A label recommendation method, device, equipment and storage medium

Abstract

The application discloses a label recommendation method and device, equipment and a storage medium. The method comprises the following steps: in the case that a target object acquires a selected label from a preset label library, acquiring a label weight time-varying function corresponding to each label in the preset label library; the label weight time-varying function satisfies that the label weight value gradually decreases and tends to a preset value over time; determining a recommendation probability of each label for the selected label based on the label weight time-varying function corresponding to each label and an initial probability of the selected label; and performing label recommendation to the target object based on the recommendation probability of each label. Through the above technical solution, the accuracy of label recommendation is improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of Internet, and particularly relates to a label recommendation method and device, equipment and a storage medium. BACKGROUND

[0002] In the Internet era, various activity marketing platforms, such as the integrated operation platform (IOP) of China Mobile, will use customer labels for precision marketing. For example, if a marketing traffic booster is needed, the marketing personnel will preferentially select the label "monthly traffic usage" to screen out users with monthly traffic usage greater than 5 GB for precision marketing, thereby achieving the purpose of improving the success rate of marketing.

[0003] However, there can be thousands of labels in the system, and the operation personnel can be in a dilemma. At this time, the labels need to be ranked to facilitate the selection of the operation personnel when planning activities. The general label ranking method can be the number of uses, user scoring ranking, etc. However, these label ranking methods have problems such as single information dimension and unreasonable label weight setting. For example, new labels are not friendly enough. Even if the quality of a new label is very high, the weight value is still low, thereby reducing the accuracy of label ranking. SUMMARY

[0004] Embodiments of the present application aim to provide a label recommendation method, device, equipment and storage medium.

[0005] The technical solution of the embodiments of the present application is as follows:

[0006] The embodiments of the present application provide an information recommendation method, applied to a label recommendation device, and the method comprises: in the case that a target object acquires a selected label from a preset label library, acquiring a label weight time-varying function corresponding to each label in the preset label library; the label weight time-varying function satisfies that the label weight value gradually decreases and tends to a preset value over time; determining a recommendation probability of each label for the selected label based on the label weight time-varying function corresponding to each label and an initial probability of the selected label; and performing label recommendation to the target object based on the recommendation probability of each label.

[0007] The embodiments of the present application provide a label recommendation device, comprising:

[0008] An acquisition module is configured to, in the case that a target object acquires a selected label from a preset label library, acquire a label weight time-varying function corresponding to each label in the preset label library; the label weight time-varying function satisfies that the label weight value gradually decreases and tends to a preset value over time;

[0009] a determination module, configured to determine a recommendation probability of each tag for the selected tag based on the tag weight time-varying function corresponding to each tag and the initial probability of the selected tag;

[0010] The recommendation module is configured to recommend tags to the target object based on the recommendation probability of each tag.

[0011] An embodiment of the present application provides a tag recommendation device, comprising: a processor, a memory, and a communication bus; the communication bus is used to implement a communication connection between the processor and the memory; the processor is used to execute a computer program stored in the memory to implement the above-mentioned tag recommendation method.

[0012] An embodiment of the present application provides a computer-readable storage medium, which stores one or more computer programs. The one or more computer programs can be executed by one or more processors to implement the above-mentioned tag recommendation method.

[0013] The embodiment of the present application provides a tag recommendation method, apparatus, device and storage medium, the method comprising: when the target object obtains the selected tag from the preset tag library, obtaining the tag weight time-varying function corresponding to each tag in the preset tag library; the tag weight time-varying function satisfies that the tag weight value gradually decreases and tends to the preset value over time; based on the tag weight time-varying function corresponding to each tag and the initial probability of the selected tag, determining the recommendation probability of each tag for the selected tag; based on the recommendation probability of each tag, recommending tags to the target object. The technical solution provided by the present application, by constructing a corresponding tag weight time-varying function for each tag, enables the weight value of each tag to be adaptively adjusted over time, which can solve the problem of emphasizing old tags over new tags in the tag weight setting, thereby improving the accuracy of tag recommendations. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] Figure 1 A flowchart of a tag recommendation process provided in an embodiment of the present application;

[0015] Figure 2 A schematic diagram of an exemplary process for constructing a time-varying function provided in an embodiment of the present application;

[0016] Figure 3 A schematic diagram of an exemplary process for determining the probability of tag recommendation provided in an embodiment of the present application;

[0017] Figure 4 A schematic diagram of the structure of a tag recommendation device provided in an embodiment of the present application;

[0018] Figure 5A structural schematic diagram of a label recommendation device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0019] The technical solutions in the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. It can be understood that the specific embodiments described herein are only used to explain the related application, but not to limit the application. In addition, it should be noted that, for the convenience of description, only the parts related to the application are shown in the drawings.

[0020] In the following description, "some embodiments" are related to a subset of all possible embodiments, but it can be understood that "some embodiments" can be the same subset or different subsets of all possible embodiments, and can be combined with each other without conflict.

[0021] Unless otherwise defined, all technical and scientific terms used in the present application have the same meaning as commonly understood by one of ordinary skill in the art to which the present application belongs. The terms used in the present application are only for the purpose of describing the present application, and are not intended to limit the present application.

[0022] The present application provides a label recommendation method, which is implemented by a label recommendation device, as shown in the following steps S101 to S102: Figure 1

[0023] In step S101, in the case that the target object acquires selected labels from a preset label library, a label weight time-varying function corresponding to each label in the preset label library is acquired; the label weight time-varying function satisfies that the label weight value gradually decreases over time and tends to a preset value.

[0024] In the embodiments of the present application, the label recommendation device is an electronic device with a label recommendation function, which can be a tablet computer, a notebook computer, a palm computer, a personal digital assistant (PDA), a desktop computer, etc., and the specific label recommendation device is not limited here.

[0025] In the embodiments of the present application, the preset label library can be a label library composed of labels representing business types, and exemplary, the labels can be "monthly usage traffic", "monthly total consumption", etc. labels related to business types.

[0026] ​In the embodiments of the present application, the target object can be a marketing personnel, an operation personnel, or any other object that needs label recommendation. For example, assuming that the target object is a marketing personnel, if the marketing personnel needs a marketing traffic booster, the marketing personnel will first select the label "monthly usage traffic", filter out users with monthly usage traffic greater than 5 GB, and then perform accurate marketing, so as to improve the success rate of marketing, reduce complaints and marketing resource consumption. However, in actual application, multiple labels need to be selected for performing one marketing activity, and the labels in the preset label library generally reach thousands, which will cause a situation of difficulty in selection. At this time, in the process of performing marketing, the label recommendation device can recommend the next label according to the label selected by the marketing personnel, for example, the marketing personnel selects label i (selected label) when selecting the label for the kth time in performing one marketing activity, and then the label recommendation device performs label recommendation for the k+1th time according to the label i selected when selecting the label for the kth time.

[0027] In the embodiments of the present application, when the target object obtains the selected label from the preset label library, the label recommendation device obtains the label weight time-varying function corresponding to each label in the preset label library, and the label weight time-varying function corresponding to each label in the preset label library satisfies that the label weight value gradually decreases over time and tends to a preset value.

[0028] For example, assuming that the preset label library includes n labels, and the label recommendation device needs to solve the problem of the existing technology that old labels are given higher weight than new labels by constructing n time-varying functions based on label creation time, i.e., label weight time-varying functions, which satisfy three constraint conditions.

[0029] Constraint condition one, time-varying function f j (t a ) is a coefficient based on label creation time, and must satisfy formula (1):

[0030] 0≤f1(t a )≤…≤f j (t a )≤…f n (t a ),j=1…n (1);

[0031] Wherein, f j (t a ) is the label weight time-varying function corresponding to label j, and t a is the creation time of the label.

[0032] Constraint condition two, time-varying function f j (t a ) will decrease with the variable t aThe more the increase, the more the n weight values generated are close to each other, and the closeness should be smooth, and at t a tends to infinity, should be equivalent to 1 / n.

[0033] Constraint three, time-varying function f j (t a ) should satisfy a

[0034] In this way, the weight value of each label can be adaptively adjusted over time, avoiding the problem of setting the weight value of the label that the old label is heavier than the new label.

[0035] Step S102, based on the label weight time-varying function corresponding to each label and the initial probability of the selected label, determining the recommendation probability of each label for the selected label.

[0036] In an embodiment of the present application, after obtaining the label weight time-varying function corresponding to each label, the label recommendation device determines the recommendation probability of each label for the selected label by using the label weight time-varying function corresponding to each label and the initial probability of the selected label. In this way, the label recommendation device can obtain the recommendation probability of each label in the preset label library for the selected label.

[0037] Step S103, based on the recommendation probability of each label, label recommendation is performed to the target object.

[0038] In an embodiment of the present application, after the label recommendation device determines the recommendation probability of each label for the selected label, label recommendation is performed to the target object based on the recommendation probability of each label.

[0039] Compared with the traditional PageRank algorithm in the prior art, the old is heavier than the new, for example, in the field of web page ranking, the PageRank algorithm is not friendly to new web pages. A new web page generally has relatively few inbound links, even if the quality of its content is high, it still needs a long time of promotion to become a high-PR-value page, thereby causing poor accuracy of recommended content.

[0040] And the present application constructs the corresponding label weight time-varying function for each label, so that the weight value of each label is adaptively adjusted over time and gradually decreases, solving the problem of setting the weight value of the label that the old label is heavier than the new label, thereby improving the accuracy of label recommendation.

[0041] In some embodiments, the label recommendation device obtains the label weight time-varying function corresponding to each label in the preset label library in the above step S101 can further include the following steps S201 to S204:

[0042] ​In the embodiment of the present application, the label recommendation device constructs a time decay function corresponding to each label according to the total number of labels included in the preset label library and the creation time of the label. For example, the time decay function is shown in formula (2):

[0043] In the embodiment of the present application, the label recommendation device constructs a time decay function corresponding to each label according to the total number of labels included in the preset label library and the creation time of the label. For example, the time decay function is shown in formula (2):

[0044]

[0045] Wherein, s j (t a ) is the time decay function corresponding to label j, n is the total number of labels included in the preset label library, t j is the system creation time of label j, t a is the current actual time of the system, i.e. the label recommendation time.

[0046] Step S202, determining a bias factor according to the total number of labels included in the preset label library.

[0047] For example, the bias factor determined by the label recommendation device according to the total number of labels included in the preset label library can be 1 / n.

[0048] Step S203, constructing a label weight time-varying function corresponding to each label according to the time decay function corresponding to each label and the bias factor.

[0049] For example, the label recommendation device constructs a label weight time-varying function corresponding to each label according to the time decay function corresponding to each label and the bias factor, as shown in formula (3):

[0050]

[0051] In the embodiment of the present application, after the label recommendation device constructs a label weight time-varying function corresponding to each label according to the time decay function corresponding to each label and the bias factor, the following steps S301 to S304 can also be included:

[0052] Step S301, obtaining a label transition matrix for the preset label library.

[0053] In the embodiments of the present application, the label recommendation device obtains the label transition matrix for the preset label library, including: obtaining label selection behavior data of each object for the preset label library; constructing a label selection feature matrix according to the label selection behavior data; determining the sum of each row element in the label selection feature matrix to obtain the total number of label selections corresponding to each row; calculating the ratio of each element in the label selection feature matrix to the total number of label selections of the corresponding row to obtain the corresponding label selection probability; and using the label selection probability corresponding to each element in the label selection feature matrix to form the label transition matrix.

[0054] Exemplarily, assuming that the preset label library includes n labels, the label recommendation device obtaining the label selection behavior data of each object for the preset label library can be the behavior data of all operation personnel when performing a marketing activity, and constructing a label selection feature matrix according to the label selection behavior data is:

[0055]

[0056] wherein, M is the label selection feature matrix, m ij is the total number of times that all operation personnel select label i and then continue to select label j according to their own experience when performing a marketing activity.

[0057] Then, each element in the matrix M is simultaneously divided by the sum of the elements in the corresponding row, that is, (label selection probability). Finally, the label transition matrix D n×n As shown in formula (5):

[0058]

[0059] It can be understood that the values of the main diagonal line of the label transition matrix are all 0, because in actual application, the operation personnel will not continue to select label i after selecting label i.

[0060] Step S302, determining a label recommendation probability function of each label for the selected label based on the label transition matrix and the time-varying function of the label weight corresponding to each label.

[0061] In the embodiments of the present application, the label recommendation device determines a label recommendation probability function of each label for a selected label based on a label transition matrix and a label weight time-varying function corresponding to each label, including: determining a first probability function of selecting each label through a first preset behavior according to a preset damping coefficient, the label transition matrix, and a probability of the selected label; determining a second probability function of selecting each label through a second preset behavior according to the preset damping coefficient, the label weight time-varying function corresponding to each label, and the probability of the selected label; and determining the sum of the first probability function and the second probability function of each label as the label recommendation probability function of each label.

[0062] Exemplarily, in actual situations, each time an activity is planned, an operator needs to perform behavior A (the first preset behavior) or behavior B (the second preset behavior); the two behaviors are respectively defined as:

[0063] Behavior A: after selecting label i, directly input label j according to the existing experience, and record the probability as α, that is, the preset damping coefficient, generally 0.85.

[0064] Behavior B: after selecting label i, do not know how to select the next label, browse all labels in the preset label library, continue to select the next label j, or do not select, which represents the end of a planning activity, and record the probability as 1-α.

[0065] P 0 (i) represents the probability of selecting label i by the operator at time t=0; P (j) represents the probability of selecting label j by the operator at time t=1. Therefore, it can be concluded that the probability of selecting label j by the operator through behavior A at time t=1 is that is, the first probability function of selecting each label through the first preset behavior. Similarly, it can be known that the probability of selecting label j by the operator through behavior B at time t=1 is that is, the second probability function of selecting each label through the second preset behavior. As described above, it can be known that the probability of selecting label j by the operator at time t=k+1 is shown in formula (6):

[0066]

[0067] Wherein, k is a natural number greater than or equal to 0; P k+1 (j) is the probability of selecting label j at the k+1th time after selecting label i at the kth time; α is a preset damping coefficient; P k (i) is the probability of selecting label i at the kth time; is the probability of selecting label j at the k+1th time through behavior A after selecting label i at the kth time; Pi,j(k+1) represents the probability of selecting label j at k+1th selection after selecting label i at kth selection by behavior B. It is noted that k represents the number of the selection of the target object in one marketing activity, rather than the actual time.

[0068] From the above discussion, Pi,j(k+1) can be equivalent to d ij Therefore, formula (6) can be rewritten as formula (7):

[0069]

[0070] Generally, the traditional PageRank algorithm is to let The advantage is that the calculation is simple and can guarantee the stability of the Markov chain, that is, the algorithm can converge, and the disadvantage is that the old is heavy and the new is light, that is, the new label is not friendly enough, and even if the quality of the new label is high, it is difficult to stand out. Therefore, the label recommendation device will build the label weight time-varying function Pi,j(k+1) can be equivalent to d That is, The above time-varying function satisfies three constraint conditions, that is:

[0071] Constraint condition one: a lighter weight is given to the old label, and a heavier weight is given to the new label, as shown in formula (8):

[0072]

[0073] Constraint condition two: with the passage of time, the weights of the old and new labels gradually converge, and in the limit case, the weights of all labels are 1 / n, which is equivalent to the original PageRank algorithm. As shown in formula (9):

[0074]

[0075] Constraint condition three: the stability of the Markov chain is satisfied, as shown in formula (10):

[0076]

[0077] As can be seen from formula (8) to formula (10), the weight given to the label is time-varying and not fixed, and with the passage of actual time, the weights of the old and new labels gradually converge, and in the limit case, the weights of all labels are constant values. Moreover, according to the above constraint conditions, a lighter weight is given to the old label, and a heavier weight is given to the new label, which can solve the defect of the traditional algorithm that the old is heavy and the new is light.

[0078] Further, the label weight time-varying function built above Pi,j(k+1) can be equivalent to d Afterwards, the label recommendation probability function can be written as formula (11):

[0079]

[0080] Step S303, the initial probability is used to iteratively solve the label recommendation probability function of each label, to obtain the recommendation probability of each label.

[0081] In the embodiment of the present application, after the label recommendation device constructs the label recommendation function, the initial probability is input into the label recommendation probability function for iterative solution, to obtain the recommendation probability of each label.

[0082] In the embodiment of the present application, the label recommendation device sets an iteration termination condition; uses the initial probability to iteratively calculate the label recommendation probability function of each label; stops the iterative calculation of the label recommendation probability function of each label when the adjacent two iteration results satisfy the iteration termination condition; and determines the result obtained by the last iteration calculation as the recommendation probability of each label.

[0083] In the embodiment of the present application, the iteration termination condition set by the label recommendation device can be: Then, the initial probability is input into the label recommendation probability function of each label for iterative calculation, to obtain the iteration result of the s-th time: After one more iteration, the iteration result of the s+1-th time is obtained: If the adjacent two iteration results satisfy the iteration termination condition, the iterative calculation of the label recommendation probability function of each label is stopped; and the result obtained by the last iteration calculation is determined as the recommendation probability of each label.

[0084] Exemplarily, the label transition matrix D n×n , the probability α, i.e. the preset damping coefficient, the initial probability the termination coefficient τ, the current system time t a , the label creation time t j , j = 1…n, and then the formula (11) is iteratively calculated, and the number of iterations is denoted by s, see formula (12):

[0085] 1, let s = 0;

[0086] 2, calculate

[0087]

[0088] 3, if , stop iteration, and output

[0089] ​4. Otherwise, let s = s + 1, Step 2 is performed.

[0090] In the embodiments of the present application, after the label recommendation device determines the recommendation probability of each label, the label recommendation to the target object includes: selecting a label that meets a preset condition from the preset label library according to the recommendation probability of each label, and determining the selected label as a target label; and recommending the target label to the target object.

[0091] In the embodiments of the present application, the preset condition can be that the labels included in the preset label library are sorted in descending order of the recommendation probability, and then the top preset number of labels are recommended as target labels to the target object, or a preset probability recommendation threshold is directly set, and the labels with a recommendation probability greater than the preset probability recommendation threshold are recommended as target labels to the target object. The preset number or the preset probability recommendation threshold can be set according to actual needs and application scenarios.

[0092] It can be understood that the present application improves the PageRank algorithm and proposes a time adaptive algorithm (time-varying function). The improved algorithm can calculate the time-varying pagerank value based on the current actual time and the label creation time, solve the problem of PageRank algorithm that old labels are valued more than new labels, and through the construction and mapping of label data, the improved algorithm is successfully applied to the field of label ranking.

[0093] The embodiments of the present application provide a label recommendation method. In the case that a target object obtains a selected label from a preset label library, a label weight time-varying function corresponding to each label in the preset label library is obtained. The label weight time-varying function satisfies that the label weight value gradually decreases and tends to a preset value as time passes. Based on the label weight time-varying function corresponding to each label and the initial probability of the selected label, a recommendation probability of each label for the selected label is determined. Based on the recommendation probability of each label, a label recommendation is made to the target object. The label recommendation method provided by the present application constructs a corresponding label weight time-varying function for each label, so that the weight value of each label is adaptively adjusted over time, which can solve the problem of old labels being valued more than new labels in the setting of label weight, and thus improve the accuracy of label recommendation.

[0094] The embodiments of the present application provide a label recommendation device, as shown in Figure 4 The label recommendation device includes:

[0095] The acquisition module 401 is configured to, in the case that a target object obtains a selected label from a preset label library, obtain a label weight time-varying function corresponding to each label in the preset label library. The label weight time-varying function satisfies that the label weight value gradually decreases and tends to a preset value as time passes.

[0096] The determining module 402 is configured to determine a recommended probability of each label for the selected label based on the time-varying function of the label weight corresponding to each label and the initial probability of the selected label.

[0097] The recommending module 403 is configured to recommend the label to the target object based on the recommended probability of each label.

[0098] In an embodiment of the present application, the obtaining module 401 is further configured to determine a time decay function corresponding to each label according to a total number of labels included in the preset label library and a creation time of the label; determine a bias factor according to the total number of labels included in the preset label library; and construct the time-varying function of the label weight corresponding to each label according to the time decay function corresponding to each label and the bias factor.

[0099] In an embodiment of the present application, the obtaining module 401 is further configured to obtain a label transition matrix for the preset label library; determine a label recommendation probability function of each label for the selected label based on the label transition matrix and the time-varying function of the label weight corresponding to each label; and the determining module 402 is further configured to iteratively solve the label recommendation probability function of each label by using the initial probability to obtain the recommended probability of each label.

[0100] In an embodiment of the present application, the obtaining module 401 is further configured to obtain label selection behavior data of each object for the preset label library; construct a label selection feature matrix according to the label selection behavior data; determine a sum of each row element in the label selection feature matrix to obtain a total number of label selections corresponding to each row; calculate a ratio of each element in the label selection feature matrix to the total number of label selections of the corresponding row to obtain a corresponding label selection probability; and use the label selection probability corresponding to each element in the label selection feature matrix to compose the label transition matrix.

[0101] In an embodiment of the present application, the determining module 402 is further configured to determine a first probability function of each label selected through a first preset behavior according to a preset damping coefficient, the label transition matrix, and the probability of the selected label; determine a second probability function of each label selected through a second preset behavior according to the preset damping coefficient, the time-varying function of the label weight corresponding to each label, and the probability of the selected label; and determine the sum of the first probability function and the second probability function of each label as the label recommendation probability function of each label.

[0102] In an embodiment of the present application, the determining module 402 is further configured to set an iteration termination condition, perform iterative calculation on the label recommendation probability function of each label using the initial probability, and stop the iterative calculation on the label recommendation probability function of each label when the results of two adjacent iterations meet the iteration termination condition, and determine the result of the last iteration as the recommendation probability of each label.

[0103] In an embodiment of the present application, the determining module 402 is further configured to select a label meeting a preset condition from a preset label library according to the recommendation probability of each label, and determine the selected label as a target label, and recommend the target label to the target object.

[0104] In an embodiment of the present application, the preset label library is a label library composed of labels representing business types.

[0105] An embodiment of the present application provides a label recommendation device, as shown in the figure, comprising a processor 501, a memory 502 and a communication bus 503. Figure 5

[0106] The communication bus 503 is configured to realize the communication connection between the processor 501 and the memory 502.

[0107] The processor 501 is configured to execute the computer program stored in the memory 502 to realize the above-mentioned label recommendation method.

[0108] An embodiment of the present application provides a computer readable storage medium, which stores one or more computer programs, and the one or more computer programs can be executed by one or more processors to realize the above-mentioned label recommendation method. The computer readable storage medium can be a volatile memory such as a random access memory (RAM), or a non-volatile memory such as a read-only memory (ROM), a flash memory, a hard disk drive (HDD) or a solid-state drive (SSD), or a device comprising one or any combination of the above memories, such as a mobile phone, a computer, a tablet device, a personal digital assistant, etc.

[0109] ​Those skilled in the art will appreciate that embodiments of the application can be readily used as a method, apparatus, or computer program product. Accordingly, the application can take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, the application can take the form of a computer program product on one or more computer readable storage media (including, but not limited to, magnetic disks under a server, optical storage media, and the like) having computer usable program code embodied in the medium.

[0110] The present application is described in reference to the flowchart and / or block diagrams of the methods, apparatus (systems) and computer program products according to embodiments of the application. It will be understood that each block of the flowchart and / or block diagrams, and combinations of blocks in the flowchart and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart and / or block diagram block or blocks. Figure 1 one or more functions specified in the flowchart and / or block diagram block or blocks. Figure 1 one or more functions specified in the flowchart and / or block diagram block or blocks.

[0111] These computer program instructions can also be stored in a computer readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer readable memory produce an article of manufacture including instructions which implement the flowchart and / or block diagram block or blocks. Figure 1 one or more functions specified in the flowchart and / or block diagram block or blocks. Figure 1 one or more functions specified in the flowchart and / or block diagram block or blocks.

[0112] These computer program instructions can also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the flowchart and / or block diagram block or blocks. Figure 1 one or more functions specified in the flowchart and / or block diagram block or blocks. Figure 1 one or more functions specified in the flowchart and / or block diagram block or blocks.

[0113] The above description is only specific embodiments of the present application, but the protection scope of the present application is not limited thereto, and any changes or substitutions easily conceived by those skilled in the art within the technical scope disclosed by the present application should be covered within the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

Claims

1. A tag recommendation method, characterized in that: The method comprises: When the target object obtains a selected tag from a preset tag library, a tag weight time-varying function corresponding to each tag in the preset tag library is obtained; the tag weight time-varying function satisfies that the tag weight value gradually decreases and tends to a preset value over time; Obtaining a label transfer matrix for the preset label library; Determining a tag recommendation probability function of each tag for the selected tag based on the tag transfer matrix and the tag weight time-varying function corresponding to each tag; Iteratively solving the tag recommendation probability function of each tag using the initial probability to obtain the recommendation probability of each tag; Based on the recommendation probability of each tag, a tag is recommended to the target object.

2. The method according to claim 1, characterized in that The method further comprises: Determining a time decay function corresponding to each tag according to the total number of tags included in the preset tag library and the creation time of the tags; Determining a bias factor according to the total number of labels included in the preset label library; The tag weight time-varying function corresponding to each tag is constructed according to the time decay function and the bias factor corresponding to each tag.

3. The method according to claim 1, characterized in that The obtaining of a user label transfer matrix for the preset label library includes: Obtaining tag selection behavior data of each object for the preset tag library; Constructing a label selection feature matrix based on the label selection behavior data; Determine the sum of the elements in each row of the label selection feature matrix to obtain the total number of label selections corresponding to each row; For each element in the label selection feature matrix, calculate the ratio of the total number of label selections in the corresponding row to obtain the corresponding label selection probability; The label selection probability corresponding to each element in the label selection feature matrix is ​​used to form the label transfer matrix.

4. The method according to claim 1, wherein The determining, based on the label transfer matrix and the label weight time-varying function corresponding to each label, a label recommendation probability function of each label for the selected label includes: Determining a first probability function for selecting each of the tags through a first preset behavior according to a preset damping coefficient, the tag transfer matrix, and the probability of the selected tag; Determining a second probability function for selecting each tag through a second preset behavior based on the preset damping coefficient, the tag weight time-varying function corresponding to each tag, and the probability of the selected tag; The sum of the first probability function and the second probability function of each tag is determined as the tag recommendation probability function of each tag.

5. The method according to claim 4, characterized in that The iteratively solving the tag recommendation probability function of each tag by using the initial probability to obtain the recommendation probability of each tag includes: Set the iteration termination condition; Using the initial probability, iteratively calculating the tag recommendation probability function of each tag; When two adjacent iteration results satisfy the iteration termination condition, stopping the iterative calculation of the tag recommendation probability function for each tag; The result obtained by the last iterative calculation is determined as the recommendation probability of each tag.

6. The method according to claim 1, characterized in that The recommending a tag to the target object based on the recommendation probability of each tag includes: Selecting a tag that meets a preset condition from the preset tag library according to the recommendation probability of each tag, and determining the selected tag as a target tag; The target tag is recommended to the target object.

7. The method according to any one of claims 1 to 6, characterized in that The preset label library is a label library composed of labels representing business types.

8. A tag recommendation device, characterized in that: include: An acquisition module is configured to acquire, when a target object acquires a selected tag from a preset tag library, a tag weight time-varying function corresponding to each tag in the preset tag library; the tag weight time-varying function satisfies that the tag weight value gradually decreases and tends to a preset value over time; A determination module is configured to obtain a label transfer matrix for the preset label library; determine a label recommendation probability function for each label for the selected label based on the label transfer matrix and the label weight time-varying function corresponding to each label; and iteratively solve the label recommendation probability function for each label using an initial probability to obtain the recommendation probability of each label; The recommendation module is configured to recommend tags to the target object based on the recommendation probability of each tag.

9. A tag recommendation device, characterized in that: include: processor, memory, and communication bus; The communication bus is used to realize the communication connection between the processor and the memory; The processor is configured to execute the computer program stored in the memory to implement the tag recommendation method according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that The computer-readable storage medium stores one or more computer programs, and the one or more computer programs can be executed by one or more processors to implement the tag recommendation method according to any one of claims 1 to 7.

Images