Object recommendation method and apparatus, storage medium, and electronic device

Abstract

Embodiments of the present application disclose an object recommendation method and device, a storage medium and an electronic device. A first keyword input by a user is received through an input box, candidate search words matched with the first keyword are determined, historical interaction information corresponding to the candidate search words on at least one user interaction behavior is determined, transaction feature information of the candidate search words is determined based on transaction information of a first candidate object similar to each candidate search word, the candidate search words are sorted based on the transaction feature information and the historical interaction information corresponding to each candidate search word, the candidate search words are displayed, a target keyword corresponding to a keyword determination event is obtained in response to the keyword determination event, a plurality of target objects matched with the target keyword are determined, and a target user is recommended based on the plurality of target objects. In this way, the object recommendation efficiency can be improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of Internet, in particular to an object recommendation method and device, a storage medium and an electronic device. BACKGROUND

[0002] With the rapid development of life and technology, people often entertain through game applications. In some UGC (User-generated Content) games, users need to search for the required components and other objects by inputting keywords.

[0003] In the research and practice of the prior art, it is found that in the existing object recommendation method, only the keywords input by the user are matched to similar objects recommended to the user, which cannot accurately and efficiently recommend objects of interest to the user, resulting in low object recommendation efficiency. SUMMARY

[0004] The object recommendation method and device, the storage medium and the electronic device provided by the embodiments of the present application can enable the user to quickly and accurately determine the target keyword to be searched, efficiently and accurately recommend objects of interest to the user, and improve the object recommendation efficiency.

[0005] The embodiments of the present application provide an object recommendation method, comprising:

[0006] receiving a first keyword input through an input box, determining a candidate search word matched with the first keyword;

[0007] determining historical interaction information corresponding to the candidate search word on at least one user interaction behavior;

[0008] determining transaction feature information of the candidate search word based on transaction information of a first candidate object similar to the candidate search word;

[0009] sorting the candidate search words based on the transaction feature information and the historical interaction information corresponding to each candidate search word;

[0010] displaying the candidate search words;

[0011] In response to a keyword determination event, obtaining a target keyword corresponding to the keyword determination event, and determining a plurality of target objects matched with the target keyword;

[0012] recommending based on the plurality of target objects.

[0013] Correspondingly, the embodiments of the present application provide an object recommendation device, comprising:

[0014] The receiving unit is configured to receive the inputted first keyword through an input box, and determine candidate search words matched with the first keyword;

[0015] The first determining unit is configured to determine historical interaction information corresponding to the candidate search words on at least one user interaction behavior;

[0016] The second determining unit is configured to determine transaction feature information of the candidate search words based on transaction information of first candidate objects similar to the candidate search words;

[0017] The sorting unit is configured to sort the candidate search words based on the transaction feature information and the historical interaction information corresponding to each of the candidate search words;

[0018] The display unit is configured to display the candidate search words;

[0019] The third determining unit is configured to, in response to a keyword determining event, acquire a target keyword corresponding to the keyword determining event, and determine a plurality of target objects matched with the target keyword;

[0020] The recommendation unit is configured to perform recommendation based on the plurality of target objects.

[0021] In addition, an embodiment of the present application further provides a computer readable storage medium, which stores a computer program, and the computer program is adapted to be loaded by a processor to execute steps in any one of the object recommendation methods provided by the embodiments of the present application.

[0022] In addition, an embodiment of the present application further provides an electronic device, which comprises a processor and a memory, the memory stores an application program, and the processor is used to run the application program in the memory to implement the object recommendation method provided by the embodiments of the present application.

[0023] An embodiment of the present application further provides a computer program product, which comprises a computer program stored in a computer readable storage medium; when a processor of an electronic device reads the computer program from the computer readable storage medium, the processor executes the computer program, so that the electronic device executes steps in the object recommendation method provided by the embodiments of the present application.

[0024] This application embodiment receives a first keyword entered through an input box, determines candidate search terms matching the first keyword, determines historical interaction information corresponding to the candidate search terms in at least one user interaction behavior, determines transaction feature information of the candidate search terms based on transaction information of first candidate objects similar to each candidate search term, sorts the candidate search terms based on the transaction feature information and historical interaction information corresponding to each candidate search term, displays the candidate search terms, and in response to a keyword determination event, obtains the target keyword corresponding to the keyword determination event and determines multiple target objects matching the target keyword, and makes recommendations based on the multiple target objects. Therefore, by identifying candidate search terms associated with the first keyword currently entered by the user in the input box, and then ranking these candidate search terms based on their historical interaction information and transaction characteristics, the system recommends these candidate search terms to the user based on the ranking results. This allows users to quickly and accurately determine their target keywords and recommend multiple target objects that match those keywords. This achieves efficient and accurate recommendation of objects that users are interested in. Furthermore, recommending search terms based on transaction characteristics increases the probability of users purchasing the recommended target objects, thereby increasing the benefits in the search scenario and ultimately improving the efficiency of object recommendation. Attached Figure Description

[0025] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0026] Figure 1 This is a schematic diagram illustrating an implementation scenario of an object recommendation method provided in an embodiment of this application;

[0027] Figure 2 This is a flowchart illustrating an object recommendation method provided in an embodiment of this application;

[0028] Figure 3 This is a schematic diagram of the search term expansion of an object recommendation method provided in an embodiment of this application;

[0029] Figure 4a This is a schematic diagram illustrating the specific architecture of an object recommendation method provided in an embodiment of this application;

[0030] Figure 4b This is another specific architectural diagram of an object recommendation method provided in an embodiment of this application;

[0031] Figure 4cis a search system architecture schematic diagram of an object recommendation method provided by an embodiment of the present application.

[0032] Figure 5 is a structural schematic diagram of an object recommendation device provided by an embodiment of the present application.

[0033] Figure 6 is a structural schematic diagram of an electronic device provided by an embodiment of the present application. DETAILED DESCRIPTION

[0034] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by a person skilled in the art without creative work fall within the scope of protection of the present application.

[0035] The embodiments of the present application provide an object recommendation method, device, storage medium and electronic device. The object recommendation device can be integrated in an electronic device, which can be a server or a terminal or the like.

[0036] The server can be an independent physical server, a server cluster or a distributed system composed of multiple physical servers, or a cloud server providing cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, network acceleration services (CDN), and basic cloud computing services such as big data and artificial intelligence platforms. The terminal can include but is not limited to a mobile phone, a computer, a smart voice interaction device, a smart home appliance, a vehicle-mounted terminal, an aircraft, etc. The terminal and the server can be directly or indirectly connected through wired or wireless communication, which is not limited in the present application.

[0037] Please refer to Figure 1 For example, the object recommendation device is integrated in an electronic device, Figure 1An implementation scenario diagram of an object recommendation method provided by an embodiment of the present application is shown in FIG. 1. In the implementation scenario, the electronic device can be a server. The electronic device can receive a first keyword that has been input through an input box, determine candidate search words that match the first keyword, determine historical interaction information corresponding to the candidate search words on at least one user interaction behavior, determine transaction feature information of the candidate search words based on transaction information of first candidate objects similar to the candidate search words, sort the candidate search words based on the transaction feature information and the historical interaction information corresponding to the candidate search words, display the candidate search words, obtain a target keyword corresponding to a keyword determination event in response to the keyword determination event, determine a plurality of target objects that match the target keyword, and perform recommendation based on the plurality of target objects.

[0038] It should be noted that Figure 1 The implementation environment scenario diagram of the object recommendation method shown in FIG. 1 is merely an example. The implementation environment scenario of the object recommendation method described in the embodiments of the present application is used to more clearly illustrate the technical solutions of the embodiments of the present application, and does not constitute a limitation on the technical solutions provided by the embodiments of the present application. It can be known by those skilled in the art that, as data processing evolves and new business scenarios appear, the technical solutions provided by the present application are also applicable to similar technical problems.

[0039] The solutions provided by the embodiments of the present application are described in detail through the following embodiments. It should be noted that the order of the following embodiments is not a limitation on the preferred order of the embodiments.

[0040] This embodiment will be described from the perspective of an object recommendation device, which can be integrated in an electronic device. The electronic device can be a terminal and / or a server, which is not limited by the present application.

[0041] Please refer to Figure 2 , Figure 2 FIG. 2 is a flowchart of an object recommendation method provided by an embodiment of the present application. The object recommendation method includes the following steps.

[0042] In step 101, a first keyword that has been input is received through an input box, and candidate search words that match the first keyword are determined.

[0043] The first keyword can be a keyword inputted by a target user in an input box. The target user can be a user currently performing a search. The input box can be an area in a user terminal interface for inputting a keyword to perform a search. For example, when the target user inputs "castle" in the input box, the first keyword can be "castle". For another example, when the target user inputs "happy" in the input box, the first keyword can be "happy".

[0044] The candidate search word can be a search word similar to the first keyword. A plurality of search words that can be searched by a user can be preset, and thus a candidate search word matching the first keyword can be searched from the search words. The candidate search word can be a search word similar to the first keyword or a search word containing the first keyword.

[0045] For example, it is assumed that a plurality of search words, including "red house", "happy child", "happy magic cube", and "happy castle", are stored in advance in a search engine corresponding to a game. When the target user inputs "happy" as the first keyword, candidate search words similar to the first keyword, such as "happy child", "happy magic cube", and "happy castle", can be searched from the plurality of search words stored in advance.

[0046] In an embodiment, a search word recommendation can be performed for a user based on a preference of the user before the user performs a search. For example, a second candidate object in which the target user is interested and a like score of the target user with respect to the second candidate object can be determined based on user association information of the target user. An association search word associated with the second candidate object can be acquired. A score of interest of the association search word can be determined based on the like score of the second candidate object associated with each association search word. The association search word can be sorted based on the score of interest of the association search word. A search word recommendation can be performed through a terminal of the target user based on a sorting result of the association search word.

[0047] The user association information can be user association information, for example, can include information such as objects that the target user has clicked to view, searched, added to the shopping cart, and purchased, and the object can be a virtual resource, for example, can be a component, a prop, a virtual map, etc. The component can include user-defined content, for example, in a UGC (User-Generated Content) game, the component can include virtual character models, castles, mineral saplings, mineral trackers, etc. Optionally, the user association information can be a user-related log of the target user. The second candidate object can be an object that the target user can be interested in based on the user association information, for example, when the target user frequently clicks on multiple types of components based on the user association information, the second candidate object can be a component of the castle type, and when the target user purchases a component of a mineral sapling based on the user association information, the second candidate object can be a component of the mineral sapling, etc. The preference score can be information indicating the degree of preference of the target user for the second candidate object, for example, when the target user clicks on component 1 three times and component 2 ten times based on the user association information, it can be shown that the target user has a higher degree of preference for component 2, and thus a higher preference score for component 2 can be calculated. The associated search term can be a search term corresponding to the second candidate object, for example, assuming that the name of the second candidate object is "Happy Castle", the associated search term corresponding to the second candidate object can be "Happy Castle", and for example, the associated search term corresponding to the second candidate object can be set in advance. The interest score can be information indicating the degree of interest of the target user in the associated search term.

[0048] Optionally, a collaborative filtering algorithm can be used to calculate the preference score of the target user for a plurality of preset search terms based on the user association information of the target user, and the search terms can be sorted according to the preference score, so that the associated search term that the target user is more likely to be interested in can be determined, and the associated search term can be pushed to the terminal of the target user for search term recommendation, thereby realizing personalized search term recommendation for the user. The user can quickly and accurately obtain the required search term by obtaining the personalized recommended associated search term before searching, thereby improving search efficiency.

[0049] In step 102, the historical interaction information corresponding to the candidate search term in at least one user interaction behavior is determined.

[0050] The user interaction behavior can be a user interaction behavior, for example, can include clicking, adding to the shopping cart, collecting, purchasing, etc. The historical interaction information can be information indicating the interaction behavior of the user based on the object recommended based on the candidate search term, for example, can include click rate, add-to-cart rate, purchase rate, etc.

[0051] The click rate can be the probability that the user clicks the object recommended based on the candidate search word after the candidate search word is searched. Specifically, the click rate can be the ratio of the number of clicks to the number of searches. The number of clicks can be the number of times the object recommended based on the candidate search word is clicked after the candidate search word is searched. The number of searches can be the number of times the candidate search word is searched as a search word.

[0052] The add-to-cart rate can be the probability that the user adds the object recommended based on the candidate search word to the shopping cart after the candidate search word is searched. Specifically, the add-to-cart rate can be the ratio of the number of add-to-cart operations to the number of searches. The number of add-to-cart operations can be the number of times the object recommended based on the candidate search word is added to the shopping cart after the candidate search word is searched. The number of searches can be the number of times the candidate search word is searched as a search word.

[0053] The purchase rate can be the probability that the user purchases the object recommended based on the candidate search word after the candidate search word is searched. Specifically, the purchase rate can be the ratio of the number of purchases to the number of searches. The number of purchases can be the number of times the object recommended based on the candidate search word is purchased after the candidate search word is searched. The number of searches can be the number of times the candidate search word is searched as a search word.

[0054] In step 103, transaction feature information of the candidate search word is determined based on transaction information of the first candidate object similar to the candidate search word.

[0055] The first candidate object can be an object similar to the candidate search term. For example, when the candidate search term is "castle", the first candidate object can be an object of various virtual castle types, such as a retro castle component, a pink castle component, a castle map, and the like. The transaction information can be information indicating the transaction attribute of the first candidate object. For example, the transaction information can include at least one of a transaction price, a buyer location, a transaction time, and a transaction frequency. The transaction price can be the price of the first candidate object. The buyer location can be the location of the user who purchased the first candidate object. The transaction time can be the time when the user purchased the first candidate object. The transaction frequency can be the number of times the first candidate object was purchased within a predetermined period of time. The transaction feature information can be information indicating the transaction feature of the first candidate object. The transaction feature can include a transaction price, a transaction concentrated area, a transaction concentrated period, and a transaction frequency. The transaction concentrated area can be the area where the buyers who purchased the first candidate object are concentrated. For example, for a first candidate object of a warm type, the locations of the buyers who purchased it are mostly in the north. For a first candidate object of a moisture-proof type, the locations of the buyers who purchased it are mostly in the south, and the like. The transaction concentrated period can be the period when the first candidate object is purchased. For example, a first candidate object of a sunny castle type can be concentrated in the morning, a first candidate object of a dark castle type can be concentrated in the evening, and the like.

[0056] The transaction feature information of the candidate search term can be determined in various ways based on the transaction information of the first candidate object similar to each candidate search term. For example, the price feature information of the candidate search term can be calculated based on the transaction price of the first candidate object similar to each candidate search term. The transaction area feature information of the candidate search term can be determined based on the buyer location of the first candidate object similar to each candidate search term. The transaction period feature information of the candidate search term can be determined based on the transaction time of the first candidate object similar to each candidate search term. The transaction frequency feature information of the candidate search term can be calculated based on the transaction frequency of the first candidate object similar to each candidate search term.

[0057] The price feature information can be information indicating a transaction price of the first candidate object, the transaction area feature information can be information indicating a transaction concentrated area of the first candidate object, the transaction time period feature information can be information indicating a transaction concentrated time period of the first candidate object, and the transaction frequency feature information can be information indicating a transaction frequency of the first candidate object. In this way, based on the transaction feature information of the first candidate object, the probability of the first candidate object being purchased in different scenarios can be determined, so that the first candidate object with a greater probability of being purchased in the current scenario can be recommended to the target user based on the scenario in which the target user is located, thereby improving the probability of the target user purchasing the first candidate object, and further improving the conversion rate of the first candidate object and the revenue brought by the search scenario.

[0058] For example, when the target user performs object search in the morning, the first candidate object with a transaction concentrated time period in the morning can be recommended to the target user, when the location of the target user is in region A, the first candidate object with a transaction concentrated area in region A can be recommended to the target user, and in addition, the first candidate object with a higher transaction frequency or a higher transaction price can also be recommended to the target user, so as to improve the probability of the target user purchasing the first candidate object, and further improve the revenue brought by the object search scenario and the object search efficiency.

[0059] In step 104, the candidate search terms are sorted based on the transaction feature information and the historical interaction information corresponding to each candidate search term.

[0060] The candidate search terms can be sorted in various ways based on the transaction feature information and the historical interaction information corresponding to each candidate search term. For example, the interaction scores of the candidate search terms can be calculated based on the historical interaction information corresponding to each candidate search term, the candidate search terms can be sorted based on the interaction scores, and the target candidate search terms can be selected from the sorted candidate search terms. The target candidate search terms can be sorted based on the transaction feature information of each target candidate search term to obtain the sorted target candidate search terms.

[0061] The interaction score can be information measuring the degree to which each candidate search term is liked by the user, and the target candidate search term can be a candidate search term that is more interesting to the user.

[0062] The interaction score of each candidate search word can be calculated in various manners based on the historical interaction information corresponding to each candidate search word. For example, each type of historical interaction information corresponding to each candidate search word can be accumulated to obtain the interaction score of each candidate search word. For example, assuming that the click rate corresponding to candidate search word a is 0.4, the add-to-cart rate is 0.2, and the purchase rate is 0.1, each type of historical interaction information is accumulated to obtain the interaction score of candidate search word a, which is 0.7. Assuming that the click rate corresponding to candidate search word b is 0.3, the add-to-cart rate is 0.1, and the purchase rate is 0.05, each type of historical interaction information is accumulated to obtain the interaction score of candidate search word b, which is 0.45.

[0063] Optionally, a weight can be set for each type of historical interaction information, so that each type of historical interaction information corresponding to each candidate search word can be weighted and summed according to the weight corresponding to each type of historical interaction information to obtain the interaction score of each candidate search word. For example, assuming that the weight corresponding to the click rate is 1, the weight corresponding to the add-to-cart rate is 1.5, and the weight corresponding to the purchase rate is 2, and assuming that the click rate corresponding to candidate search word a is 0.4, the add-to-cart rate is 0.2, and the purchase rate is 0.1, each type of historical interaction information is weighted and summed based on the weight to obtain the interaction score of candidate search word a, which is 0.4*1+0.2*1.5+0.1*2=0.9. Assuming that the click rate corresponding to candidate search word b is 0.3, the add-to-cart rate is 0.1, and the purchase rate is 0.05, each type of historical interaction information is weighted and summed based on the weight to obtain the interaction score of candidate search word b, which is 0.3*1+0.1*1.5+0.05*2=0.55.

[0064] After the interaction score of each candidate search word is calculated based on the historical interaction information corresponding to each candidate search word, the candidate search words can be sorted based on the interaction score, and the target candidate search word can be selected from the sorted candidate search words. The candidate search words can be sorted based on the interaction score, and the target candidate search word can be selected from the sorted candidate search words in various manners. For example, the candidate search words can be sorted in descending order according to the interaction score, so that the target candidate search word can be obtained by cutting off the sorted candidate search words according to the required number of target candidate search words.

[0065] The target candidate search word can be sorted based on the transaction feature information of each target candidate search word. For example, when the transaction feature information is a transaction price, the target candidate search word can be sorted from high to low according to the transaction price to obtain the sorted target candidate search word. When the transaction feature information is a transaction concentrated area, the target candidate search word can be sorted from near to far according to the distance between the transaction concentrated area and the location of the user to obtain the sorted target candidate search word. When the transaction feature information is a transaction concentrated period, the target candidate search word can be sorted from near to far according to the time difference between the transaction concentrated period and the search time of the user to obtain the sorted target candidate search word. When the transaction feature information is a transaction frequency, the target candidate search word can be sorted from high to low according to the transaction frequency to obtain the sorted target candidate search word.

[0066] In this way, by sorting the candidate search words according to the transaction feature information and the historical interaction information corresponding to each candidate search word and displaying the candidate search words to the user, the user can quickly and intuitively obtain the candidate search words that the user may be interested in, thereby improving the possibility of the user purchasing the objects recommended based on the candidate search words, improving the search conversion rate, and further improving the revenue in the search scenario.

[0067] In step 105, the candidate search word is displayed.

[0068] For example, the candidate search word can be displayed on the terminal of the target user for the target user to select.

[0069] For example, the candidate search word can be displayed on the terminal of the target user for the target user to select.

[0070] Optionally, the candidate search word can also include a popular search word.

[0071] Optionally, after the candidate search word is sorted based on the transaction feature information and the historical interaction information of each target candidate search word to obtain the sorted target candidate search word, the sorted target candidate search word can be displayed on the terminal of the target user.

[0072] For example, please refer to Figure 3 , Figure 3is a search word expansion diagram of an object recommendation method provided by an embodiment of the present application. When a target user inputs a first keyword "hello" in an input box, candidate search words corresponding to the first keyword, such as "hello city" and "hello gentleman", can be sorted and then issued to a terminal of the target user, so that the candidate search words associated with the first keyword "hello" input by the target user, such as "hello city" and "hello gentleman", can be displayed in a search interface of the terminal. Therefore, the target user can quickly obtain the search word required for searching based on the associated candidate search word without completely inputting the required keyword, thereby improving the object search efficiency.

[0073] In step 106, in response to the keyword determination event, a target keyword corresponding to the keyword determination event is obtained, and a plurality of target objects matching the target keyword are determined.

[0074] The keyword determination event can be an event of determining a keyword required for searching by the target user, for example, can include that the target user inputs a keyword in an input box and triggers a search operation, or the target user selects a candidate search word from candidate search words associated with a first keyword based on the first keyword. The target keyword can be a search word triggered by the target user for searching. The target object can be an object similar to the target keyword.

[0075] In an embodiment, in the recall stage, initial objects of multiple matching types can be obtained according to multiple matching dimensions, and then in the offline sorting stage, business indicators are considered, the initial objects are sorted according to the estimated revenue corresponding to the initial objects, so as to improve the game revenue in the search scene.

[0076] Specifically, before the target keyword corresponding to the keyword determination event is obtained and the plurality of target objects matching the target keyword are determined in response to the keyword determination event, initial objects matching a plurality of preset search words are obtained based on at least one matching dimension; the initial objects matching the search words on each matching dimension are sorted based on a priority corresponding to the matching dimension, to obtain sorted initial objects; a second purchase probability and a price corresponding to the initial objects are determined; second estimated revenue information of the initial objects is calculated based on the second purchase probability and the price; and the sorted initial objects under each matching dimension are sorted based on the second estimated revenue information, to obtain candidate objects corresponding to each search word.

[0077] The step of determining the plurality of target objects matching the target keyword can include: filtering a target search word matching the target keyword from the search words; and determining the plurality of target objects matching the target keyword based on the candidate objects corresponding to the target search word.

[0078] The matching dimensions can include complete matching, partial matching, and similar matching, and different matching dimensions correspond to different matching degrees. Each matching dimension can be pre-set with a corresponding priority. The higher the matching degree, the higher the priority. For example, the priority of complete matching is higher than that of partial matching, and the priority of partial matching is higher than that of similar matching. The initial object can be an object matched with each search term based on each matching dimension. The second purchase probability can be the probability that the initial object is purchased, i.e., the probability that the initial object is purchased by all users. The second estimated revenue information can be information indicating the revenue that the initial object can bring. The target search term can be a search term in the pre-set search terms that matches the target keyword determined by the target user.

[0079] The initial object matched with the pre-set search terms based on at least one matching dimension can be obtained in various ways. For example, an offline recall algorithm can be used to perform content recall based on semantic matching. The semantic of the search term is extracted by performing word segmentation and embedding on the search term, so as to match the search term with the object to be recalled. Specifically, for complete matching, for example, from mineral to mineral tracker, simple string matching can be used. If the name of the object to be recalled is completely matched with or contains the search term, the object can be recalled as an initial object. For partial matching, for example, from mineral tracker to mineral sapling, word segmentation can be performed based on the search item pool, and the importance of word segmentation is calculated. The lower the frequency, the higher the importance. The search term and the search item are segmented, and if there is the same segmentation, it is marked as matching. The importance of word segmentation is added to obtain a matching degree index. The matching degree index is sorted and truncated to recall. For similar matching, for example, from the Summer Palace to the Forbidden City, a language representation model (Bidirectional Encoder Representations from Transformers, referred to as Bert) can be used to calculate the vectorized expression of the object name and the search term, so as to calculate the inner product of the vectorized expression of the search term and the object name, i.e., the matching degree of the search term and the object. Based on the matching degree, the corresponding initial object is obtained by truncation recall.

[0080] The initial objects matched by the search word on each matching dimension can be sorted based on the priority of the matching dimension corresponding to the search word in various ways. For example, assuming that 100 initial objects are matched based on the matching dimension of complete matching corresponding to the search word c, 500 initial objects are matched based on the matching dimension of partial matching corresponding to the search word c, and 1000 initial objects are matched based on the matching dimension of similar matching corresponding to the search word c, the initial objects of the search word c on each matching dimension can be sorted according to the priority of the matching dimension corresponding to the search word, wherein the initial objects matched based on the matching dimension of complete matching are ranked from 1 to 100, the initial objects matched based on the matching dimension of partial matching are ranked from 101 to 600, and the initial objects matched based on the matching dimension of similar matching are ranked from 601 to 1600.

[0081] Optionally, the second purchase probability can include at least one of a probability that the initial object is purchased in a search scenario, a probability that the initial object is purchased in all scenarios, and a probability that a similar object corresponding to the initial object is purchased. The similar object can be an object similar to the initial object.

[0082] The second purchase probability corresponding to the initial object and the price can be determined in various ways. For example, the relevant logs of all users can be obtained, the second purchase probability corresponding to each initial object can be counted according to the relevant logs, and the price of each initial object can be determined.

[0083] The second estimated revenue information of the initial object can be calculated based on the second purchase probability and the price in various ways. For example, the product of the second purchase probability and the price corresponding to each initial object can be calculated to obtain the second estimated revenue information of each initial object.

[0084] The manner of sorting the sorted initial objects under each matching dimension based on the second estimated revenue information to obtain the candidate objects corresponding to each search term can be various, for example, the sorted initial objects under each matching dimension are sorted from high to low according to the corresponding second estimated revenue information to obtain the candidate objects corresponding to each search term. For example, assuming that the initial objects of the search term c under each matching dimension are sorted according to the priority corresponding to each matching dimension, the initial objects matched based on the complete matching matching dimension are ranked 1-100, the initial objects matched based on the partial matching matching dimension are ranked 101-600, and the initial objects matched based on the similar matching matching dimension are ranked 601-1600, then the arrangement order of each initial object in the ranking 1-100 under the complete matching matching dimension can be adjusted according to the second estimated revenue information corresponding to each initial object under the complete matching matching dimension, the arrangement order of each initial object in the ranking 101-600 under the partial matching matching dimension can be adjusted according to the second estimated revenue information corresponding to each initial object under the partial matching matching dimension, and the arrangement order of each initial object in the ranking 601-1600 under the similar matching matching dimension can be adjusted according to the second estimated revenue information corresponding to each initial object under the similar matching matching dimension.

[0085] In an embodiment, taking an object as a component in a UGC game as an example, refer to Figure 4a , Figure 4a is a specific architecture diagram of an object recommendation method provided by the embodiment of the application. The search term recall content, the component recall content, the component feature, and the player feature can be stored in the data layer corresponding to the game. The search term recall content can be the recalled search term, the component recall content can be the recalled component, the component feature can be the object feature, and the player feature can be the user feature. In this way, the candidate search term can be obtained by scoring the search term based on the collaborative filtering algorithm and recalling the content based on the rule based on the data stored in the data layer, thereby improving the efficiency of search term determination. In addition, the search term association and sorting can be performed according to the keyword input by the user. Specifically, the sorting can be performed based on the click rate corresponding to the search term, or the sorting can be performed based on the purchase rate after the jump, and the search term input and acquisition efficiency can be further improved based on the rule mixed recall content. In addition, the recalled components can be sorted offline, for example, the sorting can be performed based on the semantic matching degree, and the sorting can be performed based on the estimated revenue corresponding to each component, thereby improving the game revenue of the search scene.

[0086] In step 107, the target objects are recommended based on the target objects.

[0087] The recommendation based on the plurality of target objects can be performed in various manners. For example, a first purchase probability of the target user for the target object can be determined; first estimated revenue information of the target object can be determined based on the first purchase probability; the target object can be sorted based on the estimated revenue information; and the target user can be recommended based on the sorted target object.

[0088] The first purchase probability of the target user for the target object can be determined in various manners. For example, a user feature of the target user and an object feature of the target object can be obtained; and the first purchase probability of the target user for the target object can be predicted based on the user feature, the target keyword, and the object feature of the target object.

[0089] The user feature can be information indicative of the target user, and past behaviors of the user constitute user tags, and the user feature can be obtained based on the tags. The object feature can be information indicative of the target object.

[0090] Optionally, the first purchase probability of the target user for the target object can be predicted based on the user feature, the target keyword, and the object feature of the target object by using an artificial intelligence model. The artificial intelligence model can include a machine learning model (such as LightGBM) or a deep learning model (such as Bert).

[0091] Optionally, if the signal transmission quality of the terminal of the target user satisfies a preset abnormal condition, a second purchase probability of the target object being purchased can be obtained as the first purchase probability of the target user for the target object.

[0092] The preset abnormal condition can be a condition for determining that the signal transmission quality of the terminal of the target user is abnormal. For example, the preset abnormal condition can include that a data packet transmission rate of the terminal of the target user is less than a preset rate. The second purchase probability includes at least one of a probability of the target object being purchased in a search scenario, a probability of the target object being purchased in all scenarios, and a probability of a similar object corresponding to the target object being purchased.

[0093] In this way, when the signal transmission quality of the terminal of the target user is poor, the second purchase probability of the target object being purchased can be obtained based on the result of offline sorting, the first estimated revenue information of the target object can be calculated based on the second purchase probability, the target object can be sorted based on the estimated revenue information, and the response time of the object search process can be shortened, thereby improving the object search rate.

[0094] In an embodiment, when the signal transmission quality of the terminal of the target user is poor, the target object can be directly recommended according to the result of offline sorting, the response time of the object search process can be shortened, and the object search experience can be improved.

[0095] The first estimated revenue information of the target object can be determined in various manners based on the first purchase probability. For example, the price of the target object can be obtained, and the product of the price of the target object and the first purchase probability can be calculated to obtain the first estimated revenue information of the target object.

[0096] Then, the target objects can be ranked from high to low based on the estimated revenue information, so that the ranked target objects can be pushed to the terminal of the target user for recommendation.

[0097] For example, taking the object as the component as an example, refer to Figure 4b , Figure 4b is another specific architecture diagram of an object recommendation method provided by the embodiment of the present application. The data layer can store component recall offline sorting content, component features and player features. The component recall offline sorting content can be candidate objects corresponding to each search word obtained in the recall stage and the offline sorting stage. In this way, after the target user determines the target keyword, an online sorting can be performed in response to a request message (i.e., an http request) from the terminal to the server end. The sorting is based on the semantic matching degree and the LightGBM model. That is, the first purchase probability of the target user for the target object can be determined based on the player features, the target keyword and the component features through the LightGBM model. The first estimated revenue information of the target object can be determined based on the first purchase probability and the component price. In this way, the target objects can be ranked from high to low based on the estimated revenue information. The target objects with higher preset revenue can be preferentially recommended to the target user based on the ranked target objects, so as to improve the game revenue brought by the search scene, and further improve the object recommendation efficiency.

[0098] In an embodiment, taking the object as the component as an example, refer to Figure 4c , Figure 4cis a search system architecture schematic diagram of an object recommendation method provided by an embodiment of the present application. In the data layer of the search system, search association logs, search-after-click-purchase logs, player-related logs, and component-related logs can be stored. Based on these logs, the purchase rate of components and the transaction feature information and historical interaction information of each search term can be determined. In this way, based on recall algorithms such as hot recall, collaborative filtering recall, and matching recall, multiple search terms and components of multiple matching dimensions can be recalled, and through feature extraction algorithms, player features of players and component features of components can be extracted based on related logs in advance, so that based on the above data, an offline sorting algorithm can be used for offline sorting, and based on statistics plus rule sorting and machine learning model estimation of a second purchase rate, sorting can be performed, so that search term recommendation results, association term recommendation results, and component recommendation results can be stored in a remote dictionary service (Remote Dictionary Server, Redis for short) database, so that when a target user determines a target keyword, online sorting algorithms can be used to perform online sorting on offline sorting results, specifically, based on statistics plus rule sorting and machine learning model estimation of a first purchase rate, sorting can be performed, and then based on an http request triggered by a terminal based on an http interface, a corresponding search term recommendation interface, a search input association interface, and a component search interface can be called to push corresponding content to a user terminal, so that search term recommendation, search term association, and component recommendation can be implemented.

[0099] In this way, an embodiment of the present application provides a fusion recommendation algorithm that uses search term recommendation, introduces input association, and covers the whole process of search. First, the search term recommendation and search input association are introduced to fully utilize the interface before and during search, and the user search path is shortened. Second, the personalized search term recommendation is used based on user historical search big data, and the user search experience is improved. In the sorting stage, the comprehensive benefits and semantic relevance of the search scene are considered, and the benefit index and search accuracy of the search scene are improved. Through the optimization practice of an embodiment of the present application based on the whole link of the search scene, more than 50% of players can improve the search experience through search term recommendation and search input association, and the search intention can be achieved without inputting all search terms, and the average number of input characters is relatively reduced by 8.1%. The introduction of big data personalized recommendation helps players to better find products of their own intentions, and after iteration, the number of searches of players is reduced compared with before, and the average number of searches per purchase is relatively reduced by 6.4%. In the sorting stage, the semantic matching and product benefit index multidimension are considered, and through strict AB testing online, the search scene benefit is improved by 5% after optimization.

[0100] In some sandbox games based on open UGC platform, creators of the entire game world continuously produce high-quality content and fresh gameplay in the game, and build an open ecosystem together with the majority of players and developers. A large number of components created by developers are an important part of the game, and the component search consumption flow accounts for a high proportion of the entire game. When searching, if there is a specific target component, the search system needs to accurately find the target component, and if there is no specific target component, the search system needs to recommend components that meet the search intent for the player. However, the early object recommendation scheme in the game and the open source general search scheme based on the search data analysis engine (Elasticsearch) do not fully utilize the display capabilities of the search entrance before and during searching, and do not consider the game revenue indicators, and cannot improve the player experience and game revenue through personalized recommendation. Therefore, the embodiments of the present application utilize player search big data, through multiple algorithm stages such as matching recall, offline sorting, online sorting, fully utilize the interface recommendation search word before and during searching, and recommend search words and components based on player historical search big data, utilize various natural language processing algorithms for more detailed semantic matching, and consider semantic relevance and revenue indicators during sorting, so as to realize search word recommendation, input association and component search recommendation, thereby constructing a game UGC content search system that integrates recommendation algorithms and covers the entire search process. The practical data indicators based on the embodiments of the present application show that the embodiments of the present application not only improve the experience of player component search, so that the player can find the required purchased components more quickly, but also recommend components with higher estimated revenue when the player's search target is ambiguous, which can significantly improve game revenue. Through strict AB testing online, the search scheme provided by the embodiments of the present application can bring more than 5% improvement to product revenue, and can effectively improve the efficiency of object search and recommendation.

[0101] From the above, the embodiment of the application receives the first keyword input by the user through the input box, determines the candidate search word matched with the first keyword, determines the historical interaction information corresponding to the candidate search word on at least one user interaction behavior, determines the transaction feature information of the candidate search word based on the transaction information of the first candidate object similar to the candidate search word, sorts the candidate search word based on the transaction feature information and the historical interaction information corresponding to each candidate search word, displays the candidate search word, acquires the target keyword corresponding to the keyword determination event in response to the keyword determination event, and determines a plurality of target objects matched with the target keyword. The recommendation is made based on the plurality of target objects. In this way, the candidate search word associated with the first keyword input by the user in the input box is determined, then the candidate search word is sorted according to the historical interaction information and the transaction feature information corresponding to the candidate search word, so that the candidate search word is recommended to the user for selection based on the sorting result, which can enable the user to quickly and accurately determine the target keyword to be searched, thereby determining a plurality of target objects matched with the target keyword and recommending the target objects to the user, achieving efficient and accurate recommendation of the objects of interest to the user. Meanwhile, the search word is recommended based on the transaction feature information, which can improve the probability of the user purchasing the recommended target object, thereby improving the revenue brought by the search scene and further improving the object recommendation efficiency.

[0102] To better implement the above method, the embodiment of the application further provides an object recommendation device, which can be integrated in an electronic device, which can be a server.

[0103] For example, as shown in Figure 5 Fig. 1 is a structural schematic diagram of the object recommendation device provided by the embodiment of the application, which can include a receiving unit 201, a first determining unit 202, a second determining unit 203, a sorting unit 204, a display unit 205, a third determining unit 206, and a recommendation unit 207, as follows:

[0104] The receiving unit 201 is configured to receive the first keyword input through the input box and determine the candidate search word matched with the first keyword.

[0105] The first determining unit 202 is configured to determine the historical interaction information corresponding to the candidate search word on at least one user interaction behavior.

[0106] The second determining unit 203 is configured to determine the transaction feature information of the candidate search word based on the transaction information of the first candidate object similar to the candidate search word.

[0107] The sorting unit 204 is configured to sort the candidate search word based on the transaction feature information and the historical interaction information corresponding to each candidate search word.

[0108] The display unit 205 is configured to display the candidate search terms.

[0109] The third determination unit 206 is configured to, in response to the keyword determination event, acquire a target keyword corresponding to the keyword determination event, and determine a plurality of target objects matched with the target keyword.

[0110] The recommendation unit 207 is configured to recommend based on the plurality of target objects.

[0111] In some embodiments, the transaction information includes at least one of a transaction price, a buyer location, a transaction time, and a transaction frequency.

[0112] The price feature information of the candidate search term is calculated based on the transaction price of the first candidate object similar to the candidate search term.

[0113] And / or, the transaction concentrated area of the first candidate object under the same candidate search term is determined based on the buyer location of the first candidate object similar to the candidate search term, and the transaction area feature information of the candidate search term is obtained based on the transaction concentrated area.

[0114] And / or, the transaction concentrated time period of the first candidate object under the same candidate search term is determined based on the transaction time of the first candidate object similar to the candidate search term, and the transaction time period feature information of the candidate search term is obtained based on the transaction concentrated time period.

[0115] And / or, the transaction frequency feature information of the candidate search term is calculated based on the transaction frequency of the first candidate object similar to the candidate search term.

[0116] In some embodiments, the historical interaction information includes at least one of a click rate, an add-to-cart rate, and a purchase rate.

[0117] In some embodiments, the sorting unit 204 is configured to:

[0118] The interaction score of each candidate search term is calculated based on the historical interaction information corresponding to the candidate search term.

[0119] The candidate search terms are sorted based on the interaction scores, and the target candidate search terms are selected from the sorted candidate search terms.

[0120] The target candidate search terms are sorted based on the transaction feature information of each target candidate search term, and the sorted target candidate search terms are obtained.

[0121] The display unit 205 is configured to:

[0122] The sorted target candidate search terms are displayed.

[0123] In some embodiments, the recommendation unit 207 comprises:

[0124] A purchase rate determination subunit configured to determine a first purchase probability of the target object by the target user;

[0125] A revenue determination subunit configured to determine first estimated revenue information of the target object based on the first purchase probability;

[0126] An object recommendation subunit configured to sort the target object based on the estimated revenue information, and recommend the target user based on the sorted target object.

[0127] In some embodiments, the purchase rate determination subunit is configured to:

[0128] Obtain user features of the target user and object features of the target object;

[0129] Predict the first purchase probability of the target object by the target user based on the user features, the target keyword, and the object features of the target object.

[0130] In some embodiments, the purchase rate determination subunit is configured to:

[0131] If the signal transmission quality of the terminal of the target user meets a preset abnormal condition, obtain a second purchase probability of the target object being purchased as the first purchase probability of the target object by the target user.

[0132] In some embodiments, the second purchase probability comprises at least one of a probability of the target object being purchased in a search scenario, a probability of the target object being purchased in all scenarios, and a probability of a similar object corresponding to the target object being purchased.

[0133] In some embodiments, the object recommendation apparatus further comprises an offline sorting unit configured to:

[0134] Obtain initial objects matched with a plurality of preset search keywords based on at least one matching dimension;

[0135] Sort the initial objects matched with the search keywords in each matching dimension based on a priority corresponding to the matching dimension to obtain sorted initial objects;

[0136] Determine a second purchase probability and a price corresponding to the initial object;

[0137] Calculate second estimated revenue information of the initial object based on the second purchase probability and the price;

[0138] Sort the sorted initial objects in each matching dimension based on the second estimated revenue information to obtain candidate objects corresponding to each search keyword;

[0139] The third determining unit 206 is configured to:

[0140] filtering target search terms matching the target keyword from the search terms;

[0141] determining a plurality of target objects matching the target keyword based on the candidate objects corresponding to the target search terms.

[0142] In some embodiments, the object recommendation apparatus further comprises a search term recommendation unit configured to:

[0143] determining a second candidate object of interest to the target user and a like score of the target user to the second candidate object based on user association information of the target user;

[0144] obtaining association search terms associated with the second candidate object;

[0145] determining an interest score of each association search term based on the like score of the second candidate object associated with the association search term;

[0146] ranking the association search terms based on the interest scores of the association search terms;

[0147] recommending the search terms through a terminal of the target user based on the ranking results of the association search terms.

[0148] In specific implementation, each of the above units can be implemented as an independent entity, or can be combined as the same or several entities, and the specific implementation of each of the above units can be referred to the method embodiments above, which will not be described herein.

[0149] From the above, the embodiment of the present application receives the first keyword input by the input box through the receiving unit 201, determines the candidate search word matched with the first keyword, the first determining unit 202 determines the historical interaction information corresponding to the candidate search word on at least one user interaction behavior, the second determining unit 203 determines the transaction feature information of the candidate search word based on the transaction information of the first candidate object similar to the candidate search word, the sorting unit 204 sorts the candidate search word based on the transaction feature information and the historical interaction information corresponding to each candidate search word, the display unit 205 displays the candidate search word, the third determining unit 206 acquires the target keyword corresponding to the keyword determination event and determines a plurality of target objects matched with the target keyword in response to the keyword determination event, and the recommendation unit 207 makes recommendations based on the plurality of target objects. In this way, by determining the candidate search word associated with the first keyword input by the user in the input box, then sorting the candidate search word according to the historical interaction information and the transaction feature information corresponding to the candidate search word, and recommending the candidate search word to the user for selection based on the sorting result, the user can quickly and accurately determine the target keyword to be searched, thereby determining a plurality of target objects matched with the target keyword and recommending the target objects to the user, achieving efficient and accurate recommendation of the objects of interest to the user. At the same time, based on the transaction feature information, the search word is recommended, which can improve the probability of the user purchasing the recommended target object, thereby improving the revenue brought by the search scene, and further improving the object recommendation efficiency.

[0150] The embodiment of the present application also provides an electronic device, such as Figure 6 As shown in the figure, it shows the structure schematic diagram of the electronic device related to the embodiment of the present application, which can be a server, specifically:

[0151] The electronic device 300 includes a processor 301 having one or more processing cores, a memory 302 having one or more computer readable storage media, and a computer program stored on the memory 302 and executable on the processor. The processor 301 is electrically connected to the memory 302. Those skilled in the art can understand that the electronic device structure shown in the figure does not constitute a limitation on the electronic device, and can include more or fewer components than the figure, or combine certain components, or different component arrangements.

[0152] The processor 301 is the control center of the electronic device 300, which connects all parts of the electronic device 300 through various interfaces and lines, executes various functions of the electronic device 300 and processes data by running or loading the software program and / or module stored in the memory 302, and calling the data stored in the memory 302, thereby overall monitoring the electronic device 300.

[0153] In the embodiments of the present application, the processor 301 in the electronic device 300 loads the instructions corresponding to the processes of one or more application programs into the memory 302 and runs the application programs stored in the memory 302 by the processor 301 to implement various functions according to the following steps:

[0154] The input box receives the inputted first keyword, and determines candidate search words matched with the first keyword;

[0155] The historical interaction information corresponding to the candidate search words on at least one user interaction behavior is determined;

[0156] The transaction feature information of the candidate search words is determined based on the transaction information of the first candidate objects similar to the candidate search words;

[0157] The candidate search words are sorted based on the transaction feature information and the historical interaction information corresponding to the candidate search words;

[0158] The candidate search words are displayed;

[0159] In response to a keyword determination event, the target keyword corresponding to the keyword determination event is obtained, and a plurality of target objects matched with the target keyword are determined;

[0160] Recommendation is made based on the plurality of target objects.

[0161] The present scheme can receive the inputted first keyword through the input box, determine the candidate search words matched with the first keyword, determine the historical interaction information corresponding to the candidate search words on at least one user interaction behavior, determine the transaction feature information of the candidate search words based on the transaction information of the first candidate objects similar to the candidate search words, sort the candidate search words based on the transaction feature information and the historical interaction information corresponding to the candidate search words, display the candidate search words, in response to a keyword determination event, obtain the target keyword corresponding to the keyword determination event, and determine a plurality of target objects matched with the target keyword, and make recommendation based on the plurality of target objects. In this way, by determining the candidate search words associated with the first keyword inputted by the user in the input box, and then sorting the candidate search words according to the historical interaction information and the transaction feature information corresponding to the candidate search words, the candidate search words are recommended to the user for selection based on the sorting result, which can enable the user to quickly and accurately determine the target keyword to be searched, and thus recommend a plurality of target objects matched with the target keyword to the user, thereby achieving efficient and accurate recommendation of the objects of interest to the user. Meanwhile, the search words are recommended based on the transaction feature information, which can improve the probability of the user purchasing the recommended target objects, thereby improving the revenue brought by the search scene, and further improving the object recommendation efficiency.

[0162] The specific implementation of each operation can refer to the foregoing embodiments, which will not be described here again.

[0163] Optionally, as shown in FIG. 3, Figure 6 The processor 301 is electrically connected with the touch display screen 303, the radio frequency circuit 304, the audio circuit 305, the input unit 306 and the power supply 307. Those skilled in the art can understand that the electronic device structure shown in FIG. 3 does not constitute a limitation on the electronic device, and can include more or fewer components than those shown, or combine certain components, or different component arrangements. Figure 6 The processor 301 is electrically connected with the touch display screen 303, the radio frequency circuit 304, the audio circuit 305, the input unit 306 and the power supply 307. Those skilled in the art can understand that the electronic device structure shown in FIG. 3 does not constitute a limitation on the electronic device, and can include more or fewer components than those shown, or combine certain components, or different component arrangements.

[0164] The touch display screen 303 can be used to display a graphical user interface and receive operation instructions generated by a user acting on the graphical user interface. The touch display screen 303 can include a display panel and a touch panel. The display panel can be used to display information input by a user or information provided to a user and various graphical user interfaces of the electronic device, which can be composed of graphics, text, icons, videos and any combination thereof. Optionally, the display panel can be configured in the form of a liquid crystal display (LCD), an organic light-emitting diode (OLED) or the like. The touch panel can be used to collect touch operations of a user thereon or adjacent thereto (such as operations of a user using a finger, a stylus or any suitable object or accessory on or adjacent to the touch panel), and generate corresponding operation instructions, and the operation instructions execute corresponding programs. Optionally, the touch panel can include two parts of a touch detection device and a touch controller. The touch detection device detects the touch position of the user and detects the signals brought by the touch operation, and transmits the signals to the touch controller; the touch controller receives the touch information from the touch detection device, and converts it into touch coordinates, and then sends it to the processor 301, and can also receive commands from the processor 301 and execute them. The touch panel can cover the display panel, and when the touch panel detects a touch operation thereon or adjacent thereto, it is transmitted to the processor 301 to determine the type of the touch event, and then the processor 301 provides corresponding visual output on the display panel according to the type of the touch event. In the embodiments of the present application, the touch panel and the display panel can be integrated into the touch display screen 303 to realize the input and output functions. However, in some embodiments, the touch panel and the touch panel can realize the input and output functions as two independent components. That is, the touch display screen 303 can also realize the input function as part of the input unit 306.

[0165] The radio frequency circuit 304 can be used to transceive radio frequency signals to establish wireless communication with network devices or other electronic devices.

[0166] The audio circuit 305 can be used to provide an audio interface between the user and the electronic device through a speaker and a microphone. The audio circuit 305 can convert received audio data into an electrical signal and transmit the electrical signal to the speaker for conversion into an audible signal output by the speaker. On the other hand, the microphone collects a sound signal and converts the sound signal into an electrical signal, which is received by the audio circuit 305 and converted into audio data. The audio data is output to the processor 301 for processing, and then transmitted to another electronic device through the radio frequency circuit 304, or output to the memory 302 for further processing. The audio circuit 305 can also include a headphone jack to provide communication between an external device and the electronic device.

[0167] The input unit 306 can be used to receive inputted digital, character information or user feature information (such as fingerprint, iris, face information, etc.), and generate keyboard, mouse, joystick, optical or trackball signal inputs related to user settings and function control.

[0168] The power supply 307 is used to supply power to various components of the electronic device 300. Optionally, the power supply 307 can be logically connected to the processor 301 through a power management system, so that the power management system can be used to manage charging, discharging, power consumption management and other functions. The power supply 307 can also include one or more direct current or alternating current power sources, recharging systems, power failure detection circuits, power converters or inverters, power status indicators, and any other components.

[0169] Although Figure 6 The electronic device 300 can also include a camera, a sensor, a wireless fidelity module, a Bluetooth module, etc., which are not shown in the embodiments.

[0170] In the above embodiments, the description of each embodiment has its own focus, and the parts not described in detail in a certain embodiment can be referred to the related description of other embodiments. It should be noted that the electronic device provided by the embodiments of the present application and the object recommendation method suitable for the above embodiments belong to the same concept, and the specific implementation process is described in detail in the above method embodiments, which will not be repeated here.

[0171] As can be seen from the above, the electronic device provided in this application embodiment can receive the input first keyword through an input box, determine candidate search terms that match the first keyword; determine the historical interaction information corresponding to the candidate search terms in at least one user interaction behavior; determine the transaction feature information of the candidate search terms based on the transaction information of the first candidate objects similar to each candidate search term; sort the candidate search terms based on the transaction feature information and historical interaction information corresponding to each candidate search term; display the candidate search terms; in response to a keyword determination event, obtain the target keyword corresponding to the keyword determination event, and determine multiple target objects that match the target keyword; and make recommendations based on the multiple target objects. Therefore, by identifying candidate search terms associated with the first keyword currently entered by the user in the input box, and then ranking these candidate search terms based on their historical interaction information and transaction characteristics, the system recommends these candidate search terms to the user based on the ranking results. This allows users to quickly and accurately determine their target keywords and recommend multiple target objects that match those keywords. This achieves efficient and accurate recommendation of objects that users are interested in. Furthermore, recommending search terms based on transaction characteristics increases the probability of users purchasing the recommended target objects, thereby increasing the benefits in the search scenario and ultimately improving the efficiency of object recommendation.

[0172] Those skilled in the art will understand that all or part of the steps in the various methods of the above embodiments can be performed by a computer program, or by a computer program controlling related hardware. The computer program can be stored in a computer-readable storage medium and loaded and executed by a processor.

[0173] Therefore, embodiments of this application provide a computer-readable storage medium storing a computer program that can be loaded by a processor to execute the steps in any of the object recommendation methods provided in embodiments of this application. For example, the computer program can execute the following steps:

[0174] The system receives the first keyword entered through the input box and determines candidate search terms that match the first keyword.

[0175] Determine the historical interaction information corresponding to candidate search terms in at least one user interaction behavior;

[0176] Based on the transaction information of the first candidate object that is similar to each candidate search term, the transaction feature information of the candidate search terms is determined;

[0177] The candidate search terms are ranked based on the transaction feature information and historical interaction information corresponding to each candidate search term;

[0178] Display candidate search terms;

[0179] In response to the keyword determination event, a target keyword corresponding to the keyword determination event is acquired, and a plurality of target objects matching the target keyword are determined;

[0180] Recommendation is performed based on the plurality of target objects.

[0181] The scheme can receive the first keyword input in the input box, determine the candidate search word matching the first keyword, determine the historical interaction information corresponding to the candidate search word on at least one user interaction behavior, determine the transaction feature information of the candidate search word based on the transaction information of the first candidate object similar to the candidate search word, sort the candidate search word based on the transaction feature information and the historical interaction information corresponding to each candidate search word, display the candidate search word, in response to the keyword determination event, acquire the target keyword corresponding to the keyword determination event, and determine a plurality of target objects matching the target keyword, and perform recommendation based on the plurality of target objects. In this way, by determining the candidate search word associated with the first keyword input by the user in the input box, and then sorting the candidate search word according to the historical interaction information and transaction feature information corresponding to the candidate search word, the candidate search word is recommended to the user for selection based on the sorting result, which can enable the user to quickly and accurately determine the target keyword to be searched, thereby determining a plurality of target objects matching the target keyword and recommending the target objects to the user, achieving efficient and accurate recommendation of objects of interest to the user to the user. At the same time, the search word is recommended based on the transaction feature information, which can improve the probability of the user purchasing the recommended target object, thereby improving the revenue brought by the search scene, and further improving the object recommendation efficiency.

[0182] The specific implementation of each operation can be referred to the foregoing embodiments, which will not be described here.

[0183] The computer readable storage medium can include a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, etc.

[0184] Due to the computer program stored in the computer readable storage medium, the steps of any object recommendation method provided by the embodiments of the present application can be performed, and thus the beneficial effects of any object recommendation method provided by the embodiments of the present application can be achieved. Details are described in the foregoing embodiments, which will not be described here.

[0185] According to an aspect of the present application, there is provided a computer program product comprising a computer program stored in a computer readable storage medium; when a processor of an electronic device reads the computer program from the computer readable storage medium, the processor executes the computer program, so that the electronic device executes the method provided in various optional implementation manners provided by the above embodiments.

[0186] The above describes in detail the object recommendation method, device, storage medium and electronic device provided by the embodiments of the present application. The principles and implementation manners of the present application are described by applying specific examples. The above embodiment descriptions are only used to help understand the method of the present application and its core idea. Meanwhile, for those skilled in the art, according to the idea of the present application, the specific implementation manners and application ranges can be changed. In summary, the content of the specification should not be understood as limiting the present application.

Claims

1. An object recommendation method characterized by comprising: The method comprises: obtaining initial objects matching a plurality of preset search terms; determining a second purchase probability and a price corresponding to the initial objects; calculating second estimated revenue information of the initial objects based on the second purchase probability and the price; determining candidate objects corresponding to each search term in the initial objects based on the second estimated revenue information; receiving an input first keyword through an input box, and determining candidate search terms matching the first keyword; determining historical interaction information corresponding to the candidate search terms on at least one user interaction behavior; determining transaction feature information of the candidate search terms based on transaction information of first candidate objects similar to the candidate search terms; sorting the candidate search terms based on the transaction feature information and the historical interaction information corresponding to each candidate search term; displaying the candidate search terms; in response to a keyword determination event, obtaining a target keyword corresponding to the keyword determination event, and screening a target search term matching the target keyword from the search terms; determining a plurality of target objects matching the target keyword based on the candidate objects corresponding to the target search term; determining a first purchase probability of a target user for the target objects; determining first estimated revenue information of the target objects based on the first purchase probability; based on the first estimated revenue information and the target objects, recommending the target user.

2. The object recommendation method according to claim 1, wherein The transaction information includes at least one of transaction price, buyer location, transaction time, and transaction frequency. The transaction information includes at least one of transaction price, buyer location, transaction time, and transaction frequency. The transaction information includes at least one of transaction price, buyer location, transaction time, and transaction frequency. The transaction information includes at least one of transaction price, buyer location, transaction time, and transaction frequency. The transaction information includes at least one of transaction price, buyer location, transaction time, and transaction frequency. The historical interaction information includes at least one of click rate, add-to-cart rate, and purchase rate.

3. The object recommendation method according to claim 1, wherein The sorting of the candidate search terms based on the transaction feature information and the historical interaction information corresponding to each candidate search term comprises:

4. The object recommendation method according to claim 1, wherein calculating an interaction score of each candidate search term based on the historical interaction information corresponding to each candidate search term; sorting the candidate search terms based on the interaction score, and screening a target candidate search term from the sorted candidate search terms; and ​ rank the target candidate search words based on the transaction feature information of each of the target candidate search words to obtain ranked target candidate search words; the method further includes: displaying the ranked target candidate search words.

5. The object recommendation method according to claim 1, wherein the method further includes: ranking the target objects based on the first estimated revenue information, and recommending the target user based on the ranked target objects.

6. The object recommendation method according to claim 1, wherein the method further includes: obtaining user features of the target user and object features of the target object; predicting a first purchase probability of the target user to purchase the target object based on the user features, the target keyword, and the object features of the target object.

7. The object recommendation method according to claim 1, wherein the method further includes: if the signal transmission quality of the terminal of the target user meets a preset abnormal condition, obtaining a second purchase probability of the target object being purchased as the first purchase probability of the target user to purchase the target object.

8. The object recommendation method according to claim 7, wherein the second purchase probability includes at least one of a probability of the target object being purchased in a search scenario, a probability of the target object being purchased in all scenarios, and a probability of a similar object corresponding to the target object being purchased.

9. The object recommendation method according to claim 1, wherein the method further includes: obtaining initial objects matching the preset plurality of search words based on at least one matching dimension; the method further includes: ranking the initial objects matching the search words in each of the matching dimensions based on a priority corresponding to the matching dimension to obtain ranked initial objects; ranking the ranked initial objects under each matching dimension based on the second estimated revenue information to obtain candidate objects corresponding to each of the search words.

10. The object recommendation method according to any one of claims 1 to 9, wherein the method further includes: determining a second candidate object of interest to the target user and a preference score of the target user to the second candidate object based on user association information of the target user; obtaining association search words associated with the second candidate object; determining an interest score of each of the association search words based on the preference score of the second candidate object associated with the association search word; ranking the association search words based on the interest scores of the association search words; performing search word recommendation through the terminal of the target user based on the ranking result of the association search words.

11. An object recommendation apparatus characterized by comprising: the method further includes: an offline ranking unit, configured to obtain initial objects matching a preset plurality of search words, determine second purchase probabilities and prices of the initial objects, calculate second estimated revenue information of the initial objects based on the second purchase probabilities and the prices, and determine candidate objects corresponding to each of the search words in the initial objects based on the second estimated revenue information. The receiving unit is configured to receive the input first keyword through an input box, and determine candidate search words matched with the first keyword; The first determining unit is configured to determine historical interaction information corresponding to the candidate search words in at least one user interaction behavior; The second determining unit is configured to determine transaction feature information of the candidate search words based on transaction information of first candidate objects similar to the candidate search words; The sorting unit is configured to sort the candidate search words based on the transaction feature information and the historical interaction information corresponding to each of the candidate search words; The display unit is configured to display the candidate search words; The third determining unit is configured to, in response to a keyword determination event, acquire a target keyword corresponding to the keyword determination event, filter out a target search word matched with the target keyword from the search words, and determine a plurality of target objects matched with the target keyword based on the candidate objects corresponding to the target search word; The recommendation unit is configured to determine a first purchase probability of a target user for the target objects, determine first estimated revenue information of the target objects based on the first purchase probability, and recommend the target user based on the first estimated revenue information and the target objects.

12. An electronic device, comprising: The device comprises a processor and a memory, wherein the memory stores a computer program, and when the computer program is executed by the processor, the processor executes the steps of the method in any one of claims 1-10.

13. A computer-readable storage medium, characterized in that, The device comprises a computer program, and when the computer program runs on an electronic device, the computer program is used to make the electronic device execute the steps of the method in any one of claims 1-10.

Images