Recommendation method, electronic equipment and storage medium

Abstract

The invention relates to artificial intelligence, and provides a recommendation method, electronic equipment and a storage medium. The recommendation method comprises the following steps: extracting key information in search information input by a user; determining a first commodity matched with the key information based on a preset search library; recalling a second commodity matched with the key information by using a commodity recommendation model; and based on the commodity information of the first commodity, the commodity information of the second commodity and the portrait information of the user, selecting a target commodity recommended to the user from the first commodity and the second commodity. According to the method, the query intention of the user can be accurately captured, so that commodities can be accurately recommended to the user.

Description

Technical Field

[0001] This application relates to the field of artificial intelligence technology, specifically to a recommendation method, electronic device, and storage medium. Background Technology

[0002] In the e-commerce sector, merchants can independently build and operate their own websites. Because these websites offer a wide variety of products, consumers often spend a significant amount of time searching for the items they need. Therefore, intelligent shopping guide systems have emerged.

[0003] Currently, the recommendation function of intelligent shopping guide systems mainly relies on matching the similarity between the user's query information and the product name. However, this method is difficult to accurately capture the user's query intent, resulting in the inability to accurately provide users with products that meet their expectations. Summary of the Invention

[0004] This application provides a recommendation method, electronic device, and storage medium to solve the technical problem of being unable to accurately provide users with goods that meet their expectations.

[0005] The first aspect of this application provides a recommendation method, the method comprising: extracting key information from search information input by a user; determining a first product matching the key information based on a preset search library; recalling a second product matching the key information using a product recommendation model; and selecting a target product to recommend to the user from the first product and the second product based on product information of the first product, product information of the second product, and user profile information.

[0006] According to an embodiment of this application, the step of extracting key information from user-input search information includes: encoding preset prompt words to obtain the encoding vector of the prompt words; encoding each information word in the search information to obtain the semantic vector of each information word; calculating a first similarity between the semantic vector of each information word and the encoding vector of the prompt words; and determining the key information from multiple information words based on the first similarity.

[0007] According to an embodiment of this application, the preset search library includes product information of multiple candidate products. The step of determining the first product matching the key information based on the preset search library includes: matching the product information of each candidate product with the key information, and determining the candidate product corresponding to the product information containing the key information as the initial screening product; calculating the relevance score between the product information of the initial screening product and the key information; and determining the first product from the initial screening products based on the relevance score.

[0008] According to an embodiment of this application, the step of using a product recommendation model to recall a second product that matches the key information includes: using the product recommendation model to calculate the semantic similarity between the product information of multiple candidate products and the key information as a second similarity; and determining the second product from the multiple candidate products based on the second similarity.

[0009] According to an embodiment of this application, the method further includes: generating user profile information; generating user profile information includes: acquiring user behavior data, the behavior data including behavioral products in which the user interacts on multiple independent websites; determining a profile tag from preset profile tags that matches the product information of the behavioral products as the profile information.

[0010] According to an embodiment of this application, the method further includes: acquiring the user's registration data on the multiple independent websites; extracting the user's user identification code from the registration data; and establishing a mapping relationship between the user identification code and the profile information.

[0011] According to an embodiment of this application, the step of selecting a target product to recommend to the user from the first product and the second product based on the product information of the first product, the product information of the second product, and the user's profile information includes: generating prompt text based on the product information of the first product, the product information of the second product, and the user's profile information; predicting the recommendation score corresponding to the first product and the recommendation score corresponding to the second product based on the prompt text using a large language model; and determining the target product from the first product and the second product based on the recommendation scores corresponding to the first product and the second product.

[0012] According to an embodiment of this application, the method further includes: generating response information for the search information based on the product information of the target product.

[0013] A second aspect of this application provides a recommendation device, comprising: an extraction unit for extracting key information from search information input by a user; a determination unit for determining a first product matching the key information based on a preset search library; a recall unit for recalling a second product matching the key information using a product recommendation model; and a selection unit for selecting a target product to recommend to the user from the first product and the second product based on product information of the first product, product information of the second product, and user profile information.

[0014] A third aspect of this application provides an electronic device, the electronic device comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the method provided in the first aspect above.

[0015] A fourth aspect of this application provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the steps of the method provided in the first aspect above.

[0016] A fifth aspect of this application provides a computer program product, the computer program product including a computer program, which, when executed by a processor, implements the steps in the method provided in the first aspect above.

[0017] In several embodiments of this application, a first product matching the literal meaning of key information can be identified through a preset search library. A second product matching the semantic information of the key information can be recalled through a product recommendation model, thereby comprehensively acquiring both the first and second products matching the key information. By combining user profile information, the user's query intent can be accurately captured, thus accurately determining the target product to recommend to the user from the first and second products. Attached Figure Description

[0018] To more clearly illustrate the technical solutions of the embodiments of this application, the drawings used in the description of the embodiments of this application will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0019] Figure 1 This is a schematic diagram illustrating an application scenario of the recommended method provided in an embodiment of this application; Figure 2 This is a flowchart of a recommended method provided in an embodiment of this application; Figure 3 This is a schematic diagram of an intelligent shopping guide page provided in an embodiment of this application; Figure 4 This is a flowchart of a method for generating user profile information according to an embodiment of this application; Figure 5 This is a flowchart of a recommended method provided in another embodiment of this application; Figure 6 This is a schematic diagram of the framework of a recommended method provided in an embodiment of this application; Figure 7 This is a functional block diagram of a recommended device provided in an embodiment of this application; Figure 8 This is a schematic diagram of the structure of an electronic device implementing the recommended method according to an embodiment of this application. Detailed Implementation

[0020] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0021] Hereinafter, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature. In the description of embodiments of this application, words such as "exemplary" or "for example" are used to indicate examples, illustrations, or descriptions. Any embodiment or design described as "exemplary" or "for example" in the embodiments of this application should not be construed as being more preferred or advantageous than other embodiments or design solutions. Specifically, the use of words such as "exemplary" or "for example" is intended to present the relevant concepts in a specific manner.

[0022] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains. The terminology used in this application's specification is for the purpose of describing particular embodiments only and is not intended to be limiting of the application. It should be understood that, unless otherwise stated, " / " in this application means "or". For example, A / B can mean A or B. "And / or" in this application is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A alone, A and B simultaneously, and B alone. "At least one" refers to one or more. "More than one" refers to two or more. For example, at least one of a, b, or c can represent: a, b, c, a and b, a and c, b and c, and a, b, and c (seven cases).

[0023] In the e-commerce sector, merchants can independently build and operate their own websites. Because these websites offer a wide variety of products, consumers often spend a significant amount of time searching for the items they need. Therefore, intelligent shopping guide systems have emerged.

[0024] Currently, the recommendation function of intelligent shopping guide systems mainly relies on matching the similarity between the user's query information and the product name. When a user enters a long text for searching, the intelligent shopping guide system struggles to accurately capture the user's query intent, resulting in an inability to accurately provide the user with products that meet their expectations.

[0025] To address the aforementioned issues, this application provides a recommendation method that can accurately determine the target products to recommend to users.

[0026] Please see Figure 1 , Figure 1 This is a schematic diagram illustrating an application scenario of the recommended method provided in an embodiment of this application. This scenario may include various electronic devices 100 and servers 200.

[0027] Electronic device 100 may be a mobile phone, tablet computer, desktop computer, laptop computer, handheld computer, notebook computer, ultra-mobile personal computer (UMPC), netbook, as well as cellular phone, personal digital assistant (PDA), augmented reality (AR) device, virtual reality (VR) device, artificial intelligence (AI) device, wearable device, in-vehicle device, smart home device and / or smart city device. This application embodiment does not impose any special restrictions on the specific type of electronic device 100.

[0028] Server 200 can be a standalone physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, content delivery networks (CDN), and big data and artificial intelligence platforms, but it is not limited to these.

[0029] It should be noted that the method in this embodiment can be executed by electronic device 100 or server 200 alone, or by both server 200 and electronic device 100. When executed by electronic device 100 or server 200 alone, the entire recommendation process can be implemented by either electronic device 100 or server 200 independently. For example, a trained product recommendation model can be obtained by fine-tuning the product recommendation model on electronic device 100. Accordingly, after training, electronic device 100 can use the trained product recommendation model to recall products that match the key information. The above process can also be executed by server 200 alone. When executed by both server 200 and electronic device 100, server 200 can train the product recommendation model and then deploy the trained model to electronic device 100, which can then recall products that match the key information. Alternatively, some parts of the recommendation process can be implemented by electronic device 100, and some parts by server 200, with both working together to implement the entire recommendation process. Specific configurations can be made according to the actual application, and no specific limitations are made here.

[0030] It should be noted that when the recommended method provided in this application embodiment is executed by either server 200 or electronic device 100 alone, the above application scenario may also include only either server 200 or electronic device 100, or server 200 and electronic device 100 may be considered as the same device. In practical applications, when the recommended method provided in this application embodiment is executed jointly by server 200 and electronic device 100, server 200 and electronic device 100 may also be the same device, that is, server 200 and electronic device 100 may be different functional modules of the same device, or virtual devices virtualized by the same physical device.

[0031] In one possible implementation, the user can provide search information input by the electronic device 100, and the server 200 can use the recommendation method of this application embodiment to determine the first product and the second product that match the key information, and select the target product to recommend to the user from the first product and the second product, and return the target product to the electronic device 100 for presentation.

[0032] In this embodiment, the electronic device 100 and the server 200 can communicate directly or indirectly through one or more networks. This network can be a wired network or a wireless network; for example, the wireless network can be a mobile cellular network or a Wireless-Fidelity (Wi-Fi) network, and of course, it can be other possible networks. This embodiment does not limit this. It should be noted that... Figure 1The examples shown are merely illustrative; in reality, the number of electronic devices and servers is unlimited and is not specifically limited in the embodiments of this application.

[0033] The recommended methods provided by the embodiments of this application are described below with reference to the accompanying drawings and the application scenarios described above. It should be noted that the application scenarios described above are only shown to facilitate understanding of the spirit and principles of the embodiments of this application, and the embodiments of this application are not limited in any way in this respect.

[0034] like Figure 2 The diagram shown is a flowchart of a recommended method provided in an embodiment of this application. The recommended method is applied in electronic devices, for example... Figure 1 Electronic device 100. Depending on different needs, the order of steps in this flowchart can be changed, and some steps can be omitted.

[0035] S201, Extract key information from the search information entered by the user.

[0036] In at least one embodiment of this application, in response to a user operation, the electronic device can display a smart shopping guide page, for example... Figure 3 The interface shown is shown below. Users can enter any search information on the smart shopping guide page. For example, a user can enter "My neck and shoulders have been uncomfortable lately, and I want to buy neck and shoulder patches. Do you have any recommendations?"

[0037] In at least one embodiment of this application, an electronic device extracts key information from search information input by a user, including: encoding preset prompt words to obtain the encoding vector of the prompt words; encoding each information word in the search information to obtain the semantic vector of each information word; calculating a first similarity between the semantic vector of each information word and the encoding vector of the prompt words; and determining key information from multiple information words based on the first similarity.

[0038] In some embodiments of this application, the electronic device can pre-set prompts, for example, the prompts can be set to: extract product names from search information. In practical applications, the prompts can be adjusted according to actual needs.

[0039] In some embodiments of this application, the electronic device may also pre-set a mapping table, which may store the correspondence between multiple preset words and word codes. The electronic device may perform feature encoding on the prompt words based on the mapping table to obtain the encoding vector of the prompt words.

[0040] In some embodiments of this application, the electronic device can perform word segmentation on search information to obtain multiple information tokens. The electronic device can also perform feature encoding on each information token based on a mapping table to obtain an encoding vector for each information token. The electronic device uses a first weight matrix to map the encoding vector of each information token to obtain a query vector corresponding to each information token, and concatenates the query vectors corresponding to multiple information tokens based on the position of each information token in the search information to obtain a query matrix corresponding to the search information. The electronic device uses a second weight matrix to map the encoding vector of each information token to obtain a key vector corresponding to each information token, and concatenates the key vectors corresponding to multiple information tokens based on the position of each information token in the search information to obtain a key matrix corresponding to the search information. The electronic device uses a third weight matrix to map the encoding vector of each information token to obtain a value vector corresponding to each information token. The first, second, and third weight matrices can be set and adjusted according to actual needs, and this application does not impose any restrictions on them.

[0041] In some embodiments of this application, the electronic device calculates the attention weight corresponding to each information word based on the query matrix and the value matrix. The attention weight corresponding to each information word can be expressed as: ,in, It can represent a query matrix. It can represent the transpose of a value matrix. It can represent the dimension of a value matrix. It can represent a normalization function.

[0042] In some embodiments of this application, the electronic device calculates the product of the attention weight corresponding to each information word and the value vector corresponding to each information word to obtain the semantic vector of each information word.

[0043] For example, suppose the search query "want to buy neck and shoulder patches" includes the information tokens "want to buy" and "neck and shoulder patches". The query vectors for "want to buy" are Q1 and Q2, respectively; the key vectors for "want to buy" are K1 and K2, respectively; the value vectors for "want to buy" are V1 and V2, respectively. Calculations show that the attention weight of "want to buy" on top of "want to buy" is 0.8, and the attention weight of "want to buy" on top of "neck and shoulder patches" is 0.2. Therefore, the semantic vector of "want to buy" can be represented as: 0.8V1 + 0.2V2.

[0044] In some embodiments of this application, the electronic device can use a similarity calculation formula to calculate the similarity between the semantic vector of each information word and the encoding vector of the prompt word as a first similarity. The similarity calculation formula may include, but is not limited to, the cosine formula.

[0045] In some embodiments of this application, a first preset threshold can be set according to actual needs. The electronic device can determine the information word corresponding to the first similarity greater than the first preset threshold as key information. In other embodiments of this application, the electronic device can also determine the information word corresponding to the highest first similarity as key information.

[0046] The embodiments of this application can accurately determine the first similarity between each information word and the prompt word by using the semantic vector of each information word and the encoding vector of the prompt word. Then, by using the first similarity, key information can be accurately extracted from multiple information words.

[0047] S202, based on a preset search library, determine the first product that matches the key information.

[0048] In at least one embodiment of this application, the preset search library may include product information for multiple candidate products, which may include all products from multiple independent websites. The product information for each candidate product may include, but is not limited to: product title, product description, product price, and product feature tags. Product feature tags may include, but are not limited to: new product, best-selling product, promotional product, and low-priced product. The product feature tags for each candidate product can be determined based on the product information of each candidate product. For example, if the product description of product A is a price reduction of XX, then the product feature tag for product A can be determined as: promotional product.

[0049] In at least one embodiment of this application, an electronic device determines a first product that matches key information based on a preset search library, including: matching the product information of each candidate product with the key information, and determining the candidate product corresponding to the product information containing the key information as a preliminary screening product; calculating the relevance score between the product information of the preliminary screening product and the key information; and determining the first product from the preliminary screening products based on the relevance score.

[0050] In some embodiments of this application, during the calculation of relevance scores, the electronic device counts the number of times key information appears in the product information of the initially screened products, and also counts the total number of words in the product information of the initially screened products. The electronic device obtains the relevance score of the initially screened products based on the ratio of the number of times key information appears to the total number of words.

[0051] In some embodiments of this application, a second preset threshold can be set according to actual needs. The electronic device can determine the initially screened products corresponding to relevance scores greater than the second preset threshold as the first products. In other embodiments of this application, the electronic device can also determine the initially screened products corresponding to the highest N relevance scores as the first products. N is a positive integer, and N can be set and adjusted according to actual needs; for example, N can be set to 30.

[0052] This application embodiment can quickly identify the initial screening product from multiple candidate products by matching the product information of each candidate product with the key information. By identifying the candidate product corresponding to the product information containing the key information as the initial screening product, it can be ensured that the initial screening product is related to the key information. Furthermore, by calculating the correlation score between the product information of the initial screening product and the key information, the initial screening product that is strongly related to the key information can be accurately identified as the first product.

[0053] S203 uses a product recommendation model to recall a second product that matches the key information.

[0054] In at least one embodiment of this application, an electronic device uses a product recommendation model to recall a second product that matches key information, including: using the product recommendation model to calculate the semantic similarity between the product information and key information of multiple candidate products as a second similarity; and determining the second product from multiple candidate products based on the second similarity.

[0055] In some embodiments of this application, the network structure of the product recommendation model may include any one of the following: a large language model, a deep neural network, a convolutional neural network, a recurrent neural network, and a long short-term memory network; this application does not impose specific limitations on this. The electronic device can train the product recommendation model using a training set until the loss value of the product recommendation model is less than a preset threshold, thus obtaining a trained product recommendation model. The electronic device can also obtain user feedback information on the product recommendation model and periodically adjust the product recommendation model based on this feedback information.

[0056] In some embodiments of this application, the electronic device can perform feature encoding on the product information of each candidate product to obtain a semantic vector corresponding to each candidate product. The electronic device can also perform feature encoding on key information to obtain a semantic vector corresponding to the key information. The method by which the electronic device performs feature encoding on the product information and key information of each candidate product can refer to the method of feature encoding on each information word in step S202, and will not be described again in this application.

[0057] In some embodiments of this application, the electronic device can use a similarity calculation formula to calculate the similarity between the semantic vector corresponding to each candidate product and the semantic vector corresponding to the key information as a second similarity.

[0058] In some embodiments of this application, the electronic device can obtain network parameters for recalling products from the product recommendation model as a third preset threshold. The electronic device can then select candidate products with a second similarity greater than the third preset threshold as second products. In other embodiments of this application, the electronic device can obtain network parameters for recalling products from the product recommendation model as a preset quantity. The electronic device can select the top M candidate products with the highest second similarity as second products, where M represents the preset quantity and is a positive integer.

[0059] This application embodiment uses a product recommendation model to accurately calculate the semantic similarity between the product information and key information of multiple candidate products as a second similarity. Then, using this second similarity, a second product is determined from multiple candidate products. This can recall products that are semantically similar to the key information but have different literal meanings, thereby comprehensively recalling products that match the key information.

[0060] S204, based on the product information of the first product, the product information of the second product, and the user's profile information, select the target product to recommend to the user from the first product and the second product.

[0061] In at least one embodiment of this application, the electronic device can obtain user registration data on multiple independent websites. The registration data may include, but is not limited to, the user's name, email address, telephone number, and address. The electronic device can extract the user's user identification code from the registration data. The user identification code can be used to uniquely identify the user; for example, the user identification code may include, but is not limited to, the user's email address, telephone number, and address.

[0062] In at least one embodiment of this application, the electronic device can generate user profile information based on user behavior data. The method by which the electronic device generates user profile information can be referred to... Figure 4 The process is illustrated. Electronic devices can also establish a mapping relationship between user identification codes and user profile information, so that user profile information can be directly obtained using the user identification code when recommending products to users later.

[0063] In at least one embodiment of this application, an electronic device selects a target product to recommend to a user from the first product and the second product based on product information of a first product, product information of a second product, and user profile information. This includes: generating prompt text based on product information of the first product, product information of the second product, and user profile information; predicting the recommendation score corresponding to the first product and the recommendation score corresponding to the second product using a large language model based on the prompt text; and determining the target product from the first product and the second product based on the recommendation scores corresponding to the first product and the second product.

[0064] In some embodiments of this application, the prompt text may include a combination of one or more of the following information: the identity information of the large language model as a product recommendation expert, the task information of the large language model generating the recommendation score of the first product and the recommendation score of the second product based on the user's profile information, and examples for the large language model to refer to. The examples for the large language model to refer to may include, but are not limited to, positive samples that meet preset requirements. Positive samples that meet preset requirements can be used to indicate products that have been recommended to the user and added to the user's shopping cart.

[0065] In some embodiments of this application, the electronic device inputs prompt text into a large language model to obtain the recommended score for the first product and the recommended score for the second product.

[0066] In some embodiments of this application, the electronic device can select the product with a recommended score greater than a fourth preset threshold from the first product and the second product as the target product. The fourth preset threshold can be set and adjusted according to actual needs. In other embodiments of this application, the electronic device can also determine the first product and / or the second product corresponding to the highest K recommended scores as the target product, where K is a positive integer and can be set and adjusted according to actual needs. This application does not impose any restrictions on this.

[0067] This application embodiment combines user profile information with a large language model to accurately predict the recommendation scores of the first and second products for the user, thereby accurately identifying target products that meet the user's expectations for recommendation, and thus increasing the merchant's total transaction volume.

[0068] In several embodiments of this application, a first product matching the literal meaning of key information can be identified through a preset search library. A second product matching the semantic information of the key information can be recalled through a product recommendation model, thereby comprehensively acquiring both the first and second products matching the key information. By combining user profile information, the user's query intent can be accurately captured, thus accurately determining the target product to recommend to the user from the first and second products.

[0069] like Figure 4 The diagram shown is a flowchart of a method for generating user profile information according to an embodiment of this application. This method for generating user profile information is applied in electronic devices, for example... Figure 1 Electronic device 100. Depending on different needs, the order of steps in this flowchart can be changed, and some steps can be omitted.

[0070] S401, Obtain user behavior data.

[0071] In at least one embodiment of this application, the electronic device can acquire user behavior data across multiple independent websites. This behavior data includes interactive behaviors related to the products the user interacts with on the multiple independent websites. These interactive behaviors may include, but are not limited to, purchasing, adding to cart, clicking, browsing, adding to favorites, and following.

[0072] S402, determine the profile tags that match the product information of the behavior product from the preset profile tags as profile information.

[0073] In at least one embodiment of this application, the electronic device may pre-set profile tags, which may include, but are not limited to, "new product sensitive" tags, "best-selling product sensitive" tags, "event sensitive" tags, "price sensitive" tags, and "price insensitive" tags.

[0074] In at least one embodiment of this application, the product information of the behavioral product may include, but is not limited to, the product title, product description, and product price. The electronic device matches preset profile tags with the product information of the behavioral product, and determines the profile tags that match the product information of the behavioral product as profile information. For example, if the price of the behavioral product is greater than a preset price, the profile information can be a "price-insensitive" tag. Similarly, if the product title of the behavioral product includes "new product," the profile information can be a "new product sensitive" tag. Furthermore, if the sales volume of the behavioral product is greater than a preset quantity, the profile information can be a "best-selling product sensitive" tag.

[0075] In at least one embodiment of this application, the electronic device can construct a mapping relationship between a user's user identification code and profile information, so that when recommending products to a user, the corresponding profile information can be quickly obtained based on the user identification code.

[0076] In several embodiments of this application, user profile information can be accurately determined from preset profile tags by using product information of products from user interaction behaviors on multiple independent websites.

[0077] like Figure 5 The diagram shown is a flowchart of a recommended method provided in another embodiment of this application. The recommended method is applied in electronic devices, for example... Figure 1 Electronic device 100. Depending on different needs, the order of steps in this flowchart can be changed, and some steps can be omitted.

[0078] S501, extract key information from the search information entered by the user.

[0079] S502, based on a preset search library, determines the first product that matches the key information.

[0080] S503 uses a product recommendation model to recall a second product that matches the key information.

[0081] S504, based on the product information of the first product, the product information of the second product, and the user's profile information, select the target product to recommend to the user from the first product and the second product.

[0082] For details on steps S501 to S504, please refer to the above text. Figure 2 The detailed descriptions of steps S201 to S204 in the previous section will not be repeated here.

[0083] S505, Generate a response message for the search information based on the product information of the target product.

[0084] In at least one embodiment of this application, the electronic device can generate response information for search information based on the product information of the target product and using a large language model.

[0085] In other embodiments of this application, the electronic device can set a preset template according to actual needs. The preset template may include: "We provide you with the following suitable products...", and this application does not limit this. The electronic device can write the product information of the target product into the preset template to obtain the response information of the search information.

[0086] In at least one embodiment of this application, the response information may be presented in the form of, but is not limited to, text or voice.

[0087] like Figure 6 The diagram shown is a schematic representation of the framework of a recommended method provided in an embodiment of this application.

[0088] In some embodiments of this application, the electronic device receives search information (also referred to as a "user query") input by the user. The electronic device can use a Large Language Model (LLM) to extract information from the user query and extract product keywords (also referred to as "key information"). The electronic device can retrieve products based on product keywords using Elasticsearch (ES) search, or it can retrieve products based on product keywords using a semantic model (also referred to as a "product recommendation model").

[0089] In some embodiments of this application, the electronic device obtains the user's historical purchase data and historical add-to-cart data, and extracts unique identity information such as email address. Based on the user's behavior in different stores, the electronic device determines the user's profile tags (also known as profile information). These user profile tags may include: "New Product Sensitive" tag, "Hot-Selling Product Sensitive" tag, "Promotion Sensitive" tag, "Price Sensitive" tag, etc.

[0090] In some embodiments of this application, the electronic device inputs the product information of the products recalled by the search engine, the product information of the products recalled by the semantic model, and the user's profile tags into the LLM for sorting to obtain the sorting result of the recalled products.

[0091] In several embodiments of this application, a first product matching the literal meaning of key information can be identified through a preset search library. A second product matching the semantic information of the key information can be recalled through a product recommendation model, thereby comprehensively acquiring both the first and second products matching the key information. By combining user profile information, the user's query intent can be accurately captured, thus accurately determining the target product to recommend to the user from the first and second products.

[0092] like Figure 7 The diagram shown is a functional block diagram of a recommendation device provided in an embodiment of this application. The recommendation device 71 includes an extraction unit 710, a determination unit 711, a recall unit 712, a selection unit 713, a generation unit 714, an acquisition unit 715, and a creation unit 716. The module / unit referred to in this application refers to a module / unit that can be processed by a processor (e.g., ...). Figure 8 A series of computer-readable instruction segments acquired by the processor 1101 shown, and capable of performing a fixed function, stored in memory (e.g., a memory processor). Figure 8 In the memory 1102 shown.

[0093] In one embodiment, the extraction unit 710 is used to extract key information from the search information input by the user; the determination unit 711 is used to determine a first product matching the key information based on a preset search library; the recall unit 712 is used to recall a second product matching the key information using a product recommendation model; and the selection unit 713 is used to select a target product to recommend to the user from the first product and the second product based on the product information of the first product, the product information of the second product, and the user's profile information.

[0094] In one embodiment, the extraction unit 710 is specifically used for: encoding preset prompt words to obtain the encoding vector of the prompt words; encoding each information word in the search information to obtain the semantic vector of each information word; calculating the first similarity between the semantic vector of each information word and the encoding vector of the prompt words; and determining key information from multiple information words based on the first similarity.

[0095] In one embodiment, the preset search library includes product information of multiple candidate products. The determining unit 711 is specifically used to: match the product information of each candidate product with key information, and determine the candidate product corresponding to the product information containing key information as the initial screening product; calculate the relevance score between the product information of the initial screening product and the key information; and determine the first product from the initial screening products based on the relevance score.

[0096] In one embodiment, the recall unit 712 is specifically used to: use a product recommendation model to calculate the semantic similarity between the product information and key information of multiple candidate products as a second similarity; and determine a second product from multiple candidate products based on the second similarity.

[0097] In one embodiment, the generation unit 714 is used to generate user profile information; specifically, the generation unit 714 is used to: acquire user behavior data, the behavior data including behavioral products in which the user interacts on multiple independent sites; and determine profile tags that match the product information of the behavioral products from preset profile tags as profile information.

[0098] In one embodiment, the generation unit 714 is used to acquire user registration data on multiple independent sites; the extraction unit 710 is also used to extract the user identification code from the registration data; and the establishment unit 716 is used to establish a mapping relationship between the user identification code and the profile information.

[0099] In one embodiment, the selection unit 713 is specifically used to: generate prompt text based on the product information of the first product, the product information of the second product, and the user's profile information; predict the recommendation score corresponding to the first product and the recommendation score corresponding to the second product based on the prompt text and using a large language model; and determine the target product from the first product and the second product based on the recommendation scores corresponding to the first product and the second product.

[0100] In one embodiment, the generation unit 714 is used to generate response information for the search information based on the product information of the target product.

[0101] In several embodiments of this application, a first product matching the literal meaning of key information can be identified through a preset search library. A second product matching the semantic information of the key information can be recalled through a product recommendation model, thereby comprehensively acquiring both the first and second products matching the key information. By combining user profile information, the user's query intent can be accurately captured, thus accurately determining the target product to recommend to the user from the first and second products.

[0102] Figure 8 This is a schematic diagram of the structure of an electronic device implementing the recommended method according to an embodiment of this application. Optionally, Figure 8 The electronic device 100 shown is used to perform Figure 2 , Figure 4 , Figure 5 The method shown.

[0103] The electronic device 100 includes at least one processor 1101, a memory 1102, and at least one network interface 1103.

[0104] Processor 1101 may be, for example, a general-purpose central processing unit (CPU), a network processor (NP), a graphics processing unit (GPU), a neural-network processing unit (NPU), a data processing unit (DPU), a microprocessor, or one or more integrated circuits for implementing the embodiments of this application. For example, processor 1101 may include an application-specific integrated circuit (ASIC), a programmable logic device (PLD), or a combination thereof. A PLD may be, for example, a complex programmable logic device (CPLD), a field-programmable gate array (FPGA), a generic array logic (GAL), or any combination thereof.

[0105] Memory 1102 may be, for example, read-only memory (ROM) or other types of static storage devices capable of storing static information and instructions; random access memory (RAM) or other types of dynamic storage devices capable of storing information and instructions; electrically erasable programmable read-only memory (EEPROM); compact disc read-only memory (CD-ROM) or other optical disc storage, optical disc storage (including compressed discs, laser discs, optical discs, digital universal discs, Blu-ray discs, etc.); magnetic disk storage media or other magnetic storage devices; or any other medium capable of carrying or storing desired program code in the form of instructions or data structures and accessible by a computer, but not limited thereto. Optionally, memory 1102 exists independently and is connected to processor 1101 via internal connection 1104. Alternatively, memory 1102 and processor 1101 may be integrated together.

[0106] Network interface 1103 uses any transceiver-like device for communicating with other devices or communication networks. Network interface 1103 includes at least one of a wired network interface or a wireless network interface. The wired network interface is, for example, an Ethernet interface. The Ethernet interface is, for example, an optical interface, an electrical interface, or a combination thereof. The wireless network interface is, for example, a wireless local area network (WLAN) interface, a cellular network interface, or a combination thereof.

[0107] In some embodiments, processor 1101 includes one or more CPUs, such as Figure 8 CPU0 and CPU1 are shown in the diagram.

[0108] In some embodiments, the electronic device 100 may optionally include a plurality of processors, such as Figure 8 The processors 1101 and 1105 shown are illustrated. Each of these processors is, for example, a single-core processor (CPU) or a multi-core processor (CPU). Here, "processor" may optionally refer to one or more devices, circuits, and / or processing cores used to process data (such as computer program instructions).

[0109] In some embodiments, the electronic device 100 further includes an internal connection 1104. The processor 1101, memory 1102, and at least one network interface 1103 are connected via the internal connection 1104. The internal connection 1104 includes pathways for transmitting information between the aforementioned components. Optionally, the internal connection 1104 is a single board or a bus. Optionally, the internal connection 1104 may be divided into an address bus, a data bus, a control bus, etc.

[0110] In some embodiments, the electronic device 100 further includes an input / output interface 1106. The input / output interface 1106 is connected to the internal connection 1104.

[0111] Optionally, the processor 1101 implements the method in the above embodiments by reading the program code 910 stored in the memory 1102, or the processor 1101 implements the method in the above embodiments by internally stored program code. When the processor 1101 implements the method in the above embodiments by reading the program code 910 stored in the memory 1102, the memory 1102 stores program code that implements the method provided in the embodiments of this application.

[0112] For more details on how processor 1101 implements the above functions, please refer to the descriptions in the previous method embodiments, which will not be repeated here.

[0113] This embodiment also provides a computer storage medium storing computer instructions. When the computer instructions are executed on an electronic device, the electronic device performs the aforementioned method steps to implement the recommended method in the above embodiment.

[0114] This embodiment also provides a computer program product that, when run on an electronic device, causes the electronic device to perform the aforementioned steps to implement the recommended method in the above embodiment.

[0115] In addition, embodiments of this application also provide an apparatus, which may specifically be a chip, component, or module. The apparatus may include a connected processor and a memory; wherein the memory is used to store computer execution instructions, and when the apparatus is running, the processor may execute the computer execution instructions stored in the memory to cause the chip to perform the recommended methods in the above-described method embodiments.

[0116] In this embodiment, the electronic device, computer storage medium, computer program product or chip are all used to execute the corresponding method provided above. Therefore, the beneficial effects that can be achieved can be referred to the beneficial effects of the corresponding method provided above, and will not be repeated here.

[0117] Through the above description of the embodiments, those skilled in the art can clearly understand that, for the sake of convenience and brevity, only the division of the above functional modules is used as an example. In actual applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the device can be divided into different functional modules to complete all or part of the functions described above.

[0118] In the several embodiments provided in this application, it should be understood that the disclosed apparatus and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of modules or units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another apparatus, or some features may be ignored or not executed. Furthermore, the mutual coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or units may be electrical, mechanical, or other forms.

[0119] The unit described as a separate component may or may not be physically separate. The component shown as a unit can be one physical unit or multiple physical units, that is, it can be located in one place or distributed in multiple different places. Some or all of the units can be selected to achieve the purpose of the solution in this embodiment according to actual needs.

[0120] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0121] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a readable storage medium. Based on this understanding, the technical solutions of the embodiments of this application, essentially or in other words, the parts that contribute to the prior art, or all or part of the technical solutions, can be embodied in the form of a software product. This software product is stored in a storage medium and includes several instructions to cause a device (which may be a microcontroller, chip, etc.) or processor to execute all or part of the steps of the methods of the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0122] The above are merely specific embodiments of this application, but the scope of protection of this application is not limited thereto. Any changes or substitutions within the technical scope disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A recommendation method, characterized in that, The method includes: Extract key information from the user's search input; Based on a preset search library, determine the first product that matches the key information; Using a product recommendation model, recall a second product that matches the key information; Based on the product information of the first product, the product information of the second product, and the user's profile information, a target product is selected from the first product and the second product to be recommended to the user.

2. The recommended method according to claim 1, characterized in that, The extraction of key information from the user-input search information includes: Encode the preset prompt words to obtain the encoding vector of the prompt words; Encode each information word in the search information to obtain the semantic vector of each information word; Calculate the first similarity between the semantic vector of each information word and the encoding vector of the prompt word; Based on the first similarity, the key information is determined from multiple information lexical units.

3. The recommended method according to claim 1, characterized in that, The preset search library includes product information for multiple candidate products. The step of determining the first product matching the key information based on the preset search library includes: The product information of each candidate product is matched with the key information, and the candidate product corresponding to the product information containing the key information is determined as the initial screening product; Calculate the relevance score between the product information of the initially screened products and the key information; Based on the correlation score, the first product is determined from the initially screened products.

4. The recommended method according to claim 1, characterized in that, The step of using a product recommendation model to recall a second product that matches the key information includes: Using the product recommendation model, the semantic similarity between the product information of multiple candidate products and the key information is calculated as a second similarity. Based on the second similarity, the second product is determined from the plurality of candidate products.

5. The recommended method according to claim 1, characterized in that, The method further includes: generating user profile information; generating user profile information includes: Acquire the user's behavioral data, which includes the products in which the user interacts with multiple independent websites; The profile information is determined from the preset profile tags that match the product information of the product in question.

6. The recommended method according to claim 5, characterized in that, The method further includes: Obtain the user's registration data on the multiple independent websites; Extract the user's user identification code from the registration data; Establish a mapping relationship between the user identification code and the profile information.

7. The recommended method according to claim 1, characterized in that, The step of selecting a target product to recommend to the user from the first product and the second product, based on the product information of the first product, the product information of the second product, and the user's profile information, includes: Based on the product information of the first product, the product information of the second product, and the user's profile information, a prompt text is generated; Based on the prompt text, a large language model is used to predict the recommendation score for the first product and the recommendation score for the second product. Based on the recommendation scores corresponding to the first product and the second product, the target product is determined from the first product and the second product.

8. The recommended method according to claim 1, characterized in that, The method further includes: Based on the product information of the target product, generate response information for the search information.

9. An electronic device, characterized in that, include: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor, when executing the computer program, implements the recommended method as described in any one of claims 1 to 8.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, implements the recommended method as described in any one of claims 1 to 8.

Images