions executed on the computer, other programmable data processing apparatus, or other device to perform the functions / actions specified in one or more boxes of a flowchart and / or block diagram.

[0184] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of the present invention. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of an instruction containing one or more executable instructions for implementing a specified logical function. In some alternative implementations, the functions marked in the blocks may occur in a different order than those marked in the drawings. For example, two consecutive 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 action, or using a combination of dedicated hardware and computer instructions. It will be known to those skilled in the art that implementation in hardware, implementation in software, and implementation using a combination of software and hardware are equivalent.

[0185] The various embodiments of the present invention have been described above. These descriptions are exemplary and not exhaustive, and are not limited to the disclosed embodiments. Many modifications and variations will be apparent to those skilled in the art without departing from the scope and spirit of the described embodiments. The terminology used herein is chosen to best explain the principles, practical application, or technical improvements to the embodiments in the market, or to enable others skilled in the art to understand the embodiments disclosed herein. The scope of the invention is defined by the appended claims.

Claims

1. A material recommendation method, comprising: Different user vectors and different material vectors are generated offline and stored in the vector management system; In response to a set trigger event, retrieve a recall strategy that matches the target scenario; The recall strategy is used to obtain user vectors and material vectors that match the target scenario from the vector management system. Based on the recall strategy, the user vector and the material vector matching the target scenario, the target material recommended to the target user is obtained.

2. The method according to claim 1, wherein, The generation of different user vectors and different material vectors includes: Obtain a training dataset; wherein each piece of training data in the training dataset includes user features and corresponding material features; The training dataset is processed to obtain user feature sets and material feature sets; The user feature set is processed by a first preset model to generate different user vectors; and the material feature set is processed by a second preset model to generate different material vectors.

3. The method according to claim 2, wherein, The acquisition of the training dataset includes: Provides a primary configuration interface; Obtain the dataset input through the first configuration interface as the training dataset.

4. The method according to claim 1, wherein, The recall strategy includes at least one of u2i, i2i, u2i2i, u2u2i, and u2tag2i.

5. The method according to claim 4, wherein, The method for obtaining a recall strategy that matches the target scenario includes: Provide a second configuration interface for configuring the recall strategy; The recall strategy input through the second configuration interface is obtained as the recall strategy that matches the target scenario.

6. The method according to claim 1, wherein, The target scenarios include any one of the following: information recommendation scenario, product recommendation scenario, and short video recommendation scenario.

7. The method according to claim 1, wherein, The step of obtaining target materials recommended for the target user based on the recall strategy, the user vector and the material vector matching the target scenario includes: Based on the recall strategy, candidate materials are obtained by matching the user vector and material vector with the target scenario; Select recommended materials that match the target user from the candidate materials; The recommended materials are then recommended to the target user.

8. The method according to claim 7, wherein, The step of selecting recommended materials that match the target user from the candidate materials includes: After performing coarse and fine sorting on the candidate materials, recommended materials matching the target user are obtained.

9. The method according to claim 7, wherein, The step of selecting recommended materials that match the target user from the candidate materials includes: After performing coarse sorting, fine sorting, and rearrangement on the candidate materials, recommended materials matching the target user are obtained.

10. The method according to claim 1, characterized in that, The set triggering event includes at least one of the following: when the set time is reached, or when a request is received from the target user.

11. A material recommendation device, comprising: The generation module is used to generate different user vectors and different material vectors offline and store them in the vector management system; The first acquisition module is used to acquire a recall strategy that matches the target scenario in response to a set trigger event; The second acquisition module is used to acquire user vectors and material vectors that match the target scenario from the vector management system through the recall strategy; The recommendation module is used to obtain target materials recommended to target users based on the recall strategy, user vectors and material vectors that match the target scenario.

12. The apparatus according to claim 11, wherein, The generation module is specifically used for: Obtain a training dataset; wherein each piece of training data in the training dataset includes user features and corresponding material features; The training dataset is processed to obtain user feature sets and material feature sets; The user feature set is processed using a first preset model to generate different user vectors; and... The material feature set is processed by a second preset model to generate different material vectors.

13. The apparatus according to claim 12, wherein, The generation module is specifically used for: Provides a primary configuration interface; Obtain the dataset input through the first configuration interface as the training dataset.

14. The apparatus according to claim 11, wherein, The recall strategy includes at least one of u2i, i2i, u2i2i, u2u2i, and u2tag2i.

15. The apparatus according to claim 14, wherein, The second acquisition module is specifically used for: Provide a second configuration interface for configuring the recall strategy; The recall strategy input through the second configuration interface is obtained as the recall strategy that matches the target scenario.

16. The apparatus according to claim 11, wherein, The target scenarios include any one of the following: information recommendation scenario, product recommendation scenario, and short video recommendation scenario.

17. The apparatus according to claim 11, wherein, The recommendation module is specifically used for: Based on the recall strategy, candidate materials are obtained by matching the user vector and material vector with the target scenario; Select recommended materials that match the target user from the candidate materials; The recommended materials are then recommended to the target user.

18. The apparatus according to claim 17, wherein, The recommendation module is specifically used for: After performing coarse and fine sorting on the candidate materials, recommended materials matching the target user are obtained.

19. The apparatus according to claim 18, wherein, The recommendation module is specifically used for: After performing coarse sorting, fine sorting, and rearrangement on the candidate materials, recommended materials matching the target user are obtained.

20. The apparatus according to claim 11, wherein, The set triggering event includes at least one of the following: when the set time is reached, or when a request is received from the target user.

21. An electronic device comprising at least one computing device and at least one storage device, wherein, The at least one storage device is used to store instructions for controlling the at least one computing device to perform the method according to any one of claims 1 to 10; or, the device implements the apparatus according to any one of claims 11 to 20 through the computing device and the storage device.

22. A computer-readable storage medium, wherein, It stores a computer program thereon, which, when executed by a processor, implements the method as described in any one of claims 1 to 10.

Images