Sample recommendation method, device, computer-readable storage medium, and electronic device

Abstract

The present disclosure relates to a sample recommendation method, device, computer-readable storage medium and electronic device, the method comprising: obtaining a set of samples to be recommended of a user to be recommended; respectively determining the first target user and the first associated user corresponding to each sample to be recommended; for each sample to be recommended, determining the vector representation of the sample to be recommended according to the matching degree between the sample to be recommended and the corresponding first target user and the first associated user; respectively determining the first similarity parameter between the representative sample and each sample to be recommended according to the vector representation of the representative sample associated with the user to be recommended and the vector representation of each sample to be recommended; determining the target sample to be recommended according to the first similarity parameter. Therefore, the sample to be recommended can be vectorized based on the social relationship of the user to be recommended, so that the scope of the recommended sample can be effectively broadened, more comprehensive content can be recommended to the user, and the accuracy of the sample recommendation can be effectively guaranteed.

Description

Technical Field

[0001] The present disclosure relates to the field of content recommendation, and in particular, to a sample recommendation method, device, computer-readable storage medium, and electronic device. Background Art

[0002] In recent years, with the development and popularization of information technology, the information on the Internet has gradually become richer. In order to improve the user experience, content that users are interested in is usually recommended to users based on their information. In the prior art, a content-based recommendation method is usually used to recommend items that users have not been exposed to based on their past browsing history, so that the recommended content is only similar content that the user has paid attention to, and the recommended content is relatively one-sided. Summary of the invention

[0003] In order to solve the problems in the background technology, the purpose of the present disclosure is to provide a relatively rich and accurate sample recommendation method, device, computer-readable storage medium and electronic device.

[0004] In order to achieve the above objective, according to a first aspect of the present disclosure, a sample recommendation method is provided, the method comprising:

[0005] Acquire a set of samples to be recommended for a user to be recommended, wherein the set includes a plurality of samples to be recommended;

[0006] Determine respectively a first target user and a first associated user corresponding to each sample to be recommended, wherein the first target user is a user who is close to the user to be recommended and best matches the corresponding sample to be recommended, and the first associated user is at least one user who is close to the first target user;

[0007] For each sample to be recommended, determine a vector representation of the sample to be recommended according to a matching degree between the sample to be recommended and the corresponding first target user and first associated user;

[0008] Determining first similarity parameters between the representative sample and each sample to be recommended, respectively, according to the vector representation of the representative sample associated with the user to be recommended and the vector representation of each sample to be recommended;

[0009] A target sample to be recommended is determined according to the first similarity parameter.

[0010] Optionally, the method further comprises:

[0011] Determine a second target user and a second associated user corresponding to the representative sample, wherein the second target user is a user who is close to the user to be recommended and most matches the representative sample, and the second associated user is at least one user who is close to the second target user;

[0012] According to the matching degree between the representative sample and the corresponding second target user and second associated user, a vector representation of the representative sample is determined.

[0013] Optionally, the method further comprises:

[0014] Determining an attention parameter of each of the samples to be recommended, wherein the attention parameter is used to characterize the attention of the sample to be recommended over time;

[0015] The determining a target sample to be recommended according to the first similarity parameter includes:

[0016] Calculating a target recommendation value according to the first similarity parameter and the attention parameter;

[0017] The target sample to be recommended is determined according to the target recommendation value.

[0018] Optionally, calculating the target recommendation value according to the first similarity parameter and the attention parameter includes:

[0019] The target recommended value is calculated by the following formula:

[0020] Score=α×Score q +β×Score s ;

[0021] Wherein, Score represents the target recommendation value;

[0022] α represents the weight of the first similarity parameter;

[0023] β represents the weight of the attention parameter, where α+β=1;

[0024] Score q represents the first similarity parameter;

[0025] Score s represents the attention parameter.

[0026] Optionally, the method further comprises:

[0027] Determining a negatively related user of the user to be recommended, wherein the negatively related user is at least one user that is dissimilar to the user to be recommended;

[0028] Determine at least one sample that the negatively correlated user is interested in as a negatively correlated sample;

[0029] Determine the second similarity parameter of each sample to be recommended and the negatively correlated sample respectively;

[0030] The calculating the target recommendation value according to the first similarity parameter and the attention parameter includes:

[0031] The target recommendation value is calculated according to the first similarity parameter, the attention parameter and the second similarity parameter by the following formula:

[0032] Score=α×Score q +β×Score s -γ×Score p ;

[0033] Wherein, Score represents the target recommendation value;

[0034] α represents the weight of the first similarity parameter;

[0035] β represents the weight of the attention parameter;

[0036] γ represents the weight of the second similarity parameter, wherein α+β+γ=1;

[0037] Score q represents the first similarity parameter;

[0038] Score s represents the attention parameter;

[0039] Score p represents the second similarity parameter.

[0040] Optionally, the attention parameter of the sample to be recommended is determined by the following formula:

[0041]

[0042] Among them, Score s represents the attention parameter;

[0043] t represents the time when the sample to be recommended is obtained;

[0044] T represents a preset time interval constant;

[0045] Represents the average number of views of all samples in the sample set;

[0046] view num Indicates the number of views of the sample to be recommended.

[0047] Optionally, the method further comprises:

[0048] Determining a negatively related user of the user to be recommended, wherein the negatively related user is at least one user that is dissimilar to the user to be recommended;

[0049] Determine at least one sample that the negatively correlated user is interested in as a negatively correlated sample;

[0050] Determine the second similarity parameter of each sample to be recommended and the negatively correlated sample respectively;

[0051] The determining a target sample to be recommended according to the first similarity parameter includes:

[0052] According to the first similarity parameter and the second similarity parameter, the target recommendation value is calculated by the following formula:

[0053] Score=α×Score q -γ×Score p ;

[0054] Wherein, Score represents the target recommendation value;

[0055] α represents the weight of the first similarity parameter;

[0056] γ represents the weight of the second similarity parameter, wherein α+γ=1;

[0057] Score q represents the first similarity parameter;

[0058] Score p represents the second similarity parameter;

[0059] The target sample to be recommended is determined according to the target recommendation value.

[0060] Optionally, the second similarity parameter between the sample to be recommended and the negatively correlated sample is determined in the following manner:

[0061]

[0062] Among them, Score p represents the second similarity parameter;

[0063] N represents the total number of negatively correlated samples;

[0064] M represents the total number of labels of the samples to be recommended and each of the negatively correlated samples;

[0065] The samples to be recommended and the negatively correlated samples are represented by label vectors, and each dimension of the label vector corresponds to a label;

[0066] D represents the label vector of the sample to be recommended, D j The value of is used to indicate whether the sample to be recommended has the j-th dimension label;

[0067] G i Represents the label vector of the i-th negatively correlated sample, G ij The value of is used to indicate whether the i-th negative correlation sample has the j-th dimension label.

[0068] According to a second aspect of the present disclosure, a sample recommendation device is provided, the device comprising:

[0069] An acquisition module, used to acquire a set of samples to be recommended for a user to be recommended, wherein the set includes a plurality of samples to be recommended;

[0070] A first determination module is used to respectively determine a first target user and a first associated user corresponding to each sample to be recommended, wherein the first target user is a user who is close to the user to be recommended and best matches the corresponding sample to be recommended, and the first associated user is at least one user who is close to the first target user;

[0071] A second determination module is used to determine, for each sample to be recommended, a vector representation of the sample to be recommended according to a matching degree between the sample to be recommended and the corresponding first target user and first associated user;

[0072] A third determination module, configured to determine first similarity parameters between the representative sample and each sample to be recommended, respectively, based on the vector representation of the representative sample associated with the user to be recommended and the vector representation of each sample to be recommended;

[0073] The fourth determination module is used to determine the target sample to be recommended according to the first similarity parameter.

[0074] Optionally, the device further comprises:

[0075] A fifth determination module, configured to determine a second target user and a second associated user corresponding to the representative sample, wherein the second target user is a user who is close to the user to be recommended and most closely matches the representative sample, and the second associated user is at least one user who is close to the second target user;

[0076] The sixth determination module is used to determine the vector representation of the representative sample according to the matching degree between the representative sample and the corresponding second target user and second associated user.

[0077] Optionally, the device further comprises:

[0078] A seventh determination module is used to determine the attention parameter of each of the samples to be recommended, wherein the attention parameter is used to characterize the attention of the sample to be recommended over time;

[0079] The fourth determination module comprises:

[0080] A first calculation submodule, used for calculating a target recommendation value according to the first similarity parameter and the attention parameter;

[0081] The first determination submodule is used to determine the target sample to be recommended according to the target recommendation value.

[0082] Optionally, the first calculation submodule is used to calculate the target recommendation value according to the first similarity parameter and the attention parameter by using the following formula:

[0083] Score=α×Score q +β×Score s ;

[0084] Wherein, Score represents the target recommendation value;

[0085] α represents the weight of the first similarity parameter;

[0086] β represents the weight of the attention parameter, where α+β=1;

[0087] Score q represents the first similarity parameter;

[0088] Score s represents the attention parameter.

[0089] Optionally, the device further comprises:

[0090] an eighth determination module, configured to determine a negatively related user of the user to be recommended, wherein the negatively related user is at least one user that is dissimilar to the user to be recommended;

[0091] A ninth determination module, configured to determine at least one sample that the negatively correlated user is interested in as a negatively correlated sample;

[0092] a tenth determination module, used to respectively determine a second similarity parameter between each sample to be recommended and the negatively correlated sample;

[0093] The first calculation submodule is used to calculate the target recommendation value according to the first similarity parameter, the attention parameter and the second similarity parameter by the following formula:

[0094] Score=α×Score q+β×Score s -γ×Score p ;

[0095] Wherein, Score represents the target recommendation value;

[0096] α represents the weight of the first similarity parameter;

[0097] β represents the weight of the attention parameter;

[0098] γ represents the weight of the second similarity parameter, wherein α+β+γ=1;

[0099] Score q represents the first similarity parameter;

[0100] Score s represents the attention parameter;

[0101] Score p represents the second similarity parameter.

[0102] Optionally, the seventh determination module determines the attention parameter of the sample to be recommended by the following formula:

[0103]

[0104] Among them, Score s represents the attention parameter;

[0105] t represents the time when the sample to be recommended is obtained;

[0106] T represents a preset time interval constant;

[0107] Represents the average number of views of all samples in the sample set;

[0108] view num Indicates the number of views of the sample to be recommended.

[0109] Optionally, the device further comprises:

[0110] an eleventh determining module, configured to determine a negatively related user of the user to be recommended, wherein the negatively related user is at least one user that is dissimilar to the user to be recommended;

[0111] A twelfth determination module, configured to determine at least one sample that the negatively correlated user is interested in as a negatively correlated sample;

[0112] A thirteenth determination module, used to respectively determine a second similarity parameter between each sample to be recommended and the negatively correlated sample;

[0113] The fourth determination module comprises:

[0114] The second calculation submodule is configured to calculate a target recommendation value according to the first similarity parameter and the second similarity parameter by using the following formula:

[0115] Score=α×Score q -γ×Score p ;

[0116] Wherein, Score represents the target recommendation value;

[0117] α represents the weight of the first similarity parameter;

[0118] γ represents the weight of the second similarity parameter, wherein α+γ=1;

[0119] Score q represents the first similarity parameter;

[0120] Score p represents the second similarity parameter;

[0121] The second determination submodule is used to determine the target sample to be recommended according to the target recommendation value.

[0122] Optionally, the second similarity parameter between the sample to be recommended and the negatively correlated sample is determined in the following manner:

[0123]

[0124] Among them, Score p represents the second similarity parameter;

[0125] N represents the total number of negatively correlated samples;

[0126] M represents the total number of labels of the samples to be recommended and each of the negatively correlated samples;

[0127] The samples to be recommended and the negatively correlated samples are represented by label vectors, and each dimension of the label vector corresponds to a label;

[0128] D represents the label vector of the sample to be recommended, D j The value of is used to indicate whether the sample to be recommended has the j-th dimension label;

[0129] G i Represents the label vector of the i-th negatively correlated sample, G ij The value of is used to indicate whether the i-th negative correlation sample has the j-th dimension label.

[0130] According to a third aspect of the present disclosure, a computer-readable storage medium is provided, on which a computer program is stored, and when the program is executed by a processor, the steps of any method described in the first aspect are implemented.

[0131] According to a fourth aspect of the present disclosure, there is provided an electronic device, including:

[0132] a memory having a computer program stored thereon;

[0133] A processor is used to execute the computer program in the memory to implement the steps of any method described in the first aspect above.

[0134] Through the above technical solution, the recommended samples can be represented by vectors based on the social relationships of the users to be recommended. Therefore, when determining the similarity between the representative sample and the sample to be recommended, the social relationships of the users to be recommended can be integrated to determine the target sample to be recommended. On the one hand, the method of determining the target sample to be recommended is no longer limited to the users to be recommended themselves, thereby effectively broadening the scope of recommended samples and facilitating the recommendation of more comprehensive content to users. On the other hand, it can also effectively ensure the accuracy of sample recommendations and enhance user experience.

[0135] Other features and advantages of the present disclosure will be described in detail in the following detailed description. BRIEF DESCRIPTION OF THE DRAWINGS

[0136] The accompanying drawings are used to provide a further understanding of the present disclosure and constitute a part of the specification. Together with the following specific embodiments, they are used to explain the present disclosure but do not constitute a limitation of the present disclosure. In the accompanying drawings:

[0137] Figure 1 is a flow chart of a sample recommendation method provided according to an embodiment of the present disclosure;

[0138] Figure 2 is a flow chart of a sample recommendation method provided according to another embodiment of the present disclosure;

[0139] Figure 3 is a flow chart of a sample recommendation method provided according to another embodiment of the present disclosure;

[0140] Figure 4 is a flow chart of a sample recommendation method provided according to another embodiment of the present disclosure;

[0141] Figure 5 is a block diagram of a sample recommendation device provided according to an embodiment of the present disclosure;

[0142] Figure 6 is a block diagram of an electronic device according to an exemplary embodiment;

[0143] Figure 7 It is a block diagram of an electronic device according to an exemplary embodiment. DETAILED DESCRIPTION

[0144] The specific implementation of the present disclosure is described in detail below in conjunction with the accompanying drawings. It should be understood that the specific implementation described herein is only used to illustrate and explain the present disclosure, and is not used to limit the present disclosure.

[0145] As described in the background art, in the existing recommendation technology, the content-based recommendation method can only recommend content similar to the content that the user has browsed before, so that the recommended content is relatively one-sided.

[0146] In order to solve the above problems, the present disclosure provides a sample recommendation method. Figure 1 As shown, it is a flow chart of a sample recommendation method provided according to an embodiment of the present disclosure, as shown Figure 1 As shown, the method includes:

[0147] In S11, a set of samples to be recommended of a user to be recommended is obtained, wherein the set includes a plurality of samples to be recommended.

[0148] Among them, the samples to be recommended may be items, texts, etc. In one example, the sample set to be recommended may include samples of the entire sample set. For example, when a user browses a library, the sample to be recommended is a text, and the sample set may be a set formed by all the texts contained in the library. For another example, when a user browses a shopping website, the sample to be recommended may be an item, and the sample set may be a set formed by all the items in the shopping website. As another example, a set formed by samples associated with users who have an association relationship with the user to be recommended may also be determined as the sample set to be recommended. Users who have an association relationship with the user to be recommended are users who have communicated with the user to be recommended or who the user to be recommended has followed. For example, in the shopping website, users who have an association relationship with the user to be recommended may be users followed by the user to be recommended, or users who are friends with the user to be recommended.

[0149] In S12, the first target user and the first associated user corresponding to each sample to be recommended are respectively determined, wherein the first target user is a user who is close to the user to be recommended and best matches the corresponding sample to be recommended, and the first associated user is at least one user who is close to the first target user.

[0150] The intimacy between users can be represented by the number of communications between users within a preset period of time, which may include one or more of the number of mutual comments, the number of mutual likes, and the number of mutual chats. The preset period of time may be a relatively recent period such as the last week or the last three days.

[0151] For example, there may be one or more users who are close to the user to be recommended, wherein the number of users who are close to the user to be recommended can be set according to actual usage scenarios.

[0152] In one embodiment, when there is only one user who is close to the user to be recommended, the user with the highest number of communications with the user to be recommended can be determined as the user who is close to the user to be recommended. Since there is only one user who is close to the user to be recommended at this time, this user can be directly determined as the first target user.

[0153] In another embodiment, there are multiple users who are close to the user to be recommended. For example, users whose number of communications with the user to be recommended exceeds a preset threshold can be determined as users who are close to the user to be recommended. For another example, the total number of users who are close to the user to be recommended can be preset. When the number of users who are close to the user to be recommended determined based on the preset threshold exceeds the total number, users whose number of communications with the user to be recommended exceeds the preset threshold can be randomly screened, or the number of communications can be sorted from high to low, with users ranked first being given priority. The above examples are only exemplary implementations of determining users who are close to the user to be recommended, and do not limit the present disclosure.

[0154] After determining the users who are close to the user to be recommended, it is necessary to calculate the matching degree between the sample to be recommended and each user who is close to the user to be recommended. For any user who is close to the user to be recommended, the average value of the similarities between the sample to be recommended and each sample associated with the user can be determined as the matching degree between the sample to be recommended and the user. When calculating the similarity between samples, the samples can be vectorized according to the labels corresponding to the samples to determine the label vector of the samples. For example, the labels corresponding to the samples can be the fields to which the samples belong, such as electronics, physics, etc., or can be the attributes corresponding to the samples, such as e-books, physical books, etc.

[0155] When determining the label vector of a sample, the corresponding labels can be removed to determine the dimension of the label vector. When the label of the sample contains a label of the corresponding dimension, the dimension value is 1, otherwise it is 0, to determine the label vector of the sample. For example, the label associated with sample 1 is {a, d, f, g, h}, and the label associated with sample 2 is {a, b, c, f}. Therefore, when sample 1 and sample 2 are vectorized, the dimension of the label vector obtained by removing the labels of sample 1 and sample 2 is {a, b, c, d, f, g, h}. Taking sample 1 as an example, the label associated with sample 1 is {a, d, f, g, h}, then in the label vector corresponding to sample 1, the dimension values ​​corresponding to labels a, d, f, g, h are 1, and the dimension values ​​corresponding to labels b and c are 0, then the label vector corresponding to sample 1 is {1, 0, 0, 1, 1, 1, 1}, and the label vector corresponding to sample 2 is {1, 1, 1, 0, 1, 0, 0}.

[0156] After determining the label vector of the sample, the similarity between the samples can be determined based on the label vector. The method of calculating the similarity between vectors is a prior art and will not be described in detail here. The user corresponding to the maximum matching degree is determined as the first target user corresponding to the sample to be recommended. Then, the first associated user can be determined based on the first target user, and the method of determining the user intimacy has been described in detail above and will not be described in detail here.

[0157] In S13, for each sample to be recommended, a vector representation of the sample to be recommended is determined according to the matching degree between the sample to be recommended and the corresponding first target user and first associated user.

[0158] Among them, the degree of match between the sample to be recommended and the first associated user can be determined, thereby determining the vector representation of the sample to be recommended. The implementation method of determining the degree of match between the sample to be recommended and the first associated user is similar to the implementation method of determining the degree of match between the sample to be recommended and the first target user described above, and will not be repeated here. When determining the degree of match between the sample to be recommended and the corresponding first target user and the first associated user, the vector formed by each degree of match between the sample to be recommended and the corresponding first target user and the first associated user can be determined as the vector representation of the sample to be recommended. For example, each dimension in the vector representation of the sample to be recommended corresponds to a degree of match, such as the degree of match between the sample to be recommended and the corresponding first target user is p1, and the degrees of match between the sample to be recommended and the corresponding first associated user are p2, p3, p4, and p5, respectively. Then, the vector representation of the sample to be recommended is {p1, p2, p3, p4, p5}.

[0159] In S14, first similarity parameters between the representative sample and each sample to be recommended are determined respectively according to the vector representation of the representative sample associated with the user to be recommended and the vector representation of each sample to be recommended.

[0160] The representative samples associated with the user to be recommended are samples that characterize the interests of the user to be recommended. In one embodiment, the representative samples may be samples that have been browsed the most times by the user to be recommended in a recent period, for example, the texts that have been browsed the most by the user to be recommended in the past week; the representative samples may also be items that have been purchased and praised by the user to be recommended in the past month. The representative samples may be one or more, and this disclosure does not limit this.

[0161] For example, the first similarity parameter may be the cosine similarity or distance between samples. When the cosine similarity between the representative sample and the sample to be recommended is larger, the similarity between the two is higher. When the distance between the representative sample and the sample to be recommended is smaller, the similarity between the two is higher.

[0162] In S15, a target sample to be recommended is determined according to the first similarity parameter.

[0163] In one embodiment, the samples to be recommended represented by the first similarity parameter may be sorted in descending order according to their similarity with the representative sample, so that the corresponding samples to be recommended may be determined as target samples to be recommended in this order for recommendation to the user.

[0164] In another embodiment, the number of recommended target samples to be recommended may be preset, so after the samples to be recommended are sorted, the samples to be recommended with the highest number of recommended samples may be directly selected in order as target samples to be recommended for recommendation to the user. The target sample to be recommended may be one or more, and the present disclosure does not limit this.

[0165] Generally speaking, close users have common interests. This disclosure is based on this idea, and recommends samples to the user to be recommended based on the social relationship of the user to be recommended. Through the social relationship of the user to be recommended, for example, based on commonly used social software (such as Weibo, WeChat, etc.), it is possible to determine the users who have communicated with the user to be recommended and commented on each other more, so as to determine the users who are close to the user to be recommended, and then recommend samples to the user to be recommended.

[0166] In the above technical scheme, by determining the first target user and the first associated user corresponding to the sample to be recommended, the vector representation of the sample to be recommended can be determined based on the matching degree between the sample to be recommended and the corresponding first target user and the first associated user. Then, according to the vector representation of the representative sample associated with the user to be recommended and the vector representation of each sample to be recommended, the first similarity parameter between the representative sample and each sample to be recommended is determined respectively, and the target sample to be recommended is determined according to the first similarity parameter, so as to recommend the sample to the user to be recommended. Through the above technical scheme, the sample to be recommended can be represented by a vector based on the social relationship of the user to be recommended. Therefore, when determining the similarity between the representative sample and the sample to be recommended, the social relationship of the user to be recommended can be integrated to determine the target sample to be recommended. On the one hand, the method of determining the target sample to be recommended is no longer limited to the user to be recommended, thereby effectively broadening the scope of the recommended sample and facilitating the recommendation of more comprehensive content to the user. On the other hand, the accuracy of the sample recommendation can be effectively guaranteed to improve the user experience.

[0167] Optionally, the method further comprises:

[0168] A second target user and a second associated user corresponding to the representative sample are determined, wherein the second target user is a user who is close to the user to be recommended and most closely matches the representative sample, and the second associated user is at least one user who is close to the second target user.

[0169] According to the matching degree between the representative sample and the corresponding second target user and second associated user, a vector representation of the representative sample is determined.

[0170] The method of determining the second target user and the second associated user corresponding to the representative sample is similar to the method of determining the first target user and the first associated user corresponding to the recommended sample described above, and will not be repeated here.

[0171] In this embodiment, by determining the vector representation of the representative sample associated with the user to be recommended, the first similarity parameter between the representative sample and each sample to be recommended can be determined respectively according to the vector representation of the representative sample associated with the user to be recommended and the vector representation of each sample to be recommended, thereby ensuring the accuracy of the first similarity parameter and providing accurate data support for determining the target sample to be recommended.

[0172] In the sample recommendation method described above, samples are recommended to the users to be recommended based on their social relationships. When recommending to the users to be recommended, in order to improve the real-time nature of sample recommendations and further expand the scope of recommendations, users can also be recommended based on the real-time attention paid to the samples. For example, if a sample has a high level of group attention in the recent period, it means that the sample may be a hot sample in the recent period, such as a new item or sudden news. The higher the level of attention paid to the sample to be recommended, the greater the possibility of its recommendation. Based on the above ideas, optionally, Figure 2 As shown, the method provided by the present disclosure also includes:

[0173] In S21, the attention parameter of each sample to be recommended is determined, wherein the attention parameter is used to characterize the attention of the sample to be recommended over time. The higher the attention parameter of the sample to be recommended, the more attention the sample to be recommended has received.

[0174] For a sample to be recommended, the higher the number of views, the higher the attention it receives, and its attention generally decreases over time. Optionally, the attention parameter of the sample to be recommended can be determined by the following formula:

[0175]

[0176] Among them, Score s represents the attention parameter;

[0177] t represents the time when the sample to be recommended is obtained;

[0178] T represents a preset time interval constant;

[0179] Represents the average number of views of all samples in the sample set;

[0180] view num Indicates the number of views of the sample to be recommended.

[0181] The step S15 of determining the target sample to be recommended according to the first similarity parameter may include:

[0182] In S22, a target recommendation value is calculated according to the first similarity parameter and the attention parameter.

[0183] Optionally, calculating the target recommendation value according to the first similarity parameter and the attention parameter includes:

[0184] The target recommended value is calculated by the following formula:

[0185] Score=α×Score q+β×Score s ;

[0186] Wherein, Score represents the target recommendation value;

[0187] α represents the weight of the first similarity parameter;

[0188] β represents the weight of the attention parameter, where α+β=1;

[0189] Score q represents the first similarity parameter;

[0190] Score s represents the attention parameter.

[0191] In the above calculation formula, the target recommendation value is determined by weighted summing the first similarity parameter and the attention parameter, so that the social relationship and group attention corresponding to the sample to be recommended can be considered simultaneously when determining the target recommendation value, thereby improving the comprehensiveness and accuracy of the target recommendation value. Among them, α and β can be set according to the actual usage scenario, and this disclosure does not limit this.

[0192] In S23, the target sample to be recommended is determined according to the target recommendation value. The method of determining the target sample to be recommended according to the target recommendation value is similar to the method of determining the target sample to be recommended according to the first similarity parameter described above, and will not be repeated here.

[0193] In the above technical solution, the target sample to be recommended can be determined based on the first similarity parameter and attention parameter of the sample to be recommended, that is, in the process of determining the target sample to be recommended for the user to be recommended, not only the social relationship of the user to be recommended is fully considered, but also the attention of the sample to be recommended itself is fully considered. In this way, the scope of sample recommendation can be further expanded, the real-time performance of sample recommendation can be improved, and samples that the user has not paid attention to can be recommended to the user, so that the determined target sample to be recommended is more comprehensive and the user experience is improved.

[0194] In addition, when recommending samples to a user to be recommended, samples that the user to be recommended is interested in are generally recommended. However, for some samples, the user does not clearly express his or her preference for the sample. Therefore, in order to better recommend samples to the user to be recommended, the present disclosure provides the following method.

[0195] Alternatively, if Figure 3 As shown, in Figure 1 On the basis of, the method further comprises:

[0196] In S31, negatively correlated users of the user to be recommended are determined, wherein the negatively correlated user is at least one user that is dissimilar to the user to be recommended. The similarity between users can be determined based on samples associated with each user, and the users can be represented by vectorizing the samples associated with the users, thereby determining the similarity between users through the vectors of the users. When the similarity between the user and the user to be recommended is less than a preset similarity threshold, it can be determined that the user is dissimilar to the user to be recommended, and the user can be determined as a negatively correlated user of the user to be recommended. The vectorized representation method is similar to the sample vectorization method described above, and will not be repeated here.

[0197] In S32, at least one sample that the negatively correlated user is interested in is determined as a negatively correlated sample. The sample that the negatively correlated user is interested in may be a sample that has been viewed more times by the negatively correlated user in a recent period of time. For example, if the total number of negatively correlated samples is preset, such as 10, the samples associated with the negatively correlated users may be determined as negatively correlated samples in the order of their total number of views from high to low, and the top 10 samples may be determined as negatively correlated samples. For another example, a negative correlation threshold may be set, and samples whose total number of views by the negatively correlated user exceeds the negative correlation threshold may be determined as negatively correlated samples. For example, the negative correlation threshold may be set to 100, and when the total number of views of a sample by the negatively correlated user exceeds 100 times, the sample may be determined as a negatively correlated sample.

[0198] In S33, the second similarity parameter between each sample to be recommended and the negatively correlated sample is determined respectively.

[0199] The samples can be vectorized according to the labels corresponding to the samples, so when determining the second similarity parameters of the recommended samples and the negatively correlated samples, calculations can be performed based on the label vectors corresponding to the samples. The specific implementation has been described in detail above and will not be repeated here.

[0200] In this embodiment, for samples that cannot determine the user's preferences, the samples to be recommended can be excluded based on the samples that users who are not similar to the user to be recommended are interested in. In the present disclosure, based on the idea that samples that users who are not similar to the current user are interested in are samples that the current user is not interested in, negatively related users are determined, and negatively related samples are determined based on the negatively related users. Based on this, when the user's preferences cannot be determined, samples that the user to be recommended is not interested in can be effectively avoided from being recommended, thereby ensuring the accuracy of sample recommendation.

[0201] Optionally, the second similarity parameter between the sample to be recommended and the negatively correlated sample is determined in the following manner:

[0202]

[0203] Among them, Score p represents the second similarity parameter;

[0204] N represents the total number of negatively correlated samples;

[0205] M represents the total number of labels of the samples to be recommended and each of the negatively correlated samples;

[0206] The samples to be recommended and the negatively correlated samples are represented by label vectors, and each dimension of the label vector corresponds to a label;

[0207] D represents the label vector of the sample to be recommended, D j The value of is used to indicate whether the sample to be recommended has the j-th dimension label;

[0208] G i Represents the label vector of the i-th negatively correlated sample, G ij The value of is used to indicate whether the i-th negative correlation sample has the j-th dimension label.

[0209] Among them, when determining the label vector of the sample, the labels corresponding to the sample to be recommended and the negatively correlated samples can be deduplicated to determine the dimension of the label vector. For example, the dimension of the determined label is {a, b, c, d, e, f, g, h, i, k, l, m}, and then the label vector of the sample can be determined based on the dimension of the label. For example, the labels contained in the sample to be recommended are {c, d, g, h, i, k}, then the label vector D of the sample to be recommended is {0, 0, 1, 1, 0, 0, 1, 1, 1, 0, 0}, D1 is 0, indicating that the sample to be recommended does not have the first dimension label a, and D3 is 1, indicating that the sample to be recommended has the third dimension label c. The method for determining the label vector of the negatively correlated sample is similar to the method for determining the label vector of the sample to be recommended, and will not be repeated here.

[0210] The step S15 of determining the target sample to be recommended according to the first similarity parameter may include:

[0211] In S34, the target recommendation value is calculated according to the first similarity parameter and the second similarity parameter. For example, the target recommendation value may be calculated by the following formula:

[0212] Score=α×Score q -γ×Score p ;

[0213] Wherein, Score represents the target recommendation value;

[0214] α represents the weight of the first similarity parameter;

[0215] γ represents the weight of the second similarity parameter, wherein α+γ=1;

[0216] Score q represents the first similarity parameter;

[0217] Score p represents the second similarity parameter.

[0218] In S35, the target sample to be recommended is determined according to the target recommendation value.

[0219] The specific implementation of S35 is similar to that of S23 and will not be described in detail here. The values ​​of α and γ can be set according to the actual usage scenario. In this embodiment, by combining the second similarity parameter, the target recommendation value corresponding to the sample that the user is not interested in can be effectively reduced, thereby ensuring the accuracy of the target recommendation value, and further ensuring the accuracy of the target sample to be recommended determined based on the target recommendation value, which meets the user's usage needs.

[0220] Optionally, in order to further improve the accuracy of sample recommendation, Figure 4 As shown, in Figure 2 On the basis of, the method further comprises:

[0221] In S41, negatively related users of the user to be recommended are determined, wherein the negatively related user is at least one user who is not similar to the user to be recommended;

[0222] In S42, at least one sample that the negatively correlated user is interested in is determined as a negatively correlated sample;

[0223] In S43, second similarity parameters between each sample to be recommended and the negatively correlated sample are determined respectively.

[0224] Among them, the specific implementation methods of S41, S42, and S43 are the same as the specific implementation methods of S31, S32, and S33, and will not be repeated here.

[0225] The step S22 of calculating the target recommendation value according to the first similarity parameter and the attention parameter may include:

[0226] In S44, the target recommendation value is calculated according to the first similarity parameter, the attention parameter and the second similarity parameter. For example, the target recommendation value may be calculated by the following formula:

[0227] Score=α×Score q +β×Score s -γ×Score p ;

[0228] Wherein, Score represents the target recommendation value;

[0229] α represents the weight of the first similarity parameter;

[0230] β represents the weight of the attention parameter;

[0231] γ represents the weight of the second similarity parameter, wherein α+β+γ=1;

[0232] Score q represents the first similarity parameter;

[0233] Score s represents the attention parameter;

[0234] Score p represents the second similarity parameter.

[0235] Therefore, by determining the negatively correlated users and negatively correlated samples of the user to be recommended, it is possible to effectively avoid recommending samples that the user is not interested in, thereby ensuring the user experience. In this embodiment, the target recommendation value is determined by fusing the first similarity parameter, the attention parameter, and the second similarity parameter to further ensure the accuracy of the target recommendation value. Among them, the values ​​of α, β, and γ can be set according to the actual usage scenario. At the same time, through the above technical solution, the scope of the target samples to be recommended can also be effectively broadened, and samples that the user has never been exposed to before but may be interested in can be recommended to the user, making the user portrait richer and more complete, further improving the user experience.

[0236] The present disclosure also provides a sample recommendation device, such as Figure 5 As shown, the device 10 comprises:

[0237] The acquisition module 100 is used to acquire a set of samples to be recommended for a user to be recommended, wherein the set includes a plurality of samples to be recommended;

[0238] A first determination module 200 is used to respectively determine a first target user and a first associated user corresponding to each sample to be recommended, wherein the first target user is a user who is close to the user to be recommended and best matches the corresponding sample to be recommended, and the first associated user is at least one user who is close to the first target user;

[0239] The second determination module 300 is used to determine, for each sample to be recommended, a vector representation of the sample to be recommended according to the matching degree between the sample to be recommended and the corresponding first target user and the first associated user;

[0240] A third determination module 400 is used to determine a first similarity parameter between the representative sample and each sample to be recommended, respectively, according to the vector representation of the representative sample associated with the user to be recommended and the vector representation of each sample to be recommended;

[0241] The fourth determination module 500 is used to determine the target sample to be recommended according to the first similarity parameter.

[0242] Optionally, the device 10 further includes:

[0243] A fifth determination module, configured to determine a second target user and a second associated user corresponding to the representative sample, wherein the second target user is a user who is close to the user to be recommended and most closely matches the representative sample, and the second associated user is at least one user who is close to the second target user;

[0244] The sixth determination module is used to determine the vector representation of the representative sample according to the matching degree between the representative sample and the corresponding second target user and second associated user.

[0245] Optionally, the device 10 further includes:

[0246] A seventh determination module is used to determine the attention parameter of each of the samples to be recommended, wherein the attention parameter is used to characterize the attention of the sample to be recommended over time;

[0247] The fourth determining module 500 includes:

[0248] A first calculation submodule, used for calculating a target recommendation value according to the first similarity parameter and the attention parameter;

[0249] The first determination submodule is used to determine the target sample to be recommended according to the target recommendation value.

[0250] Optionally, the first calculation submodule is used to calculate the target recommendation value according to the first similarity parameter and the attention parameter by using the following formula:

[0251] Score=α×Score q +β×Score s ;

[0252] Wherein, Score represents the target recommendation value;

[0253] α represents the weight of the first similarity parameter;

[0254] β represents the weight of the attention parameter, where α+β=1;

[0255] Score qrepresents the first similarity parameter;

[0256] Score s represents the attention parameter.

[0257] Optionally, the device 10 further includes:

[0258] an eighth determination module, configured to determine a negatively related user of the user to be recommended, wherein the negatively related user is at least one user that is dissimilar to the user to be recommended;

[0259] A ninth determination module, configured to determine at least one sample that the negatively correlated user is interested in as a negatively correlated sample;

[0260] a tenth determination module, used to respectively determine a second similarity parameter between each sample to be recommended and the negatively correlated sample;

[0261] The first calculation submodule is used to calculate the target recommendation value according to the first similarity parameter, the attention parameter and the second similarity parameter by the following formula:

[0262] Score=α×Score q +β×Score s -γ×Score p ;

[0263] Wherein, Score represents the target recommendation value;

[0264] α represents the weight of the first similarity parameter;

[0265] β represents the weight of the attention parameter;

[0266] γ represents the weight of the second similarity parameter, wherein α+β+γ=1;

[0267] Score q represents the first similarity parameter;

[0268] Score s represents the attention parameter;

[0269] Score p represents the second similarity parameter.

[0270] Optionally, the seventh determination module determines the attention parameter of the sample to be recommended by the following formula:

[0271]

[0272] Among them, Score s represents the attention parameter;

[0273] t represents the time when the sample to be recommended is obtained;

[0274] T represents a preset time interval constant;

[0275] Represents the average number of views of all samples in the sample set;

[0276] view num Indicates the number of views of the sample to be recommended.

[0277] Optionally, the device 10 further includes:

[0278] an eleventh determining module, configured to determine a negatively related user of the user to be recommended, wherein the negatively related user is at least one user that is dissimilar to the user to be recommended;

[0279] A twelfth determination module, configured to determine at least one sample that the negatively correlated user is interested in as a negatively correlated sample;

[0280] A thirteenth determination module, used to respectively determine a second similarity parameter between each sample to be recommended and the negatively correlated sample;

[0281] The fourth determining module 500 includes:

[0282] The second calculation submodule is configured to calculate a target recommendation value according to the first similarity parameter and the second similarity parameter by using the following formula:

[0283] Score=α×Score q -γ×Score p ;

[0284] Wherein, Score represents the target recommendation value;

[0285] α represents the weight of the first similarity parameter;

[0286] γ represents the weight of the second similarity parameter, wherein α+γ=1;

[0287] Score q represents the first similarity parameter;

[0288] Score p represents the second similarity parameter;

[0289] The second determination submodule is used to determine the target sample to be recommended according to the target recommendation value.

[0290] Optionally, the second similarity parameter between the sample to be recommended and the negatively correlated sample is determined in the following manner:

[0291]

[0292] Among them, Score p represents the second similarity parameter;

[0293] N represents the total number of negatively correlated samples;

[0294] M represents the total number of labels of the samples to be recommended and each of the negatively correlated samples;

[0295] The samples to be recommended and the negatively correlated samples are represented by label vectors, and each dimension of the label vector corresponds to a label;

[0296] D represents the label vector of the sample to be recommended, D j The value of is used to indicate whether the sample to be recommended has the j-th dimension label;

[0297] G i Represents the label vector of the i-th negatively correlated sample, G ij The value of is used to indicate whether the i-th negative correlation sample has the j-th dimension label.

[0298] Regarding the device in the above embodiment, the specific manner in which each module performs operations has been described in detail in the embodiment of the method, and will not be elaborated here.

[0299] Figure 6 FIG. 7 is a block diagram of an electronic device 700 according to an exemplary embodiment. Figure 6 As shown, the electronic device 700 may include: a processor 701 and a memory 702. The electronic device 700 may also include one or more of a multimedia component 703, an input / output (I / O) interface 704, and a communication component 705.

[0300] The processor 701 is used to control the overall operation of the electronic device 700 to complete all or part of the steps in the above-mentioned sample recommendation method. The memory 702 is used to store various types of data to support the operation of the electronic device 700. For example, these data may include instructions for any application or method used to operate on the electronic device 700, and application-related data, such as contact data, sent and received messages, pictures, audio, video, etc. The memory 702 can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (Static Random Access Memory, referred to as SRAM), electrically erasable programmable read-only memory (Electrically Erasable Programmable Read-Only Memory, referred to as EEPROM), erasable programmable read-only memory (Erasable Programmable Read-Only Memory, referred to as EPROM), programmable read-only memory (Programmable Read-Only Memory, referred to as PROM), read-only memory (Read-Only Memory, referred to as ROM), magnetic memory, flash memory, magnetic disk or optical disk. The multimedia component 703 may include a screen and an audio component. The screen may be, for example, a touch screen, and the audio component is used to output and / or input audio signals. For example, the audio component may include a microphone, which is used to receive external audio signals. The received audio signal may be further stored in the memory 702 or sent through the communication component 705. The audio component also includes at least one speaker for outputting audio signals. The I / O interface 704 provides an interface between the processor 701 and other interface modules, and the above-mentioned other interface modules may be keyboards, mice, buttons, etc. These buttons may be virtual buttons or physical buttons. The communication component 705 is used for wired or wireless communication between the electronic device 700 and other devices. Wireless communication, such as Wi-Fi, Bluetooth, Near Field Communication (NFC), 2G, 3G or 4G, or a combination of one or more of them, so the corresponding communication component 705 may include: Wi-Fi module, Bluetooth module, NFC module.

[0301] In an exemplary embodiment, the electronic device 700 can be implemented by one or more application specific integrated circuits (ASIC), digital signal processors (DSP), digital signal processing devices (DSPD), programmable logic devices (PLD), field programmable gate arrays (FPGA), controllers, microcontrollers, microprocessors or other electronic components to execute the above-mentioned sample recommendation method.

[0302] In another exemplary embodiment, a computer-readable storage medium including program instructions is also provided, and when the program instructions are executed by a processor, the steps of the above-mentioned sample recommendation method are implemented. For example, the computer-readable storage medium can be the above-mentioned memory 702 including program instructions, and the above-mentioned program instructions can be executed by the processor 701 of the electronic device 700 to complete the above-mentioned sample recommendation method.

[0303] Figure 7 1 is a block diagram of an electronic device 1900 according to an exemplary embodiment. For example, the electronic device 1900 may be provided as a server. Figure 7 The electronic device 1900 includes a processor 1922, which may be one or more, and a memory 1932 for storing a computer program executable by the processor 1922. The computer program stored in the memory 1932 may include one or more modules each corresponding to a set of instructions. In addition, the processor 1922 may be configured to execute the computer program to perform the above-mentioned sample recommendation method.

[0304] In addition, the electronic device 1900 may further include a power supply component 1926 and a communication component 1950, wherein the power supply component 1926 may be configured to perform power management of the electronic device 1900, and the communication component 1950 may be configured to implement communication of the electronic device 1900, for example, wired or wireless communication. In addition, the electronic device 1900 may further include an input / output (I / O) interface 1958. The electronic device 1900 may operate based on an operating system stored in the memory 1932, such as Windows Server™, Mac OS X™, Unix™, Linux™, etc.

[0305] In another exemplary embodiment, a computer-readable storage medium including program instructions is also provided, and when the program instructions are executed by a processor, the steps of the above-mentioned sample recommendation method are implemented. For example, the computer-readable storage medium can be the above-mentioned memory 1932 including program instructions, and the above-mentioned program instructions can be executed by the processor 1922 of the electronic device 1900 to complete the above-mentioned sample recommendation method.

[0306] The preferred embodiments of the present disclosure are described in detail above in conjunction with the accompanying drawings; however, the present disclosure is not limited to the specific details in the above embodiments. Within the technical concept of the present disclosure, a variety of simple modifications can be made to the technical solution of the present disclosure, and these simple modifications all fall within the protection scope of the present disclosure.

[0307] It should also be noted that the various specific technical features described in the above specific embodiments can be combined in any suitable manner without contradiction. In order to avoid unnecessary repetition, the present disclosure will not further describe various possible combinations.

[0308] In addition, various embodiments of the present disclosure may be arbitrarily combined, and as long as they do not violate the concept of the present disclosure, they should also be regarded as the contents disclosed by the present disclosure.

Claims

1. A sample recommendation method, characterized in that: The method comprises: Acquire a set of samples to be recommended for a user to be recommended, wherein the set includes a plurality of samples to be recommended; Determine respectively a first target user and a first associated user corresponding to each sample to be recommended, wherein the first target user is a user who is close to the user to be recommended and best matches the corresponding sample to be recommended, and the first associated user is at least one user who is close to the first target user; For each sample to be recommended, determine a vector representation of the sample to be recommended according to a matching degree between the sample to be recommended and the corresponding first target user and first associated user; Determining first similarity parameters between the representative sample and each sample to be recommended, respectively, according to the vector representation of the representative sample associated with the user to be recommended and the vector representation of each sample to be recommended, wherein the representative sample associated with the user to be recommended is a sample that represents the interest direction of the user to be recommended; Determine a target sample to be recommended according to the first similarity parameter; The matching degree between the user and the sample to be recommended is the average value of the similarities between the sample to be recommended and each sample associated with the user.

2. The method according to claim 1, characterized in that The method further comprises: Determine a second target user and a second associated user corresponding to the representative sample, wherein the second target user is a user who is close to the user to be recommended and most matches the representative sample, and the second associated user is at least one user who is close to the second target user; According to the matching degree between the representative sample and the corresponding second target user and second associated user, a vector representation of the representative sample is determined.

3. The method according to claim 1, characterized in that The method further comprises: Determining an attention parameter of each of the samples to be recommended, wherein the attention parameter is used to characterize the attention of the sample to be recommended over time; The determining a target sample to be recommended according to the first similarity parameter includes: Calculating a target recommendation value according to the first similarity parameter and the attention parameter; The target sample to be recommended is determined according to the target recommendation value.

4. The method according to claim 3, characterized in that The calculating the target recommendation value according to the first similarity parameter and the attention parameter comprises: The target recommended value is calculated by the following formula: ； in, represents the target recommended value; represents the weight of the first similarity parameter; represents the weight of the attention parameter, where ; represents the first similarity parameter; represents the attention parameter.

5. The method according to claim 3, characterized in that: The method further comprises: Determining a negatively related user of the user to be recommended, wherein the negatively related user is at least one user that is dissimilar to the user to be recommended; Determine at least one sample that the negatively correlated user is interested in as a negatively correlated sample; Determine the second similarity parameter of each sample to be recommended and the negatively correlated sample respectively; The calculating the target recommendation value according to the first similarity parameter and the attention parameter includes: The target recommendation value is calculated according to the first similarity parameter, the attention parameter and the second similarity parameter by the following formula: ； in, represents the target recommended value; represents the weight of the first similarity parameter; represents the weight of the attention parameter; represents the weight of the second similarity parameter, wherein, ; represents the first similarity parameter; represents the attention parameter; represents the second similarity parameter.

6. The method according to any one of claims 3 to 5, characterized in that: The attention parameter of the sample to be recommended is determined by the following formula: in, represents the attention parameter; Indicates the time when the sample to be recommended is obtained; Indicates a preset time interval constant; Represents the average number of views of all samples in the sample set; Indicates the number of views of the sample to be recommended.

7. The method according to claim 1, characterized in that The method further comprises: Determining a negatively related user of the user to be recommended, wherein the negatively related user is at least one user that is dissimilar to the user to be recommended; Determine at least one sample that the negatively correlated user is interested in as a negatively correlated sample; Determine the second similarity parameter of each sample to be recommended and the negatively correlated sample respectively; The determining a target sample to be recommended according to the first similarity parameter includes: According to the first similarity parameter and the second similarity parameter, the target recommendation value is calculated by the following formula: ； in, represents the target recommended value; represents the weight of the first similarity parameter; represents the weight of the second similarity parameter, wherein, ; represents the first similarity parameter; represents the second similarity parameter; The target sample to be recommended is determined according to the target recommendation value.

8. The method according to claim 5 or 7, characterized in that: The second similarity parameter between the sample to be recommended and the negatively correlated sample is determined in the following manner: in, represents the second similarity parameter; represents the total number of negatively correlated samples; Represents the total number of labels of the sample to be recommended and each of the negatively correlated samples; The samples to be recommended and the negatively correlated samples are represented by label vectors, and each dimension of the label vector corresponds to a label; represents the label vector of the sample to be recommended, The value of is used to indicate whether the sample to be recommended has j Dimension label; Indicates i The label vector of negatively correlated samples, The numerical value is used to represent the i Whether there is the first j Dimension label.

9. A sample recommendation device, characterized in that: The device comprises: An acquisition module, used to acquire a set of samples to be recommended for a user to be recommended, wherein the set includes a plurality of samples to be recommended; A first determination module is used to respectively determine a first target user and a first associated user corresponding to each sample to be recommended, wherein the first target user is a user who is close to the user to be recommended and best matches the corresponding sample to be recommended, and the first associated user is at least one user who is close to the first target user; A second determination module is used to determine, for each sample to be recommended, a vector representation of the sample to be recommended according to a matching degree between the sample to be recommended and the corresponding first target user and first associated user; A third determination module is used to determine the first similarity parameter between the representative sample and each sample to be recommended according to the vector representation of the representative sample associated with the user to be recommended and the vector representation of each sample to be recommended, wherein the representative sample associated with the user to be recommended is a sample that represents the interest direction of the user to be recommended; A fourth determination module, configured to determine a target sample to be recommended according to the first similarity parameter; The matching degree between the user and the sample to be recommended is the average value of the similarities between the sample to be recommended and each sample associated with the user.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the steps of the method described in any one of claims 1 to 8 are implemented.

11. An electronic device, characterized in that: include: a memory having a computer program stored thereon; A processor, configured to execute the computer program in the memory to implement the steps of the method according to any one of claims 1 to 8.

Images