Article recommendation method and device

Abstract

The invention provides an article recommendation method and device, and the method comprises the steps: screening out a first article sequence corresponding to an initial article set from an article set under a first category, adding first articles in the first article sequence into the initial article set, and obtaining an intermediate article set corresponding to each first article; screening out second article sequences corresponding to the intermediate article sets from the article sets under the second category, and adding second articles in the second article sequences into the corresponding intermediate article sets to obtain updated intermediate article sets; and according to each updated intermediate item set corresponding to the initial item set, determining a target item set corresponding to the initial item set so as to perform item recommendation according to each target item set. According to the embodiment, dependence on user behavior data can be reduced, deep association between the articles can be mined, and the quality and efficiency of article recommendation can be improved.

Description

Technical Field

[0001] The present invention relates to the field of computer technology, and in particular to an item recommendation method and device. Background Art

[0002] In the field of e-commerce, by recommending to users a combination of items associated with a target item, the sales of various items can be promoted. For example, when a user is browsing a top, a pair of jeans, a pair of sneakers, etc. can be recommended to him. However, the existing technology usually relies on user behavior data to recommend items, and the quality and efficiency of item recommendations based on new users or new items are low. In addition, the method of recommending items using pre-set recommendation rules in the existing technology has great deficiencies in item association mining, and it is difficult to meet the personalized needs of users, resulting in low quality of item recommendations. Summary of the invention

[0003] In view of this, the embodiments of the present disclosure at least provide an item recommendation method and device, which can reduce the dependence on user behavior data, explore the deep connections between items, and improve the quality and efficiency of item recommendations.

[0004] In a first aspect, an embodiment of the present disclosure provides an item recommendation method, comprising:

[0005] Selecting a first item sequence corresponding to the initial item set from the item set under the first category, and adding the first items in the first item sequence to the initial item set to obtain an intermediate item set corresponding to each first item;

[0006] Selecting second item sequences corresponding to each intermediate item set from the item set under the second category, and adding the second items in the second item sequences to the corresponding intermediate item sets, to obtain updated intermediate item sets;

[0007] According to each updated intermediate item set corresponding to the initial item set, a target item set corresponding to the initial item set is determined, so as to perform item recommendation according to each target item set.

[0008] Optionally, determining a target item set corresponding to the initial item set according to each updated intermediate item set corresponding to the initial item set includes:

[0009] Determine whether the lengths of each updated intermediate item set corresponding to the initial item set meet preset requirements;

[0010] In response to the length of the updated intermediate item set satisfying a preset requirement, using the updated intermediate item set as a target item set corresponding to the initial item set;

[0011] In response to the length of the updated intermediate item set not meeting the preset requirement, third item sequences corresponding to each updated intermediate item set are screened out from the item set under the new second category, and the third items in the third item sequences are respectively added to the corresponding updated intermediate item set. The above steps are repeated until the length of the updated intermediate item set meets the preset requirement, so as to determine the target item set corresponding to the initial item set.

[0012] Optionally, selecting a first item sequence corresponding to the initial item set from the item set under the first category includes:

[0013] determining a first characteristic of items in the initial set of items;

[0014] Determine a second feature of each item in the set of items under the first category based on the item information corresponding to the set of items under the first category;

[0015] The similarity between the first feature and each second feature is calculated, and a first item sequence is generated based on items corresponding to the second features whose similarity with the first feature exceeds a preset threshold.

[0016] Optionally, the method further comprises:

[0017] Using a pre-trained feature similarity model, a first item sequence corresponding to the initial item set is selected from the item set under the first category;

[0018] The feature similarity model is trained based on the following steps:

[0019] Each item in a predetermined item collocation sequence is used as a starting query sample; the item collocation sequence is obtained based on the historical behavior data of the target user;

[0020] Based on the items in the item collocation sequence except the initial query sample, a first positive sample set corresponding to each initial query sample is formed, and a first negative sample set corresponding to each initial query sample is determined in the candidate item pool; the candidate item pool does not include the items in the item collocation sequence;

[0021] The first positive sample items in the first positive sample set are added to the corresponding initial query samples respectively to obtain intermediate query samples corresponding to each first positive sample item, and based on the intermediate query samples, a second positive sample set and a second negative sample set corresponding to the intermediate query samples are determined;

[0022] Add the second positive sample items in the second positive sample set to the corresponding intermediate query samples respectively to obtain updated intermediate query samples;

[0023] Based on the updated intermediate query samples corresponding to the item collocation sequence, a training data set of the feature similarity model is determined to obtain a feature similarity model based on the training data set; the training data set includes a first positive sample set, a first negative sample set, a second positive sample set and a second negative sample set.

[0024] Optionally, the method further comprises:

[0025] Determine whether the sample length of the updated intermediate query sample meets the preset requirement;

[0026] In response to the sample length of the updated intermediate query sample meeting the preset requirement, the first positive sample set, the first negative sample set, the second positive sample set and the second negative sample set are determined as training data sets, so as to obtain a feature similarity model based on the training data sets;

[0027] In response to the sample length of the updated intermediate query sample not meeting the preset requirement, based on the item collocation sequence and the updated intermediate query sample, a third positive sample set and a third negative sample set corresponding to the updated intermediate query sample are determined, and the third positive sample items in the third positive sample set are respectively added to the corresponding updated intermediate query sample, and the above steps are repeated until the sample length of the updated intermediate query sample meets the preset requirement, so as to determine the training data set and train to obtain the feature similarity model.

[0028] Optionally, determining the first positive sample set, the first negative sample set, the second positive sample set, and the second negative sample set as a training data set, and obtaining a feature similarity model by training based on the training data set, includes:

[0029] The training data set is input into the feature similarity model, and the first goal is to make the feature similarity between the first positive sample set and the starting query sample approach to the first threshold, and the feature similarity between the second positive sample set and the intermediate query sample approach to the first threshold; the second goal is to make the feature similarity between the first negative sample set and the starting query sample approach to the second threshold, and the feature similarity between the second negative sample set and the intermediate query sample approach to the second threshold; the feature similarity model is obtained by training; the first threshold and the second threshold are pre-set.

[0030] In a second aspect, an embodiment of the present disclosure provides an item recommendation device, comprising:

[0031] A screening module, configured to screen out first item sequences corresponding to the initial item set from the item set under the first category, and add the first items in the first item sequences to the initial item set respectively, to obtain an intermediate item set corresponding to each first item;

[0032] An updating module, configured to select second item sequences corresponding to each intermediate item set from the item set under the second category, and add the second items in the second item sequences to the corresponding intermediate item sets, respectively, to obtain an updated intermediate item set;

[0033] The recommendation module is used to determine the target item set corresponding to the initial item set according to each updated intermediate item set corresponding to the initial item set, so as to make item recommendations according to each target item set.

[0034] In a third aspect, an embodiment of the present disclosure further provides an electronic device, comprising: a processor, a memory and a bus, wherein the memory stores machine-readable instructions executable by the processor, and when the computer device is running, the processor and the memory communicate through the bus, and when the machine-readable instructions are executed by the processor, the above-mentioned first aspect, or the steps in any optional implementation of the first aspect are performed.

[0035] In a fourth aspect, an embodiment of the present disclosure further provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the steps of the above-mentioned first aspect, or any optional implementation of the first aspect are executed.

[0036] In a fifth aspect, an embodiment of the present disclosure further provides a computer program product, including a computer program, which implements the method of any of the above embodiments when executed by a processor.

[0037] Any of the above aspects or any implementation methods of any of the aspects, by gradually constructing a multi-level item association set related to the initial item set, reduces the direct dependence on user behavior data, and at the same time mines the deep association between items. Specifically, the first item sequence associated with the initial item set is screened out from the item set under the first category, and it is added to the initial item set one by one to generate an intermediate item set related to each first item. This process is based on the attribute association of the preset first category and the item itself, not the user behavior data, and makes full use of the intrinsic attributes of the item. In addition, by screening out the second item sequence related to each intermediate item set from the item set under the second category, and adding these items to the corresponding intermediate item set, the associated items corresponding to the initial item set are further enriched. Based on all updated intermediate item sets, a target item set associated with the initial item set can be generated, and items are recommended to users based on the target item set. Since the embodiment of the present disclosure focuses on the attribute relationship between items in different categories and association levels, rather than relying on user behavior data such as clicks, browsing or purchases, potential item associations can still be mined in the case of sparse data or insufficient user behavior samples. At the same time, through layer-by-layer screening and association expansion, the randomness or one-sidedness of the recommendation results is avoided, thereby improving the accuracy and coverage of the recommendations, and being able to efficiently capture the associations between items, thereby improving the quality and efficiency of item recommendations.

[0038] The effects of the above-mentioned item recommendation device, electronic device and storage medium can be found in the description of the above-mentioned item recommendation method, which will not be repeated here. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] In order to more clearly illustrate the technical solutions of the embodiments of the present disclosure, the following is a brief introduction to the drawings required for use in the embodiments. The drawings herein are incorporated into the specification and constitute a part of the specification. These drawings illustrate embodiments consistent with the present disclosure and are used together with the specification to illustrate the technical solutions of the present disclosure. It should be understood that the following drawings only illustrate certain embodiments of the present disclosure and should not be regarded as limiting the scope. For ordinary technicians in this field, other relevant drawings can also be obtained based on these drawings without creative work.

[0040] Figure 1 A flowchart of an item recommendation method provided by an embodiment of the present disclosure is shown;

[0041] Figure 2 A schematic diagram of a flow chart of an item recommendation method provided by an embodiment of the present disclosure is shown;

[0042] Figure 3 A schematic diagram of a flow chart of a method for determining a training data set for a feature similarity model provided in an embodiment of the present disclosure is shown;

[0043] Figure 4 A schematic diagram of an item recommendation device provided by an embodiment of the present disclosure is shown;

[0044] Figure 5 An exemplary system architecture is shown in which embodiments of the present disclosure may be applied;

[0045] Figure 6 A schematic diagram of the structure of a computer system of a terminal device or a server used to implement an embodiment of the present disclosure is shown. DETAILED DESCRIPTION

[0046] In order to make the purpose, technical scheme and advantages of the embodiments of the present disclosure clearer, the technical scheme in the embodiments of the present disclosure will be clearly and completely described below in conjunction with the drawings in the embodiments of the present disclosure. Obviously, the described embodiments are only part of the embodiments of the present disclosure, rather than all of the embodiments. The components of the embodiments of the present disclosure generally described and shown in the drawings here can be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present disclosure provided in the drawings is not intended to limit the scope of the present disclosure for protection, but merely represents the selected embodiments of the present disclosure. Based on the embodiments of the present disclosure, all other embodiments obtained by those skilled in the art without making creative work belong to the scope of protection of the present disclosure.

[0047] It should be noted that in the technical solution of the present invention, the collection, use, storage, sharing and transfer of user personal information involved are in compliance with the provisions of relevant laws and regulations, and it is necessary to inform the user and obtain the user's consent or authorization. When applicable, the user's personal information is de-identified and / or anonymized and / or encrypted.

[0048] The raising of the above problems and the solutions are the results obtained by the inventor after practice and careful research. The discovery process of the above problems and the solutions proposed for the above problems should be the contributions made by the inventor to the present disclosure during the disclosure process.

[0049] The technical solutions in the present disclosure will be clearly and completely described below in conjunction with the drawings in the present disclosure. Obviously, the described embodiments are only part of the embodiments of the present disclosure, rather than all of the embodiments. The components of the present disclosure generally described and shown in the drawings here can be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present disclosure provided in the drawings is not intended to limit the scope of the present disclosure claimed for protection, but merely represents the selected embodiments of the present disclosure. Based on the embodiments of the present disclosure, all other embodiments obtained by those skilled in the art without making creative work belong to the scope of protection of the present disclosure.

[0050] It should be noted that similar reference numerals and letters denote similar items in the following drawings, and therefore, once an item is defined in one drawing, further definition and explanation thereof is not required in subsequent drawings.

[0051] To facilitate understanding of this embodiment, a method for recommending items disclosed in the embodiment of the present disclosure is first introduced in detail. The execution subject of the method for recommending items provided in the embodiment of the present disclosure is generally a computer device with certain computing capabilities, and the computer device includes, for example: a terminal device or a server or other processing device, and the terminal device may be a user equipment (UE), a mobile device, a user terminal, a terminal, a cellular phone, a cordless phone, a personal digital assistant (PDA), a handheld device, a computing device, a vehicle-mounted device, a wearable device, etc. In some possible implementations, the method for recommending items may be implemented by a processor calling a computer-readable instruction stored in a memory.

[0052] See also Figure 1 FIG. 1 is a flowchart of an item recommendation method provided by an embodiment of the present disclosure, and the method includes S101 to S103, wherein:

[0053] S101: Filtering out first item sequences corresponding to an initial item set from an item set under a first category, and adding first items in the first item sequences to the initial item set respectively, to obtain an intermediate item set corresponding to each first item.

[0054] In the embodiment of the present disclosure, the initial item set may include target items to be associated with the target items; for example, if the user is currently browsing item A, item A may be used as the initial item set, and the recommended item combinations associated with item A may be determined in subsequent steps. In specific implementations, item recommendations may be made based on any number of target items, that is, the number of items in the initial item set may be one or more, and the specific items and item quantity in the initial item set may be set according to actual conditions and needs.

[0055] In the disclosed embodiment, in order to facilitate the selection of items from different categories, the categories here can adopt three-level categories. The third-level category refers to the third level of classification in the classification system divided according to the hierarchical structure, which is a further subdivision based on the first-level category and the second-level category. Generally, the deeper the classification level, the more specific and refined the category. Specifically, the first-level category represents a large category, representing the main classification of items or content, such as electronic products, fresh products, etc.; the second-level category is subdivided on the basis of the first-level category, such as mobile phones, milk, etc.; the third-level category is a further subdivision under the second-level category, which is more specific, such as smart phones, high-calcium milk, etc.

[0056] In a specific implementation, a category sequence can be predetermined, and the category sequence can include multiple categories of items, indicating that it is expected to select items from these categories for matching recommendations. For the category sequence [category1, category2, category3, category4], category1 can represent the item category of the target item in the initial item set. When there are multiple target items in the initial item set, the corresponding item category can also be added to the category sequence. Category2-4 can represent the expectation to select items that match the target items in the initial item set from these three categories of items for item recommendation. In the above example, category2 is the first category, that is, the first category can be the first category in the category sequence other than the target item in the initial item set. Exemplarily, assuming that the target item in the initial item set is a newly released smartphone of brand A, the corresponding category sequence can be [smartphone, mobile phone case, tempered glass film, mobile phone stand], indicating that for the smartphone, it is expected to select items that match the target item from the three categories of mobile phone case, tempered glass film, and mobile phone stand.

[0057] It should be noted that the first category, the second category, etc. in the embodiment of the present disclosure can be predetermined, for example, the category sequence corresponding to the target items of different categories can be stored in the storage location according to the predetermined rules, and when the category sequence corresponding to the current target item needs to be determined, it can be directly obtained from the storage location; or it can also be determined based on big data, for example, the current big data shows that users who have purchased smart phones usually buy mobile phone cases, then the mobile phone case can be used as one of the categories corresponding to the smart phone, so as to select items that match the smart phone from this category; when the big data shows that the needs of users in the current scene have changed, for example, after a period of time, the big data shows that users who have purchased smart phones no longer buy mobile phone cases, but mobile phone bags, at this time, according to the new data, the category corresponding to the smart phone can be updated to the mobile phone bag, so as to select items that match the smart phone from the mobile phone bag category for recommendation. The above-mentioned method of determining the first category, the second category, etc. corresponding to the target item is only used as an example of possible implementation methods in the embodiment of the present disclosure, and does not constitute an improper limitation of the present invention. In actual applications, it can be set according to actual conditions. The embodiment of the present disclosure does not make specific limitations on this, and its function can be realized. In addition, the embodiments of the present disclosure do not impose specific limitations on the number of categories in the category sequence, the corresponding relationship between the categories and the target items in the initial items, etc. In actual applications, they can be set according to actual conditions.

[0058] In the embodiment of the present disclosure, after the initial item set and the first sequence are determined, the first item sequence corresponding to the initial item set can be screened out from the item set under the first category. Specifically, the first item sequence corresponding to the initial item set can be screened out from the item set under the first category, including: determining the first feature of the items in the initial item set; determining the second feature of each item in the item set under the first category based on the item information corresponding to the item set under the first category; calculating the similarity between the first feature and each second feature, and generating the first item sequence based on the items corresponding to the second feature whose similarity to the first feature exceeds a preset threshold.

[0059] In a specific implementation, the first feature of the items in the initial item set can be extracted first, and then the item set under the first category can be screened out from the full item pool according to the first category, and then the second feature of each item in the item set under the first category can be extracted. Then the similarity between the second feature and the first feature of different first items can be calculated respectively, and the first item sequence corresponding to the initial item set can be generated based on the items corresponding to the n second features with the highest similarity. It should be noted that any method can be used to extract the above-mentioned first feature and second feature; for example, features can be manually designed according to domain knowledge or business rules to extract feature information of dimensions such as keywords, price ranges, brands, and materials from the product description; open source text processing models can also be used to extract the features of the item based on item information. The above is only an example of some feasible feature extraction methods in the embodiments of the present disclosure, and does not constitute an improper limitation on the present invention. In practical applications, the method of extracting item features can be determined according to actual conditions and needs. The embodiments of the present disclosure do not make specific limitations on this, and the function can be realized.

[0060] For example, assuming that the initial item set includes item A, the item set under the first category is [ ], after feature extraction, the feature of item A in the initial item set is , the corresponding features of each item in the item set under the first category are , , and ; At this time, we can calculate and The similarity between and The similarity between and The similarity between and The similarity between . Assume and The similarity between them is 0.78. and The similarity between them is 0.4. and The similarity between them is 0.52. and The similarity between them is 0.83; if the goal is to take the two first items with the highest similarity to generate the first item sequence, then the first item sequence corresponding to the initial item set obtained in this step can be [ ]. It should be noted that the above selection of the n items with the highest similarity to generate the first item sequence corresponding to the initial item set is only used as an example of a possible implementation method in the embodiment of the present disclosure. In actual applications, a threshold can also be set, and all items in the item set under the first category that exceed the threshold are used to generate the first item sequence; or, weights can be set for each item in the item set under the first category according to the user's preference. For example, the user data corresponding to the current user indicates that the user prefers a certain specific pattern. Then, when selecting the first item to generate the first item sequence, the weight value of the item with the corresponding pattern can be adjusted to a higher value. In this way, even if the similarity between the item and the items in the initial item set is not within the selection range, it can be placed in the first item sequence. The above content is only used as an example of a feasible implementation method in the embodiment of the present disclosure, and does not constitute an improper limitation on the present invention. In actual applications, it can be selected and set based on actual conditions and needs. The embodiment of the present disclosure does not make specific limitations on this, and its function can be realized.

[0061] In the disclosed embodiment, a pre-trained feature similarity model can also be used to extract the second feature of each first item in the item set under the first category. In a specific implementation, the input information of the feature similarity model can be obtained based on the SKU (Stock Keeping Unit) of the target item in the initial item set. For example, the prompt word and the SKU of the target item can be input into the feature similarity model so that the feature similarity model can obtain the first feature of the target item in the initial item set based on the instruction; the prompt word here can be "generate a feature vector for this item to be used for matching related items". When the initial item set includes more than one target item, these target items can appear in the form of an item list, and the prompt word at this time can be "generate a feature vector for this item list to be used for matching related items". The training method of the above-mentioned feature similarity model will be described in S201~S205 below, and will not be repeated here.

[0062] Figure 2 The flowchart of the item recommendation method provided by the embodiment of the present disclosure is as follows. Figure 2As shown, after obtaining the first item sequence, each first item in the first item sequence can be added to the initial item set to obtain an intermediate item set corresponding to each first item. For example, continuing the example described above, when determining that the first item sequence is [ ], then there are two corresponding intermediate item sets, namely [A ] and [A ].

[0063] S102: Filter out second item sequences corresponding to each intermediate item set from the item set under the second category, and add the second items in the second item sequences to the corresponding intermediate item sets, to obtain updated intermediate item sets.

[0064] In this step, after the intermediate item set is obtained, each second item sequence corresponding to each intermediate item set can be screened out from the item set under the second category. The screening method is similar to the method described above for screening out the first item sequence corresponding to the initial item set from the item set under the first category, and will not be repeated here. It should be noted that the first category and the second category are different, so the items that match the target product in the initial item set are selected from different categories for item recommendation.

[0065] In the specific implementation, Figure 2 As shown, for each intermediate item set, steps similar to those in S101 can be repeated to select the second item sequence corresponding to each intermediate item set. ] The second item sequence determined is [ ], then after adding the second item in the second item sequence to the intermediate item set, we can get the intermediate item set [A ] corresponds to the two updated intermediate item sets [A ] and [A ] Similarly, assuming that for the intermediate item set [A ] The second item sequence determined is [ ], then with the intermediate item set [A ]The corresponding updated intermediate item set is [A ]、[A ] and [A ]. It should be noted that the second item sequences determined for different intermediate item sets may include the same second items. For repeated second items, the weight of the item can be increased when the target item set is finally generated, indicating the strong correlation between the item and the target item in the initial item set.

[0066] S103: Determine a target item set corresponding to the initial item set according to each updated intermediate item set corresponding to the initial item set, so as to make item recommendations according to each target item set.

[0067] In this step, after obtaining the updated intermediate item set, the length of the updated intermediate item set can be first judged, and the target item set can be determined according to the judgment result. Specifically, according to each updated intermediate item set corresponding to the initial item set, the target item set corresponding to the initial item set is determined, including: judging whether the length of each updated intermediate item set corresponding to the initial item set meets the preset requirements; in response to the length of the updated intermediate item set meeting the preset requirements, the updated intermediate item set is used as the target item set corresponding to the initial item set; in response to the length of the updated intermediate item set not meeting the preset requirements, each third item sequence corresponding to each updated intermediate item set is selected from the item set under the new second category, and the third items in the third item sequence are respectively added to the corresponding updated intermediate item set, and the above steps are repeated until the length of the updated intermediate item set meets the preset requirements, so as to determine the target item set corresponding to the initial item set.

[0068] In a specific implementation, when its length meets the preset requirements, each updated intermediate item set can be used as the final target item set corresponding to the initial item set. The preset requirements here can be determined in advance, can be determined according to the number of different categories in the category sequence, or can be determined according to the specific quantity requirements in the current item recommendation scenario, and the embodiment of the present disclosure does not specifically limit this. For example, assuming that the predetermined category sequence is [category1, category2, category3], then the preset length of the target item set can be set to 3, that is, the target item set needs to include three items, of which the first is the target item in the initial item set, the second item represents the item selected in category2 that matches the target item, and the third item represents the item selected in category3 that matches the target item, and finally the target item set corresponding to the initial item set is obtained. When the length of the updated intermediate item set does not meet the preset requirements, the screening steps described above can be repeated to screen out the third item sequences corresponding to each updated intermediate item set from the item set under the new second category, and the third items of the third item sequence are added to the corresponding updated intermediate item set. The steps of selecting the third item sequence from the item set under the new second category and adding each third item in each third item sequence to the corresponding updated intermediate item set are similar to the steps described above and will not be repeated here. After adding the third items in each third item sequence to the corresponding updated intermediate item set, it can be determined again whether the length of the updated intermediate item set meets the preset requirements, until the obtained updated intermediate item set meets the preset requirements, it can be determined as the target item set corresponding to the initial item set. It should be noted that the new second category described above is different from the first category and the original second category. Assuming that the predetermined category sequence is [category1, category2, category3, category4], then the first category is category2, the second category is category3, and the new second category is category4.

[0069] In another possible implementation, in addition to using the length of the updated intermediate item set as the basis for determining whether to determine it as the final target item set, other methods may be used to determine whether to determine the updated intermediate item set as the final target item set. For example, a standard item set may be preset. When the updated intermediate item set includes an item in the standard item set, the process of screening out a new item sequence is stopped, and the current updated intermediate item set is determined as the target item set. Otherwise, the process of screening the item sequence and adding the items therein to the corresponding updated intermediate item set may be continued. The disclosed embodiment does not specifically limit the preset requirements that the updated intermediate item set needs to meet. In actual applications, it can be set according to actual conditions to achieve its function.

[0070] A method for training a feature similarity model provided by an embodiment of the present disclosure is described below. The method includes S201 to S205, wherein:

[0071] S201: taking each item in a predetermined item collocation sequence as a starting query sample; the item collocation sequence is obtained based on historical behavior data of the target user.

[0072] In the embodiments of the present disclosure, before training the feature similarity model, it is necessary to first determine the item matching sequence corresponding to the target user. In a specific implementation, the item matching sequence can be determined by the target user's browsing, clicking, exposure records and other behavioral data; for example, the target user first browsed a smartphone and then browsed a mobile phone case, then the smartphone and the mobile phone case can form the item matching sequence corresponding to the target user. In addition, the item matching sequence corresponding to the target user can also be determined based on the attributes of the items. For example, if the target user browsed a sweatshirt, then the item matching sequence corresponding to the target user can be generated based on items such as jeans and hats that are also clothing. The above method for determining the item matching sequence is only exemplified as a possible implementation method in the embodiments of the present disclosure and does not constitute an improper limitation on the present invention.

[0073] Figure 3 A flow chart of a training data set for determining a feature similarity model provided by an embodiment of the present disclosure. Figure 3 As shown in the figure, assuming that the item collocation sequence corresponding to the target user is [A, B, C, D], then A, B, C, and D can be used as the starting query samples respectively.

[0074] S202: Based on the items in the item collocation sequence except the initial query sample, a first positive sample set corresponding to each initial query sample is formed, and a first negative sample set corresponding to each initial query sample is determined in a candidate item pool; the candidate item pool does not include the items in the item collocation sequence.

[0075] Continuing with the example in the previous text, after obtaining four starting query samples, you can start traversing from the first item A in sequence, or you can traverse the four starting query samples in parallel at the same time. The embodiment of the present disclosure does not make specific restrictions on this. Taking the starting query sample [A] as an example, the traversal process of other starting query samples is the same as the traversal process of the starting query sample, and the embodiment of the present disclosure will not be repeated here. Specifically, you can first construct the prompt word "generate a representation for this list of items for matching related items", and take item A as the starting query sample. The corresponding first positive sample set is [B, C, D]. The negative sample set can be extracted from the full pool of items except for the items in the item matching sequence. The extraction method and quantity can be set according to the actual situation. The embodiment of the present disclosure does not make specific restrictions on this. In this way, the first positive sample set [B, C, D] and the first negative sample set [rand sample] corresponding to the starting query sample [A] can be obtained.

[0076] S203: Add the first positive sample items in the first positive sample set to the corresponding starting query samples respectively to obtain intermediate query samples corresponding to each first positive sample item, and based on the intermediate query samples, determine the second positive sample set and the second negative sample set corresponding to the intermediate query samples.

[0077] In this step, the decibels of each first positive sample item in the first positive sample set can be added to the corresponding initial query sample. For example, still taking the initial query sample [A] as an example, Figure 3 As shown, the intermediate query samples [A, B], [A, C] and [A, D] can be obtained, and then for each intermediate query sample, the second positive sample set and the second negative sample set corresponding to each intermediate query sample can be obtained. Taking the intermediate query sample [A, B] as an example, the corresponding second positive sample set can be [C, D]; similarly, the second positive sample sets corresponding to the intermediate query samples [A, C] and [A, D] are [B, D] and [B, C] respectively. The method for determining the second negative sample set corresponding to each intermediate query sample is similar to the method for determining the first negative sample set mentioned above, and will not be repeated here.

[0078] S204: The second positive sample items in the second positive sample set are respectively added to the corresponding intermediate query samples to obtain updated intermediate query samples.

[0079] In this step, after obtaining the second positive sample set, the second positive sample items in the second positive sample set can be added to the corresponding intermediate query samples to obtain updated intermediate query samples. For example, taking the intermediate query sample [A, B] as an example, the corresponding second positive sample set is [C, D]. Each second positive sample item in it is added to the intermediate query sample, and two updated intermediate query samples [A, B, C] and [A, B, D] are obtained.

[0080] S205: Based on the updated intermediate query samples corresponding to the item matching sequence, determine a training data set of the feature similarity model to obtain a feature similarity model based on the training data set; the training data set includes a first positive sample set, a first negative sample set, a second positive sample set and a second negative sample set.

[0081] In the disclosed embodiment, after obtaining the updated intermediate query sample, it can be determined whether the sample length of the updated intermediate query sample meets the preset requirements to determine the training data set used to train the feature similarity model. Specifically, the method also includes: determining whether the sample length of the updated intermediate query sample meets the preset requirements; in response to the sample length of the updated intermediate query sample meeting the preset requirements, determining the first positive sample set, the first negative sample set, the second positive sample set and the second negative sample set as the training data set to obtain the feature similarity model based on the training data set training; in response to the sample length of the updated intermediate query sample not meeting the preset requirements, based on the item collocation sequence and the updated intermediate query sample, determining the third positive sample set and the third negative sample set corresponding to the updated intermediate query sample, respectively adding the third positive sample items in the third positive sample set to the corresponding updated intermediate query sample, and repeating the above steps until the sample length of the updated intermediate query sample meets the preset requirements to determine the training data set, and train to obtain the feature similarity model.

[0082] In a specific implementation, after obtaining the updated intermediate query sample, it is possible to first determine whether the sample length of the updated intermediate query sample meets the preset requirements. The preset requirements here can be determined based on the length of the item matching sequence, or can be determined based on actual conditions and needs. The present disclosed embodiment does not specifically limit the specific content of the preset requirements. In the case where the sample length of the updated intermediate query sample meets the preset requirements, the first positive sample set, the second positive sample set, the first negative sample set and the second negative sample set obtained previously can be directly used as the training data set for the feature similarity model. In the case where the updated intermediate query sample does not meet the preset requirements, the third positive sample set and the third negative sample set corresponding to the updated intermediate query sample are determined in a similar manner to that described above. The specific determination method is similar to the method for determining the second positive sample set and the second negative sample set in the previous text, and will not be repeated here. After determining the third positive sample set, the third positive sample items in the third positive sample set can be added to the corresponding updated intermediate query samples respectively, and then it can be determined again whether the sample length of the updated intermediate query samples after adding meets the preset requirements, until the obtained updated intermediate query samples meet the preset requirements, the training data set for training the feature similarity model can be determined.

[0083] In the embodiment of the present disclosure, after the above-mentioned step of determining the training data set is completed for each starting query sample, a complete training data set for training the feature similarity model can be obtained. Assuming that the training data set finally obtained includes a first positive sample set, a first negative sample set, a second positive sample set and a second negative sample set, a feature similarity model can be obtained by training based on the first positive sample set, the first negative sample set, the second positive sample set and the second negative sample set. Specifically, the first positive sample set, the first negative sample set, the second positive sample set and the second negative sample set are determined as the training data set, so as to obtain the feature similarity model based on the training data set, including: inputting the training data set into the feature similarity model, making the feature similarity between the first positive sample set and the starting query sample approach the first threshold, and making the feature similarity between the second positive sample set and the intermediate query sample approach the first threshold as the first goal; making the feature similarity between the first negative sample set and the starting query sample approach the second threshold, and making the feature similarity between the second negative sample set and the intermediate query sample approach the second threshold as the second goal; training to obtain the feature similarity model; the first threshold and the second threshold are pre-set.

[0084] In a specific implementation, after the first positive sample set, the first negative sample set, the second positive sample set and the second negative sample set are input into the feature similarity model, the feature results extracted by the model can be obtained, and the similarities between the first positive sample set, the first negative sample set, the second positive sample set, the second negative sample set and the corresponding starting query sample can be calculated. For the first positive sample set and the second positive sample set, the model can be trained so that their corresponding similarities approach 1; for the first negative sample set and the second negative sample set, the model can be trained so that their corresponding similarities approach 0, thereby training to obtain the final feature similarity model. It should be noted that the first threshold value that the positive sample set tends to and the second threshold value that the negative sample set tends to can be set according to actual conditions. The above description of the first threshold value as 1 and the second threshold value as 0 is only given as an example of a possible implementation method and does not constitute an improper limitation on the present invention.

[0085] According to a second aspect of the embodiment of the present disclosure, Figure 4 As shown, an item recommendation device 400 is provided, comprising:

[0086] A screening module 401 is used to screen out first item sequences corresponding to the initial item set from the item set under the first category, and add the first items in the first item sequence to the initial item set to obtain an intermediate item set corresponding to each first item;

[0087] An updating module 402 is used to select second item sequences corresponding to each intermediate item set from the item set under the second category, and add the second items in the second item sequences to the corresponding intermediate item sets to obtain updated intermediate item sets;

[0088] The recommendation module 403 is used to determine the target item set corresponding to the initial item set according to each updated intermediate item set corresponding to the initial item set, so as to make item recommendations according to each target item set.

[0089] Optionally, the recommendation module 403 is specifically used for:

[0090] Determine whether the lengths of each updated intermediate item set corresponding to the initial item set meet preset requirements;

[0091] In response to the length of the updated intermediate item set satisfying a preset requirement, using the updated intermediate item set as a target item set corresponding to the initial item set;

[0092] In response to the length of the updated intermediate item set not meeting the preset requirement, third item sequences corresponding to each updated intermediate item set are screened out from the item set under the new second category, and the third items in the third item sequences are respectively added to the corresponding updated intermediate item set. The above steps are repeated until the length of the updated intermediate item set meets the preset requirement, so as to determine the target item set corresponding to the initial item set.

[0093] Optionally, the screening module 401 is specifically used for:

[0094] determining a first characteristic of items in the initial set of items;

[0095] Determine a second feature of each item in the set of items under the first category based on the item information corresponding to the set of items under the first category;

[0096] The similarity between the first feature and each second feature is calculated, and a first item sequence is generated based on items corresponding to the second features whose similarity with the first feature exceeds a preset threshold.

[0097] Optionally, the screening module 401 is specifically used for:

[0098] Using a pre-trained feature similarity model, a first item sequence corresponding to the initial item set is selected from the item set under the first category;

[0099] The device also includes a model training module 404, which is specifically used to:

[0100] Each item in a predetermined item collocation sequence is used as a starting query sample; the item collocation sequence is obtained based on the historical behavior data of the target user;

[0101] Based on the items in the item collocation sequence except the initial query sample, a first positive sample set corresponding to each initial query sample is formed, and a first negative sample set corresponding to each initial query sample is determined in the candidate item pool; the candidate item pool does not include the items in the item collocation sequence;

[0102] The first positive sample items in the first positive sample set are added to the corresponding initial query samples respectively to obtain intermediate query samples corresponding to each first positive sample item, and based on the intermediate query samples, a second positive sample set and a second negative sample set corresponding to the intermediate query samples are determined;

[0103] Add the second positive sample items in the second positive sample set to the corresponding intermediate query samples respectively to obtain updated intermediate query samples;

[0104] Based on the updated intermediate query samples corresponding to the item collocation sequence, a training data set of the feature similarity model is determined to obtain a feature similarity model based on the training data set; the training data set includes a first positive sample set, a first negative sample set, a second positive sample set and a second negative sample set.

[0105] Optionally, the model training module 404 is further used to:

[0106] Determine whether the sample length of the updated intermediate query sample meets the preset requirement;

[0107] In response to the sample length of the updated intermediate query sample meeting the preset requirement, the first positive sample set, the first negative sample set, the second positive sample set and the second negative sample set are determined as training data sets, so as to obtain a feature similarity model based on the training data sets;

[0108] In response to the sample length of the updated intermediate query sample not meeting the preset requirement, based on the item collocation sequence and the updated intermediate query sample, a third positive sample set and a third negative sample set corresponding to the updated intermediate query sample are determined, and the third positive sample items in the third positive sample set are respectively added to the corresponding updated intermediate query sample, and the above steps are repeated until the sample length of the updated intermediate query sample meets the preset requirement, so as to determine the training data set and train to obtain the feature similarity model.

[0109] Optionally, the model training module 404 is specifically used for:

[0110] The training data set is input into the feature similarity model, and the first goal is to make the feature similarity between the first positive sample set and the starting query sample approach to the first threshold, and the feature similarity between the second positive sample set and the intermediate query sample approach to the first threshold; the second goal is to make the feature similarity between the first negative sample set and the starting query sample approach to the second threshold, and the feature similarity between the second negative sample set and the intermediate query sample approach to the second threshold; the feature similarity model is obtained by training; the first threshold and the second threshold are pre-set.

[0111] According to a third aspect of an embodiment of the present disclosure, there is provided an electronic device for item recommendation, comprising: one or more processors; a storage device for storing one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors implement the method provided by the first aspect of the embodiment of the present disclosure.

[0112] According to a fourth aspect of an embodiment of the present disclosure, a computer-readable medium is provided, on which a computer program is stored, and when the program is executed by a processor, the method provided by the first aspect of the embodiment of the present disclosure is implemented.

[0113] According to a fifth aspect of an embodiment of the present invention, a computer program product is provided, including a computer program, and the computer program implements the method of any of the above embodiments when executed by a processor.

[0114] Figure 5 An exemplary system architecture 500 is shown to which the item recommendation method or item recommendation apparatus implemented by the present disclosure can be applied.

[0115] like Figure 5 As shown, system architecture 500 may include terminal devices 501, 502, 503, a network 504 and a server 505. Network 504 is used to provide a medium for communication links between terminal devices 501, 502, 503 and server 505. Network 504 may include various connection types, such as wired, wireless communication links or optical fiber cables, etc.

[0116] Users can use terminal devices 501, 502, 503 to interact with server 505 through network 504 to receive or send messages, etc. Various communication client applications can be installed on terminal devices 501, 502, 503, such as shopping applications, web browser applications, search applications, instant messaging tools, email clients, social platform software, etc. (only for example).

[0117] The terminal devices 501 , 502 , and 503 may be various electronic devices having a display screen and supporting web browsing, including but not limited to smart phones, tablet computers, laptop computers, desktop computers, and the like.

[0118] The server 505 may be a server that provides various services, such as a backend management server (for example only) that provides support for shopping websites browsed by users using the terminal devices 501, 502, and 503. The backend management server may process the received item recommendation request and feed back the processing result (for example only) to the terminal device.

[0119] It should be noted that the method for recommending items provided in the embodiment of the present invention is generally executed by the server 505, and accordingly, the device for recommending items is generally set in the server 505. The method for recommending items provided in the embodiment of the present invention can also be executed by the terminal devices 501, 502, 503, and accordingly, the device for recommending items can be set in the terminal devices 501, 502, 503.

[0120] It should be understood that Figure 5 The number of terminal devices, networks and servers in the embodiment is only for illustration. Any number of terminal devices, networks and servers may be provided according to implementation requirements.

[0121] Reference below Figure 6, which shows a schematic diagram of the structure of a computer system 600 of a terminal device suitable for implementing an embodiment of the present invention. Figure 6 The terminal device shown is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present invention.

[0122] like Figure 6 As shown, the computer system 600 includes a central processing unit (CPU) 601, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 602 or a program loaded from a storage part 608 into a random access memory (RAM) 603. In the RAM 603, various programs and data required for the operation of the system 600 are also stored. The CPU 601, the ROM 602, and the RAM 603 are connected to each other through a bus 604. An input / output (I / O) interface 605 is also connected to the bus 604.

[0123] The following components are connected to the I / O interface 605: an input section 606 including a keyboard, a mouse, etc.; an output section 607 including a cathode ray tube (CRT), a liquid crystal display (LCD), etc., and a speaker, etc.; a storage section 608 including a hard disk, etc.; and a communication section 609 including a network interface card such as a LAN card, a modem, etc. The communication section 609 performs communication processing via a network such as the Internet. A drive 610 is also connected to the I / O interface 605 as needed. A removable medium 611, such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, etc., is installed on the drive 610 as needed, so that a computer program read therefrom is installed into the storage section 608 as needed.

[0124] In particular, according to the embodiments disclosed in the present invention, the process described above with reference to the flowchart can be implemented as a computer software program. For example, the embodiments disclosed in the present invention include a computer program product, which includes a computer program carried on a computer-readable medium, and the computer program contains program code for executing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from the network through the communication part 609, and / or installed from the removable medium 611. When the computer program is executed by the central processing unit (CPU) 601, the above-mentioned functions defined in the system of the present invention are executed.

[0125] It should be noted that the computer-readable medium shown in the present invention may be a computer-readable signal medium or a computer-readable storage medium or any combination of the above two. The computer-readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or device, or any combination of the above. More specific examples of computer-readable storage media may include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In the present invention, a computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in combination with an instruction execution system, device or device. In the present invention, a computer-readable signal medium may include a data signal propagated in a baseband or as part of a carrier wave, which carries a computer-readable program code. This propagated data signal may take a variety of forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above. Computer-readable signal media may also be any computer-readable medium other than computer-readable storage media, which may send, propagate or transmit a program for use by or in conjunction with an instruction execution system, apparatus or device. The program code contained on the computer-readable medium may be transmitted using any appropriate medium, including but not limited to: wireless, wire, optical cable, RF, etc., or any suitable combination of the above.

[0126] The flow chart and block diagram in the accompanying drawings illustrate the possible architecture, function and operation of the system, method and computer program product according to various embodiments of the present invention. In this regard, each box in the flow chart or block diagram can represent a module, a program segment, or a part of a code, and the above-mentioned module, program segment, or a part of a code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in a different order from the order marked in the accompanying drawings. For example, two boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram or flow chart, and the combination of the boxes in the block diagram or flow chart can be implemented with a dedicated hardware-based system that performs a specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0127] The modules involved in the embodiments of the present invention may be implemented by software or hardware. The modules described may also be set in a processor. For example, a processor includes a screening module, an updating module, and a recommendation module, wherein the names of these modules do not constitute limitations on the modules themselves in certain cases. For example, the updating module may also be described as "screening out a first item sequence corresponding to an initial item set from an item set under a first category, adding the first items in the first item sequence to the initial item set, and obtaining an intermediate item set module corresponding to each first item".

[0128] As another aspect, the present invention also provides a computer-readable medium, which may be included in the device described in the above embodiment; or may exist independently and not be assembled into the device. The above computer-readable medium carries one or more programs. When the above one or more programs are executed by a device, the device implements the following method: select a first item sequence corresponding to the initial item set from the item set under the first category, add the first items in the first item sequence to the initial item set respectively, and obtain an intermediate item set corresponding to each first item; select each second item sequence corresponding to each intermediate item set from the item set under the second category, add the second items in the second item sequence to the corresponding intermediate item set respectively, and obtain an updated intermediate item set; determine the target item set corresponding to the initial item set according to each updated intermediate item set corresponding to the initial item set, so as to recommend items according to each target item set.

[0129] Finally, it should be noted that the above embodiments are only specific implementation methods of the present disclosure, which are used to illustrate the technical solutions of the present disclosure, rather than to limit them. The protection scope of the present disclosure is not limited thereto. Although the present disclosure is described in detail with reference to the above embodiments, ordinary technicians in the field should understand that any technician familiar with the technical field can still modify the technical solutions recorded in the above embodiments within the technical scope disclosed in the present disclosure, or can easily think of changes, or make equivalent replacements for some of the technical features therein; and these modifications, changes or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present disclosure, and should be included in the protection scope of the present disclosure. Therefore, the protection scope of the present disclosure should be based on the protection scope of the claims.

Claims

1. An item recommendation method, characterized in that: include: Selecting a first item sequence corresponding to the initial item set from the item set under the first category, and adding the first items in the first item sequence to the initial item set to obtain an intermediate item set corresponding to each first item; Selecting second item sequences corresponding to the intermediate item sets from the item set under the second category, and adding the second items in the second item sequences to the corresponding intermediate item sets to obtain updated intermediate item sets; According to each updated intermediate item set corresponding to the initial item set, a target item set corresponding to the initial item set is determined, so as to perform item recommendation according to each target item set.

2. The method according to claim 1, characterized in that: Determining a target item set corresponding to the initial item set according to each updated intermediate item set corresponding to the initial item set includes: Determining whether the lengths of each updated intermediate item set corresponding to the initial item set meet preset requirements; In response to the length of the updated intermediate item set satisfying a preset requirement, using the updated intermediate item set as a target item set corresponding to the initial item set; In response to the length of the updated intermediate item set not meeting the preset requirement, third item sequences corresponding to each updated intermediate item set are screened out from the item set under the new second category, and the third items in the third item sequences are respectively added to the corresponding updated intermediate item set. The above steps are repeated until the length of the updated intermediate item set meets the preset requirement, so as to determine the target item set corresponding to the initial item set.

3. The method according to claim 1, characterized in that A first item sequence corresponding to the initial item set is selected from the item set under the first category, including: determining a first characteristic of an item in the initial set of items; Determining a second feature of each item in the set of items under the first category based on the item information corresponding to the set of items under the first category; The similarity between the first feature and each of the second features is calculated, and a first item sequence is generated based on items corresponding to second features whose similarity with the first feature exceeds a preset threshold.

4. The method according to claim 1, characterized in that: The method further comprises: Using a pre-trained feature similarity model, a first item sequence corresponding to the initial item set is selected from the item set under the first category; The feature similarity model is trained based on the following steps: Each item in a predetermined item collocation sequence is used as a starting query sample; the item collocation sequence is obtained based on historical behavior data of the target user; Based on the items in the item collocation sequence other than the initial query sample, a first positive sample set corresponding to each initial query sample is formed, and a first negative sample set corresponding to each initial query sample is determined in a candidate item pool; the candidate item pool does not include the items in the item collocation sequence; Add the first positive sample items in the first positive sample set to the corresponding initial query samples respectively to obtain intermediate query samples corresponding to each first positive sample item, and determine a second positive sample set and a second negative sample set corresponding to the intermediate query samples based on the intermediate query samples; Add the second positive sample items in the second positive sample set to the corresponding intermediate query samples respectively to obtain updated intermediate query samples; Based on the updated intermediate query sample corresponding to the item collocation sequence, a training data set of the feature similarity model is determined to obtain the feature similarity model through training based on the training data set; the training data set includes the first positive sample set, the first negative sample set, the second positive sample set and the second negative sample set.

5. The method according to claim 4, characterized in that The method further comprises: Determining whether the sample length of the updated intermediate query sample meets a preset requirement; In response to the sample length of the updated intermediate query sample meeting a preset requirement, determining the first positive sample set, the first negative sample set, the second positive sample set, and the second negative sample set as a training data set, so as to obtain the feature similarity model through training based on the training data set; In response to the sample length of the updated intermediate query sample not meeting the preset requirement, based on the item collocation sequence and the updated intermediate query sample, a third positive sample set and a third negative sample set corresponding to the updated intermediate query sample are determined, and the third positive sample items in the third positive sample set are respectively added to the corresponding updated intermediate query sample, and the above steps are repeated until the sample length of the updated intermediate query sample meets the preset requirement, so as to determine the training data set and train to obtain the feature similarity model.

6. The method according to claim 5, characterized in that Determining the first positive sample set, the first negative sample set, the second positive sample set, and the second negative sample set as a training data set, so as to obtain the feature similarity model through training based on the training data set, including: The training data set is input into the feature similarity model, and the first goal is to make the feature similarity between the first positive sample set and the starting query sample approach to a first threshold, and the feature similarity between the second positive sample set and the intermediate query sample approach to the first threshold; the second goal is to make the feature similarity between the first negative sample set and the starting query sample approach to a second threshold, and the feature similarity between the second negative sample set and the intermediate query sample approach to a second threshold; the feature similarity model is obtained by training; the first threshold and the second threshold are pre-set.

7. An item recommendation device, characterized in that: include: A screening module, configured to screen out first item sequences corresponding to the initial item set from the item set under the first category, and add the first items in the first item sequences to the initial item set respectively, to obtain an intermediate item set corresponding to each first item; An updating module, configured to select second item sequences corresponding to the intermediate item sets from the item set under the second category, and add the second items in the second item sequences to the corresponding intermediate item sets to obtain updated intermediate item sets; The recommendation module is used to determine the target item set corresponding to the initial item set according to each updated intermediate item set corresponding to the initial item set, so as to make item recommendations according to each target item set.

8. An electronic device, characterized in that: include: one or more processors; a storage device for storing one or more programs, When the one or more programs are executed by the one or more processors, the one or more processors implement the method according to any one of claims 1 to 6.

9. A computer readable medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the method according to any one of claims 1 to 6 is implemented.

10. A computer program product, comprising a computer program, which, when executed by a processor, implements the method according to any one of claims 1 to 6.

Images