Label Recommendation Method, Label Recommendation Device, Electronic Device, and Storage Medium

Abstract

The present disclosure provides a tag recommendation method, a tag recommendation device, an electronic device, and a computer-readable storage medium, belonging to the field of computer technology. The method includes: obtaining a plurality of tags under a target category and the associated users of each of the tags; determining the tag with the largest number of associated users among the plurality of tags as a reference tag, and the other tags as popularity tags; adding a target tag combination with a confidence level reaching a first threshold between the reference tag and each of the popularity tags to an associated tag set, where the target tag combination is a popularity tag and the corresponding reference tag; recommending at least one popularity tag in the associated tag set to the associated users of the reference tag corresponding to the popularity tag. The present disclosure can improve the hit rate of tag recommendation for user needs, enhance the recommendation effect, and reduce the labor cost.

Description

Background Art

[0002] With the increasingly widespread popularization and application of the Internet in various industries, enterprises in multiple fields such as e-commerce, Internet finance, life services, and games are committed to better recommending products or services to users through the Internet to explore user needs, increase user traffic, and improve service quality.

[0003] Most of the existing product (or service) recommendation methods rely on operators to manually count and predict the potential needs of users to recommend corresponding products, and the recommendation activities for each product are usually configured and carried out separately. Therefore, the existing methods have relatively high labor costs, and the results of manual statistics and predictions usually have a low hit rate for the actual needs of users, resulting in the product recommendation not achieving the expected effect; in addition, the products recommended by the existing methods have a relatively high similarity with the products already purchased by users, which limits the exploration of user needs to a relatively small range and further affects the effect of product recommendation.

[0004] It should be noted that the information disclosed in the above background art section is only used to enhance the understanding of the background of the present disclosure, and thus may include information that does not constitute the prior art known to those of ordinary skill in the art. Summary of the Invention

[0005] The present disclosure provides a tag recommendation method, a tag recommendation device, an electronic device, and a computer-readable storage medium, thereby at least to a certain extent overcoming the problems of poor effect and high labor cost of the existing product recommendation methods.

[0006] Other features and advantages of the present disclosure will become apparent through the following detailed description, or will be learned in part through the practice of the present disclosure.

[0007] According to an aspect of the present disclosure, a tag recommendation method is provided, including: obtaining a plurality of tags under a target category and the associated users of each of the tags; determining the tag with the largest number of associated users among the plurality of tags as a reference tag, and other tags as popularity tags; adding the target tag combinations with the confidence level between the reference tag and each of the popularity tags reaching a first threshold to an associated tag set, where the target tag combinations are popularity tags and the corresponding reference tags; and recommending at least one popularity tag in the associated tag set to the associated users of the reference tag corresponding to the popularity tag.

[0008] In an exemplary embodiment of the present disclosure, after adding the target label combinations with the confidence between the reference label and each of the popularity labels reaching a first threshold to the associated label set, the method further includes: removing the reference label from the multiple labels, and determining the popularity label with the highest confidence between the multiple labels and the removed reference label as the new reference label; determining the confidence between the new reference label and the remaining popularity labels, and adding the target label combinations with the confidence reaching the first threshold to the associated label set; repeating the above steps until only one label remains among the multiple labels.

[0009] In an exemplary embodiment of the present disclosure, before obtaining the multiple labels under the target category and the associated users of each of the labels, the method further includes: obtaining initial labels, and clustering the initial labels to obtain multiple categories; using any one of the multiple categories as the target category.

[0010] In an exemplary embodiment of the present disclosure, clustering the initial labels to obtain multiple categories includes: counting the support of the label combinations formed by any N labels in the initial labels, where N is an integer greater than 1; counting the label combinations with the support reaching a second threshold, and classifying the label combinations having at least one common label into one category to obtain the multiple categories.

[0011] In an exemplary embodiment of the present disclosure, using any one of the multiple categories as the target category includes: counting the total number of associated users of each category, and calculating the average value of the label-associated users of each category; sorting in descending order according to the average value of the label-associated users of each category, and sequentially using each category as the target category.

[0012] In an exemplary embodiment of the present disclosure, before obtaining the multiple labels under the target category and the associated users of each of the labels, the method further includes: removing the labels with the number of associated users lower than a third threshold from the multiple labels.

[0013] In an exemplary embodiment of the present disclosure, adding the target label combinations with the confidence between the reference label and each of the popularity labels reaching a first threshold to the associated label set includes: adding the target label combinations with the confidence reaching the first threshold and the confidence of the target label combinations to the associated label set; recommending at least one popularity label in the associated label set to the associated users of the reference label corresponding to the popularity label includes: sequentially recommending each popularity label to the associated users of the reference label corresponding to the popularity label in descending order of the confidence of each target label combination in the associated label set.

[0014] According to one aspect of the present disclosure, there is provided a tag recommendation device, including: a tag information acquisition module configured to acquire a plurality of tags under a target category and the number of associated users of each of the tags; a reference tag determination module configured to determine the tag with the largest number of associated users among the plurality of tags as a reference tag, and determine other tags as popularity tags; a confidence determination module configured to add a target tag combination with a confidence level reaching a first threshold between the reference tag and each of the popularity tags to an associated tag set, where the target tag combination is a popularity tag and the corresponding reference tag; and a popularity tag recommendation module configured to recommend at least one popularity tag in the associated tag set to the associated users of the reference tag corresponding to the popularity tag.

[0015] According to one aspect of the present disclosure, there is provided an electronic device, including: a processor; and a memory configured to store executable instructions of the processor; wherein the processor is configured to execute the method described in any one of the above via executing the executable instructions.

[0016] According to one aspect of the present disclosure, there is provided a computer-readable storage medium having a computer program stored thereon, where the computer program, when executed by a processor, implements the method described in any one of the above.

[0017] The exemplary embodiments of the present disclosure have the following beneficial effects:

[0018] After acquiring the tag set of the target category and the associated users of each tag, a reference tag is determined according to the number of associated users, and the confidence level between the reference tag and other popularity tags is calculated, and target tag combinations with a relatively high degree of association are screened out to form an associated tag set, and then tag recommendation is performed according to the tag combination situation in the associated tag set. On the one hand, through the calculation and screening of the confidence level of the target tag combination, the association between tags can be discovered, and tag recommendation can be performed according to the association situation, which can improve the hit rate of tag recommendation for the actual needs of users and enhance the recommendation effect. On the other hand, based on the tag information in the acquired target category, the present exemplary embodiment can automatically select a reference tag, calculate the confidence level, generate an associated tag set, and finally automatically perform tag recommendation according to the associated tag set, thereby realizing automatic tag recommendation and saving labor costs. On the further hand, by calculating and mining the association relationship of the tags under the target category, the scope of tag recommendation can be expanded, thereby further enhancing the recommendation effect.

[0019] It should be understood that the above general description and the following detailed description are only exemplary and explanatory, and cannot limit the present disclosure. Description of the Drawings

[0020] The accompanying drawings here are incorporated into the specification and form a part of this specification, showing embodiments consistent with the present disclosure, and are used together with the specification to explain the principles of the present disclosure. Obviously, the accompanying drawings in the following description are only some embodiments of the present disclosure, and those of ordinary skill in the art can obtain other drawings based on these drawings without creative efforts.

[0021] Figure 1 Schematically showing a flowchart of a tag recommendation method in this exemplary embodiment;

[0022] Figure 2 Schematically showing a flowchart of another tag recommendation method in this exemplary embodiment;

[0023] Figure 3 Schematically showing a sub - flowchart of a tag recommendation method in this exemplary embodiment;

[0024] Figure 4 Schematically showing a flowchart of yet another tag recommendation method in this exemplary embodiment;

[0025] Figure 5 Schematically showing a structural block diagram of a tag recommendation device in this exemplary embodiment;

[0026] Figure 6 Schematically showing an electronic device for implementing the above - mentioned method in this exemplary embodiment;

[0027] Figure 7 Schematically showing a computer - readable storage medium for implementing the above - mentioned method in this exemplary embodiment. Detailed implementation manners

[0028] Example embodiments will now be described more fully with reference to the accompanying drawings. However, the example embodiments can be implemented in various forms and should not be construed as limited to the examples set forth herein; rather, these embodiments are provided so that this disclosure will be more thorough and complete, and will fully convey the concept of the example embodiments to those skilled in the art. The features, structures, or characteristics described may be combined in any suitable manner in one or more embodiments.

[0029] The exemplary embodiment of the present disclosure first provides a tag recommendation method. Among them, a tag refers to a form of content presentation of an Internet product or service. The actual product or service can be mapped to tags composed of simple texts. Recommending tags to users means recommending the products or services represented by the tags to users.

[0030] Refer to Figure 1 As shown, this tag recommendation method may include steps S110 to S140:

[0031] S110. Obtain multiple tags under the target category and the associated users of each tag.

[0032] Among them, the target category refers to a set of specific types of tags. For example, tags such as "apple", "watermelon", and "banana" belong to the "fruit" category. The associated users of a tag refer to users whose behaviors have a certain association with the product of the tag. For example, they can be users who have purchased the product of the tag or users who have collected, browsed, searched, or commented on the product of the tag. The present disclosure does not make special limitations on this.

[0033] S120. Determine the tag with the largest number of associated users among the above multiple tags as the reference tag, and determine other tags as popularity tags.

[0034] After obtaining the associated users of each tag, the number of associated users of each tag can be counted. Among them, the tag with the largest number of associated users can be used as the reference tag, and other tags can be used as popularity tags. In this exemplary embodiment, the reference tag and the popularity tag are a set of relative concepts. The reference tag can be regarded as the tag with the highest popularity in the current target category. Other tags can be based on this tag to calculate the relative popularity (i.e., the degree of association). Therefore, other tags can be called popularity tags.

[0035] S130. Add the target tag combinations with the confidence level reaching the first threshold between the reference tag and each popularity tag to the associated tag set, where the target tag combination is the popularity tag and the corresponding reference tag.

[0036] Confidence level is a concept in the association rule. In this exemplary embodiment, calculate the confidence level of reference tag -> popularity tag, and its meaning is as follows:

[0037]

[0038] Among them, B refers to the set of associated users of the reference label, H refers to the set of associated users of each popularity label, Confidence refers to the confidence level, and Count refers to the number of elements in the set. It can be seen from formula (1) that the confidence level actually refers to the proportion of the number of users who are associated with both the reference label and the popularity label to the number of users associated with the reference label. A high confidence level indicates a high degree of association between the reference label and the popularity label. According to this method, the confidence level between the reference label and each popularity label can be calculated, and then the reference label-popularity label combinations with relatively high confidence levels can be screened out through a first threshold, that is, the target label combinations. The first threshold is a confidence level screening criterion set according to experience. The reference label-popularity label combinations with a confidence level lower than the first threshold have a lower degree of association. In this exemplary embodiment, it can be considered that they belong to weak association combinations and are not adopted; the target label combinations that reach the first threshold (i.e., are greater than or equal to the first threshold) can be considered strong association combinations, and these combinations are formed into an associated label set for use in subsequent steps.

[0039] S140. Recommend at least one popularity label in the associated label set to the associated users of the reference label corresponding to the popularity label.

[0040] The associated label set contains many target label combinations, and each target label combination marks which label is the reference label and which label is the popularity label. For a target label combination, there are usually a part of users who are only associated with the reference label of the combination and not with the popularity label of the combination. Therefore, the popularity label can be recommended to this part of users.

[0041] In this exemplary embodiment, the above-mentioned recommendation of the popularity label can be implemented for any one or more target label combinations in the associated label set, or a part of the target label combinations can be screened out through certain conditions to implement the above-mentioned recommendation of the popularity label. For example, the target label combinations with a relatively large number of associated users or the target label combinations with a relatively large difference in the number of associated users between the reference label and the popularity label are screened out, etc. The above-mentioned recommendation of the popularity label can also be implemented for all the target label combinations in the associated label set. The present disclosure does not make any special limitations on this.

[0042] Based on the above description, in this exemplary embodiment, after obtaining the tag set of the target category and the associated users of each tag, the reference tag is determined according to the number of associated users, and the confidence between the reference tag and other popular tags is calculated. The target tag combinations with relatively high correlation are screened out to form an associated tag set, and then tag recommendations are made according to the tag combination situation in the associated tag set. On the one hand, through the calculation and screening of the confidence of the target tag combinations, the associations between tags can be discovered, and tag recommendations can be made according to the association situation, which can improve the hit rate of tag recommendations for the actual needs of users and enhance the recommendation effect. On the other hand, based on the tag information in the obtained target category, this exemplary embodiment can automatically select the reference tag, calculate the confidence, generate the associated tag set, and finally automatically make tag recommendations according to the associated tag set, thus realizing the automation of tag recommendations and saving labor costs. On the other hand, by calculating and mining the association relationships of the tags under the target category, the scope of tag recommendations can be expanded, thereby further enhancing the recommendation effect.

[0043] In one exemplary embodiment, as shown in Figure 2 After step S130, the tag recommendation method may further include the following steps:

[0044] S131. Remove the reference tag from the above-mentioned multiple tags, and determine the popular tag with the highest confidence between the above-mentioned multiple tags and the removed reference tag as the new reference tag;

[0045] S132. Determine the confidence between the new reference tag and the remaining popular tags, and add the target tag combinations with the confidence reaching the first threshold to the associated tag set;

[0046] S133. Repeat steps S131 and S132 until only one tag remains among the above-mentioned multiple tags.

[0047] In other words, through step S130, the target tag combinations formed by the reference tag and the popular tags with strong associations under the current reference tag are screened out. Then a new reference tag can be replaced. The specific replacement process is as described in step S131. The previous reference tag can be removed from the target category, and the popular tag with the highest confidence with the previous reference tag can be determined as the new reference tag, and the other popular tags remain as popular tags.

[0048] For example, if the number of labels in the target category at the initial stage is L, first determine a reference label B1, and the remaining L-1 labels are popularity labels. In the first round of confidence calculation, B1 calculates confidence with the L-1 popularity labels respectively, for a total of L-1 calculations. The target label combinations with confidence reaching the first threshold can be combined to form an associated label set; in the second round of confidence calculation, B1 can be removed, and among the remaining L-1 popularity labels, the popularity label with the highest confidence with B1 is determined as the new reference label B2, and the remaining L-2 labels are still popularity labels. B2 calculates confidence with them respectively, for a total of L-2 calculations. The target label combinations with confidence reaching the first threshold can be added to the above-mentioned associated label set; in the third round of confidence calculation, B2 can be removed, and there are a total of L-2 labels remaining in the target category. Determine the new reference label B3 according to the above method, calculate the confidence, and add the strongly associated target label combinations to the above-mentioned associated label set. It can be seen that with each round of re-determining the reference label and calculating the confidence, the number of labels in the target category becomes smaller and smaller, and the number of target label combinations in the associated label set usually becomes larger and larger. When it comes to only two labels B L-1 and B L left in the target category, the B L-2 with higher confidence with the previous reference label B L-1 is determined as the new reference label, calculate the confidence Confidence(B L-1 ->B L ) and determine whether to add this combination to the associated label set; then in the last round, B L-1 is removed from the target category, and only one label B L is left, and the confidence cannot be calculated continuously, so the above loop process ends.

[0049] Through the above loop process, the confidence calculation between any two labels in the target category is actually completed, and the strongly associated label combinations are screened out through the first threshold. Finally, an associated label set is obtained, and subsequent label recommendations can be carried out. Thus, sufficient association mining of the labels in the target category is realized, and based on this, label recommendations can achieve a relatively sufficient recommendation effect.

[0050] Figure 3 Exemplarily shows the processes of steps S131 and S132. The target category includes a total of 6 labels A, B, C, D, E, and F. Among them, A has the largest number of associated users. First, it is determined as the reference label, and the confidence between A and the labels B to F is calculated respectively. The confidence is represented by b1, c1, d1, etc. In this exemplary embodiment, the first threshold can be set to 0.3, and the target label combinations with confidence reaching 0.3 are added to the associated label set, such as Figure 3As shown in the list on the right side in the figure, the confidence levels of A - E are lower than 0.3, so they are not added. Remove A from the target categories and enter the second round of calculation. It can be determined that the label D with the highest confidence level with A is used as the reference label. Calculate the confidence levels of D with B, C, E, and F respectively, and add the target label combinations that reach 0.3 to the associated label set. Among them, the confidence level of D - C is lower than 0.3, so it is not added. Remove D from the target categories again and enter the third round of calculation. It can be determined that the label B with the highest confidence level with D is used as the reference label. Calculate the confidence levels of B with C, E, and F respectively, and add the target label combinations that reach 0.3 to the associated label set. Remove B from the target categories again and enter the fourth round of calculation. It can be determined that the label F with the highest confidence level with B is used as the reference label. Calculate the confidence levels of F with C and E respectively, and add the target label combinations that reach 0.3 to the associated label set. Remove F from the target categories again and enter the fifth round of calculation. There are still labels C and E left. Use the label C with a higher confidence level with F as the reference label and calculate the confidence level of C -> E. Since it is higher than 0.3, it is added to the associated label set. Remove C from the target categories, and only one label E remains, and the loop process ends, obtaining Figure 3 the associated label set shown in the figure. According to Figure 3 the associated label set, recommend label D to the associated users of label A, recommend label B to the associated users of label A... recommend label E to the associated users of label C, thus completing the process of label recommendation.

[0051] In an exemplary embodiment, based on Figure 2 the label recommendation method flow shown in the figure, when looping through steps S131 and S132, the number of loops can be set, for example, set to M. Then, determine a reference label in each round, and by calculating the confidence levels of this reference label with other popular labels, screen out the target label combinations and add them to the associated label set. A total of M rounds are performed, and M reference labels are determined in sequence. After that, no matter how many labels are left among the initial multiple labels, the loop ends, and step S140 is performed according to the obtained associated label set.

[0052] Figure 4 shows a flowchart of a label recommendation method in this exemplary embodiment. Refer to Figure 4 shown in the figure, the associated user information of each label in the target category can be obtained through the following process: Extract and count the user behavior records of each label from the user behavior log, and perform data validity verification to screen out the valid data and generate the associated user information of each label.

[0053] The above - mentioned target category is a set of a specific type of labels. The label recommendation method can also be extended to multiple categories or even all labels in the application scenario. Therefore, in an exemplary embodiment, refer to Figure 4As shown in the figure, the label recommendation method may further include the following steps:

[0054] S108. Obtain initial labels and cluster the initial labels to obtain multiple categories.

[0055] S109. Use any one of the multiple categories as the target category.

[0056] Among them, according to the specific application scenario, the initial labels can be all or part of the labels within a specific application, or all or part of the labels for a specific field, etc. The present disclosure does not make special limitations on this. The labels can be clustered by the K-means algorithm, or other clustering methods can be adopted. After obtaining multiple categories, steps S110-S140 can be applied to any one of the categories to implement label recommendation under that category, or steps S110-S140 can be applied to each category respectively, so as to implement label recommendation within the entire label range, making the method of this embodiment more general.

[0057] Furthermore, the above-mentioned clustering of the initial labels to obtain multiple categories can be specifically implemented through the following steps:

[0058] Statistically analyze the support of the label combinations formed by any N labels in the initial labels, where N is an integer greater than 1.

[0059] Statistically analyze the label combinations whose support reaches the second threshold, and classify the label combinations with at least one common label among these label combinations into one category to obtain the multiple categories in step S108.

[0060] Among them, support is a concept in association rules. In this exemplary embodiment, the support of a label combination containing N labels can be calculated by the following method:

[0061]

[0062] Among them, A 0 is the set of associated users of all labels, and A 1 , A 2 …A N are respectively the sets of associated users of the N labels for calculating support, and Support is the support. It can be seen from formula (2) that the meaning of support is the proportion of the number of users who are associated with N labels at the same time to the total number of users. N can be an integer greater than 1, such as 2, 3, 4, etc. If N is 2, the support of each two-label combination in all labels can be statistically analyzed. If N is 3, the support of each three-label combination can be statistically analyzed. The value of N can be set according to experience. When the general relevance of the labels is strong, N can be set to a larger value, and vice versa, it can be set to a smaller value.

[0063] The support degree can reflect the correlation degree among the tags within a tag combination. Therefore, it can be measured by a second threshold. A tag combination that reaches the second threshold is considered to have a relatively high correlation degree and is an effective tag combination, which enters the subsequent clustering step. The second threshold can be set according to experience. When N is relatively large, the second threshold can be set relatively low.

[0064] When classifying tag combinations with at least one common tag into one category, the following two specific methods can be used:

[0065] (1). Suppose there are three tag combinations, (A 1 、A 2 …A N ), (B 1 、B 2 …B N ) and (C 1 、C 2 …C N ). If there are tags A i 、B j 、C k respectively in the three tag combinations, and A i = B j = C k , that is, these three tags are the same tag, then these three tag combinations can be classified into one category.

[0066] (2). Suppose there are three tag combinations, (A 1 、A 2 …A N ), (B 1 、B 2 …B N ) and (C 1 、C 2 …C N ). If there are tags A i 、B j respectively in tag combinations A and B, and A i = B j , and there are tags B k 、C l respectively in tag combinations B and C, and B k = C l , that is, there is the same tag in tag combinations A and B, and there is also the same tag in tag combinations B and C, but there is no same tag in tag combinations A and C. In this case, these three tag combinations can also be classified into one category, that is, when a tag combination has any common tag with an already classified tag combination, this tag combination can be classified into the already classified tag combination.

[0067] This embodiment does not make a special limitation on which of the above methods is specifically adopted. By classifying the tag combinations with the support degree reaching the second threshold, multiple categories can be obtained, so as to facilitate subsequent tag recommendation for each category.

[0068] It should be noted that in practical applications, the values of the above N and the second threshold can be adjusted according to the result feedback. For example, when the number of finally obtained categories is too small, the value of N can be appropriately reduced or the second threshold can be decreased; when the number of tags in each category is too small, the value of N can be appropriately reduced, etc.

[0069] In an exemplary embodiment, step S109 can be specifically implemented through the following steps:

[0070] Count the total number of associated users in each category, and calculate the average value of the tag-associated users in each category.

[0071] Sort in descending order according to the average value of the tag-associated users, and successively take each category as the target category.

[0072] Among them, the total number of associated users in each category refers to the sum of the associated users of all tags in this category, and repeated users are also counted repeatedly. For example, if user A is associated with both tag A and tag B, it is counted as 2 in the total number of associated users; the average value of the tag-associated users is the average value obtained by dividing the total number of associated users in this category by the total number of tags in this category. The average value of the tag-associated users can reflect the popularity status of a category. The lower the average value, the lower the overall popularity of the tags in this category. For categories with low popularity, the potential for association mining is usually high. Therefore, they can be preferentially determined as the target category, and steps S110 to S140 can be executed to achieve tag recommendation and optimize the allocation of resources; for categories with high popularity, since the base of associated users is large, tag recommendation can have an effect on a large number of users. Therefore, they can also be preferentially determined as the target category and steps S110 to S140 can be executed; this disclosure does not make a special limitation on the specific order.

[0073] In other embodiments, it is also possible to count the variance of the number of associated users of each tag in each category. A category with a large variance indicates that the associated user groups of the tags in it are quite different, and steps S110 to S140 can be preferentially executed to achieve tag recommendation, which can achieve a better recommendation effect.

[0074] In an exemplary embodiment, referring to Figure 4 as shown, after step S110, the tag recommendation method may further include the following step S111:

[0075] S111. Remove the tags with the number of associated users lower than the third threshold from the above-mentioned multiple tags.

[0076] Among them, the number of associated users reflects the popularity of each tag, so it can be distinguished by the third threshold. Tags with an associated number of users less than the third threshold are usually less popular. The third threshold can be set based on experience or the characteristics of the application scenario and the target category. For tags with lower popularity, the confidence between them and the benchmark tag is usually low, and it is difficult to form a target tag combination. In order to reduce the amount of confidence calculation in subsequent steps, the tags with lower popularity can be removed.

[0077] Furthermore, step S111 may specifically include the following steps:

[0078] Mark the tags whose associated users in the target category are less than a third threshold as cold tags, and mark the tags whose associated users reach the third threshold as hot tags;

[0079] Recommend cold tags to users associated with hot tags;

[0080] Remove the cold labels from the target category.

[0081] Among them, the labels in the target category are distinguished by a third threshold, and the labels whose associated user number reaches the third threshold are hot labels, and the labels whose associated user number is less than the third threshold are cold labels. For cold labels, the subsequent confidence calculation process may not be performed, and the cold labels may be directly recommended to the associated users of the hot labels, and may also be recommended to the associated users of other cold labels to increase the popularity of the cold labels as much as possible. Subsequently, the cold labels are removed from the target category, so that only the hot labels remain in the target category. For the hot labels, the subsequent steps S120 to S140 are performed to achieve the recommendation of the hot labels. Therefore, this embodiment is equivalent to classifying the labels of the target category into hot and cold categories. Different label recommendation mechanisms are adopted for cold labels and hot labels, respectively, which can better adapt to the characteristics of each type of label and achieve better recommendation effects.

[0082] It should be noted that the hot tags are relative to the cold tags, and are different from the heat tags in step S120. In this exemplary embodiment, after the cold tags are removed from the target category, the hot tags are left. In step S120, a reference tag is selected from the hot tags, and the remaining hot tags become heat tags.

[0083] Furthermore, the third threshold value may be set and optimized by the following steps:

[0084] If the cold label ratio of the target category exceeds the preset range, the third threshold is adjusted, and the labels in the target category are classified into cold labels and hot labels again using the third threshold.

[0085] Among them, the preset range is the normal range of the ratio of cold labels to all labels of the target category. When there are too many cold labels, the number of hot labels is small, and the association between hot labels cannot be effectively mined. In addition, the label recommendation volume is too high, which may produce meaningless recommendations. In this case, the third threshold can be appropriately lowered. When there are too few cold labels, for hot labels with a small number of associated users, the confidence calculation and label recommendation process of hot labels may not be able to achieve full label recommendation, affecting the effect. In this case, the third threshold can be appropriately increased. Controlling the cold label ratio within an appropriate range can achieve a better balance between the label recommendation volume and the recommendation effect, and achieve a high cost-effectiveness of label recommendation.

[0086] In an exemplary embodiment, step S130 may include the following steps:

[0087] The target label combination whose confidence reaches the first threshold and the confidence of the target label combination are added to the associated label set.

[0088] Accordingly, step S140 may include the following steps:

[0089] According to the order of confidence of each target tag combination in the associated tag set from high to low, each hot tag is recommended to the associated user of the reference tag corresponding to the hot tag.

[0090] Refer to the above Figure 3 As shown, in the associated tag set, in addition to recording the benchmark tag and hot tag of each target tag combination, the confidence of each target tag combination can also be recorded. After obtaining a complete set of associated tags, each combination can be sorted from high to low according to the confidence, and the hot tags in the top-ranked target tag combination are preferentially recommended to the associated users of their corresponding benchmark tags. Since the confidence reflects the degree of association between the benchmark tag and the hot tag in each target tag combination, for combinations with a high degree of association, a higher hit rate can usually be achieved when recommending tags. Therefore, recommending tags in each target tag combination in sequence according to the order of confidence can further optimize resource allocation and achieve better recommendation effects.

[0091] Further, after all the hot tags in the associated tag set have been recommended, after a certain period of time, the tag recommendation may have an effect on users, that is, users have performed corresponding consumption, following, commenting, favoriting and other behaviors according to the recommended tags, and established new associations with the tags, then the number of associated users of each tag has changed greatly. In this case, steps S110 to S140 can be repeated to start the tag recommendation process again. That is, the entire tag recommendation process can be carried out cyclically. An interval time can be set, and after the previous process is completely finished, the next tag recommendation process starts after this interval time. A cycle can also be set, and the above-mentioned tag recommendation process is carried out once within each cycle. Thus, long-term tag recommendation and product promotion can be achieved, and user traffic can be increased.

[0092] An exemplary embodiment of the present disclosure further provides a tag recommendation device. Refer to Figure 5 As shown, the device 500 may include: a tag information acquisition module 510, configured to acquire a plurality of tags under a target category and the number of associated users of each tag; a reference tag determination module 520, configured to determine the tag with the largest number of associated users among the above-mentioned plurality of tags as the reference tag, and other tags as hot tags; a confidence determination module 530, configured to add a target tag combination whose confidence between the reference tag and each hot tag reaches a first threshold to the associated tag set, where the target tag combination is a hot tag and the corresponding reference tag; a hot tag recommendation module 540, configured to recommend at least one hot tag in the associated tag set to the associated users of the reference tag corresponding to the hot tag.

[0093] In an exemplary embodiment, the confidence determination module may include: a confidence calculation unit, configured to add a target tag combination whose confidence between the reference tag and each hot tag reaches a first threshold to the associated tag set; a reference tag change unit, configured to remove the reference tag from the plurality of tags, and determine the hot tag with the highest confidence between the plurality of tags and the removed reference tag as the new reference tag; the confidence calculation unit is further configured to determine the confidence between the new reference tag and the remaining hot tags, and add the target tag combination whose confidence reaches the first threshold to the associated tag set; the confidence determination module may further include: a scheduling processing unit, configured to schedule the reference tag change unit to repeatedly remove the reference tag from the plurality of tags, and determine the hot tag with the highest confidence between the plurality of tags and the removed reference tag as the new reference tag, and schedule the confidence calculation unit to repeatedly determine the confidence between the new reference tag and the remaining hot tags, and add the target tag combination whose confidence reaches the first threshold to the associated tag set until only one tag remains among the above-mentioned plurality of tags.

[0094] In an exemplary embodiment, the tag recommendation device may further include: a target category determination module, configured to obtain initial tags, cluster the initial tags to obtain multiple categories, and use any one of the multiple categories as the target category.

[0095] In an exemplary embodiment, the target category determination module may include: a tag clustering unit, configured to count the support degrees of tag combinations formed by any N tags in the initial tags, where N is an integer greater than 1, count the tag combinations whose support degrees reach a second threshold, and classify the tag combinations having at least one common tag into one category to obtain multiple categories.

[0096] In an exemplary embodiment, the target category determination module may include: a category sorting unit, configured to count the total number of associated users of each category, calculate the mean value of the tag-associated users of each category, sort them in descending order of the mean value of the tag-associated users, and use each category as the target category in turn.

[0097] In an exemplary embodiment, the tag information acquisition module may also be configured to remove the tags whose associated user numbers are lower than a third threshold from the above-mentioned multiple tags.

[0098] In an exemplary embodiment, the confidence determination module may be configured to add the target tag combinations whose confidence degrees reach a first threshold and the confidence degrees of the target tag combinations to the associated tag set; the hot tag recommendation module may be configured to sequentially recommend each hot tag to the associated users of the reference tag corresponding to the hot tag in descending order of the confidence degrees of the target tag combinations in the associated tag set.

[0099] The specific details of the above-mentioned modules / units have been described in detail in the embodiments of the method part, and thus will not be elaborated here.

[0100] The exemplary embodiment of the present disclosure also provides an electronic device capable of implementing the above method.

[0101] Those skilled in the art to which the present disclosure pertains can understand that various aspects of the present disclosure can be implemented as a system, a method, or a program product. Therefore, various aspects of the present disclosure can be specifically implemented in the following forms, namely: a complete hardware implementation manner, a complete software implementation manner (including firmware, microcode, etc.), or an implementation manner combining hardware and software aspects, which can be collectively referred to as "circuit", "module", or "system" here.

[0102] Next, refer to Figure 6 to describe the electronic device 600 according to this exemplary embodiment of the present disclosure. Figure 6 The shown electronic device 600 is only an example and should not impose any limitation on the functions and usage scope of the embodiments of the present disclosure.

[0103] As shown Figure 6 in FIG. 1, the electronic device 600 is presented in the form of a general-purpose computing device. The components of the electronic device 600 may include, but are not limited to: at least one of the above-mentioned processing units 610, at least one of the above-mentioned storage units 620, a bus 630 connecting different system components (including the storage unit 620 and the processing unit 610), and a display unit 640.

[0104] Among them, the storage unit stores program code, and the program code can be executed by the processing unit 610, so that the processing unit 610 executes the steps according to various exemplary embodiments of the present disclosure described in the "Exemplary Method" section of the present specification above. For example, the processing unit 610 may execute Figure 1 the steps S110 to S140 shown in FIG. 2, etc.

[0105] The storage unit 620 may include a readable medium in the form of a volatile storage unit, such as a random access storage unit (RAM) 621 and / or a cache storage unit 622, and may further include a read-only storage unit (ROM) 623.

[0106] The storage unit 620 may further include a program / utility 624 having a set (at least one) of program modules 625. Such program modules 625 include, but are not limited to: an operating system, one or more application programs, other program modules, and program data. Each or some combination of these examples may include the implementation of a network environment.

[0107] The bus 630 may represent one or more of several types of bus structures, including a storage unit bus or a storage unit controller, a peripheral bus, a graphics acceleration port, a processing unit, or a local bus using any bus structure in a variety of bus structures.

[0108] The electronic device 600 can also communicate with one or more external devices 800 (such as a keyboard, a pointing device, a Bluetooth device, etc.), and can also communicate with one or more devices that enable a user to interact with the electronic device 600, and / or communicate with any device that enables the electronic device 600 to communicate with one or more other computing devices (such as a router, a modem, etc.). Such communication can be carried out through the input / output (I / O) interface 650. Moreover, the electronic device 600 can also communicate with one or more networks (such as a local area network (LAN), a wide area network (WAN), and / or a public network, such as the Internet) through the network adapter 660. As shown in the figure, the network adapter 660 communicates with other modules of the electronic device 600 through the bus 630. It should be understood that, although not shown in the figure, other hardware and / or software modules can be used in combination with the electronic device 600, including but not limited to: microcode, device drivers, redundant processing units, external disk drive arrays, RAID systems, tape drives, and data backup storage systems, etc.

[0109] Through the description of the above embodiments, those skilled in the art can easily understand that the exemplary embodiments described herein can be implemented by software, or can be implemented by the way of software combined with necessary hardware. Therefore, the technical solutions according to the embodiments of the present disclosure can be embodied in the form of a software product, and the software product can be stored in a non-volatile storage medium (which can be a CD-ROM, a USB flash drive, a mobile hard disk, etc.) or on a network, including several instructions to enable a computing device (which can be a personal computer, a server, a terminal device, or a network device, etc.) to execute the method according to the exemplary embodiments of the present disclosure.

[0110] The exemplary embodiments of the present disclosure also provide a computer-readable storage medium, on which a program product capable of implementing the above method of this specification is stored. In some possible implementation manners, various aspects of the present disclosure can also be implemented in the form of a program product, which includes program code. When the program product runs on a terminal device, the program code is used to enable the terminal device to execute the steps according to various exemplary embodiments of the present disclosure described in the above "Exemplary Method" section of this specification.

[0111] Referring Figure 7 As shown, a program product 700 for implementing the above method according to the exemplary embodiments of the present disclosure is described. It can adopt a portable compact disc read-only memory (CD-ROM) and include program code, and can run on a terminal device, such as a personal computer. However, the program product of the present disclosure is not limited to this. In this document, the readable storage medium can be any tangible medium that contains or stores a program, and the program can be used by or in combination with an instruction execution system, device, or device.

[0112] The program product may employ any combination of one or more readable media. The readable media can be a readable signal medium or a readable storage medium. A readable storage medium can be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination of the foregoing. More specific examples (a non-exhaustive list) of the readable storage medium include: an electrical connection having one or more wires, a portable disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0113] A computer-readable signal medium may include a data signal propagated in a baseband or as part of a carrier wave, in which the readable program code is carried. Such a propagated data signal may take many forms, including but not limited to an electromagnetic signal, an optical signal, or any suitable combination of the foregoing. The readable signal medium may also be any readable medium other than the readable storage medium, which can send, propagate, or transmit a program for use by or in connection with an instruction execution system, apparatus, or device.

[0114] The program code contained on the readable medium may be transmitted with any appropriate medium, including but not limited to wireless, wired, optical fiber cable, RF, etc., or any suitable combination of the foregoing.

[0115] The program code for performing the operations of the present disclosure may be written in any combination of one or more programming languages, including object-oriented programming languages such as Java, C++, etc., and also including conventional procedural programming languages such as the "C" language or similar programming languages. The program code may be executed entirely on the user's computing device, partially on the user's device, executed as a stand-alone software package, partially on the user's computing device and partially on a remote computing device, or entirely on the remote computing device or server. In the case of a remote computing device, the remote computing device may be connected to the user's computing device through any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computing device (e.g., through the Internet using an Internet service provider).

[0116] In addition, the above drawings are only schematic illustrations of the processes included in the method according to the exemplary embodiments of the present disclosure, rather than for limiting purposes. It is easy to understand that the processes shown in the above drawings do not indicate or limit the chronological order of these processes. Additionally, it is also easy to understand that these processes may be executed synchronously or asynchronously, for example, in multiple modules.

[0117] It should be noted that although several modules or units of the device for action execution are mentioned in the above detailed description, such division is not mandatory. In fact, according to the exemplary embodiments of the present disclosure, the features and functions of two or more modules or units described above can be embodied in one module or unit. Conversely, the features and functions of one module or unit described above can be further divided and embodied by multiple modules or units.

[0118] Other embodiments of the present disclosure will be readily apparent to those skilled in the art upon consideration of the specification and practice of the invention disclosed herein. This application is intended to cover any variations, uses, or adaptations of the present disclosure that follow the general principles of the present disclosure and include common general knowledge or conventional technical means in the technical field not disclosed herein. The specification and examples are only to be considered as exemplary, and the true scope and spirit of the present disclosure are pointed out by the claims.

[0119] It should be understood that the present disclosure is not limited to the exact structures described above and shown in the drawings, and various modifications and changes can be made without departing from its scope. The scope of the present disclosure is only limited by the appended claims.

Claims

1. A label recommendation method, characterized in that, it includes: Obtain multiple labels under the target category and the associated users of each of the labels; Determine the label with the largest number of associated users among the multiple labels as the reference label, and the other labels as the popularity labels; Add the target label combinations with a confidence level reaching the first threshold between the reference label and each of the popularity labels to the associated label set, where the target label combinations are the popularity labels and the corresponding reference labels; Recommend at least one popularity label in the associated label set to the associated users of the reference label corresponding to the popularity label; Wherein, after obtaining multiple labels under the target category and the associated users of each of the labels, the method further includes: Mark the labels with the number of associated users less than the third threshold in the target category as cold labels, and the labels with the number of associated users reaching the third threshold as hot labels; Recommend the cold labels to the associated users of the hot labels; Remove the cold labels from the target category.

2. The method according to claim 1, characterized in that, after adding the target label combinations with a confidence level reaching the first threshold between the reference label and each of the popularity labels to the associated label set, the method further includes: Remove the reference label from the multiple labels, and determine the popularity label with the highest confidence level between the multiple labels and the removed reference label as the new reference label; Determine the confidence level between the new reference label and the remaining popularity labels, and add the target label combinations with a confidence level reaching the first threshold to the associated label set; Repeat the above steps until only one label remains among the multiple labels.

3. The method according to claim 1, characterized in that, before obtaining multiple labels under the target category and the associated users of each of the labels, the method further includes: Obtain the initial labels and cluster the initial labels to obtain multiple categories; Use any one of the multiple categories as the target category.

4. The method according to claim 3, characterized in that, clustering the initial labels to obtain multiple categories includes: Count the support degrees of the label combinations formed by any N labels in the initial labels, where N is an integer greater than 1; Count the label combinations with the support degree reaching the second threshold, and classify the label combinations with at least one common label into one category to obtain the multiple categories.

5. The method according to claim 3, characterized in that, using any one of the multiple categories as the target category includes: Count the total number of person-times of the associated users of each category, and calculate the average value of the label-associated users of each category; Sort according to the level of the average value of the label-associated users, and use each category as the target category in turn.

6. The method according to claim 1, characterized in that, the method further includes: If the cold label ratio of the target category exceeds the preset range, adjust the third threshold, and classify the labels in the target category into cold labels and hot labels again through the third threshold.

7. The method according to claim 1, characterized in that, Combining and adding the target labels whose confidence level between the reference label and each heat label reaches the first threshold to the associated label set includes: Adding the target label combination whose confidence level reaches the first threshold and the confidence level of the target label combination to the associated label set; Recommending at least one heat label in the associated label set to the associated users of the reference label corresponding to the heat label includes: Sequentially recommending each heat label to the associated users of the reference label corresponding to the heat label in the order from high to low of the confidence levels of each target label combination in the associated label set.

8. A label recommendation device Characterized in that It includes: A label information acquisition module, configured to acquire multiple labels under a target category and the number of associated users of each label; A reference label determination module, configured to determine the label with the largest number of associated users among the multiple labels as the reference label, and other labels as heat labels; A confidence level determination module, configured to add the target label combinations whose confidence level between the reference label and each heat label reaches the first threshold to the associated label set, where the target label combination is a heat label and the corresponding reference label; A heat label recommendation module, configured to recommend at least one heat label in the associated label set to the associated users of the reference label corresponding to the heat label; Wherein, the label information acquisition module is further configured to: after acquiring multiple labels under a target category and the number of associated users of each label, mark the labels with the number of associated users less than the third threshold in the target category as cold labels, and mark the labels with the number of associated users reaching the third threshold as hot labels; recommend the cold labels to the associated users of the hot labels; and remove the cold labels from the target category.

9. An electronic device Characterized in that It includes: A processor; And A memory, configured to store the executable instructions of the processor; Wherein, the processor is configured to execute the method according to any one of claims 1-7 by executing the executable instructions.

10. A computer-readable storage medium, on which a computer program is stored Characterized in that The computer program, when executed by a processor, implements the method according to any one of claims 1-7.

Images