Article recommendation method and device and electronic equipment

Abstract

The invention provides an article recommendation method and device and electronic equipment, and relates to the technical field of information recommendation, and the article recommendation method comprises the following steps: obtaining historical behavior information of a target user and a pre-generated demand scene map, the demand scene map comprises first demand keywords corresponding to a plurality of crowds and first article categories corresponding to the first demand keywords, and the first demand keywords are used for describing one category of shopping demands of the corresponding crowds; according to the historical behavior information, determining a target crowd to which the target user belongs from the plurality of crowds, and determining a target demand keyword corresponding to the target user from first demand keywords corresponding to the target crowd; and according to a first article category corresponding to the target demand keyword, determining a to-be-recommended article of the target user. The object is recommended to the user according to the demand of the user, the accurate effect of the recommendation result is improved, and the user experience is improved.

Description

Technical Field

[0001] The present application relates to the technical field of information recommendation, and in particular to an item recommendation method, device and electronic device. Background Art

[0002] Item recommendations are widely used in convenience stores, supermarkets, or e-commerce websites. They can help users get the items they want and help merchants sell more items.

[0003] Related technologies usually use the following methods to recommend items: 1. Item recommendation based on i2i, specifically, based on the items clicked by the user, recommend the same or similar items to the user through collaborative filtering methods such as co-click and co-purchase; 2. Item recommendation based on vector recall technology, specifically, through a deep learning model, obtain the user's vector representation based on the items purchased by the user, obtain the item's vector representation based on the item's features, and then obtain the similarity between the user and the item, and then recommend to the user items that have a high similarity with the user's vector representation; 3. Item recommendation based on a graph model, specifically, construct a graph model between items, obtain the item's vector representation, and recommend to the user items that are similar to the items purchased or clicked by the user through the similarity between the vector representations.

[0004] The above-mentioned item recommendation methods in the related art are all based on the similarity between items to recommend items to users. It is easy for the recommended items to not meet the user's needs, the accuracy of the recommendation results is poor, and the user experience is poor. Summary of the invention

[0005] The present application aims to solve one of the technical problems in the related art at least to some extent.

[0006] The present application proposes an item recommendation method, device and electronic device to solve the technical problems of poor accuracy of recommendation results and poor user experience in the item recommendation method in the related art.

[0007] The first aspect of the present application proposes an item recommendation method, comprising: obtaining historical behavior information of a target user and a pre-generated demand scenario map, wherein the demand scenario map includes first demand keywords corresponding to multiple groups of people and first item categories corresponding to the first demand keywords, and the first demand keywords are used to describe a type of shopping demand of the corresponding group of people; determining a target group to which the target user belongs from the multiple groups of people based on the historical behavior information, and determining a target demand keyword corresponding to the target user from the first demand keyword corresponding to the target group; and determining an item to be recommended for the target user based on the first item category corresponding to the target demand keyword.

[0008] The second aspect of the present application proposes an item recommendation device, including: a first acquisition module, used to obtain historical behavior information of a target user and a pre-generated demand scenario map, wherein the demand scenario map includes first demand keywords corresponding to multiple groups of people and first item categories corresponding to the first demand keywords, and the first demand keywords are used to describe a type of shopping demand of the corresponding group of people; a first determination module, used to determine the target group to which the target user belongs from the multiple groups of people based on the historical behavior information, and determine the target demand keywords corresponding to the target user from the first demand keywords corresponding to the target group; a second determination module, used to determine the target user's to-be-recommended items based on the first item category corresponding to the target demand keywords.

[0009] The third aspect embodiment of the present application proposes an electronic device, comprising: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the item recommendation method proposed in the first aspect embodiment of the present application.

[0010] The fourth aspect embodiment of the present application proposes a non-transitory computer-readable storage medium storing computer instructions, wherein the computer instructions are used to enable the computer to execute the item recommendation method proposed in the first aspect embodiment of the present application.

[0011] The fifth aspect embodiment of the present application proposes a computer program product, including a computer program, which, when executed by a processor, implements the item recommendation method proposed in the first aspect embodiment of the present application.

[0012] The item recommendation method, device and electronic device provided in the embodiments of the present application obtain the historical behavior information of the target user and a pre-generated demand scenario map, wherein the demand scenario map includes first demand keywords corresponding to multiple groups of people and first item categories corresponding to the first demand keywords, and the first demand keywords are used to describe a type of shopping demand of the corresponding group of people. According to the historical behavior information, the target group to which the target user belongs is determined from multiple groups of people, and the target demand keywords corresponding to the target user are determined from the first demand keywords corresponding to the target group. According to the first item category corresponding to the target demand keywords, the target user's to-be-recommended items are determined, thereby achieving the goal of recommending items to users according to their needs, improving the accuracy of the recommendation results, and improving the user experience.

[0013] Additional aspects and advantages of the present application will be given in part in the description below, and in part will become apparent from the description below, or will be learned through the practice of the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] The above and / or additional aspects and advantages of the present application will become apparent and easily understood from the following description of the embodiments in conjunction with the accompanying drawings, in which:

[0015] Figure 1 A flowchart of the item recommendation method provided in Example 1 of the present application;

[0016] Figure 2 A schematic diagram of an item recommendation page provided in Example 1 of the present application;

[0017] Figure 3 A flowchart of the item recommendation method provided in Example 2 of the present application;

[0018] Figure 4 A partial schematic diagram of the demand scenario map provided in Example 2 of the present application;

[0019] Figure 5 A flowchart of the item recommendation method provided in Example 3 of the present application;

[0020] Figure 6 This is a schematic diagram of the structure of the item recommendation device provided in the fourth embodiment of the present application;

[0021] Figure 7 This is a schematic diagram of the structure of the item recommendation device provided in Example 5 of the present application;

[0022] Figure 8 A block diagram of an exemplary electronic device suitable for implementing the embodiments of the present application is shown. DETAILED DESCRIPTION

[0023] Embodiments of the present application are described in detail below, and examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are intended to be used to explain the present application, and should not be construed as limiting the present application.

[0024] It should be noted that the acquisition, storage and application of user personal information involved in the technical solution of this application are in compliance with the provisions of relevant laws and regulations and do not violate public order and good morals.

[0025] The item recommendation methods in related technologies are all based on the similarity between items to recommend items to users. This may easily lead to the problem that the recommended items do not meet the user's needs, such as recommending items that the user has already purchased, and the accuracy of the recommendation results is poor, resulting in a poor user experience.

[0026] The present application aims to solve the technical problems of poor accuracy of recommendation results and poor user experience in the item recommendation method in the related art, and proposes an item recommendation method, device, electronic device, storage medium and computer program product.

[0027] Among them, the item recommendation method includes: obtaining the historical behavior information of the target user and a pre-generated demand scenario map, wherein the demand scenario map includes a first demand keyword corresponding to multiple groups of people and a first item category corresponding to the first demand keyword, and the first demand keyword is used to describe a type of shopping demand of the corresponding group of people; according to the historical behavior information, determining the target group to which the target user belongs from multiple groups of people, and determining the target demand keyword corresponding to the target user from the first demand keyword corresponding to the target group; according to the first item category corresponding to the target demand keyword, determining the target user's to-be-recommended item. The item recommendation method, device, and electronic device provided by the embodiments of the present application can recommend items to users according to their needs, improve the accuracy of the recommendation results, and improve the user experience.

[0028] The following describes the item recommendation method, device, electronic device, storage medium and computer program product of the embodiments of the present application with reference to the accompanying drawings.

[0029] First, the item recommendation method provided in the embodiment of the present application is described.

[0030] It should be noted that the item recommendation method provided in the embodiment of the present application is executed by an item recommendation device. The item recommendation device can be an electronic device, or it can be configured in an electronic device, so as to recommend items to users according to their needs by executing the item recommendation method provided in the embodiment of the present application, thereby improving the accuracy of the recommendation results and improving the user experience. The item recommendation device can be applied to an e-commerce platform, for example, it can be used to determine the items to be recommended displayed on the homepage of the platform when the user logs in to the e-commerce platform, and it can also be applied to other platforms, and this application does not limit this. The embodiment of the present application is described by taking the application of the item recommendation device to an e-commerce platform as an example.

[0031] Among them, the electronic device can be a personal computer (PC for short), a cloud device, a mobile device, a server, etc. The mobile device can be, for example, a mobile phone, a tablet computer, a personal digital assistant, a wearable device, a car-mounted device, or any other hardware device, and this application does not impose any restrictions on this.

[0032] Figure 1 This is a flow chart of the item recommendation method provided in Example 1 of the present application. Figure 1 As shown, the item recommendation method may include the following steps 101-103.

[0033] Step 101, obtaining historical behavior information of target users and a pre-generated demand scenario map, wherein the demand scenario map includes first demand keywords corresponding to multiple groups of people and first item categories corresponding to the first demand keywords, and the first demand keyword is used to describe a type of shopping demand of the corresponding group of people.

[0034] The target user is the user for whom item recommendation is to be performed.

[0035] The historical behavior information of the target user is the operational behavior information of the target user in the past period of time, for example, it may include information reported by the target user on the e-commerce platform or registered membership information, information about items clicked by the target user on the e-commerce platform in the past period of time, information about items purchased by the target user on the e-commerce platform in the past period of time, etc. This application does not impose any restrictions on this.

[0036] The demand scenario map includes first demand keywords corresponding to multiple groups of people and first item categories corresponding to each first demand keyword.

[0037] Among them, the crowd refers to a group of people with similar shopping needs, such as the mother and baby crowd, the home improvement crowd, etc. Among them, the shopping needs of the mother and baby crowd are to buy mother and baby items, and the shopping needs of the home improvement crowd are to buy home decoration items.

[0038] In the embodiment of the present application, multiple groups can be obtained based on the information reported by each user on the e-commerce platform or the registered member information. For example, multiple users who have submitted data related to mothers and babies on the e-commerce platform can be determined as the mother and baby group, and multiple users who have registered as home improvement members on the e-commerce platform can be determined as the home improvement group.

[0039] Among them, each group of people may correspond to one or more first demand keywords, and the first demand keywords are used to describe a type of shopping demand of the corresponding group of people. It should be noted that the type of shopping demand described by each first demand keyword is a more specific shopping demand divided from the shopping needs of the corresponding group of people. For example, the first demand keywords corresponding to the home improvement group may include "kitchen storage", "bedroom storage", etc. Among them, "kitchen storage" describes the shopping demand of the home improvement group: to buy items with kitchen storage function; "bedroom storage" describes the shopping demand of the home improvement group: to buy items with kitchen storage function. Among them, the first demand keyword can be a verb, gerund, etc., and this application does not impose any restrictions on this.

[0040] In the demand scenario graph, each first demand keyword has one or more corresponding first item categories, wherein the multiple first item categories are categories of the same level. Items under the first item category can meet the shopping needs described by the corresponding first demand keyword.

[0041] For example, suppose that the items in the e-commerce platform are divided into multiple first-level categories, each first-level category is divided into multiple second-level categories, and each second-level category is divided into multiple third-level categories. Among them, the first item category can be a fine-grained third-level category. For example, the first demand keyword "kitchen storage" has the following multiple first item categories of the same level: 1692_pot rack, 30761_multi-function storage rack, 30766_steaming grid / rack / cage / drawer, 30757_dish rack, 6217_seasoning utensils, 30756_rag rack / kitchen cleaning storage rack, 30759_chopstick cage rack, 16821_kitchen storage utensils, 30758_cutting board / pot lid rack, 30753_spice rack, 30752_micro Oven rack / cooking rack, 30754_knife rack, 11977_kitchen rack, 30755_sink / drain rack, 6215_fresh-keeping box, etc.; "Bedroom storage" has the following multiple first item categories of the same level: 13782_storage cabinet, 13785_storage rack / basket, 13783_storage box, 1660_washing / ironing, 13781_storage bag / bag, 21989_storage box, 21493_partition storage, 13784_dust cover. Among them, the number before "_" indicates the identification of the corresponding first item category. For example, "1692" means that the identification of the first item category "pot rack" is "1692".

[0042] In the embodiment of the present application, a demand scenario map can be generated in advance, and then item recommendations can be made for target users based on the pre-generated demand scenario map.

[0043] Step 102: determining a target group to which the target user belongs from multiple groups according to the historical behavior information, and determining a target demand keyword corresponding to the target user from the first demand keyword corresponding to the target group.

[0044] As a possible implementation method, the target group to which the target user belongs can be determined from multiple groups based on the information reported by the target user on the e-commerce platform or the registered member information.

[0045] For example, assuming that the target user submits data related to mothers and babies on the e-commerce platform, it can be determined based on the data that the target user belongs to the mother and baby group among the multiple groups included in the demand scenario map.

[0046] As another possible implementation method, the target group to which the target user belongs may be determined from multiple groups based on information about items that the target user has clicked on the e-commerce platform in the past period of time.

[0047] For example, assuming that the target user has frequently clicked on items related to home improvement on the e-commerce platform in the past period of time, it can be determined based on this information that the target user belongs to the home improvement group among the multiple groups included in the demand scenario map.

[0048] Furthermore, the target demand keyword corresponding to the target user can be determined from the first demand keyword corresponding to the target population based on the information of the items clicked by the target user on the e-commerce platform in the past period of time. Among them, the first item category to which the items clicked by the target user on the e-commerce platform in the past period of time belong can be first determined, and the first demand keyword corresponding to the first item category in the demand scenario graph can be determined as the target demand keyword corresponding to the target user.

[0049] For example, assuming that the first demand keywords corresponding to the home improvement crowd include "kitchen storage", "bedroom storage", etc., it has been determined through the above method that the target user belongs to the home improvement crowd, and the items that the target user has clicked on the e-commerce platform in the past period of time include pot racks, dish racks, and food storage boxes. Since the pot rack, dish rack, and food storage box belong to the "1692_pot rack", "30757_dish rack", and "6215_food storage box" categories respectively, and the first demand keyword corresponding to these first item categories in the demand scenario map is "kitchen storage", it can be determined that the target demand keyword corresponding to the target user is "kitchen storage".

[0050] Step 103 , determining the items to be recommended for the target user according to the first item category corresponding to the target demand keyword.

[0051] In an embodiment of the present application, a first item category corresponding to a target demand keyword may be determined from a demand scenario graph, and items under the first item category may be determined as items to be recommended for the target user.

[0052] For example, continuing with the above example, assuming that the target user belongs to the home improvement crowd, and the items that the target user has clicked on the e-commerce platform in the past period of time include pot racks, dish racks, and food storage boxes, and the target demand keyword corresponding to the target user is "kitchen storage", and because "kitchen storage" has the following multiple first item categories of the same level: 1692_pot rack, 30761_multi-function storage rack, 30766_steaming grid / rack / cage / drawer, 30757_dish rack, 621 7_Seasoning utensils, 30756_Dough rack / kitchen cleaning storage rack, 30759_Chopstick cage rack, 16821_Kitchen storage utensils, 30758_Cutting board / pot lid rack, 30753_Seasoning rack, 30752_Microwave rack / oven rack, 30754_Knife rack, 11977_Kitchen storage rack, 30755_Sink / drain rack, 6215_Fresh-keeping box, etc., then items under any category of the above multiple first item categories can be determined as items to be recommended.

[0053] The items under the first item category corresponding to the target demand keyword may be sorted, and the items with the highest sorting are determined as the items to be recommended to the target user. The sorting rules may be set as needed, and this application does not limit this.

[0054] It should be noted that the target user's to-be-recommended items may be items belonging to the same first item category as the item clicked by the target user. For example, if the item clicked by the user is a pot rack of a certain brand, the item to be recommended may be a pot rack of another brand. Alternatively, the target user's to-be-recommended items may also be items corresponding to the target demand keyword and belonging to a different first item category than the item clicked by the target user. For example, if the target demand keyword is “kitchen storage” and the item clicked by the user is a pot rack of a certain brand, the item to be recommended may be a rack under the “30761_Multi-function storage rack” category corresponding to “kitchen storage”, etc.

[0055] Therefore, by determining the target population to which the target user belongs and the corresponding target demand keywords based on the demand scenario map according to the target user's historical behavior information, and then determining the target user's recommended items according to the first item category corresponding to the target demand keywords, it is possible to recommend items to the target user based on their needs. This can largely avoid the problem of recommending items that the target user has already purchased, and the novelty of the recommended items is improved to be more in line with the user's needs, thereby improving the accuracy of the recommendation results and the user experience.

[0056] In addition, in a possible implementation form, while recommending the recommended items to the target user, Figure 2 As shown in the item recommendation page, the target user's target group and target need keywords are displayed, so that the target user can understand the reason for recommending the item to him through the target need keywords, so that the target user can clearly perceive that the item recommendation device can meet his shopping needs more humanely, thereby increasing the target user's click rate on the recommended items and improving the user experience. Figure 2 The “moms” in the oval frame are the target group of target users. Figure 2 The “painting tools”, “children’s tableware” and “baby snacks” framed in the oval frame are the target demand keywords corresponding to the target users.

[0057] In summary, the item recommendation method provided in the embodiment of the present application obtains the historical behavior information of the target user and a pre-generated demand scenario map, wherein the demand scenario map includes first demand keywords corresponding to multiple groups of people and first item categories corresponding to the first demand keywords, and the first demand keywords are used to describe a type of shopping demand of the corresponding group of people. According to the historical behavior information, the target group to which the target user belongs is determined from multiple groups of people, and the target demand keywords corresponding to the target user are determined from the first demand keywords corresponding to the target group. According to the first item category corresponding to the target demand keywords, the target user's to-be-recommended items are determined, thereby achieving the goal of recommending items to users according to their needs, improving the accuracy of the recommendation results, and improving the user experience.

[0058] Combine the following Figure 3 , the process of generating a demand scenario graph in an embodiment of the present application is explained.

[0059] Figure 3 This is a flow chart of the item recommendation method provided in Example 2 of this application. Figure 3 As shown, the item recommendation method may include the following steps 301-307.

[0060] Step 301 , for each group of people, according to the historical behavior information of the group of people, the target item attributes that affect the shopping decision of the group of people are obtained, and the attribute values ​​corresponding to the target item attributes are obtained.

[0061] The target item attribute is an item attribute that affects people's shopping decisions among multiple item attributes. The item attribute is an attribute of an item, such as "applicable age", "applicable gender", "applicable scene", etc. The attribute value corresponding to the item attribute is a specific attribute value of the item, for example, the attribute value corresponding to "applicable age" may include "infant", "child", etc., and the attribute value corresponding to "applicable gender" may include "male" and "female".

[0062] Among them, the attribute value corresponding to the target item attribute may include one or more, and this application does not impose any restrictions on this.

[0063] Taking mothers and infants as an example, since "applicable age" and "applicable gender" are the main decision-making factors that mothers and infants consider when shopping, it can be determined that the target item attributes that affect the shopping decisions of mothers and infants include "applicable age" and "applicable gender", and the attribute values ​​corresponding to "applicable age" and "applicable gender" are obtained. Among them, the attribute values ​​corresponding to "applicable age" include, for example, "infants" and "children", and the attribute values ​​corresponding to "applicable gender" include, for example, "male" and "female".

[0064] Step 302: Obtain a second demand keyword for describing the shopping demand of a group of people and a first item category corresponding to the second demand keyword.

[0065] Among them, the second demand keyword has the same function as the first demand keyword, and both are used to describe a type of shopping demand of the group of people. In the embodiment of the present application, the difference between the first demand keyword and the second demand keyword is that the first demand keyword describes a more fine-grained shopping demand of the group of people compared with the second demand keyword. For example, the second demand keyword corresponding to the home improvement group may include "storage", which describes the shopping demand of the home improvement group: buying items with storage functions. The first demand keyword corresponding to the home improvement group may include "kitchen storage", "bedroom storage", etc. Among them, "kitchen storage" describes the shopping demand of the home improvement group: buying items with kitchen storage functions; "bedroom storage" describes the shopping demand of the home improvement group: buying items with kitchen storage functions. Among them, the second demand keyword can be a verb, a gerund, etc., and this application does not limit this.

[0066] The first item category corresponding to the second demand keyword is the item category to which items that can meet the shopping demand described by the second demand keyword belong.

[0067] In the embodiment of the present application, the items that can meet the shopping needs described by the second demand keyword can be determined first, and then the first item category to which these items belong can be determined as the first item category corresponding to the second demand keyword. The first item category corresponding to the second demand keyword is a category of the same level as the first item category corresponding to the first demand keyword.

[0068] For example, assuming that the items in the e-commerce platform are divided into multiple first-level categories, each first-level category is divided into multiple second-level categories, and each second-level category is divided into multiple third-level categories, then the first item category corresponding to the first demand keyword and the first item category corresponding to the second demand keyword can both be fine-grained third-level categories.

[0069] Among them, step 302 can be executed before step 301 or after step 301, and only needs to be executed before step 303. This application does not limit the execution timing of step 302.

[0070] Step 303, based on the attribute value corresponding to the target item attribute and the second demand keyword, determine the first demand keyword corresponding to the population, and based on the attribute value corresponding to the target item attribute and the first item category corresponding to the second demand keyword, determine the first item category corresponding to the first demand keyword.

[0071] As a possible implementation method, step 303 can be implemented in the following way: combining the attribute value corresponding to the target item attribute with the second demand keyword to obtain the first demand keyword corresponding to the population; for any first demand keyword corresponding to the population, based on the coverage rate of the attribute values ​​constituting the first demand keyword in the items under the target item category, the target item category is screened to obtain the first item category corresponding to the first demand keyword, wherein the target item category is any first item category corresponding to the second demand keyword constituting the first demand keyword.

[0072] For each group of people, at least one attribute value corresponding to at least one target item attribute and at least one second demand keyword may be combined to obtain a first demand keyword corresponding to the group of people.

[0073] The coverage rate of the attribute value in the items under the target item category may be the ratio of the number of items under the target item category having the attribute value to the total number of items under the target item category. For example, taking the coverage rate of the attribute value "children" of "applicable age" in the items under the category "7057: baby diapers" as an example, since the applicable age of the items under the category "7057: baby diapers" is generally infants, the ratio of the number of items under the category "7057: baby diapers" having the attribute value "children" to the total number of items under the category "7057: baby diapers" is substantially 0.

[0074] In an embodiment of the present application, for any first demand keyword corresponding to a group of people, any first item category corresponding to the second demand keyword constituting the first demand keyword is referred to as a target item category. When the coverage rate of the attribute values ​​constituting the first demand keyword in the items under the target item category is lower than a preset coverage rate threshold, the target item category can be deleted, thereby determining the remaining first item categories in all the first item categories corresponding to the second demand keyword as the first item category corresponding to the first demand keyword.

[0075] For example, assuming that the target item attributes corresponding to the mother and baby population include "applicable age" and "applicable season", the attribute values ​​corresponding to "applicable age" include "infant" and "child", etc., and the attribute values ​​corresponding to "applicable season" include "spring", "summer", "autumn", and "winter", and the second demand keyword includes "protecting the baby and sleeping well", and the first item categories corresponding to "protecting the baby and sleeping well" include the following categories: 1563: baby strollers; 6317: guardrails; 13290: baby sleeping bags / blankets; 11227: underwear; 14938: home clothes; 7057: baby diapers; 14 930: jumpsuit / crawling suit; 15273: blanket; 15257: fiber quilt; 13535: children's mattress; 1564: crib; 15269: ice silk mat; 1559: daily care; 15261: flower and grass pillow; 7052: infant formula; 15276: pillow towel and pillowcase; 15614: baby pillow core / pillowcase; 15274: mattress / mattress; 15613: baby mat / mosquito net; 15609: baby diaper / towel; 15251: three-piece suit; 15262: fiber pillow; 15283: pillow; 15267: straw mat / rattan mat.

[0076] Taking "summer" as an example, "baby", "summer" and "baby sleeping" can be combined to obtain the first demand keyword "baby-summer-baby sleeping", and "children", "summer" and "baby sleeping" can be combined to obtain the first demand keyword "children-summer-baby sleeping". Among them, assuming that the coverage threshold is 60%, since the coverage of "children" in the items under the categories of "1563: baby stroller", "13290: baby sleeping bag / blanket", "7057: baby diapers", "14930: jumpsuit / crawling clothes", "1564: crib", "15269: ice silk mat", "7052: infant milk powder", "15276: pillowcase and pillowcase", and "15609: baby diaper / towel" are all very low, these first item categories can be deleted to obtain the first demand keywords corresponding to the mother and baby population and the first item categories corresponding to the first demand keywords as shown in Table 1 below.

[0077] Table 1 The first demand keywords and the corresponding first item categories for mothers and infants

[0078]

[0079] Step 304: Generate a demand scenario graph based on the first demand keywords corresponding to the multiple groups of people and the first item categories corresponding to the first demand keywords.

[0080] Thus, a demand scenario map can be generated, which includes first demand keywords corresponding to multiple groups of people and first item categories corresponding to the first demand keywords. Then, item recommendations can be made for users based on the demand scenario map.

[0081] Among them, part of the content in the demand scenario map can be Figure 4 The form shown.

[0082] Step 305, obtaining historical behavior information of the target user and a pre-generated demand scenario map, wherein the demand scenario map includes first demand keywords corresponding to multiple groups of people and first item categories corresponding to the first demand keywords, and the first demand keyword is used to describe a type of shopping demand of the corresponding group of people.

[0083] Step 306: Determine the target group to which the target user belongs from multiple groups according to the historical behavior information, and determine the target demand keyword corresponding to the target user from the first demand keyword corresponding to the target group.

[0084] Step 307 , determining the items to be recommended for the target user according to the first item category corresponding to the target demand keyword.

[0085] The specific implementation process and principle of steps 305-307 may refer to the description of other embodiments and will not be repeated here.

[0086] In summary, the item recommendation method provided in the embodiment of the present application obtains the target item attributes that affect the shopping decisions of the group of people according to the historical behavior information of the group of people, obtains the attribute values ​​corresponding to the target item attributes, obtains the second demand keyword used to describe the shopping needs of the group of people and the first item category corresponding to the second demand keyword, determines the first demand keyword corresponding to the group of people based on the attribute values ​​corresponding to the target item attributes and the second demand keyword, and determines the first item category corresponding to the first demand keyword based on the attribute values ​​corresponding to the target item attributes and the first item category corresponding to the second demand keyword, generates a demand scenario map based on the first demand keywords corresponding to multiple groups of people and the first item categories corresponding to the first demand keywords, and realizes demand The scenario graph is generated by obtaining the historical behavior information of the target user and a pre-generated demand scenario graph, wherein the demand scenario graph includes a first demand keyword corresponding to multiple groups of people and a first item category corresponding to the first demand keyword, and the first demand keyword is used to describe a type of shopping demand of the corresponding group of people. According to the historical behavior information, the target group to which the target user belongs is determined from multiple groups of people, and the target demand keyword corresponding to the target user is determined from the first demand keyword corresponding to the target group. According to the first item category corresponding to the target demand keyword, the target user's to-be-recommended items are determined, thereby recommending items to the target user according to the pre-generated demand scenario graph and the target user's needs, thereby improving the accuracy of the recommendation results and the user experience.

[0087] The process of generating a demand scenario graph in the embodiment of the present application is further explained below in conjunction with the accompanying drawings.

[0088] Figure 5 This is a flow chart of the item recommendation method provided in Example 3 of this application. Figure 5 As shown, before step 101, the item recommendation method may include the following steps 501-507.

[0089] Step 501 : for each group of people, determine a second item category related to the group of people from a plurality of candidate item categories according to historical behavior information of the group of people, wherein the subcategory of the candidate item category is a category at the same level as the first item category.

[0090] The candidate item category is a predetermined item category, and the subcategory of the candidate item category is a category at the same level as the first item category. The second item category is at the same level as the candidate item category.

[0091] Among them, the number of second item categories related to the crowd can be one or more, which can be set according to the needs of the actual scene, and this application does not impose any restrictions on this.

[0092] For example, assuming that the items in the e-commerce platform are divided into multiple first-level categories, each first-level category is divided into multiple second-level categories, and each second-level category is divided into multiple third-level categories, then the first item category can be a fine-grained third-level category, and the second item category can be a coarse-grained second-level category or a first-level category.

[0093] As a possible implementation method, for each group of people, the second item category related to the group of people can be obtained based on the historical behavior information of the group of people through the term frequency-inverse document frequency TF-IDF algorithm. Among them, the TF-IDF algorithm is a common weighting technology used in information retrieval and data mining.

[0094] That is, step 501 can be implemented in the following manner: for each candidate item category, according to the historical behavior information of the crowd, the order quantity of the crowd for the items in the candidate item category and the total order quantity of the crowd are obtained; the total order quantity of all users and the total order quantity of all users for the items in the candidate item category are obtained; according to the order quantity of the crowd for the items in the candidate item category, the total order quantity of the crowd, the total order quantity of all users and the total order quantity of all users for the items in the candidate item category, the term frequency-inverse document frequency TF-IDF score of the candidate item category is determined; according to the TF-IDF score of each candidate item category, a second item category related to the crowd is determined from multiple candidate item categories.

[0095] Among them, the historical behavior information of the crowd is the operational behavior information of each user in the crowd in the past period of time, such as the information reported by each user on the e-commerce platform or the registered membership information, the information of the items clicked by each user on the e-commerce platform in the past period of time, the information of the items purchased by each user on the e-commerce platform in the past period of time, etc. This application does not impose any restrictions on this.

[0096] The term frequency-inverse document frequency TF-IDF score of each candidate item category can be obtained by the following formula (1):

[0097] TF-idf 1= user_cid1_ord / user_all_ord*log(cnt / cid1_cnt+1) (1)

[0098] Among them, TF-idf1 represents the TF-IDF score of the candidate item category; user_cid1_ord represents the number of orders placed by the group for items under the candidate item category; user_all_ord represents the total number of orders placed by the group; cnt represents the total number of orders placed by all users; cid1_cnt represents the total number of orders placed by all users for items under the candidate item category. Among them, cid1_cnt+1 can avoid the denominator being 0.

[0099] Specifically, the candidate item categories may be ranked in descending order of TF-IDF scores, and one or more candidate item categories ranked first may be determined as the second item category related to the crowd.

[0100] For example, the second item category related to mothers and babies may include a “mother and baby” category, a “toys and musical instruments” category, and a “books” category.

[0101] Step 502: Acquire multiple item attributes and corresponding attribute values ​​of items under the second item category.

[0102] Specifically, the set of item attributes and corresponding attribute values ​​of all items under the second item category may be determined as multiple item attributes and corresponding attribute values ​​of the items under the second item category. Among them, for the same item attribute, the set of attribute values ​​of all items under the second item category for the item attribute may be determined as the attribute value corresponding to the item attribute. For example, if the attribute value of the item attribute "applicable age" for a certain item under the second item category is "infant", and the attribute value of the item attribute "applicable age" for another item under the second item category is "child", then the attribute values ​​corresponding to the item attribute "applicable age" of the items under the second item category include "infant" and "child".

[0103] For example, taking an item under the "mother and baby" category, the following item attributes and corresponding attribute values ​​of the item can be obtained: brand: AAA; item name: AAA milk powder; number: 123456; gross weight: 480.00g; origin: BB country; stage: stage 1; packaging form: canned; net content: 0-400g; applicable age: 0-6 months. Then, the set of item attributes and corresponding attribute values ​​of all items under the "mother and baby" category can be determined as multiple item attributes and corresponding attribute values ​​of items under the "mother and baby" category.

[0104] It is understandable that the attribute values ​​of the same item attributes in different categories may be different. For example, the item attribute "applicable age" may have the following attribute values ​​in different categories: "1-3 years old", "1-4 years old", "2-5 years old", "2 years old", "over 3 years old", etc. In order to avoid duplication or contradiction in the attribute values ​​determined for the same item attribute, the different attribute values ​​corresponding to the same item attribute can be unified in advance to obtain non-overlapping and complete attribute values. For example, "1-3 years old", "1-4 years old", "2-5 years old", "2 years old", and "over 3 years old" can be unified into "1 year old", "2 years old", "3-6 years old", and "6-12 years old".

[0105] Step 503, according to the multiple item attributes and the corresponding attribute values, a target item attribute that influences the shopping decision of the crowd is obtained from the multiple item attributes, and the attribute value corresponding to the target item attribute is obtained.

[0106] As a possible implementation manner, step 503 can be implemented in the following manner: for each item attribute, in the case where there are multiple attribute values ​​corresponding to the item attribute, obtain the coverage rate of the item attribute in the items under the second item category, and the proportion of each attribute value corresponding to the item attribute in the items under the second item category in all attribute values ​​of the item attribute; according to the coverage rate of each item attribute in the items under the second item category, and the proportion of each attribute value corresponding to each item attribute in the items under the second item category in all attribute values ​​of the item attribute, obtain the target item attribute that affects the shopping decision of the crowd from the multiple item attributes.

[0107] The coverage rate of a certain item attribute in the items under the second item category may be a ratio of the number of items under the second item category having the item attribute to the total number of items under the second item category.

[0108] For example, taking the second item category "household supplies" as an example, the coverage rate of the item attribute "applicable population" in the items under the "household supplies" category is the ratio of the number of items with the item attribute "applicable population" under the "household supplies" category to the total number of items under the "household supplies" category.

[0109] Among them, the proportion of a certain attribute value corresponding to a certain item attribute in the items under the second item category in all attribute values ​​of the item attribute is the ratio of the number of items under the second item category that have the attribute value corresponding to the item attribute to the number of items under the second item category that have the item attribute.

[0110] For example, taking the second item category "household items", the item attribute "applicable population", and the attribute values ​​corresponding to the "applicable population" including "general", "children", and "adults" as an example, the ratio of the number of items under the "household items" category that have the attribute value "general" corresponding to the item attribute "applicable population" to the number of items under the "household items" category that have the item attribute "applicable population" can be determined as the proportion of the attribute value "general" corresponding to the item attribute "applicable population" in all attribute values ​​of the item attribute "applicable population" in the items under the "household items" category. Similarly, the ratio of the number of items under the "household items" category that have the attribute value "children" corresponding to the item attribute "applicable population" to the number of items under the "household items" category that have the item attribute "applicable population" can be determined as the proportion of the attribute value "children" corresponding to the item attribute "applicable population" in all attribute values ​​of the item attribute "applicable population" in the items under the "household items" category. The ratio of the number of items under the "household supplies" category that have the attribute value "adult" corresponding to the item attribute "applicable population" to the number of items under the "household supplies" category that have the item attribute "applicable population" can be determined as the proportion of the attribute value "adult" corresponding to the item attribute "applicable population" in all attribute values ​​of the item attribute "applicable population" in the items under the "household supplies" category.

[0111] As a possible implementation, the item attributes whose corresponding coverage is greater than a preset threshold and whose corresponding attribute values ​​account for a relatively uniform proportion of all attribute values ​​can be determined as target item attributes. The preset threshold can be set as required, and this application does not limit this.

[0112] For example, taking the second item category "household goods", the item attribute "applicable population", and the attribute values ​​corresponding to "applicable population" include "general", "children", and "adults" as an example, the coverage rate of the item attribute "applicable population" in the items under the "household goods" category can be determined by the above method, and the proportion of each attribute value "general", "children", and "adults" corresponding to the item attribute "applicable population" in all attribute values ​​of the item attribute "applicable population" in the items under the "household goods" category can be determined. When the proportions corresponding to "general", "children", and "adults" are 92%, 3%, and 5% respectively, since the proportion corresponding to "general" is much different from the proportions corresponding to "children" and "adults", the item attribute "applicable population" is not used as the target item attribute.

[0113] As a possible implementation method, the item attributes whose corresponding coverage rate among the item attributes is greater than a preset threshold and whose corresponding attribute value accounts for less than the preset attribute value threshold in all attribute values ​​can be determined as target item attributes. The preset attribute value threshold can be set as needed, and this application does not limit this.

[0114] For example, taking the second item category "household items", the item attribute "applicable scenarios", and the attribute values ​​corresponding to "applicable scenarios" include "bedroom", "living room", and "kitchen" as an example, assuming that the preset attribute value threshold is 50%, the coverage rate of the item attribute "applicable scenarios" in the items under the "household items" category can be determined in the above manner, and the proportion of each attribute value "bedroom", "living room", and "kitchen" corresponding to the item attribute "applicable scenarios" in all attribute values ​​of the item attribute "applicable scenarios" in the items under the "household items" category can be determined. When the corresponding proportions of "bedroom", "living room", and "kitchen" are 45%, 30%, and 25% respectively, since the proportion of each attribute value in all attribute values ​​does not exceed 50%, the item attribute "applicable scenarios" can be used as the target item attribute.

[0115] Furthermore, after the attributes of the target item are determined, the attributes of the target item may be provided to business experts for review to improve the accuracy of the attributes of the target item.

[0116] Step 504: Obtain a second demand keyword for describing the shopping demand of a group of people and a first item category corresponding to the second demand keyword.

[0117] As a possible implementation method, starting from the item titles of items frequently purchased by the crowd, the second demand keyword used to describe the shopping needs of the crowd and the first item category corresponding to the second demand keyword can be obtained.

[0118] That is, step 504 can be implemented in the following manner: obtain multiple active item titles under the second item category related to the crowd; input each active item title into a pre-trained language model to obtain a vector representation of each active item title; cluster each active item title according to the vector representation of each active item title to obtain at least one title set; extract demand keywords from the active item titles included in the at least one title set to obtain second demand keywords.

[0119] Among them, the active item titles are the item titles that have been clicked by users in the recent period.

[0120] Among them, the pre-trained language model can be BERT (Bidirectional Encoder Representations from Transformers, a bidirectional encoding representation language model) or other models that can obtain vector representations, and this application does not limit this. The pre-trained language model in the embodiment of the present application can be a BERT model that has been fine-tuned based on e-commerce data.

[0121] In the embodiment of the present application, the DBSCAN (Density-Based Spatial Clustering of Applications with Noise) method can be used to cluster the titles of each active item. Since the DBSCAN method can automatically determine the number of clusters of active item titles and has a certain outlier detection capability, the accuracy of clustering can be improved.

[0122] It should be noted that due to the large number of active item titles and the large dimension of the vector representation output by the BERT model, the clustering speed may be slow and the clustering may be uneven in size, such as a class with smaller categories. To avoid this problem, the vector representation of the active item titles output by the pre-trained language model can be reduced in dimension, and then the active item titles can be clustered based on the reduced vector representation to obtain at least one title set.

[0123] Among them, the UMAP (Uniform Manifold Approximation Projection) algorithm can be used to reduce the dimension of the vector representation output by the pre-trained language model. Using the UMAP algorithm for dimensionality reduction can better preserve the global structure of the data, and the dimensionality reduction efficiency is high, which can be better combined with the clustering algorithm.

[0124] In an embodiment of the present application, the KeyBert model can be used to extract the demand keywords from the active item titles included in each title set to obtain the second demand keywords. The KeyBert model is a small and easy-to-use keyword extraction technology that uses BERT embedding to create keywords and keyword phrases that are most similar to the document. One or more second demand keywords can be extracted from each title set.

[0125] In addition, in order to ensure the accuracy of the second demand keywords, the extracted second demand keywords may also be provided to manual review and rewriting to obtain the final second demand keywords.

[0126] In the embodiment of the present application, after determining the second demand keyword, the items that can meet the shopping needs described by the second demand keyword can be determined, and then the first item category to which these items belong can be determined as the first item category corresponding to the second demand keyword.

[0127] For example, the first item categories corresponding to the second demand keywords “health monitoring” and “anti-lost” as shown in Table 2 below can be determined in the above manner.

[0128] Table 2 Second demand keywords and corresponding first item categories

[0129]

[0130] As another possible implementation manner, starting from the search words input by the crowd, the second demand keyword used to describe the shopping demand of the crowd and the first item category corresponding to the second demand keyword can be obtained.

[0131] That is, step 504 can be implemented in the following manner: obtain multiple search terms searched by the crowd within a preset time period, and the search frequency of the search terms within the preset time period is greater than a frequency threshold; for each search term, input the item title under the search term whose click volume exceeds the click volume threshold into the pre-trained language model to obtain a vector representation of the item title, and obtain the vector representation of the search term based on the vector representation of the item title and the click volume of the item title; cluster the search terms based on the vector representation of each search term to obtain at least one search term set; extract demand keywords for the search terms included in at least one search term set to obtain a second demand keyword.

[0132] Among them, the preset time period can be set according to the needs of the actual scenario, and this application does not impose any restrictions on this.

[0133] There may be multiple item titles whose click volume exceeds a click volume threshold under each search term, and each item title may be input into a pre-trained language model to obtain a vector representation of the item title.

[0134] The vector representation of the search term can be obtained by the following formula (2):

[0135]

[0136] Where E represents the vector representation of the search term; c i represents the number of clicks on the i-th item title among the item titles whose clicks on the search term exceed the click threshold; C represents the total number of clicks on the item titles whose clicks on the search term exceed the click threshold; e i Indicates the vector representation of the i-th item title among the item titles whose click volume under the search term exceeds the click volume threshold.

[0137] In the embodiment of the present application, the DBSCAN method can be used to cluster the search terms. Since the DBSCAN method can automatically determine the number of clusters of the search terms and has a certain outlier detection capability, the accuracy of clustering can be improved.

[0138] It should be noted that due to the large number of search terms and the large dimension of the vector representation output by the BERT model, the clustering speed may be slow and the clustering class size may be uneven, such as a class with smaller categories. To avoid this problem, the vector representation of the search terms can be reduced in dimension first, and then the search terms can be clustered based on the reduced vector representation to obtain at least one search term set.

[0139] Among them, the UMAP algorithm can be used to reduce the dimension of the vector representation of the search term. The UMAP algorithm can better preserve the global structure of the data, and the dimensionality reduction efficiency is high, which can be better combined with the clustering algorithm.

[0140] In the embodiment of the present application, the TF-IDF algorithm can be used to extract the demand keywords from the search words included in each search word set to obtain the second demand keywords. One or more second demand keywords can be extracted from each search word set.

[0141] For each search word set, the TF-IDF score of each search word therein may be calculated, and one or more search words with the highest score may be used as the second demand keyword.

[0142] The TF-IDF score of each search term in each search term set can be obtained by the following formula (3):

[0143] TF-idf 2=N1 / N2*log(N3 / N4) (3)

[0144] Among them, TF-idf 2 represents the TF-IDF score of each search term in the search term set; N1 represents the number of the search term in the search term set; N2 represents the total number of search terms in the search term set; N3 represents the total number of all search terms in all search term sets; N4 represents the number of the search term in all search term sets.

[0145] In addition, in order to ensure the accuracy of the second demand keywords, the extracted second demand keywords may also be provided to manual review and rewriting to obtain the final second demand keywords.

[0146] In the embodiment of the present application, after the second demand keyword is determined, the first item category to which the items whose click volume exceeds the click volume threshold under the second demand keyword belong can be determined as the first item category corresponding to the second demand keyword.

[0147] As another possible implementation method, the above two methods can be combined to obtain the second demand keyword used to describe the shopping needs of the crowd and the first item category corresponding to the second demand keyword based on the item titles of the items frequently purchased by the crowd and the search terms input by the crowd.

[0148] It should be noted that when using the above two methods to obtain the second demand keywords used to describe the shopping needs of the crowd and the first item category corresponding to the second demand keywords, there may be a problem that the second demand keywords determined by the above two methods are similar. In the embodiment of the present application, similar second demand keywords can also be merged.

[0149] Specifically, at least two second demand keywords may be merged according to the overlap degree of the first item categories corresponding to the at least two second demand keywords. For example, when the overlap degree of the first item categories corresponding to the at least two second demand keywords exceeds a preset overlap threshold, the at least two second demand keywords may be merged. Any second demand keyword before the merge is used as the merged second demand keyword.

[0150] The degree of overlap of the first item categories corresponding to at least two second demand keywords is the ratio of the number of the same first item categories corresponding to the at least two second demand keywords to the total number of the first item categories corresponding to the at least two second demand keywords.

[0151] For example, assuming that starting from the item titles of items that people often buy, the second demand keywords and the first item categories corresponding to the second demand keywords are obtained as shown in the second row of Table 3 below, and starting from the search terms input by the people, the second demand keywords and the first item categories corresponding to the second demand keywords are obtained as shown in the third row of Table 3 below. Since the first item categories corresponding to the two second demand keywords have a high degree of overlap, the two second demand keywords can be merged, for example, the merged second demand keywords are determined to be "exquisite sleep" or "storage", and the union of the first item categories corresponding to the two second demand keywords is determined as the first item category corresponding to the merged second demand keywords.

[0152] Table 3 Second demand keywords and corresponding first item categories

[0153]

[0154] Step 505: Combine the attribute value corresponding to the attribute of the target item with the second demand keyword to obtain the first demand keyword corresponding to the group of people.

[0155] Step 506, for any first demand keyword corresponding to the population, based on the coverage rate of the attribute values ​​constituting the first demand keyword in the items under the target item category, the target item category is screened to obtain the first item category corresponding to the first demand keyword, wherein the target item category is any first item category corresponding to the second demand keyword constituting the first demand keyword.

[0156] It should be noted that steps 501-506 are operations performed on each group of people.

[0157] Step 507: Generate a demand scenario graph based on the first demand keywords corresponding to the multiple groups of people and the first item categories corresponding to the first demand keywords.

[0158] The specific implementation process and principle of steps 505-507 may refer to the description of other embodiments and will not be repeated here.

[0159] Thus, a demand scenario map can be generated, which includes first demand keywords corresponding to multiple groups of people and first item categories corresponding to the first demand keywords. Then, based on the demand scenario map, item recommendations can be made for users, thereby improving the accuracy of recommendation results and user experience.

[0160] In actual applications, item recommendations can be made for users in different types of applications based on the demand scenario graph. For example, when a user logs into a certain application, one or more target demand keywords can be displayed on the homepage according to the target group to which the user belongs and the corresponding target demand keywords, and after the user clicks on a target demand keyword, an aggregate page of multiple items under the first item category corresponding to the target demand keyword can be displayed.

[0161] In actual applications, the business side of the e-commerce platform can also be guided to optimize the supply of materials (such as short videos and topics) based on the demand scenario map. For example, the supply and conversion efficiency of different groups of people corresponding to different materials can be analyzed, and targeted optimization can be performed for materials with low conversion rates for a certain group of people, and timely supplementation can be performed for materials with low supply for a certain group of people.

[0162] Figure 6 This is a schematic diagram of the structure of the item recommendation device provided in Example 4 of the present application.

[0163] like Figure 6 As shown, the item recommendation device 600 may include: a first acquisition module 610 , a first determination module 620 and a second determination module 630 .

[0164] The first acquisition module 610 is used to acquire historical behavior information of target users and a pre-generated demand scenario map, wherein the demand scenario map includes first demand keywords corresponding to multiple groups of people and first item categories corresponding to the first demand keywords, and the first demand keywords are used to describe a type of shopping demand of the corresponding group of people;

[0165] The first determination module 620 is used to determine the target group to which the target user belongs from multiple groups according to the historical behavior information, and determine the target demand keyword corresponding to the target user from the first demand keyword corresponding to the target group;

[0166] The second determination module 630 is used to determine the to-be-recommended items for the target user according to the first item category corresponding to the target demand keyword.

[0167] It should be noted that the item recommendation device provided in the embodiment of the present application can execute the item recommendation method of the aforementioned embodiment. The item recommendation device can be an electronic device or can be configured in an electronic device to recommend items to users according to their needs by executing the item recommendation method provided in the embodiment of the present application, thereby improving the accuracy of the recommendation results and improving the user experience.

[0168] Among them, the electronic device can be a PC, cloud device, mobile device, server, etc., and the mobile device can be, for example, a mobile phone, tablet computer, personal digital assistant, wearable device, vehicle-mounted device, or any other hardware device, and this application does not impose any restrictions on this.

[0169] It should be noted that the explanations in the aforementioned item recommendation method embodiment are also applicable to the item recommendation device of this embodiment and will not be repeated here.

[0170] The item recommendation device of the embodiment of the present application obtains the historical behavior information of the target user and a pre-generated demand scenario map, wherein the demand scenario map includes first demand keywords corresponding to multiple groups of people and first item categories corresponding to the first demand keywords, and the first demand keywords are used to describe a type of shopping demand of the corresponding group of people. According to the historical behavior information, the target group to which the target user belongs is determined from multiple groups of people, and the target demand keywords corresponding to the target user are determined from the first demand keywords corresponding to the target group. According to the first item category corresponding to the target demand keywords, the target user's to-be-recommended items are determined, thereby achieving the goal of recommending items to users according to their needs, improving the accuracy of recommendation results, and improving user experience.

[0171] Figure 7 This is a schematic diagram of the structure of the item recommendation device provided in Example 5 of the present application.

[0172] like Figure 7As shown, the item recommendation device 700 may include: a first acquisition module 710, a first determination module 720 and a second determination module 730. Figure 7 The first acquisition module 710, the first determination module 720 and the second determination module 730 are shown in FIG. Figure 6 The first acquisition module 610, the first determination module 620 and the second determination module 630 shown have the same structure and function.

[0173] like Figure 7 As shown, the item recommendation device 700 may further include:

[0174] The second acquisition module 740 is used to acquire, for each group of people, target item attributes that affect the shopping decision of the group of people according to the historical behavior information of the group of people, and acquire attribute values ​​corresponding to the target item attributes;

[0175] The third acquisition module 750 is used to acquire a second demand keyword used to describe the shopping demand of a group of people and a first item category corresponding to the second demand keyword;

[0176] The third determination module 760 is used to determine the first demand keyword corresponding to the group of people based on the attribute value corresponding to the target item attribute and the second demand keyword, and determine the first item category corresponding to the first demand keyword based on the attribute value corresponding to the target item attribute and the first item category corresponding to the second demand keyword;

[0177] The generation module 770 is used to generate a demand scenario graph based on the first demand keywords corresponding to multiple groups of people and the first item categories corresponding to the first demand keywords.

[0178] In a possible implementation of the embodiment of the present application, the second acquisition module 740 includes:

[0179] a determination unit, configured to determine, based on historical behavior information of the crowd, a second item category related to the crowd from a plurality of candidate item categories, wherein a subcategory of the candidate item category is a category at the same level as the first item category;

[0180] A first acquisition unit, used to acquire multiple item attributes and corresponding attribute values ​​of items under the second item category;

[0181] The second acquisition unit is used to acquire target item attributes that influence the shopping decision of a group of people from the multiple item attributes according to the multiple item attributes and the corresponding attribute values.

[0182] In a possible implementation manner of the embodiment of the present application, the determining unit is specifically configured to:

[0183] For each candidate item category, based on the historical behavior information of the crowd, obtain the crowd's order quantity for items in the candidate item category and the crowd's total order quantity;

[0184] Obtain the total order quantity of all users and the total order quantity of all users for items under the candidate item category;

[0185] Determine the term frequency-inverse document frequency (TF-IDF) score of the candidate item category based on the number of orders placed by the crowd for items under the candidate item category, the total number of orders placed by the crowd, the total number of orders placed by all users, and the total number of orders placed by all users for items under the candidate item category;

[0186] According to the TF-IDF score of each candidate item category, a second item category related to the crowd is determined from the multiple candidate item categories.

[0187] In a possible implementation manner of the embodiment of the present application, the second acquisition unit is specifically configured to:

[0188] For each item attribute, if there are multiple attribute values ​​corresponding to the item attribute, obtain the coverage rate of the item attribute in the items under the second item category, and the proportion of each attribute value corresponding to the item attribute in all attribute values ​​of the item attribute in the items under the second item category;

[0189] According to the coverage rate of each item attribute in the items under the second item category and the proportion of each attribute value corresponding to each item attribute in the items under the second item category in all attribute values ​​of the item attributes, target item attributes that influence people's shopping decisions are obtained from multiple item attributes.

[0190] In a possible implementation of the embodiment of the present application, the third acquisition module 750 is specifically configured to:

[0191] Get multiple active item titles under a second item category related to the crowd;

[0192] Input each active item title into the pre-trained language model to obtain a vector representation of each active item title;

[0193] Clustering the titles of each active item according to the vector representation of the title of each active item to obtain at least one title set;

[0194] Demand keywords are extracted from active item titles included in at least one title set to obtain second demand keywords.

[0195] In a possible implementation of the embodiment of the present application, the third acquisition module 750 is specifically configured to:

[0196] Acquire multiple search terms searched by a group of people within a preset time period, where the search frequency of the search terms within the preset time period is greater than a frequency threshold;

[0197] For each search term, the item titles whose click volume exceeds the click volume threshold under the search term are input into the pre-trained language model to obtain the vector representation of the item title, and the vector representation of the search term is obtained based on the vector representation of the item title and the click volume of the item title;

[0198] Clustering the search terms according to the vector representation of each search term to obtain at least one search term set;

[0199] Demand keywords are extracted from the search words included in at least one search word set to obtain second demand keywords.

[0200] In a possible implementation of the embodiment of the present application, the third determining module 760 is specifically configured to:

[0201] Combine the attribute value corresponding to the target item attribute with the second demand keyword to obtain the first demand keyword corresponding to the group of people;

[0202] For any first demand keyword corresponding to a population, the target item category is screened based on the coverage rate of the attribute values ​​constituting the first demand keyword in the items under the target item category to obtain the first item category corresponding to the first demand keyword, wherein the target item category is any first item category corresponding to the second demand keyword constituting the first demand keyword.

[0203] It should be noted that the item recommendation device provided in the embodiment of the present application can execute the item recommendation method of the aforementioned embodiment. The item recommendation device can be an electronic device or can be configured in an electronic device to recommend items to users according to their needs by executing the item recommendation method provided in the embodiment of the present application, thereby improving the accuracy of the recommendation results and improving the user experience.

[0204] Among them, the electronic device can be a PC, cloud device, mobile device, server, etc., and the mobile device can be, for example, a mobile phone, tablet computer, personal digital assistant, wearable device, vehicle-mounted device, or any other hardware device, and this application does not impose any restrictions on this.

[0205] It should be noted that the explanations in the aforementioned item recommendation method embodiment are also applicable to the item recommendation device of this embodiment and will not be repeated here.

[0206] The item recommendation device of the embodiment of the present application obtains the historical behavior information of the target user and a pre-generated demand scenario map, wherein the demand scenario map includes first demand keywords corresponding to multiple groups of people and first item categories corresponding to the first demand keywords, and the first demand keywords are used to describe a type of shopping demand of the corresponding group of people. According to the historical behavior information, the target group to which the target user belongs is determined from multiple groups of people, and the target demand keywords corresponding to the target user are determined from the first demand keywords corresponding to the target group. According to the first item category corresponding to the target demand keywords, the target user's to-be-recommended items are determined, thereby achieving the goal of recommending items to users according to their needs, improving the accuracy of recommendation results, and improving user experience.

[0207] In order to implement the above embodiments, the present application also proposes an electronic device, comprising: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the item recommendation method proposed in any of the aforementioned embodiments of the present application.

[0208] In order to implement the above embodiments, the present application also proposes a non-transitory computer-readable storage medium storing computer instructions, wherein the computer instructions are used to enable the computer to execute the item recommendation method proposed in any of the aforementioned embodiments of the present application.

[0209] In order to implement the above embodiments, the present application also proposes a computer program product, including a computer program, which, when executed by a processor, implements the item recommendation method proposed in any of the above embodiments of the present application.

[0210] Figure 8 A block diagram of an exemplary electronic device suitable for implementing the embodiments of the present application is shown. Figure 8 The electronic device 800 shown is merely an example and should not bring any limitation to the functions and scope of use of the embodiments of the present application.

[0211] like Figure 8 As shown, the electronic device 800 is in the form of a general purpose computing device. The components of the electronic device 800 may include, but are not limited to: one or more processors or processing units 16, a system memory 28, and a bus 18 connecting different system components (including the system memory 28 and the processing unit 16).

[0212] The bus 18 represents one or more of several types of bus structures, including a memory bus or memory controller, a peripheral bus, a graphics acceleration port, a processor or a local bus using any of a variety of bus structures. For example, these architectures include but are not limited to Industry Standard Architecture (ISA) bus, Micro Channel Architecture (MAC) bus, Enhanced ISA bus, Video Electronics Standards Association (VESA) local bus and Peripheral Component Interconnection (PCI) bus.

[0213] The electronic device 800 typically includes a variety of computer system readable media. These media can be any available media that can be accessed by the electronic device 800, including volatile and non-volatile media, removable and non-removable media.

[0214] The memory 28 may include computer system readable media in the form of volatile memory, such as random access memory (RAM) 30 and / or cache memory 32. The electronic device 800 may further include other removable / non-removable, volatile / non-volatile computer system storage media. By way of example only, the storage system 34 may be used to read and write non-removable, non-volatile magnetic media ( Figure 8 not shown, usually called a "hard drive"). Although Figure 8 Not shown, a disk drive for reading and writing a removable non-volatile disk (e.g., a "floppy disk"), and an optical disk drive for reading and writing a removable non-volatile optical disk (e.g., a compact disc read only memory (CD-ROM), a digital versatile disc read only memory (DVD-ROM), or other optical media) may be provided. In these cases, each drive may be connected to the bus 18 via one or more data medium interfaces. The memory 28 may include at least one program product having a set (e.g., at least one) of program modules that are configured to perform the functions of the various embodiments of the present application.

[0215] A program / utility 40 having a set (at least one) of program modules 42 may be stored, for example, in the memory 28, such program modules 42 including, but not limited to, an operating system, one or more application programs, other program modules, and program data, each of which or some combination may include an implementation of a network environment. The program modules 42 generally perform the functions and / or methods of the embodiments described herein.

[0216] The electronic device 800 may also communicate with one or more external devices 8 (e.g., keyboards, pointing devices, displays 24, etc.), one or more devices that enable a user to interact with the electronic device 800, and / or any device that enables the electronic device 800 to communicate with one or more other computing devices (e.g., a network card, a modem, etc.). Such communication may be performed through an input / output (I / O) interface 22. Furthermore, the electronic device 800 may also communicate with one or more networks (e.g., a local area network (LAN), a wide area network (WAN), and / or a public network, such as the Internet) through a network adapter 20. As shown, the network adapter 20 communicates with other modules of the electronic device 800 through a bus 18. It should be understood that although Figure 8 Not shown, other hardware and / or software modules may be used in conjunction with electronic device 800, including but not limited to: microcode, device drivers, redundant processing units, external disk drive arrays, RAID systems, tape drives, and data backup storage systems.

[0217] The processing unit 16 executes various functional applications and data processing by running programs stored in the system memory 28, such as implementing the methods mentioned in the above embodiments.

[0218] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "example", "specific example", or "some examples" etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present application. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner. In addition, those skilled in the art may combine and combine the different embodiments or examples described in this specification and the features of the different embodiments or examples, without contradiction.

[0219] In addition, the terms "first" and "second" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined as "first" and "second" may explicitly or implicitly include at least one of the features. In the description of this application, the meaning of "plurality" is at least two, such as two, three, etc., unless otherwise clearly and specifically defined.

[0220] Any process or method description in a flowchart or otherwise described herein may be understood to represent a module, fragment or portion of code comprising one or more executable instructions for implementing the steps of a custom logical function or process, and the scope of the preferred embodiments of the present application includes alternative implementations in which functions may not be performed in the order shown or discussed, including performing functions in a substantially simultaneous manner or in the reverse order depending on the functions involved, which should be understood by technicians in the technical field to which the embodiments of the present application belong.

[0221] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as an ordered list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by an instruction execution system, device or apparatus (such as a computer-based system, a system including a processor, or other system that can fetch instructions from an instruction execution system, device or apparatus and execute the instructions), or in combination with these instruction execution systems, devices or apparatuses. For the purpose of this specification, "computer-readable medium" can be any device that can contain, store, communicate, propagate or transmit a program for use by an instruction execution system, device or apparatus, or in combination with these instruction execution systems, devices or apparatuses. More specific examples of computer-readable media (a non-exhaustive list) include the following: an electrical connection with one or more wires (electronic device), a portable computer disk box (magnetic device), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), a fiber optic device, and a portable compact disk read-only memory (CDROM). In addition, the computer-readable medium may even be paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium and then editing, interpreting or processing in other suitable ways if necessary, and then stored in a computer memory.

[0222] It should be understood that the various parts of the present application can be implemented by hardware, software, firmware or a combination thereof. In the above-mentioned embodiments, multiple steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, as in another embodiment, it can be implemented by any one of the following technologies known in the art or their combination: a discrete logic circuit having a logic gate circuit for implementing a logic function for a data signal, a dedicated integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

[0223] A person skilled in the art may understand that all or part of the steps in the method for implementing the above-mentioned embodiment may be completed by instructing related hardware through a program, and the program may be stored in a computer-readable storage medium, which, when executed, includes one or a combination of the steps of the method embodiment.

[0224] In addition, each functional unit in each embodiment of the present application may be integrated into a processing module, or each unit may exist physically separately, or two or more units may be integrated into one module. The above-mentioned integrated module may be implemented in the form of hardware or in the form of a software functional module. If the integrated module is implemented in the form of a software functional module and sold or used as an independent product, it may also be stored in a computer-readable storage medium.

[0225] The storage medium mentioned above may be a read-only memory, a magnetic disk or an optical disk, etc. Although the embodiments of the present application have been shown and described above, it can be understood that the above embodiments are exemplary and cannot be understood as limiting the present application. A person of ordinary skill in the art may change, modify, replace and modify the above embodiments within the scope of the present application.

Claims

1. An item recommendation method, characterized in that: The method comprises: Acquire historical behavior information of target users and a pre-generated demand scenario map, wherein the demand scenario map includes first demand keywords corresponding to multiple groups of people and first item categories corresponding to the first demand keywords, and the first demand keywords are used to describe a type of shopping demand of the corresponding group of people; Determine, from the multiple groups, a target group to which the target user belongs according to the historical behavior information, and determine, from the first demand keyword corresponding to the target group, a target demand keyword corresponding to the target user; Determine the to-be-recommended item for the target user according to the first item category corresponding to the target demand keyword.

2. The method according to claim 1, characterized in that The method further comprises: For each of the groups of people, obtaining target item attributes that affect the shopping decisions of the group of people based on the historical behavior information of the group of people, and obtaining attribute values ​​corresponding to the target item attributes; Acquire a second demand keyword for describing the shopping demand of the group of people and a first item category corresponding to the second demand keyword; Determine the first demand keyword corresponding to the group based on the attribute value corresponding to the target item attribute and the second demand keyword, and determine the first item category corresponding to the first demand keyword based on the attribute value corresponding to the target item attribute and the first item category corresponding to the second demand keyword; The demand scenario graph is generated based on the first demand keywords corresponding to the plurality of groups of people and the first item categories corresponding to the first demand keywords.

3. The method according to claim 2, characterized in that The step of obtaining target item attributes that influence the shopping decisions of the group of people based on the historical behavior information of the group of people includes: Determine, according to the historical behavior information of the group of people, a second item category related to the group of people from a plurality of candidate item categories, wherein a subcategory of the candidate item category is a category at the same level as the first item category; Obtaining multiple item attributes and corresponding attribute values ​​of items under the second item category; According to the multiple item attributes and corresponding attribute values, a target item attribute that influences the shopping decision of the group of people is obtained from the multiple item attributes.

4. The method according to claim 3, characterized in that The determining, based on the historical behavior information of the group of people, a second item category related to the group of people from a plurality of candidate item categories includes: For each of the candidate item categories, obtaining the order quantity of the items under the candidate item category by the group of people and the total order quantity of the group of people according to the historical behavior information of the group of people; Obtaining a total order quantity of all users and a total order quantity of items under the candidate item category by all users; Determine a TF-IDF score of the candidate item category according to the order quantity of the group of people for items under the candidate item category, the total order quantity of the group of people, the total order quantity of all users, and the total order quantity of all users for items under the candidate item category; According to the TF-IDF score of each of the candidate item categories, a second item category related to the group of people is determined from the multiple candidate item categories.

5. The method according to claim 3, characterized in that: The step of obtaining target item attributes that influence the shopping decision of the group of people from the multiple item attributes according to the multiple item attributes and the corresponding attribute values ​​includes: For each of the item attributes, if there are multiple attribute values ​​corresponding to the item attribute, obtain the coverage rate of the item attribute in the items under the second item category, and the proportion of each of the attribute values ​​corresponding to the item attribute in the items under the second item category in all the attribute values ​​of the item attribute; According to the coverage rate of each of the item attributes in the items under the second item category, and the proportion of each of the attribute values ​​corresponding to each of the item attributes in the items under the second item category in all the attribute values ​​of the item attributes, the target item attributes that influence the shopping decision of the group of people are obtained from the multiple item attributes.

6. The method according to claim 3, characterized in that: The obtaining of a second demand keyword for describing the shopping demand of the group of people and a first item category corresponding to the second demand keyword includes: Obtain a plurality of active item titles under a second item category related to the population; Inputting each of the active item titles into a pre-trained language model to obtain a vector representation of each of the active item titles; Clustering the active item titles according to the vector representation of the active item titles to obtain at least one title set; Demand keywords are extracted from the active item titles included in the at least one title set to obtain the second demand keywords.

7. The method according to claim 3, characterized in that The obtaining of a second demand keyword for describing the shopping demand of the group of people and a first item category corresponding to the second demand keyword includes: Acquire multiple search terms searched by the group of people within a preset time period, wherein the search frequency of the search terms within the preset time period is greater than a frequency threshold; For each of the search terms, the title of an item whose click volume exceeds a click volume threshold under the search term is input into a pre-trained language model to obtain a vector representation of the item title, and the vector representation of the search term is obtained according to the vector representation of the item title and the click volume of the item title; Clustering the search terms according to the vector representation of each search term to obtain at least one search term set; Demand keywords are extracted from the search words included in the at least one search word set to obtain the second demand keywords.

8. The method according to any one of claims 2 to 7, characterized in that: The determining, based on the attribute value corresponding to the target item attribute and the second demand keyword, the first demand keyword corresponding to the group, and the determining, based on the attribute value corresponding to the target item attribute and the first item category corresponding to the second demand keyword, the first item category corresponding to the first demand keyword, includes: Combining the attribute value corresponding to the attribute of the target item with the second demand keyword to obtain the first demand keyword corresponding to the group of people; For any first demand keyword corresponding to the population, the target item category is screened based on the coverage rate of the attribute values ​​constituting the first demand keyword in the items under the target item category to obtain the first item category corresponding to the first demand keyword, wherein the target item category is any first item category corresponding to the second demand keyword constituting the first demand keyword.

9. An item recommendation device, characterized in that: The device comprises: A first acquisition module is used to acquire historical behavior information of target users and a pre-generated demand scenario map, wherein the demand scenario map includes first demand keywords corresponding to multiple groups of people and first item categories corresponding to the first demand keywords, and the first demand keywords are used to describe a type of shopping demand of the corresponding group of people; A first determination module is used to determine the target group to which the target user belongs from the multiple groups according to the historical behavior information, and to determine the target demand keyword corresponding to the target user from the first demand keyword corresponding to the target group; The second determination module is used to determine the to-be-recommended items for the target user according to the first item category corresponding to the target demand keyword.

10. An electronic device, characterized in that: include: at least one processor; as well as a memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the method according to any one of claims 1 to 8.

11. A non-transitory computer-readable storage medium storing computer instructions, characterized in that: The computer instructions are used to cause the computer to execute the method according to any one of claims 1 to 8.

Images