Search recommendation method, apparatus and device

Abstract

The application discloses a search recommendation method, device and equipment. The method receives a search request of a user, the search request carrying a user identifier and a search word; determines whether a historical user behavior corresponding to the user identifier is related to the search word according to the search request; if there is no historical user behavior, or there is a historical user behavior but the historical user behavior is not related to the search word, a user group corresponding to the user identifier is determined; and a recommendation result corresponding to the search word is determined according to the user group. In this way, the commodity interaction behavior of the user group to which the low-activity user (including a new user) belongs and which is related to the search is used to recall the commodity that the low-activity user who has no or few commodity interaction behaviors related to the search likes, so that a personalized search recommendation mode for a general group is realized, and therefore, the conversion efficiency and the user experience can be improved.

Description

Technical Field

[0001] This application relates to the field of e-commerce technology, specifically to a method and apparatus for searching and recommending business objects, as well as electronic devices. Background Technology

[0002] With the increasing maturity of the internet and information technology, e-commerce has been widely applied in the socio-economic field. Personalized recall is an important means to improve conversion efficiency and enhance user experience in e-commerce platform search scenarios.

[0003] E-commerce platforms include cross-border e-commerce platforms (such as AliExpress of Alibaba) and non-cross-border e-commerce platforms (such as Taobao of Alibaba). Currently, cross-border e-commerce platforms typically employ the same personalized search strategy as non-cross-border e-commerce platforms. This strategy determines personalized product search results corresponding to the current search term for users who have recently interacted with products related to the current search, based on that user's recent product interaction data. For users who have not recently interacted with products related to the current search, or have had little to no product interaction, the personalized search strategy is not implemented; instead, product search results are directly determined based solely on the current search term. Because users on non-cross-border e-commerce platforms have higher user activity, this strategy achieves higher accuracy in recommending business targets.

[0004] However, in the process of developing this invention, the inventors discovered that the existing technology has at least the following problems: the solution can only provide personalized recall services for users with a certain level of activity, and is therefore more suitable for product search scenarios on non-cross-border e-commerce platforms. However, due to the characteristics of cross-border e-commerce platforms, such as wide user distribution, large differences, low activity levels, and varying privacy protection policies in different countries, a large number of international users do not have relevant product interaction behavior. Therefore, this solution cannot recall products that these users prefer, which will affect both user experience and the transaction rate and UV conversion rate of cross-border e-commerce platforms. In summary, how to provide personalized product search recommendation services for low-activity international users to improve the accuracy of search results, thereby improving user experience, transaction rate, and UV conversion rate, has become an urgent problem that needs to be solved by those skilled in the art. Summary of the Invention

[0005] This application provides a search recommendation method to address the problem that existing technologies cannot provide personalized product search recommendation services for inactive international users. This application also provides a search recommendation device and an electronic device.

[0006] This application provides a search recommendation method, including:

[0007] Receive a user's search request, the search request carrying a user identifier and search terms;

[0008] Based on the search request, determine whether the historical user behavior corresponding to the user identifier is related to the search term;

[0009] If there is no historical user behavior, or if there is historical user behavior but it is not related to the search term, determine the user group corresponding to the user identifier;

[0010] The recommended results corresponding to the search terms are determined based on the user group.

[0011] Optionally, determining the recommended results corresponding to the search term based on the user group includes:

[0012] Obtain the first user behavior related to the search term from the historical user behavior corresponding to the user group;

[0013] Determine the recommendation result for the first user behavior;

[0014] The recommendation result of the first user behavior is used as the recommendation result corresponding to the search term.

[0015] Optionally, obtaining the first user behavior related to the search term from the historical user behavior corresponding to the user group includes:

[0016] Obtain the probability distribution of behaviors related to the search term in the historical user behavior of the user group;

[0017] The behavior with the highest probability is selected as the first user behavior from the behavior probability distribution.

[0018] Optionally, obtaining the first user behavior related to the search term from the historical user behavior corresponding to the user group includes:

[0019] Obtain evaluation metrics related to the search term from the historical user behavior of the user group, including but not limited to: total transaction volume, transaction conversion rate, and repurchase rate;

[0020] Select the user behavior corresponding to the evaluation indicator that reaches or exceeds the target value as the first user behavior.

[0021] Optionally, obtaining the first user behavior related to the search term from the historical user behavior corresponding to the user group includes:

[0022] Obtain behavioral metrics related to the search term from the historical user behavior of the user group, including but not limited to: adding to cart, favorites, browsing, forwarding, recommending, and transactions;

[0023] Select the user behavior that meets the aforementioned behavioral indicators as the first user behavior.

[0024] Optionally, determining the user group corresponding to the user identifier includes:

[0025] Based on the user identifier, multiple users who share at least one of the following characteristics are identified as a user group: geographical location, country or region, ethnicity, age group, and hobbies.

[0026] This application provides a search recommendation method, including:

[0027] Send the user's search request to the server, the search request carrying the user identifier and search terms;

[0028] Display the recommended results corresponding to the search terms sent back by the server;

[0029] The server determines the recommendation result using the following steps: determining whether the historical user behavior corresponding to the user identifier is related to the search term; if there is no historical user behavior, or if there is historical user behavior but it is not related to the search term, determining the user group corresponding to the user identifier; and determining the recommendation result corresponding to the search term based on the user group.

[0030] This application provides a search recommendation device, including:

[0031] A request receiving unit is used to receive a user's search request, wherein the search request carries a user identifier and search terms;

[0032] The judgment unit is used to determine, based on the search request, whether the historical user behavior corresponding to the user identifier is related to the search term;

[0033] The user group determination unit is used to determine the user group corresponding to the user identifier if there is no historical user behavior, or if there is historical user behavior but it is not related to the search term.

[0034] The recommendation result determination unit is used to determine the recommendation result corresponding to the search term based on the user group.

[0035] This application provides a search recommendation device, including:

[0036] The request sending unit is used to send a user's search request to the server, wherein the search request carries a user identifier and search terms;

[0037] The recommendation results display unit is used to display the recommendation results corresponding to the search terms sent back by the server;

[0038] The server determines the recommendation result using the following steps: determining whether the historical user behavior corresponding to the user identifier is related to the search term; if there is no historical user behavior, or if there is historical user behavior but it is not related to the search term, determining the user group corresponding to the user identifier; and determining the recommendation result corresponding to the search term based on the user group.

[0039] This application provides an electronic device, including:

[0040] Processor; and

[0041] The memory stores a program for implementing the search recommendation method. After the device is powered on and the program for the method is run by the processor, it performs the following steps: receiving a user's search request, the search request carrying a user identifier and a search term; determining whether the historical user behavior corresponding to the user identifier is related to the search term based on the search request; if there is no historical user behavior, or if there is historical user behavior but it is not related to the search term, determining the user group corresponding to the user identifier; and determining the recommendation result corresponding to the search term based on the user group.

[0042] This application provides an electronic device, including:

[0043] Processor; and

[0044] The memory stores a program for implementing the search recommendation method. After the device is powered on and the program for the method is run by the processor, it performs the following steps: sending a user's search request to the server, the search request carrying a user identifier and a search term; displaying the recommendation results corresponding to the search term returned by the server; wherein the server determines the recommendation results using the following steps: determining whether the historical user behavior corresponding to the user identifier is related to the search term; if there is no historical user behavior, or if there is historical user behavior but it is not related to the search term, determining the user group corresponding to the user identifier; and determining the recommendation results corresponding to the search term based on the user group.

[0045] This application also provides a computer-readable storage medium storing instructions that, when executed on a computer, cause the computer to perform the various methods described above.

[0046] This application also provides a computer program product including instructions that, when run on a computer, cause the computer to perform the various methods described above.

[0047] Compared with the prior art, this application has the following advantages:

[0048] The search recommendation method provided in this application receives a user's search request, which carries a user identifier and search terms; determines whether the historical user behavior corresponding to the user identifier is related to the search terms based on the search request; if there is no historical user behavior, or if there is historical user behavior but it is not related to the search terms, determines the user group corresponding to the user identifier; and determines the recommendation results corresponding to the search terms based on the user group. This processing method enables the recall of preferred search results for low-activity users (including new users) who do not have or have little historical interaction behavior related to the current search based on the historical interaction behavior of the user group to which they belong, thus realizing a personalized search recommendation method for a broad group. Since the group-based personalization has significant convergence in certain dimensions of international user grouping, extending user behavior entities (such as user click-to-purchase behavior) to user group behavior entities can provide personalized search recommendation services for low-activity users, thereby improving conversion efficiency and user experience. In addition, since group behavior (such as national user behavior) is used, the risk of privacy leakage can be effectively reduced. Attached Figure Description

[0049] Figure 1 A flowchart illustrating an embodiment of the search recommendation method provided in this application;

[0050] Figure 2 This application provides an illustration of an application scenario for the search recommendation method provided in this application.

[0051] Figure 3 This application provides a schematic diagram illustrating personalized recall of an embodiment of the search recommendation method. Detailed Implementation

[0052] Many specific details are set forth in the following description to provide a full understanding of this application. However, this application can be implemented in many other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of this application; therefore, this application is not limited to the specific embodiments disclosed below.

[0053] This application provides a search recommendation method and apparatus, as well as an electronic device. The various solutions are described in detail in the following embodiments.

[0054] First Embodiment

[0055] Please refer to Figure 1 This is a flowchart illustrating an embodiment of the search recommendation method of this application. The executing entity of this method includes, but is not limited to, a server, and can also be any device capable of implementing the method. In this embodiment, the method may include the following steps:

[0056] Step S101: Receive a user's search request, which carries a user identifier and search terms.

[0057] The search request can be for a product object, and the search terms can be product names, brand names, etc. The search request can also be for a news object, and the search terms can be content related to the news headline. The search request can also be for other business objects. This embodiment uses a product object search as an example to illustrate the method.

[0058] like Figure 2 As shown, the execution entity of the method provided in this embodiment is a server. This server can be deployed on a cross-border e-commerce platform. International buyers on this platform interact with the server through their clients. The server and client can be connected via a network, such as the client connecting to the internet via Wi-Fi. Buyers open the product search page of the e-commerce platform through their clients and enter a search term (such as "Apple") in the search box. The server determines the personalized product search results corresponding to the search term for the user and returns the search results to the client. The client displays the personalized search results for the user to view, so that the user can easily find products of interest and place an order.

[0059] The server-side component can be deployed on a cloud server or a dedicated server for implementing a cross-border e-commerce platform system, and can be deployed in a data center. The server can be a cluster of servers or a single server.

[0060] The client includes, but is not limited to, mobile communication devices, namely, mobile phones or smartphones, as well as personal computers, tablets, iPads and other terminal devices.

[0061] After receiving a search request, the next step can be taken to determine whether the historical user behavior corresponding to the user identifier is related to the search term.

[0062] Step S103: Determine whether the historical user behavior corresponding to the user identifier is related to the search term based on the search request.

[0063] The method provided in this embodiment can store user behavior logs on the server side. These log files may include historical user behavior data, such as which search terms the user searched and what interactive behaviors the user exhibited with the product objects retrieved by the server. These interactive behaviors include, but are not limited to: adding product objects to the shopping cart, adding product objects to favorites, browsing product objects, and making transactions with product objects, etc.

[0064] Table 1 shows the correspondence between user, search term and product object interaction data in this embodiment.

[0065]

[0066] Table 1. Correspondence between user, search term, and product interaction data.

[0067] As shown in Table 1, a user can perform different interactive behaviors on multiple product objects obtained from a single search, such as clicking to browse the details page of some product objects, clicking to purchase a product object, adding a product object to the shopping cart, etc. These behavioral data can be recorded in a log file.

[0068] In step S103, based on the user identifier carried in the search request, it is determined whether the historical user behavior corresponding to the user identifier is related to the search term. Specifically, this can be achieved by querying the user behavior log to see if the current search term is included in the user's historical search terms. If it is, it can be further queried whether the search term corresponds to the user's historical behavior. If it does, it can be determined that the historical user behavior corresponding to the user identifier is related to the search term, and then the prior art's personalized recall logic can be entered. Based on the user's historical user behavior related to the search term, the recommendation result is determined. This does not trigger the general group personalized recall logic provided in this application embodiment, and the process ends.

[0069] If the server determines that the user has no recent relevant searches or behaviors, it can proceed to the next step to determine the user group corresponding to the user identifier.

[0070] Step S105: If there is no historical user behavior, or if there is historical user behavior but it is not related to the search term, determine the user group corresponding to the user identifier.

[0071] The method provided in this application embodiment can provide personalized recommendations to users without historical user behavior. Users without historical user behavior can be new users of the cross-border e-commerce platform or existing users with historical user behavior, but whose historical behavior is irrelevant to the search terms. When a user has no historical user behavior, or has historical user behavior but it is irrelevant to the search terms, a generalized personalized search recommendation method is used to recommend products to the user. To implement the generalized personalized search recommendation method, the user group corresponding to the user identifier must first be determined.

[0072] The method can determine multiple users who share at least one of the following characteristics based on the user identifier: geographical location, country or region, ethnicity, age group, and hobbies, to form a user group.

[0073] In one example, a user's group can be determined from multiple dimensions, such as national group, regional group, ethnic group, age group, hobby group, and other grouping methods. Therefore, the user can belong to multiple user groups.

[0074] In practice, users' geographical location, ethnicity, and age group can be determined based on their registration information, historical logistics addresses, etc.; users' interests can be determined based on their historical transaction behavior; and then, they can be grouped according to various information, such as different age groups corresponding to different user groups, different interests corresponding to different user groups, different ethnicities corresponding to different user groups, and so on.

[0075] In cross-border e-commerce platforms, user behavior exhibits significant regional characteristics, with marked differences across different regions. Therefore, country-based grouping is suitable. To determine the country group to which a user belongs, the method can employ at least the following approaches:

[0076] Method 1: Determine the user's country of origin based on their registration information, historical shipping addresses, and other relevant data. For example, determine the user's country of origin based on their nationality information during registration.

[0077] Method 2: Determine the user's country / group based on user login information. For example, the user's country / group can be determined based on the IP address information used to log in via their client.

[0078] Method 3: Determine the user's country / group based on the client version information. For example, if the user's AliExpress client is the US version, then the user belongs to the US group.

[0079] By using methods two and three, the user's country of origin can be directly determined based on the user's login information or client version information. This allows for the identification of a new user's country of origin even if they have not entered their nationality information or engaged in any transactions. Consequently, personalized product search and recommendation services can be provided to new international users.

[0080] In another example, information from multiple dimensions can be combined to determine the user's group, such as grouping users aged 20-30 in the United States into one group, users aged 20-30 in France into another group, and so on.

[0081] In this embodiment, after a user's group is determined, the user grouping information can be stored so that the user's group can be directly determined based on the user grouping information later. Table 2 shows the user grouping information in this embodiment.

[0082]

[0083] Table 2. User Grouping Information

[0084] As shown in Table 2, a user can have multiple group information. The data in Table 2 can be determined using the method described above.

[0085] In practice, information such as the user's geographical location, country or region, ethnicity, age group, and hobbies can be stored according to the user identifier. In this way, multiple users with the same geographical location, country or region, ethnicity, age group, and hobbies can be identified as a user group.

[0086] Once the user group is identified, the next step is to determine the recommended results corresponding to the search term based on the user group.

[0087] Step S107: Determine the recommended results corresponding to the search term based on the user group.

[0088] User groups typically have historical user behavior related to the search terms. Based on the historical user behavior of user groups, products related to the search terms are determined to be recommended to the users. This method is a personalized search recommendation method for a broad user group.

[0089] In this embodiment, step S107 may include the following sub-steps:

[0090] Step S1071: Obtain the first user behavior related to the search term from the historical user behavior corresponding to the user group.

[0091] The historical user behavior corresponding to a user group may include first user behavior related to the search term, second user behavior related to the search term, and third user behavior unrelated to the search term. The method can determine the recommendation result based on the first user behavior. This first user behavior may involve product objects related to the search term that are of interest to most users in the group; based on the concept of broad group recommendation, the current user will also be interested in these product objects. In specific implementation, the first user behavior related to the search term can be determined from Table 1 above, based on the historical search terms and historical user behavior corresponding to each user in the user group.

[0092] For example, for a new user searching for products related to "apple", the server can first determine the new user's country based on the IP address information of the new user's client login; then, it can obtain all users in that country from Table 2 to form a user group; then, from Table 1, it can determine the first user behavior related to the search term "apple" based on the historical search terms and historical user behavior of each user in that country; finally, it can determine the recommendation results based on the first user behavior.

[0093] In one example, step S1071 can be implemented as follows: obtain the probability distribution of behaviors related to the search term in the historical user behaviors corresponding to the user group; select the behavior with the highest probability in the probability distribution as the first user behavior. Among some users in the user group who have historically conducted related searches, the various historical user behaviors related to the search term may have different probabilities of occurrence. Some behaviors have a high probability of occurrence, such as most people selecting a certain product, or most people adding a certain product to their shopping cart. These selection behaviors or adding to the shopping cart behaviors are the first user behaviors.

[0094] In another example, step S1071 can be implemented as follows: obtain the evaluation index related to the search term in the historical user behavior of the user group; select the user behavior corresponding to the evaluation index that reaches or exceeds the target value as the first user behavior.

[0095] The evaluation metrics include, but are not limited to, total transaction volume, conversion rate, and repurchase rate. Among users who have historically conducted related searches, their various historical user behaviors related to the search terms often differ in terms of evaluation metrics. For example, if the total transaction volume, conversion rate, or repurchase rate for certain product items is high, then the historical user behavior related to these product items is considered the primary user behavior.

[0096] It should be noted that the method provided in this application, when calculating metrics such as conversion rate, only needs to use transaction logs based on the grouping of product transactions (e.g., country), without needing to know personal information, thus posing no risk of privacy leakage. In specific implementation, the evaluation metrics can be determined through offline statistical methods of grouped behavior. This approach can effectively improve search recommendation speed, thereby enhancing user experience.

[0097] In another example, step S1071 can be implemented as follows: obtain the behavioral indicators related to the search term in the historical user behavior of the user group; select the user behavior that meets the behavioral indicators as the first user behavior.

[0098] The behavioral metrics include, but are not limited to: adding to cart, saving to favorites, browsing, forwarding, recommending, and trading. Among users who have historically conducted related searches, these users will engage in interactive behaviors towards products they are interested in, while they may not interact with products they are not interested in. In other words, if the above behavioral metrics indicate that a user is interested in the relevant products, then actions such as adding to cart, saving to favorites, browsing, forwarding, recommending, and trading will be considered the primary user behavior.

[0099] In this embodiment, the user's corresponding group is first determined using the method described in step S105, such as country grouping, ethnic grouping, etc.; then, high-frequency differential behaviors within the user group to which the user belongs can be used as personalized product recall for that user. For example, if the conversion rate of a certain product among users in a certain group (e.g., 80%) is significantly higher than the conversion rate among global users (e.g., 30%), then that product is considered a personalized product for that group. More intuitively, the method provided in this application embodiment recalls related products through keyword search, and the general personalization framework recalls top products in the group, such as the products with the highest sales in the United States under the keyword "Apple".

[0100] In one example, the server is specifically used to determine the data values ​​of preset indicators of product objects within the user group in the interaction behavior data; and to determine the recommendation results based on the data values.

[0101] The preset indicators include those that can distinguish high-frequency differences in the online shopping behavior of products across different user groups, such as conversion rates. In this embodiment, the business items with the highest conversion rates are used as the recommendation results.

[0102] In one example, to provide personalized product search recommendations to inactive international users, the following approach could be used: the top-selling products by country could be used as grouping triggers. However, personalized recall is primarily about reflecting the differences between users. Directly using best-selling products might result in similar products across different countries, making it difficult to highlight these differences, thus this method is less effective.

[0103] In this embodiment, cross-entropy is used to determine the evaluation metric. Specifically, the server determines a first ratio between the total transaction volume of a candidate product (a product appearing in the interaction behavior data of a user's user group) within that user group and the total global transaction volume of the candidate product (all user groups within a certain grouping dimension); and a second ratio between the total transaction volume of product objects within the user group and the total global transaction volume of product objects; the cross-entropy of the first and second ratios is used as the evaluation metric. This approach reflects the differences between users, resulting in higher accuracy of search recommendation results. A detailed description of this implementation is as follows.

[0104] Suppose we have a sample set with two probability distributions p and q, where p is the true distribution and q is the false distribution. If the expected code length required to identify a sample, measured according to the true distribution p, is: H(p) = -∑p(i)logp(i), then if we use the false distribution q to represent the average code length from the true distribution p, it should be: H(p) = -∑p(i)logq(i). H(p,q) is called cross-entropy.

[0105] For convenience, the following explanations are all under the same category. Let n represent the total global transaction volume. c This represents the total volume of transactions in that country. i Let n represent the total global transaction volume of product i. ic This represents the total volume of transactions for product i in country c. The conversion rate is calculated using the following formula:

[0106]

[0107] In practice, for products with a certain sales volume, a preset number of products with the highest conversion rates (such as the top 30 products with the highest conversion rates) can be selected as candidate business objects for general personalized triggers.

[0108] Next, the recall and refinement scoring processes can be performed. The method provided in this application is designed to find potential personalized products for users who have not undergone personalized recall. Therefore, the aforementioned generalized personalized recall trigger only serves as a supplement to the true personalized recall trigger. It can be observed that the more user behavior there is, the more obvious their intentions become; therefore, the impact of the generalized personalized trigger should be reduced accordingly. Therefore, in the refinement scoring stage, a decay function for the generalized personalized trigger score is used on top of the generalized personalized score. The generalized personalized trigger only takes effect when user behavior is minimal.

[0109] Once the server determines the user group to which the target user belongs, it can determine the user group's interaction behavior data related to the target search term based on the historical interaction behavior data shown in Table 1. In practice, the server can first determine all user information for that user group based on the target user's user group information; then, it can determine the interaction behavior data of these users related to the target search term from the historical interaction behavior data.

[0110] It is important to emphasize that historical user behavior refers specifically to user behavior related to the target search term. For example, if the target search term is "Apple," then historical user behavior related to this term could include: historical user behavior related to "Apple," historical user behavior related to "iPhone," and so on. In other words, historical user behavior related to the target search term can include: historical interaction data of the user's group corresponding to the target search term, and historical interaction data of the user's group corresponding to other search terms similar to the target search term. Since identifying other search terms similar to the target search term is a relatively mature existing technology, it will not be elaborated upon here.

[0111] In practice, this involves identifying user group's interaction data with product objects related to the target search term within a specific timeframe. For example, identifying user group's interaction data with product objects related to the target search term over the past month or week.

[0112] Step S1072: Determine the recommendation result of the first user behavior.

[0113] In this embodiment, after the server determines the interaction behavior data of product objects related to the target search term for the user group to which the target user belongs, it can determine the target product objects corresponding to the target search term to recommend to the target user based on this data. These product objects may be product objects that the user is interested in. In specific implementation, the product objects involved in the first user behavior can be used as the recommendation result of the first user behavior.

[0114] Step S1073: Use the recommendation result of the first user behavior as the recommendation result corresponding to the search term.

[0115] Please refer to Figure 3This is a schematic diagram illustrating personalized recall in an embodiment of the search recommendation method of this application. In this embodiment, a user enters the search scenario of a cross-border e-commerce platform and inputs the search keyword "Apple". If the server determines that the user has recently searched for related keywords and has had product interaction behavior (such as clicking, purchasing, etc.), it enters the personalized recall logic of the prior art. Based on the user's personal product object interaction behavior data related to the target search term, the recommendation result is determined. This may not trigger the general group personalized recall logic provided in this application embodiment, and the process ends. If the server determines that the user has no recent related searches and behaviors, it can first determine the user's corresponding group through the above method, such as country grouping, style grouping, etc. Then, the high-frequency difference behavior within the user group to which the user belongs can be used as the user's personalized product recall. For example, if the conversion rate of a certain product among users in a certain group is significantly higher than the conversion rate among global users, then that product is used as the personalized product for that group. More intuitively, the method provided in this application embodiment will recall related products through keyword search, and the general personalization framework will recall the top products of the group, such as the products with the most transactions in the United States under the keyword "Apple".

[0116] In this embodiment, after sending the recommendation results back to the client, the server is also used to determine the user's interaction behavior data with the recommendation results; and to store the correspondence between the user, the search term, and the user's interaction behavior data with the recommendation results. For example, this correspondence is stored in Table 1 above.

[0117] In one example, the server also stores the correspondence between user group information, search terms, and product object interaction behavior data, but may not store the correspondence between users and user groups. Specifically, it is used to determine the product object interaction behavior data related to the target search term for the user's user group based on the correspondence between the user group information, search terms, and product object interaction behavior data. Table 3 shows the correspondence between user group information, search terms, and product object interaction behavior data in this embodiment.

[0118]

[0119]

[0120] Table 3. Correspondence between user group information, search terms, and product object interaction data.

[0121] This approach utilizes group behavior (such as national user behavior) to avoid recording individual user group information. It also allows for the exclusion of recording individual user search behavior, effectively preventing the leakage of personal information and complying with the privacy protection policies of different countries. Therefore, it can effectively reduce the risk of privacy leaks.

[0122] In practice, after online A / B testing, using the method provided in the embodiments of this application, the UV (unique visitor) conversion rate can be increased by approximately 0.58%, and the UV value rate can be increased by approximately 1.89%.

[0123] As can be seen from the above embodiments, the search recommendation method provided in this application receives a user's search request, which carries a user identifier and search terms; determines whether the historical user behavior corresponding to the user identifier is related to the search terms based on the search request; if there is no historical user behavior, or if there is historical user behavior but it is not related to the search terms, determines the user group corresponding to the user identifier; and determines the recommendation results corresponding to the search terms based on the user group. This processing method enables the recall of preferred search results for low-activity users (including new users) who do not have or have little historical interaction behavior related to the current search based on the historical interaction behavior of the user group to which the low-activity user belongs, thus realizing a personalized search recommendation method for a broad group. Since the group-based personalization has significant convergence in certain dimensions of international user grouping, extending user behavior entities (such as user click-to-purchase behavior) to user group behavior entities can provide personalized search recommendation services for low-activity users, thereby improving conversion efficiency and user experience. In addition, since group behavior (such as national user behavior) is used, the risk of privacy leakage can be effectively reduced.

[0124] Second Embodiment

[0125] Corresponding to the search recommendation method described above, this application also provides a search recommendation apparatus. This apparatus corresponds to an embodiment of the method described above.

[0126] The parts of this embodiment that are the same as those in the first embodiment will not be repeated here; please refer to the corresponding parts in Embodiment 1. A search recommendation device provided in this application includes:

[0127] A request receiving unit is used to receive a user's search request, wherein the search request carries a user identifier and search terms;

[0128] The judgment unit is used to determine, based on the search request, whether the historical user behavior corresponding to the user identifier is related to the search term;

[0129] The user group determination unit is used to determine the user group corresponding to the user identifier if there is no historical user behavior, or if there is historical user behavior but it is not related to the search term.

[0130] The recommendation result determination unit is used to determine the recommendation result corresponding to the search term based on the user group.

[0131] Third Embodiment

[0132] This application also provides an electronic device. Since the device embodiments are substantially similar to the method embodiments, the description is relatively simple; relevant details can be found in the description of the method embodiments. The device embodiments described below are merely illustrative.

[0133] This embodiment provides an electronic device, which includes a processor and a memory. The memory stores a program for implementing a search recommendation method. After the device is powered on and the program for the method is run by the processor, it performs the following steps: receiving a user's search request, the search request carrying a user identifier and a search term; determining whether the historical user behavior corresponding to the user identifier is related to the search term based on the search request; if there is no historical user behavior, or if there is historical user behavior but it is not related to the search term, determining the user group corresponding to the user identifier; and determining the recommendation result corresponding to the search term based on the user group.

[0134] Fourth embodiment

[0135] Corresponding to the search recommendation system described above, this application also provides a search recommendation method. The executing entity of this method includes, but is not limited to, a client, and can also be any device capable of implementing the method. The parts of this embodiment that are the same as those in the first embodiment will not be repeated here; please refer to the corresponding parts in Embodiment 1.

[0136] In this embodiment, the search recommendation method includes the following steps:

[0137] Step 1: Send the user's search request to the server, the search request carrying the user identifier and search terms;

[0138] Step 2: Display the recommended results corresponding to the search terms sent back by the server.

[0139] The server determines the recommendation result using the following steps: determining whether the historical user behavior corresponding to the user identifier is related to the search term; if there is no historical user behavior, or if there is historical user behavior but it is not related to the search term, determining the user group corresponding to the user identifier; and determining the recommendation result corresponding to the search term based on the user group.

[0140] Fifth Embodiment

[0141] In the above embodiments, a search recommendation method is provided. Correspondingly, this application also provides a search recommendation apparatus. This apparatus corresponds to the embodiments of the above method.

[0142] The parts of this embodiment that are the same as those in the first embodiment will not be repeated here; please refer to the corresponding parts in Embodiment 1. A search recommendation device provided in this application includes:

[0143] The request sending unit is used to send a user's search request to the server, wherein the search request carries a user identifier and search terms;

[0144] The recommendation results display unit is used to display the recommendation results corresponding to the search terms sent back by the server;

[0145] The server determines the recommendation result using the following steps: determining whether the historical user behavior corresponding to the user identifier is related to the search term; if there is no historical user behavior, or if there is historical user behavior but it is not related to the search term, determining the user group corresponding to the user identifier; and determining the recommendation result corresponding to the search term based on the user group.

[0146] Sixth Embodiment

[0147] This application also provides an embodiment of an electronic device. Since the device embodiment is substantially similar to the method embodiment, it is described simply; relevant details can be found in the description of the method embodiment. The device embodiment described below is merely illustrative.

[0148] This embodiment provides an electronic device, comprising: a processor and a memory; the memory for storing a program implementing a search recommendation method. After the device is powered on and the program is run by the processor, it performs the following steps: sending a user's search request to a server, the search request carrying a user identifier and a search term; displaying recommendation results corresponding to the search term returned by the server; wherein the server determines the recommendation results using the following steps: determining whether the historical user behavior corresponding to the user identifier is related to the search term; if there is no historical user behavior, or if there is historical user behavior but it is not related to the search term, determining the user group corresponding to the user identifier; and determining the recommendation results corresponding to the search term based on the user group.

[0149] Although this application discloses preferred embodiments as described above, it is not intended to limit this application. Any person skilled in the art can make possible changes and modifications without departing from the spirit and scope of this application. Therefore, the scope of protection of this application should be determined by the scope defined in the claims of this application.

[0150] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0151] Memory may include non-persistent storage in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM. Memory is an example of computer-readable media.

[0152] 1. Computer-readable media includes both permanent and non-permanent, removable and non-removable media that can store information by any method or technology. Information can be computer-readable instructions, data structures, modules of programs, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic magnetic disk storage or other magnetic storage devices, or any other non-transferable medium that can be used to store information accessible by a computing device. As defined herein, computer-readable media does not include non-transitory computer-readable media, such as modulated data signals and carrier waves.

[0153] 2. Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

Claims

1. A search recommendation method characterized by comprising: The application is applied to a cross-border e-commerce platform, and comprises the following steps: receiving a search request of a user, wherein the search request carries a user identifier and a search word; determining whether historical user behaviors corresponding to the user identifier are related to the search word according to the search request, wherein the historical user behaviors corresponding to the user identifier refer to interaction behaviors of the user on recalled commodity objects; if there are no historical user behaviors or there are historical user behaviors but the historical user behaviors are not related to the search word, determining a user group corresponding to the user identifier, wherein the user group is determined by a country or region to which the user belongs; determining a recommendation result corresponding to the search word according to historical user behaviors of the user group.

2. The method of claim 1, wherein, The step of determining the recommendation result corresponding to the search word according to the historical user behaviors of the user group comprises the following steps: obtaining a first user behavior related to the search word in the historical user behaviors of the user group; determining a recommendation result of the first user behavior; taking the recommendation result of the first user behavior as the recommendation result corresponding to the search word.

3. The method of claim 2, wherein, The step of obtaining the first user behavior related to the search word in the historical user behaviors of the user group comprises the following steps: obtaining a behavior probability distribution related to the search word in the historical user behaviors of the user group; selecting a behavior with the highest probability in the behavior probability distribution as the first user behavior.

4. The method of claim 2, wherein, The step of obtaining the first user behavior related to the search word in the historical user behaviors of the user group comprises the following steps: obtaining an evaluation index related to the search word in the historical user behaviors of the user group, wherein the evaluation index comprises but is not limited to the following: total transaction amount, transaction conversion rate, repurchase rate; selecting a user behavior corresponding to an evaluation index reaching or exceeding a target value as the first user behavior.

5. The method of claim 2, wherein, The step of obtaining the first user behavior related to the search word in the historical user behaviors of the user group comprises the following steps: obtaining a behavior index related to the search word in the historical user behaviors of the user group, wherein the behavior index comprises but is not limited to the following: adding a shopping cart, collecting, browsing, forwarding, recommending, and trading; selecting a user behavior satisfying the behavior index as the first user behavior.

6. The method according to any one of claims 1-5, characterized in that, The step of determining the user group corresponding to the user identifier comprises the following steps: determining at least one same item of a user in the user identifier, wherein the at least one same item comprises the following: a geographical position of the user, a country or region to which the user belongs, a nationality of the user, an age range of the user, and a hobby of the user, and the at least one same item is used as a user group.

7. A search recommendation method characterized by comprising: The application is applied to a cross-border e-commerce platform, and comprises the following steps: sending a search request of a user to a server, wherein the search request carries a user identifier and a search word; displaying a recommendation result corresponding to the search word returned by the server; The service end determines the recommendation result by the following steps: determining whether the historical user behavior corresponding to the user identifier is related to the search word, the historical user behavior corresponding to the user identifier being an interaction behavior of the user on a recalled commodity object; if there is no historical user behavior, or there is historical user behavior but the historical user behavior is not related to the search word, determining a user group corresponding to the user identifier, the user group being determined by a country or region to which the user belongs; and determining the recommendation result corresponding to the search word according to the historical user behavior of the user group.

8. A search recommendation apparatus characterized by comprising: The application is applied to a cross-border e-commerce platform, and comprises: a request receiving unit configured to receive a search request of a user, the search request carrying a user identifier and a search word; a judging unit configured to determine, according to the search request, whether historical user behavior corresponding to the user identifier is related to the search word, the historical user behavior corresponding to the user identifier being an interaction behavior of the user on a recalled commodity object; a user group determining unit configured to determine, if there is no historical user behavior, or there is historical user behavior but the historical user behavior is not related to the search word, a user group corresponding to the user identifier, the user group being determined by a country or region to which the user belongs; a recommendation result determining unit configured to determine a recommendation result corresponding to the search word according to the historical user behavior of the user group.

9. A search recommendation apparatus characterized by comprising: The application is applied to a cross-border e-commerce platform, and comprises: a request sending unit configured to send a search request of a user to a service end, the search request carrying a user identifier and a search word; a recommendation result display unit configured to display the recommendation result corresponding to the search word returned by the service end; The service end determines the recommendation result by the following steps: determining whether the historical user behavior corresponding to the user identifier is related to the search word, the historical user behavior corresponding to the user identifier being an interaction behavior of the user on a recalled commodity object; if there is no historical user behavior, or there is historical user behavior but the historical user behavior is not related to the search word, determining a user group corresponding to the user identifier, the user group being determined by a country or region to which the user belongs; and determining the recommendation result corresponding to the search word according to the historical user behavior of the user group.

10. An electronic device, comprising: comprise: a processor; and a memory configured to store a program for implementing a search recommendation method, the device being powered and running the program of the method by the processor, and performing the following steps: receiving a search request of a user, the search request carrying a user identifier and a search word; determining, according to the search request, whether historical user behavior corresponding to the user identifier is related to the search word, the historical user behavior corresponding to the user identifier being an interaction behavior of the user on a recalled commodity object; if there is no historical user behavior, or there is historical user behavior but the historical user behavior is not related to the search word, determining a user group corresponding to the user identifier, the user group being determined by a country or region to which the user belongs; and determining a recommendation result corresponding to the search word according to the historical user behavior of the user group.

Images