Method and device for pushing target resources

Abstract

The present invention discloses a method and device for pushing target resources, and relates to the field of computer technology. The method comprises: vectorizing the feature data of a first type of target resource to obtain a feature vector of the first type of target resource; calculating the similarity between the feature vector of the first type of target resource and the feature vector of the second type of target resource, and screening out the second type of target resource similar to the first type of target resource according to the similarity calculation result; determining the quality evaluation value of the first type of target resource according to the quality evaluation value of the second type of target resource similar to the first type of target resource; and pushing the target resource according to the quality evaluation value of the first type of target resource. Through the above steps, the effect of resource push can be improved, the occupation of system computing resources by resource push can be reduced, and the user experience of the person being pushed can be improved.

Description

Technical Field

[0001] The present invention relates to the field of computer technology, and in particular to a method and device for pushing target resources. Background Art

[0002] With the development of Internet technology, it is becoming more and more common to push (or place) resources online and offline. For example, e-commerce platforms can push resources such as coupons and advertisements through various channels, and video website platforms can push resources such as videos through various channels. When pushing resources, how to evaluate the quality of resources is a key issue, which often affects the effect of resource push. For example, low accuracy in resource quality assessment will lead to unsatisfactory resource push effects, which not only reduces the user experience of the push recipients, but also causes a waste of manpower and material resources required for resource push.

[0003] In the existing technology, most resource quality assessment models use a single machine learning model and only output an overall score, without breaking it down into specific channel dimensions. Even if the channel dimension is taken into account, the existing technology still requires training different quality assessment models based on different channels.

[0004] In the process of implementing the present invention, the inventors found that there are at least the following problems in the prior art: First, there are huge differences when pushing resources on different channels. If the channel factor is not considered and only an overall score is used, the accuracy of resource quality assessment will be greatly reduced, thereby reducing the effect of subsequent resource push. Second, when pushing resources, if different resource quality assessment models are established according to different channels, it is necessary to train corresponding models for different channels, which not only occupies a large amount of system computing resources and wastes a lot of human and material resources, but also has unsatisfactory quality assessment results, thereby leading to unsatisfactory resource push results. Summary of the invention

[0005] In view of this, the present invention provides a method and device for pushing target resources, which can improve the effect of resource pushing, reduce the occupation of system computing resources by resource pushing, and improve the user experience of the pushed user.

[0006] To achieve the above objective, according to one aspect of the present invention, a method for pushing a target resource is provided.

[0007] The target resource pushing method of the present invention includes: vectorizing the feature data of the first category of target resources to obtain the feature vector of the first category of target resources; wherein the first category of target resources are target resources that have not been pushed; calculating the similarity between the feature vector of the first category of target resources and the feature vector of the second category of target resources, and screening out the second category of target resources similar to the first category of target resources according to the similarity calculation result; wherein the second category of target resources are target resources that have been pushed; determining the quality assessment value of the first category of target resources according to the quality assessment value of the second category of target resources similar to the first category of target resources; and pushing the target resources according to the quality assessment value of the first category of target resources.

[0008] Optionally, the calculating the similarity between the feature vector of the first category target resource and the feature vector of the second category target resource, and screening out the second category target resources similar to the first category target resources according to the similarity calculation result includes: calculating the similarity between the feature vector of the first category target resource and the feature vector of the second category target resource under the target channel; when the maximum value of the similarity under the target channel meets a preset condition, taking the second category target resource corresponding to the maximum value of the similarity as the second category target resource similar to the first category target resource.

[0009] Optionally, the calculating the similarity between the feature vector of the first category target resource and the feature vector of the second category target resource, and screening out the second category target resources similar to the first category target resources according to the similarity calculation result also includes: when the maximum similarity value under the target channel does not meet a preset condition, calculating the similarity between the feature vector of the first category target resource and the second category target resource under all channels, and taking the second category target resource corresponding to the maximum similarity value under all channels as the second category target resource similar to the first category target resource.

[0010] Optionally, determining the quality assessment value of the first category target resources based on the quality assessment value of the second category target resources similar to the first category target resources includes: obtaining a pre-stored actual click rate or actual conversion rate of the second category target resources similar to the first category target resources, and using the actual click rate or actual conversion rate as the quality assessment value of the first category target resources.

[0011] Optionally, pushing target resources according to the quality evaluation values ​​of the first category of target resources includes: filtering out resources to be recommended from the target resources according to the quality evaluation values ​​of the first category of target resources, and pushing the resources to be recommended; or, determining a display order of target resources according to the quality evaluation values ​​of the first category of target resources, and pushing the target resources according to the display order.

[0012] Optionally, the feature data of the first category of target resources includes: item information associated with the first category of target resources; the vectorization processing of the feature data of the first category of target resources to obtain the feature vector of the first category of target resources includes: vectorization processing of the item information associated with the first category of target resources based on a word embedding model to obtain the feature vector of each item information associated with the first category of target resources; fusion processing of the feature vectors of each item information associated with the first category of target resources to obtain the feature vector of the first category of target resources.

[0013] Optionally, the method further includes: after the first-category target resource is pushed, updating its quality assessment value according to its actual click rate or actual conversion rate in the target channel.

[0014] To achieve the above objective, according to another aspect of the present invention, a device for pushing a target resource is provided.

[0015] The push device for target resources of the present invention includes: a vectorization processing module, which is used to perform vectorization processing on the feature data of a first type of target resources to obtain a feature vector of the first type of target resources; wherein the first type of target resources are target resources that have not been pushed; a screening module, which is used to calculate the similarity between the feature vector of the first type of target resources and the feature vector of the second type of target resources, and screen out a second type of target resources similar to the first type of target resources according to the similarity calculation result; wherein the second type of target resources are target resources that have been pushed; a determination module, which is used to determine the quality assessment value of the first type of target resources according to the quality assessment value of the second type of target resources similar to the first type of target resources; and a push module, which is used to push the target resources according to the quality assessment value of the first type of target resources.

[0016] To achieve the above objective, according to another aspect of the present invention, an electronic device is provided.

[0017] The electronic device of the present invention comprises: one or more processors; and a storage device for storing one or more programs; when the one or more programs are executed by the one or more processors, the one or more processors implement the method for pushing target resources of the present invention.

[0018] To achieve the above objective, according to yet another aspect of the present invention, a computer readable medium is provided.

[0019] The computer-readable medium of the present invention stores a computer program, and when the program is executed by a processor, the method for pushing target resources of the present invention is implemented.

[0020] An embodiment of the above invention has the following advantages or beneficial effects: by vectorizing the feature data of the first category of target resources to obtain the feature vector of the first category of target resources, calculating the similarity between the feature vector of the first category of target resources and the feature vector of the second category of target resources, screening out the second category of target resources similar to the first category of target resources according to the similarity calculation result, determining the quality assessment value of the first category of target resources according to the quality assessment value of the second category of target resources similar to the first category of target resources, and pushing the target resources according to the quality assessment value of the first category of target resources. These steps can improve the effect of resource push, reduce the occupation of system computing resources by resource push, and improve the user experience of the person being pushed.

[0021] The further effects of the above-mentioned non-conventional optional manner will be described below in conjunction with the specific implementation manner. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] The accompanying drawings are used to better understand the present invention and do not constitute an improper limitation of the present invention.

[0023] Figure 1 is a schematic diagram of the main flow of the method for pushing target resources according to the first embodiment of the present invention;

[0024] Figure 2 is a schematic diagram of the main flow of a method for pushing target resources according to a second embodiment of the present invention;

[0025] Figure 3 is a schematic diagram of the main flow of a coupon push method according to a third embodiment of the present invention;

[0026] Figure 4 is a schematic diagram of main modules of a device for pushing target resources according to a fourth embodiment of the present invention;

[0027] Figure 5 is an exemplary system architecture diagram to which embodiments of the present invention may be applied;

[0028] Figure 6 It is a schematic diagram of the structure of a computer system suitable for implementing an electronic device of an embodiment of the present invention. DETAILED DESCRIPTION

[0029] The following is a description of exemplary embodiments of the present invention in conjunction with the accompanying drawings, including various details of the embodiments of the present invention to facilitate understanding, which should be considered as merely exemplary. Therefore, it should be recognized by those of ordinary skill in the art that various changes and modifications may be made to the embodiments described herein without departing from the scope and spirit of the present invention. Similarly, for clarity and conciseness, the description of well-known functions and structures is omitted in the following description.

[0030] It should be pointed out that the various embodiments of the present invention and the technical features in the embodiments may be combined with each other without affecting the implementation of the present invention.

[0031] Figure 1 FIG. 1 is a schematic diagram of the main flow of the method for pushing target resources according to the first embodiment of the present invention. Figure 1 As shown, the method for pushing target resources in an embodiment of the present invention includes:

[0032] Step S101: vectorizing feature data of a first type of target resource to obtain a feature vector of the first type of target resource.

[0033] Among them, the first type of target resources are target resources that have not been pushed. In specific implementation, the meaning of the target resources is different in different application scenarios. For example, in the coupon push scenario, the target resource may be a coupon, and accordingly, the first type of target resource may be a coupon that has not been pushed; in the advertisement push scenario, the target resource may be an advertisement; in the article push scenario, the target resource may be an article; in the short video push scenario, the target resource may be a short video.

[0034] Exemplarily, in step S101, the feature data of the first type of target resources may be vectorized based on the neural network model to obtain a feature vector of the first type of target resources. In specific implementations, the feature data of the target resources may also be different in different application scenarios. For example, in a coupon push scenario, the item information to which the coupon is applicable may be used as the feature data of the target resource; in an article push scenario, the corpus information of the article may be used as the feature data of the target resource.

[0035] Step S102: Calculate the similarity between the feature vector of the first-category target resource and the feature vector of the second-category target resource, and select the second-category target resource similar to the first-category target resource according to the similarity calculation result.

[0036] Among them, the second type of target resources are target resources that have been pushed. For example, in the coupon push scenario, the first type of target resources can be new, unpushed coupons, and the second type of target resources can be existing, pushed coupons; in the advertisement push scenario, the first type of target resources can be new, unpushed advertisements, and the second type of target resources can be existing, pushed advertisements; in the article push scenario, the first type of target resources can be new, unpushed articles, and the second type of target resources can be existing, pushed articles; in the short video push scenario, the first type of target resources can be new, unpushed short videos, and the second type of target resources can be existing, pushed short videos.

[0037] In an optional example, the cosine similarity calculation formula may be used to calculate the similarity between the feature vector of the first type of target resource and the feature vector of the second type of target resource:

[0038]

[0039] Among them, y represents cosine similarity, A i The i-th element in the feature vector representing the first type of target resource, B i The i-th element in the feature vector representing the second type of target resource.

[0040] It should be noted that, without affecting the implementation of the present invention, other methods may be used to calculate the similarity between the feature vector of the first type of target resource and the feature vector of the second type of target resource. For example, the Euclidean distance between the two vectors may be used to represent the similarity of the two vectors, and then the similarity between the feature vector of the first type of target resource and the feature vector of the second type of target resource may be calculated based on the Euclidean distance formula.

[0041] In an optional example, after calculating the similarity between the feature vector of the first-category target resource and the feature vectors of each second-category target resource under the target channel, the feature vector of the second-category target resource corresponding to the maximum similarity can be used as the second-category target resource similar to the first-category target resource. For example, assuming that the feature vector of the first-category target resource (such as target resource A) has the greatest similarity with the feature vector of the second-category target resource (such as target resource F), then target resource F is used as the second-category target resource similar to target resource A.

[0042] Step S103: Determine the quality evaluation value of the first-category target resource according to the quality evaluation value of the second-category target resource that is similar to the first-category target resource.

[0043] In an optional example, the quality evaluation values ​​of each second-category target resource may be pre-stored, for example, in a database table. Furthermore, in this step, the database table may be queried according to the identifier of the second-category target resource similar to the first-category target resource to obtain the quality evaluation value of the second-category target resource similar to the first-category target resource. Afterwards, the obtained quality evaluation value of the second-category target resource similar to the first-category target resource may be used as the quality evaluation value of the first-category target resource.

[0044] The quality evaluation value may be a click-through rate, conversion rate, or other quality characterization index of the target resource. For example, when the quality evaluation value is the click-through rate of the target resource, in step S103, the actual click-through rate of the second type of target resource similar to the first type of target resource that is pre-stored may be obtained, and the actual click-through rate may be used as the quality evaluation value of the first type of target resource.

[0045] In another optional example, the quality evaluation value of the second type of target resources similar to the first type of target resources may also be calculated in real time, and the quality evaluation value may be used as the quality evaluation value of the first type of target resources.

[0046] Step S104: Pushing target resources according to the quality evaluation value of the first-category target resources.

[0047] In an optional example, step S104 includes: selecting resources to be recommended from the target resources according to the quality evaluation values ​​of the first-category target resources, and pushing the resources to be recommended. Specifically, in this optional example, for a set consisting of multiple first-category target resources, or a set consisting of some first-category target resources and some second-category target resources, the first M target resources with the largest quality evaluation values ​​in the set, or target resources with quality evaluation values ​​greater than a preset threshold value, can be used as resources to be recommended, and the resources to be recommended can be pushed.

[0048] In another optional example, step S104 includes: determining the display order of the target resources according to the quality evaluation value of the first-category target resources, and pushing the target resources according to the display order. Specifically, in this optional example, for a set consisting of multiple first-category target resources, or a set consisting of some first-category target resources and some second-category target resources, the target resources in the set can be sorted according to the quality evaluation value, and the sorting result is used as the display order of the target resources, and then the target resources are displayed according to the display order when pushing.

[0049] In the embodiment of the present invention, the above steps overcome the defects of using regression models to evaluate resource quality in the prior art, improve the effect of resource push, reduce the occupation of system computing resources by resource push, and improve the user experience of the push recipient.

[0050] Figure 2 FIG. 2 is a schematic diagram of the main flow of the method for pushing target resources according to the second embodiment of the present invention. Figure 2 As shown, the method for pushing target resources in an embodiment of the present invention includes:

[0051] Step S201: performing vectorization processing on feature data of a first type of target resources to obtain feature vectors of the first type of target resources.

[0052] Among them, the first type of target resources are target resources that have not been pushed. In specific implementation, the meaning of the target resources is different in different application scenarios. For example, in the coupon push scenario, the target resource may be a coupon, and accordingly, the first type of target resource may be a coupon that has not been pushed; in the advertisement push scenario, the target resource may be an advertisement; in the article push scenario, the target resource may be an article; in the short video push scenario, the target resource may be a short video.

[0053] Exemplarily, when the feature data of the first category of target resources includes item information associated with the first category of target resources, step S201 specifically includes: vectorizing the item information associated with the first category of target resources based on a word embedding model to obtain feature vectors of each item information associated with the first category of target resources; and fusing the feature vectors of each item information associated with the first category of target resources to obtain the feature vector of the first category of target resources.

[0054] The word embedding model may be a word2vec algorithm model or other models that can map a word or phrase into a vector in a real number field.

[0055] Among them, the fusion processing of the feature vectors of each item information associated with the first type of target resources (wherein the "feature vector of item information" can also be referred to as "item feature vector") can be: summing or averaging the elements of the corresponding dimensions in each item feature vector.

[0056] In specific implementation, the characteristic data of the target resource may be different in different application scenarios. For example, in the coupon push scenario, the item information to which the coupon is applicable may be used as the characteristic data of the target resource; in the article push scenario, the corpus information of the article may be used as the characteristic data of the target resource.

[0057] Step S202: Calculate the similarity between the feature vector of the first-category target resource and the feature vector of the second-category target resource under the target channel.

[0058] When pushing or placing target resources, various push channels are involved. For example, when the target resource is a coupon, its push channels include the "payment settlement page channel", "clear shopping cart channel", "member channel" channel on the e-commerce platform, and other channels that can be used for coupon push, such as live video broadcasting.

[0059] In an embodiment of the present invention, for various push channels, a data pool under the corresponding channel is constructed to store information about target resources that have been pushed under the channel, that is, information about second-category target resources under the channel. For example, the data pool constructed under the corresponding channel includes: a data pool corresponding to channel 1, a data pool corresponding to channel 2, and a data pool corresponding to channel 3. In addition, in an embodiment of the present invention, a data pool under all channels is also constructed to store information about target resources that have been pushed under all channels, that is, information about second-category target resources under all channels. Among them, the information about second-category target resources stored in the data pool may include: feature data of second-category target resources, and / or feature vectors of second-category target resources.

[0060] In an embodiment of the present invention, in order to reduce the amount of calculation, the target channel of the first type of target data can be determined first, and then the similarity between the feature vector of the first type of target resource and the feature vector of each second type of target resource in the data pool under the target channel is calculated.

[0061] Step S203: Determine whether the maximum similarity value under the target channel meets a preset condition.

[0062] Exemplarily, the preset condition may be: the maximum similarity value is greater than or equal to a preset threshold. The preset threshold may be flexibly set according to the application scenario. For example, the preset threshold may be set to 0.6 or other values. If the judgment result of step S203 is yes, step S204 is executed; if the judgment result of step S203 is no, step S205 is executed.

[0063] Step S204: taking the second-category target resource corresponding to the maximum similarity value as the second-category target resource similar to the first-category target resource.

[0064] Further, after step S204, step S206 may be executed.

[0065] Step S205: Calculate the similarity between the feature vector of the first-category target resource and the second-category target resource under all channels, and use the second-category target resource corresponding to the maximum similarity under all channels as the second-category target resource similar to the first-category target resource.

[0066] Further, after step S205, step S206 may be executed.

[0067] Step S206: Determine the quality assessment value of the first-category target resource according to the quality assessment value of the second-category target resource that is similar to the first-category target resource.

[0068] The quality assessment value may be a click-through rate, a conversion rate (such as a purchase conversion rate), or other quality characterization indicators of the target resource. For example, when the quality assessment value is the click-through rate of the target resource, in step S206, the actual click-through rate of the second type of target resource similar to the first type of target resource stored in advance may be obtained, and the actual click-through rate may be used as the quality assessment value of the first type of target resource. For example, when the quality assessment value is the conversion rate of the target resource, in step S206, the actual conversion rate of the second type of target resource similar to the first type of target resource stored in advance may be obtained, and the actual conversion rate may be used as the quality assessment value of the first type of target resource.

[0069] Step S207: Pushing target resources according to the quality evaluation value of the first type of target resources.

[0070] In an optional example, step S207 includes: selecting resources to be recommended from the target resources according to the quality evaluation values ​​of the first-category target resources, and pushing the resources to be recommended. Specifically, in this optional example, for a set consisting of multiple first-category target resources, or a set consisting of some first-category target resources and some second-category target resources, the first M target resources with the largest quality evaluation values ​​in the set, or target resources with quality evaluation values ​​greater than a preset threshold value, can be used as resources to be recommended, and the resources to be recommended can be pushed to the user.

[0071] In another optional example, step S207 includes: determining the display order of the target resources according to the quality evaluation value of the first-category target resources, and pushing the target resources according to the display order. Specifically, in this optional example, for a set consisting of multiple first-category target resources, or a set consisting of some first-category target resources and some second-category target resources, the target resources in the set can be sorted according to the quality evaluation value, and the sorting result is used as the display order of the target resources, and then the target resources are displayed according to the display order when pushing to the user.

[0072] Furthermore, the method of the embodiment of the present invention may further include the following steps: after the first type of target resource is pushed, its quality evaluation value is updated according to its actual click rate or actual conversion rate in the target channel. Through this step, the quality evaluation value can accurately reflect the current quality of the target resource, thereby helping to improve the push effect of the target resource.

[0073] In the embodiment of the present invention, the above steps overcome the defects of low accuracy and consumption of system computing resources in the prior art of using regression models to evaluate resource quality, as well as the resulting defects of poor resource push effect and low user experience, thereby improving the effect of resource push, reducing the occupation of system computing resources by resource push, and improving the user experience of the person being pushed.

[0074] Figure 3 FIG. 2 is a schematic diagram of the main flow of the coupon push method according to the third embodiment of the present invention. Figure 3 As shown, the coupon push method of the embodiment of the present invention includes:

[0075] Step S301: vectorize the items browsed by the user based on the word2vec algorithm to obtain the feature vector corresponding to each item.

[0076] Exemplarily, in this step, a sequence of items browsed by users on the website within a period of time (such as the past year, or the past six months, etc.) can be selected, and each item in the sequence can be vectorized based on the word2vec algorithm (a word embedding model) to obtain a feature vector corresponding to each item.

[0077] Step S302: performing fusion processing on the feature vectors corresponding to the items to which the new coupon is applicable, so as to obtain the feature vector of the new coupon.

[0078] The fusion process may be: summing or averaging the elements of the corresponding dimensions in the feature vectors of the items to which the new coupon applies. For example, assuming that the new coupon is applicable to item 1, item 2, and item 3, and the feature vector of item 1 is represented by sku 1 =[x 11 ,x 12 ,…,x 1j ], the feature vector of item 2 is represented by sku 2 =[x 21 ,x 22 ,…,x 2j ], the feature vector of item 3 is represented by sku 3 =[x 31 ,x 32 ,…,x 3j ], the three feature vectors can be averaged to obtain the feature vector of the new coupon, which can be expressed as coupon = [avg (x 11 +x 21 +x 31 ),avg(x 12 +x 22 +x 32 ),…avg(x1j +x 2j +x 3j )].

[0079] Furthermore, before the feature vectors corresponding to the items to which the new coupon applies are merged, the items to which the new coupon applies can be filtered according to the sales volume of the items, and the items with high sales volume can be retained. For example, the top Q items with the highest sales volume among all the items to which the new coupon applies can be retained, and then the feature vectors of the top Q items can be merged to obtain the feature vector of the new coupon.

[0080] Step S303: Calculate the similarity between the feature vector of the new coupon and the feature vector of the existing coupon under the target channel.

[0081] When pushing or placing coupons, various push channels are involved. For example, the push channels for coupons may include the "payment settlement page channel", "clear shopping cart channel", "member channel" channel on the e-commerce platform, and other channels that can be used for coupon push such as live video broadcasting.

[0082] In an embodiment of the present invention, for various push channels, a data pool under the corresponding channel is constructed to store the information of existing and pushed coupons under the channel, that is, the information of the second-category coupons under the channel. For example, the data pool constructed under the corresponding channel includes: a data pool corresponding to channel 1, a data pool corresponding to channel 2, and a data pool corresponding to channel 3. In addition, in an embodiment of the present invention, a data pool under all channels is also constructed to store the information of coupons that have been pushed under all channels, that is, the information of the second-category coupons under all channels. Among them, the information of the second-category coupons stored in the data pool may include: feature data of the second-category coupons, and / or feature vectors of the second-category coupons.

[0083] In the embodiment of the present invention, in order to reduce the amount of calculation, the target channel of the new coupon may be determined first, and then the similarity between the feature vector of the new coupon and the feature vectors of each existing coupon in the data pool under the target channel may be calculated.

[0084] Step S304: Determine whether the maximum similarity value is greater than a preset threshold.

[0085] The preset threshold value can be flexibly set according to the application scenario. For example, the preset threshold value can be set to 0.6 or other values. If the judgment result of step S304 is yes, step S305 is executed; if the judgment result of step S304 is no, step S306 and step S307 are executed.

[0086] Step S305: Determine the quality evaluation value of the new coupon according to the quality evaluation value of the existing coupon corresponding to the maximum similarity value under the target channel.

[0087] In an optional example, the quality evaluation values ​​of each existing coupon may be stored in advance, for example, in a database table. Furthermore, in this step, the database table may be queried according to the identifier of the existing coupon corresponding to the maximum similarity under the target channel to obtain the quality evaluation value of the existing coupon that is most similar to the new coupon. Afterwards, the quality evaluation value of the existing coupon that is most similar to the new coupon may be used as the quality evaluation value of the first type of target resource.

[0088] The quality evaluation value may be the click-through rate, conversion rate, or other quality characterization indicators of the target resource. For example, when the quality evaluation value is the click-through rate of the target resource, in step S305, the actual click-through rate of the existing coupon corresponding to the maximum similarity value under the pre-stored target channel may be obtained, and the actual click-through rate may be used as the quality evaluation value of the new coupon.

[0089] In another optional example, the quality evaluation value of the existing coupon most similar to the new coupon under the target channel may also be calculated in real time, and the quality evaluation value may be used as the quality evaluation value of the new coupon.

[0090] Further, after step S305, step S308 may be executed.

[0091] Step S306: Calculate the similarity between the feature vector of the new coupon and the feature vector of the existing coupons under all channels.

[0092] Step S307: Determine the quality evaluation value of the new coupon according to the quality evaluation value of the existing coupon corresponding to the maximum similarity value under all channels.

[0093] Further, after step S307, step S308 may be executed.

[0094] Step S308: Push the new coupon according to the quality evaluation value of the new coupon.

[0095] In an optional example, coupons to be recommended may be screened out from the coupons according to the quality evaluation value of the new coupons, and the coupons to be recommended may be pushed. Specifically, in this optional example, for a set consisting of multiple new coupons, or a set consisting of some new coupons and some existing coupons, the first M coupons with the largest quality evaluation values ​​in the set, or coupons with quality evaluation values ​​greater than a preset threshold value, may be used as coupons to be recommended, and the coupons to be recommended may be pushed to the user. In addition, when pushing the coupons to be recommended, the display order of the coupons to be recommended may also be determined according to the quality evaluation value, and the coupons may be pushed according to the display order.

[0096] In another optional example, the display order of the coupons may be determined according to the quality evaluation value of the new coupons, and the coupons may be pushed according to the display order. Specifically, in this optional example, for a set consisting of multiple new coupons, or a set consisting of some new coupons and some existing coupons, the coupons in the set may be sorted according to the quality evaluation value, and the result of the sorting may be used as the display order of the coupons, and then the coupons may be displayed according to the display order when pushed to the user. For example, the coupons may be sorted in descending order of the quality evaluation value, and the result of the sorting may be used as the display order of the coupons.

[0097] Furthermore, the method of the embodiment of the present invention may further include the following steps: after the new coupon is pushed, its quality evaluation value is updated according to its actual click rate or actual conversion rate in the target channel.

[0098] Compared with the prior art, the coupon push method provided by the embodiment of the present invention has at least the following advantages:

[0099] 1. Different from the existing method of using a single machine learning model to evaluate the quality of target resources, in the embodiment of the present invention, the feature data of the new coupon is vectorized by the word2vec algorithm for processing text in the NLP field, and the feature vector of the new coupon is obtained by fusion processing. Then, the quality evaluation value of the new coupon is determined by calculating the similarity between the feature vectors of the new coupon and the existing coupon. This helps to overcome the defects of low accuracy and consumption of system computing resources in the prior art of using regression models to evaluate the quality of coupons, as well as the defects of poor coupon push effect and low user experience caused by the above, improves the effect of coupon push, reduces the occupation of system computing resources by coupon push, and improves the user experience of the pushed person.

[0100] 2. In the embodiment of the invention, the first Q items with the highest sales volume are retained to represent the coupons, and the method of generating sentence vectors using word vectors in the NLP field is used to average the item vectors corresponding to each coupon, and finally obtain the feature vector value of the coupon. This not only solves the problem of the source of the coupon feature vector, but also greatly improves the representativeness and value accuracy of the coupon feature vector.

[0101] 3. By dividing the single-channel coupon pool and the omni-channel coupon pool (or "full coupon pool") and combining the coupon similarity model, the modeling problem of the channel dimension coupon quality evaluation model is solved. In addition, only one model needs to be developed to evaluate the quality of coupons under each channel, without the need to develop different quality evaluation models for different channels.

[0102] 4. After a new coupon is released, its quality assessment value is regularly updated with the actual conversion rate (or "real conversion rate") or actual click-through rate (or "real click-through rate") of the coupon in the target channel, so that the quality assessment value can accurately reflect the current quality of the coupon, thereby helping to improve the push effect of the coupon.

[0103] Figure 4 FIG. 4 is a schematic diagram of the main modules of a device for pushing target resources according to a fourth embodiment of the present invention. Figure 4 As shown, the target resource pushing device 400 according to the embodiment of the present invention includes: a vectorized processing module 401 , a screening module 402 , a determination module 403 , and a pushing module 404 .

[0104] The vectorization processing module 401 is used to perform vectorization processing on the feature data of the first type of target resources to obtain the feature vector of the first type of target resources.

[0105] Among them, the first type of target resources are target resources that have not been pushed. In specific implementation, the meaning of the target resources is different in different application scenarios. For example, in the coupon push scenario, the target resource may be a coupon, and accordingly, the first type of target resource may be a coupon that has not been pushed; in the advertisement push scenario, the target resource may be an advertisement; in the article push scenario, the target resource may be an article; in the short video push scenario, the target resource may be a short video.

[0106] Exemplarily, the vectorization processing module 401 can perform vectorization processing on the feature data of the first type of target resources based on the neural network model to obtain the feature vector of the first type of target resources. In specific implementation, the feature data of the target resources may also be different in different application scenarios. For example, in the coupon push scenario, the item information to which the coupon is applicable can be used as the feature data of the target resource; in the article push scenario, the corpus information of the article can be used as the feature data of the target resource.

[0107] The screening module 402 is used to calculate the similarity between the feature vector of the first-category target resource and the feature vector of the second-category target resource, and screen out the second-category target resource similar to the first-category target resource according to the similarity calculation result.

[0108] Among them, the second type of target resources are target resources that have been pushed. For example, in the coupon push scenario, the first type of target resources can be new, unpushed coupons, and the second type of target resources can be existing, pushed coupons; in the advertisement push scenario, the first type of target resources can be new, unpushed advertisements, and the second type of target resources can be existing, pushed advertisements; in the article push scenario, the first type of target resources can be new, unpushed articles, and the second type of target resources can be existing, pushed articles; in the short video push scenario, the first type of target resources can be new, unpushed short videos, and the second type of target resources can be existing, pushed short videos.

[0109] In an optional example, the screening module 402 may use a cosine similarity calculation formula to calculate the similarity between the feature vector of the first type of target resource and the feature vector of the second type of target resource:

[0110]

[0111] Among them, y represents cosine similarity, A i The i-th element in the feature vector representing the first type of target resource, B i The i-th element in the feature vector representing the second type of target resource.

[0112] It should be noted that, without affecting the implementation of the present invention, the screening module 402 may also use other methods to calculate the similarity between the feature vector of the first type of target resource and the feature vector of the second type of target resource. For example, the Euclidean distance between the two vectors may be used to represent the similarity of the two vectors, and then the similarity between the feature vector of the first type of target resource and the feature vector of the second type of target resource may be calculated based on the Euclidean distance formula.

[0113] In an optional example, after calculating the similarity between the feature vector of the first-category target resource and the feature vectors of each second-category target resource under the target channel, the screening module 402 may use the feature vector of the second-category target resource corresponding to the maximum similarity as the second-category target resource similar to the first-category target resource. For example, assuming that the feature vector of the first-category target resource (such as target resource A) has the greatest similarity with the feature vector of the second-category target resource (such as target resource F), target resource F is regarded as the second-category target resource similar to target resource A.

[0114] The determination module 403 is used to determine the quality evaluation value of the first-category target resource according to the quality evaluation value of the second-category target resource similar to the first-category target resource.

[0115] In an optional example, the quality evaluation values ​​of each second-category target resource may be pre-stored, for example, they may be stored in a database table. Further, the determination module 403 may query the database table according to the identifier of the second-category target resource similar to the first-category target resource to obtain the quality evaluation value of the second-category target resource similar to the first-category target resource. Afterwards, the determination module 403 may use the obtained quality evaluation value of the second-category target resource similar to the first-category target resource as the quality evaluation value of the first-category target resource.

[0116] The quality evaluation value may be a click-through rate, conversion rate, or other quality characterization index of the target resource. For example, when the quality evaluation value is the click-through rate of the target resource, the determination module 403 may obtain the actual click-through rate of the second type of target resource that is similar to the first type of target resource and store it in advance, and use the actual click-through rate as the quality evaluation value of the first type of target resource.

[0117] In another optional example, the determination module 403 may also calculate in real time the quality evaluation value of the second type of target resources that are similar to the first type of target resources, and use the quality evaluation value as the quality evaluation value of the first type of target resources.

[0118] The push module 404 is configured to push target resources according to the quality evaluation value of the first type of target resources.

[0119] In an optional example, the push module 404 may filter out the resources to be recommended from the target resources according to the quality evaluation values ​​of the first-category target resources, and push the resources to be recommended. Specifically, in this optional example, for a set consisting of multiple first-category target resources, or a set consisting of some first-category target resources and some second-category target resources, the first M target resources with the largest quality evaluation values ​​in the set, or target resources with quality evaluation values ​​greater than a preset threshold value, may be selected as resources to be recommended, and the resources to be recommended may be pushed.

[0120] In another optional example, the push module 404 may determine the display order of the target resources according to the quality evaluation value of the first-category target resources, and push the target resources according to the display order. Specifically, in this optional example, for a set consisting of multiple first-category target resources, or a set consisting of some first-category target resources and some second-category target resources, the target resources in the set may be sorted according to the quality evaluation value, and the sorting result is used as the display order of the target resources, and then the target resources are displayed according to the display order when pushing.

[0121] In the embodiment of the present invention, the above device overcomes the defect of using regression model to evaluate resource quality in the prior art, improves the effect of resource push, reduces the occupation of system computing resources by resource push, and improves the user experience of the push recipient.

[0122] Figure 5 An exemplary system architecture 500 is shown to which a method for pushing a target resource or an apparatus for pushing a target resource according to an embodiment of the present invention can be applied.

[0123] like Figure 5 As shown, system architecture 500 may include terminal devices 501, 502, 503, a network 504 and a server 505. Network 504 is used to provide a medium for communication links between terminal devices 501, 502, 503 and server 505. Network 504 may include various connection types, such as wired, wireless communication links or optical fiber cables, etc.

[0124] Users can use terminal devices 501, 502, 503 to interact with server 505 through network 504 to receive or send messages, etc. Various communication client applications can be installed on terminal devices 501, 502, 503, such as shopping applications, web browser applications, search applications, instant messaging tools, email clients, social platform software, etc.

[0125] The terminal devices 501 , 502 , and 503 may be various electronic devices having a display screen and supporting web browsing, including but not limited to smart phones, tablet computers, laptop computers, desktop computers, and the like.

[0126] Server 505 may be a server that provides various services, such as a background management server that provides support for shopping websites browsed by users using terminal devices 501, 502, and 503. The background management server may perform quality assessment on target resources and push target resources to the terminal device according to the quality assessment value.

[0127] It should be noted that the method for pushing target resources provided in the embodiment of the present invention may be executed by the server 505 , and accordingly, the device for pushing target resources is disposed in the server 505 .

[0128] It should be understood that Figure 5 The number of terminal devices, networks and servers in the embodiment is only for illustration. Any number of terminal devices, networks and servers may be provided according to implementation requirements.

[0129] Reference below Figure 6 , which shows a schematic diagram of the structure of a computer system 600 of an electronic device suitable for implementing an embodiment of the present invention. Figure 6 The computer system shown is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present invention.

[0130] like Figure 6 As shown, the computer system 600 includes a central processing unit (CPU) 601, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 602 or a program loaded from a storage part 608 into a random access memory (RAM) 603. In the RAM 603, various programs and data required for the operation of the system 600 are also stored. The CPU 601, the ROM 602, and the RAM 603 are connected to each other via a bus 604. An input / output (I / O) interface 605 is also connected to the bus 604.

[0131] The following components are connected to the I / O interface 605: an input section 606 including a keyboard, a mouse, etc.; an output section 607 including a cathode ray tube (CRT), a liquid crystal display (LCD), etc., and a speaker, etc.; a storage section 608 including a hard disk, etc.; and a communication section 609 including a network interface card such as a LAN card, a modem, etc. The communication section 609 performs communication processing via a network such as the Internet. A drive 610 is also connected to the I / O interface 605 as needed. A removable medium 611, such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, etc., is installed on the drive 610 as needed, so that a computer program read therefrom is installed into the storage section 608 as needed.

[0132] In particular, according to the embodiments disclosed in the present invention, the process described above with reference to the flowchart can be implemented as a computer software program. For example, the embodiments disclosed in the present invention include a computer program product, which includes a computer program carried on a computer-readable medium, and the computer program includes a program code for executing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from the network through the communication part 609, and / or installed from the removable medium 611. When the computer program is executed by the central processing unit (CPU) 601, the above-mentioned functions defined in the system of the present invention are executed.

[0133] It should be noted that the computer-readable medium shown in the present invention may be a computer-readable signal medium or a computer-readable storage medium or any combination of the above two. The computer-readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or device, or any combination of the above. More specific examples of computer-readable storage media may include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In the present invention, a computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in combination with an instruction execution system, device or device. In the present invention, a computer-readable signal medium may include a data signal propagated in a baseband or as part of a carrier wave, which carries a computer-readable program code. This propagated data signal may take a variety of forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above. Computer-readable signal media may also be any computer-readable medium other than computer-readable storage media, which may send, propagate or transmit a program for use by or in conjunction with an instruction execution system, apparatus or device. The program code contained on the computer-readable medium may be transmitted using any appropriate medium, including but not limited to: wireless, wire, optical cable, RF, etc., or any suitable combination of the above.

[0134] The flow chart and block diagram in the accompanying drawings illustrate the possible architecture, function and operation of the system, method and computer program product according to various embodiments of the present invention. In this regard, each box in the flow chart or block diagram can represent a module, a program segment, or a part of a code, and the above-mentioned module, program segment, or a part of a code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in a different order from the order marked in the accompanying drawings. For example, two boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram or flow chart, and the combination of the boxes in the block diagram or flow chart can be implemented with a dedicated hardware-based system that performs a specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0135] The modules involved in the embodiments of the present invention may be implemented by software or hardware. The modules described may also be set in a processor. For example, they may be described as: a processor includes a vectorization processing module, a screening module, a determination module, and a push module. The names of these modules do not, in some cases, constitute limitations on the modules themselves. For example, the vectorization processing module may also be described as a "module for performing vectorization processing on the characteristic data of the first type of target resources."

[0136] As another aspect, the present invention also provides a computer-readable medium, which may be included in the device described in the above embodiment; or it may exist independently and not be assembled into the device. The above computer-readable medium carries one or more programs, and when the above one or more programs are executed by a device, the device includes: vectorizing the feature data of the first type of target resources to obtain the feature vector of the first type of target resources; wherein the first type of target resources are target resources that have not been pushed; calculating the similarity between the feature vector of the first type of target resources and the feature vector of the second type of target resources, and filtering out the second type of target resources similar to the first type of target resources according to the similarity calculation result; wherein the second type of target resources are target resources that have been pushed; determining the quality evaluation value of the first type of target resources according to the quality evaluation value of the second type of target resources similar to the first type of target resources; and pushing the first type of target resources according to the quality evaluation value of the first type of target resources.

[0137] According to the technical solution of the embodiment of the present invention, the effect of resource push can be improved, the occupation of system computing resources by resource push can be reduced, and the user experience of the person being pushed can be improved.

[0138] The above specific implementations do not constitute a limitation on the protection scope of the present invention. It should be understood by those skilled in the art that various modifications, combinations, sub-combinations and substitutions may occur depending on design requirements and other factors. Any modification, equivalent substitution and improvement made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A method for pushing a target resource, characterized in that: The method comprises: Performing vectorization processing on the feature data of the first type of target resources to obtain feature vectors of the first type of target resources; wherein the first type of target resources are target resources that have not been pushed; Calculate the similarity between the feature vector of the first-category target resource and the feature vector of the second-category target resource, and select the second-category target resource similar to the first-category target resource according to the similarity calculation result; wherein the second-category target resource is the target resource that has been pushed; specifically comprising: calculating the similarity between the feature vector of the first-category target resource and the feature vector of the second-category target resource under the target channel; when the maximum value of the similarity under the target channel does not meet the preset condition, calculate the similarity between the feature vector of the first-category target resource and the second-category target resource under all channels, and use the second-category target resource corresponding to the maximum value of the similarity under all channels as the second-category target resource similar to the first-category target resource; Determining the quality assessment value of the first type of target resource according to the quality assessment value of the second type of target resource similar to the first type of target resource; The target resources are pushed according to the quality evaluation values ​​of the first type of target resources.

2. The method according to claim 1, characterized in that The calculating the similarity between the feature vector of the first-category target resource and the feature vector of the second-category target resource, and selecting the second-category target resource similar to the first-category target resource according to the similarity calculation result, further includes: When the maximum similarity value under the target channel meets a preset condition, the second-category target resource corresponding to the maximum similarity value is used as the second-category target resource similar to the first-category target resource.

3. The method according to claim 1, characterized in that Determining the quality evaluation value of the first-category target resource according to the quality evaluation value of the second-category target resource similar to the first-category target resource comprises: The actual click rate or actual conversion rate of the pre-stored second-category target resource similar to the first-category target resource is obtained, and the actual click rate or actual conversion rate is used as the quality evaluation value of the first-category target resource.

4. The method according to claim 1, characterized in that: The pushing of the target resource according to the quality evaluation value of the first type of target resource includes: Selecting resources to be recommended from the target resources according to the quality evaluation values ​​of the first type of target resources, and pushing the resources to be recommended; or The display order of the target resources is determined according to the quality evaluation values ​​of the first-category target resources, and the target resources are pushed according to the display order.

5. The method according to any one of claims 1 to 4, characterized in that: The feature data of the first type of target resources includes: item information associated with the first type of target resources; the vectorization processing of the feature data of the first type of target resources to obtain the feature vector of the first type of target resources includes: Based on the word embedding model, the item information associated with the first category of target resources is vectorized to obtain a feature vector of each item information associated with the first category of target resources; the feature vectors of each item information associated with the first category of target resources are fused to obtain a feature vector of the first category of target resources.

6. The method according to claim 1, characterized in that The method further comprises: After the first-category target resource is pushed, its quality evaluation value is updated according to its actual click rate or actual conversion rate in the target channel.

7. A device for pushing target resources, characterized in that: The device comprises: A vectorization processing module, used for performing vectorization processing on feature data of a first type of target resource to obtain a feature vector of the first type of target resource; wherein the first type of target resource is a target resource that has not been pushed; A screening module is used to calculate the similarity between the feature vector of the first-category target resource and the feature vector of the second-category target resource, and screen out the second-category target resource similar to the first-category target resource according to the similarity calculation result; wherein the second-category target resource is a target resource that has been pushed; and is specifically used to: calculate the similarity between the feature vector of the first-category target resource and the feature vector of the second-category target resource under the target channel; when the maximum value of the similarity under the target channel does not meet the preset condition, calculate the similarity between the feature vector of the first-category target resource and the second-category target resource under all channels, and use the second-category target resource corresponding to the maximum value of the similarity under all channels as the second-category target resource similar to the first-category target resource; A determination module, configured to determine the quality assessment value of the first type of target resource according to the quality assessment value of the second type of target resource similar to the first type of target resource; A push module is used to push target resources according to the quality evaluation value of the first type of target resources.

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

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

Images