Advertisement display method, device, electronic device and readable storage medium

Abstract

The embodiments of the present disclosure provide an advertisement display method, device, electronic device and readable storage medium, the method comprising: determining a threshold value of an advertisement position corresponding to a user request according to a user request and an advertisement recall list corresponding to the user request, estimating the click rate of each advertisement in the advertisement recall list, obtaining an estimated click rate of each advertisement, taking advertisements with estimated click rates exceeding the threshold value as candidate advertisements, and sorting and displaying the candidate advertisements. In the embodiments of the present disclosure, since the threshold value of different advertisement positions is determined for each user, and each advertisement is screened according to the threshold value and the click rate of each advertisement, the problem of the same designated advertisement access threshold in the prior art leading to the delivery of high-priced, low-click rate advertisements, which affects the user experience, is avoided, and the accuracy of advertisement display is improved at the user granularity.

Description

Technical Field

[0001] Embodiments of the present disclosure relate to the field of network technology, and in particular, to an advertisement display method, device, electronic device, and readable storage medium. Background Art

[0002] With the rapid development of Internet technology, the proportion of advertisers choosing to place advertisements on the Internet has gradually increased. In order to improve user experience and merchant revenue, methods based on ad click-through rate estimation are currently widely used to rank advertisements.

[0003] Specifically, the click-through rate of the candidate ads is first estimated through a logistic regression algorithm, and then the quality (Quality) of the candidate ads is calculated based on the estimated click-through rate. Finally, the candidate ads are sorted and displayed in reverse order according to Bid×Quality, where Bid is the bid of the ad. The higher the bid and the higher the quality, the higher the ranking of the ad.

[0004] However, based on the current advertising ranking method, for advertisements with particularly high bids, even if the quality is not high (low click-through rate), they can still be ranked at a higher position, resulting in the inability to filter out advertisements of poor quality, which in turn affects the user experience. Summary of the invention

[0005] Embodiments of the present disclosure provide an advertisement display method, device, electronic device, and readable storage medium to improve the accuracy of advertisement display.

[0006] According to a first aspect of an embodiment of the present disclosure, there is provided an advertisement display method, the method comprising:

[0007] Determining a threshold value of an advertisement slot corresponding to the user request according to the user request and an advertisement recall list corresponding to the user request;

[0008] Estimating the click rate of each advertisement in the advertisement recall list to obtain an estimated click rate of each advertisement;

[0009] Advertisements with estimated click-through rates exceeding the threshold value are taken as candidate advertisements, and the candidate advertisements are ranked and displayed.

[0010] According to a second aspect of an embodiment of the present disclosure, there is provided an advertisement display device, the device comprising:

[0011] A threshold determination module, configured to determine a threshold value of an advertisement slot corresponding to a user request according to a user request and an advertisement recall list corresponding to the user request;

[0012] An advertisement screening module, used to estimate the click rate of each advertisement in the advertisement recall list to obtain an estimated click rate of each advertisement;

[0013] The sorting and displaying module is used to take the advertisements whose estimated click-through rates obtained by the advertisement screening module exceed the threshold value as candidate advertisements, and to sort and display the candidate advertisements.

[0014] According to a third aspect of an embodiment of the present disclosure, there is provided an electronic device, including:

[0015] A processor, a memory, and a computer program stored in the memory and executable on the processor, wherein the processor implements the aforementioned advertisement display method when executing the program.

[0016] According to a fourth aspect of an embodiment of the present disclosure, a readable storage medium is provided. When instructions in the storage medium are executed by a processor of an electronic device, the electronic device is enabled to execute the aforementioned advertisement display method.

[0017] The embodiments of the present disclosure provide an advertisement display method, device, electronic device and readable storage medium, the method comprising: determining a threshold value of an advertisement position corresponding to a user request according to a user request and an advertisement recall list corresponding to the user request, estimating the click rate of each advertisement in the advertisement recall list, obtaining an estimated click rate of each advertisement, taking advertisements whose estimated click rates exceed the threshold value as candidate advertisements, and sorting and displaying the candidate advertisements. In the embodiments of the present disclosure, since the threshold value of different advertisement positions is determined for each user, and each advertisement is screened according to the threshold value and the click rate of each advertisement, the problem of the prior art that a uniformly specified advertisement access threshold leads to the delivery of high-priced, low-click rate advertisements, which affects the user experience, is avoided, and the accuracy of advertisement display is improved at the user granularity. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] In order to more clearly illustrate the technical solutions of the embodiments of the present disclosure, the drawings required for use in the description of the embodiments of the present disclosure will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments of the embodiments of the present disclosure. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0019] Figure 1 A flowchart showing the steps of an advertisement display method in an embodiment of the present disclosure is shown;

[0020] Figure 2 A structural diagram of an advertisement display device in one embodiment of the present disclosure is shown;

[0021] Figure 3 A structural diagram of an electronic device provided by an embodiment of the present disclosure is shown. DETAILED DESCRIPTION

[0022] The following will be combined with the drawings in the embodiments of the present disclosure to clearly and completely describe the technical solutions in the embodiments of the present disclosure. Obviously, the described embodiments are part of the embodiments of the present disclosure, rather than all of the embodiments. Based on the embodiments in the embodiments of the present disclosure, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the embodiments of the present disclosure.

[0023] Embodiment 1

[0024] Reference Figure 1 , which shows a flowchart of the steps of an advertisement display method in an embodiment of the present disclosure, including:

[0025] Step 101: Determine a threshold value of an advertisement position corresponding to a user request according to a user request and an advertisement recall list corresponding to the user request;

[0026] Step 102: Estimating the click rate of each advertisement in the advertisement recall list to obtain an estimated click rate of each advertisement;

[0027] Step 103: Advertisements whose estimated click-through rates exceed the threshold value are selected as candidate advertisements, and the candidate advertisements are sorted and displayed.

[0028] The advertisement display method disclosed in the present invention can be applied to terminals, and the terminals specifically include but are not limited to: smart phones, tablet computers, e-book readers, MP3 (Moving Picture Experts Group Audio Layer III) players, MP4 (Moving Picture Experts Group Audio Layer IV) players, laptop computers, car computers, desktop computers, set-top boxes, smart TVs, wearable devices, etc.

[0029] It should be noted that the user request in the present disclosure may be every search request for the user, wherein it may be a search request triggered by a preset operation when the user enters a search term in a search engine, a shopping website search bar, a social platform search bar, etc., or it may be a search request initiated by the user for a consumption scenario such as nearby, a designated area such as home or company, etc. The advertisement recall list may include advertisements associated with the search terms in the user search request, and a list of candidate advertisements determined according to the search terms.

[0030] Usually, when receiving a user request, if the user request is a search request, multiple natural results can be determined based on the search terms in the search request, and an ad recall list can also be determined based on the search terms in the search request. Taking the search scenario in a search engine as an example, the user enters the search term "headphones" in the search bar and clicks the search button to initiate a search request. Based on the search term "headphones", multiple natural results can be determined, such as encyclopedia entries on headphones, headphone-related forums, "headphones"-related forums, "headphones"-related Q&A, etc. Based on the search term "headphones", an ad recall list can be determined, including shopping Tmall headphone shopping ads, JD headphone shopping ads, Suning.com headphone shopping ads, etc.

[0031] When displaying natural results and advertisements, the display positions are usually distinguished, so that natural results and advertisements are mixed, such as displaying natural results in the first position, advertisements in the second position, natural results in the third position, advertisements in the fourth position, etc. At this time, the second and fourth positions are confirmed as advertisement positions. The embodiment of the present disclosure can determine the number and position of advertisements based on the historical information of the user in the user request, such as the historical number of advertisement clicks, the historical number of advertisement conversions, and the historical number of advertisement closings, so as to determine the advertisement position corresponding to the user request, or a fixed advertisement position can be uniformly set, and the present disclosure does not make specific restrictions on this.

[0032] In the disclosed embodiment, a threshold value is used to represent the entry click rate threshold of the advertisement in the advertisement slot corresponding to the user request. Generally speaking, there is a certain corresponding relationship between the click rate of an advertisement and its quality. The better the quality of the advertisement, the higher its click rate, and the lower the quality of the advertisement, the lower its click rate. By determining the threshold value of the corresponding advertisement slot based on the user request and the advertisement recall list, advertisements with low click rates and low quality can be filtered out, thereby ensuring the quality of the advertisements, avoiding high-bid and low-quality advertisements from being listed in the front, which affects the accuracy of advertisement recommendations, and improving the user experience. At the same time, by determining the threshold value of the corresponding advertisement slot based on the user request, advertisements can be filtered to different degrees according to the situations of different users, thereby further improving the accuracy of advertisement recommendations.

[0033] Therefore, it is necessary to estimate the click-through rate of each advertisement in the advertisement recall list, where the click-through rate is the number of times the advertisement / natural result is clicked by users and the number of times the advertisement / natural result is exposed to users. The click-through rate of each advertisement in the advertisement recall list is estimated, and then compared with the threshold value of the corresponding ad position. Ads with estimated click-through rates lower than the threshold value are filtered out, and ads with estimated click-through rates higher than the threshold value are selected as candidate ads for the corresponding ad position. At this time, different ad positions may have different candidate ads, and candidate ads in different ad positions may overlap.

[0034] In the embodiments of the present disclosure, candidate advertisements in the same advertisement slot may be sorted and displayed during the display process, such as displaying the first advertisement among the candidate advertisements in the advertisement slot on the first page, and displaying the second advertisement among the candidate advertisements in the advertisement slot on the next page. For advertisements that have been displayed, when it is the candidate advertisement's turn to be displayed again in other advertisement slots, the next advertisement may be skipped for display, or the next advertisement may be skipped for display after being repeated a preset number of times. Optionally, the candidate advertisements may be sorted from high to low according to the estimated click-through rate, or the candidate advertisements may be scored for quality and sorted from high to low according to the score. The present disclosure does not impose any specific limitation on the manner in which the candidate advertisements are sorted and displayed.

[0035] In an optional embodiment of the present disclosure, determining the threshold value of the advertisement position corresponding to the user request according to the user request and the advertisement recall list corresponding to the user request includes:

[0036] Step S11, determining feature data related to the user request according to the user request and the advertisement recall list;

[0037] Step S12: inputting the characteristic data into a threshold coefficient model, and outputting a threshold coefficient of the advertisement space corresponding to the user request through the threshold coefficient model;

[0038] Step S13: estimating the click rate of the advertisement position corresponding to the user request to obtain the estimated click rate of the advertisement position;

[0039] Step S14: Determine the threshold value of the advertisement position according to the threshold coefficient of the advertisement position and the estimated click rate of the position of the advertisement position.

[0040] User characteristic data is data related to predicted user behavior determined based on user requests and an advertisement recall list. In an optional embodiment of the present disclosure, the characteristic data may include an estimated click-through rate, an estimated conversion rate, an estimated transaction amount corresponding to each advertisement in the advertisement recall list, and at least one of the search terms, cities, and categories corresponding to the user request, that is, predicting the probability of a user clicking on an advertisement, the probability of entering a corresponding website after clicking an advertisement, the probability of entering a corresponding website and purchasing a corresponding product, etc., or predicting the user's search intent, etc.

[0041] The disclosed embodiment can input feature data into a threshold coefficient model to obtain a threshold coefficient for an ad slot corresponding to a user request. Since different ad slots may receive different degrees of user attention, different ad slots may have different threshold coefficients. When the threshold coefficient model receives feature data corresponding to the same user request and an ad recall list, it can output threshold coefficients corresponding to different ad slots based on the input feature data, so as to set different threshold values ​​for different ad slots and filter the ad recall list to different degrees. It can also output threshold coefficients for all ad slots in a unified manner, so as to perform unified threshold value calculations for different ad slots and improve the efficiency of ad screening.

[0042] At the same time, due to the different positions of ad slots, the degree of user attention they receive varies, and therefore, the click-through rates of ad slots also vary. For example, when the ad slots are in the second, fifth, and seventh positions on the search page, users may prefer to click on the ad first and then browse other natural results. In this case, the click-through rate of the second-ranked ad slot will be higher; users may prefer to browse some natural results first and then click on the ad for comparison. In this case, the click-through rate of the fifth-ranked ad slot will be higher; users may prefer to browse all natural results and then click on the ad for comparison. In this case, the click-through rate of the seventh-ranked ad slot will be higher. Therefore, the user's historical click data can be obtained based on the user's request, thereby determining the estimated click-through rate of each position of the ad slot corresponding to the current user's request.

[0043] In the disclosed embodiment, the threshold value of an ad slot can be determined by a threshold coefficient and an estimated click-through rate of each ad slot. The threshold coefficient is a threshold coefficient of the ad slot determined by a threshold coefficient model based on the feature data corresponding to the current user request and the current advertisement recall list. The estimated click-through rate of the ad slot is a prediction of the click-through rate of the ad slot based on the historical click data corresponding to the user request. Therefore, the estimated click-through rate of the ad slot can be corrected by the threshold coefficient to obtain an accurate threshold value corresponding to the ad slot.

[0044] In addition, in addition to calculating the threshold value corresponding to the ad position in the above manner, the distance between the merchant corresponding to the advertisement and the user, the star rating of the merchant corresponding to the advertisement, etc. can also be used as screening conditions for candidate advertisements in the advertisement recall list, thereby further ensuring the quality of candidate advertisements and improving user experience.

[0045] In an optional embodiment of the present disclosure, the threshold coefficient model is trained by the following steps:

[0046] Step S21, determining historical feature data related to the historical user request according to the historical user request and the historical advertisement recall list corresponding to the historical user request;

[0047] Step S22: using the historical feature data as input to an initial threshold coefficient model, and outputting an estimated threshold coefficient of a historical advertisement position corresponding to the historical user request through the initial threshold coefficient model;

[0048] Step S23, determining an estimated threshold value of the historical advertisement position according to the estimated threshold coefficient of the historical advertisement position and the estimated click rate of the historical position of the historical advertisement position;

[0049] Step S24, determining the cumulative value corresponding to the historical advertising position according to the actual click value after the historical advertising position adopts the estimated threshold value, the estimated click value before the historical advertising position adopts the estimated threshold value, the number of ads that can be exposed in the historical advertising recall list, and the number of ads that are actually exposed in the historical advertising recall list;

[0050] Step S25: using the accumulated value as the reward function of the initial threshold coefficient model, training the initial threshold coefficient model, and adjusting the model parameters of the initial threshold coefficient model to obtain a trained threshold coefficient model.

[0051] In an optional embodiment of the present disclosure, the actual click value is calculated based on the preset click value of the natural result, the single click cost of the advertisement, and the conversion coefficient between the advertisement revenue and the advertisement click volume;

[0052] The estimated click value is determined based on the estimated click rate of the historical advertising position, and the estimated click rate of the historical advertising position meets a preset threshold.

[0053] In the present disclosure, deep reinforcement learning (DDPG) can be used for modeling, and historical user requests and historical advertisement recall lists corresponding to the historical requests are collected to determine historical feature data of the historical user requests, and the historical feature data is input into the initial threshold coefficient model, and then the estimated threshold coefficient of the historical advertisement position corresponding to the historical user request is output, and then the estimated click-through rate of the historical position of the historical advertisement position, or the actual click-through rate of the historical position of the historical advertisement position, is used to determine the estimated threshold value of the advertisement position. Optionally, historical data such as historical user requests and historical advertisement recall lists can be exported in real time or offline by online logs, and the logs can be cleaned to remove outliers, duplicate data, useless data or data that may be cheating in the log generation process, and the data obtained that is useful for building the model is normalized, and / or data transformation and combination are performed, and then the data is input into the initial threshold coefficient model for model training.

[0054] After determining the estimated threshold value, the historical advertisement recall list of the historical advertisement position can be screened according to the estimated threshold value to determine the historical candidate advertisement corresponding to the historical advertisement position. The actual click value of the historical advertisement position after adopting the estimated threshold value can be determined through the historical candidate advertisement. In the disclosed embodiment, the click value is an indicator used to measure the estimated user click or non-click advertisement after adjusting the threshold value. Since the platform generates revenue by placing advertisements, but it may affect the user experience, the click value needs to achieve a balance between the user experience and the advertising platform. While measuring the user experience by estimating the click rate, in order to optimize the click rate while balancing the platform's revenue, the click value of the advertisement can be marked.

[0055] Among them, under the condition that the design of the advertising mechanism is relatively reasonable, without further improving the advertising coverage, there is a certain exchange relationship between the decline in user experience and platform revenue. For example, after certain adjustments such as increasing the threshold value, the recommendation is more accurate, the user click-through rate will increase, but the platform revenue will decrease, and vice versa, the user click-through rate will decrease, but the platform revenue will increase. In addition, for click-based advertising, in a single click, since the user's click on the advertisement brings revenue, its click value should be higher than the user's click value on the natural results. Similarly, for the same ad click, the click with a higher single-click billing has a higher click value.

[0056] Based on the above conditions, the click value of natural results can be set to 1, and the click value of advertisements can be set to 1+m*cpc, where cpc represents the cost per click of the advertisement. Optionally, in order to avoid a large difference between the click value of advertisements and the click value of natural results, the conversion relationship coefficient m can be multiplied before cpc, and the value range of m can be made to be (0,1), thereby balancing the click value of natural results and the click value of advertisements. m can be regarded as the conversion relationship coefficient between platform revenue and user experience.

[0057] In order to calculate the exchange relationship coefficient m, we can set up an experimental group including natural results and advertisements and a control group with only natural results, and assume that the value brought by each user request is equal on average. Therefore, we can get the following formula:

[0058]

[0059] Among them, SumClick represents the total number of clicks; SumRevenue represents the revenue generated by clicking on ads; SumRequest represents the number of ad requests, subscript e represents the experimental group, and subscript b represents the control group. The left formula in formula (1) represents the ratio of the click-through rate in the traffic containing ads and natural results and the average advertising revenue generated by each user request to the number of ad requests, and the right formula represents the ratio of the click-through rate of only natural results to the number of ad requests. In order to ensure that the two sides of the formula are within the same dimensional range after adding the platform revenue, so as to calculate m, it is assumed that the values ​​of the two sides are equal and the formula (1) is simplified to obtain:

[0060]

[0061] In the above formula, Ctr b Indicates the click-through rate of natural results in the control group, Ctr c It represents the click-through rate of the experimental group's advertisements. That is, from the above content, we can know that the exchange relationship coefficient m can be expressed by the ratio of the difference between the click-through rate of natural results and the click-through rate of advertisements to the revenue from clicking on advertisements. Thus, we can obtain the preset click value of natural results, the cost per click of advertisements, and the exchange relationship between advertisement revenue and advertisement clicks, and obtain the actual click value of the advertisement position.

[0062] In the disclosed embodiment, in order to compare whether the click value of the historical advertising position is improved after adopting the adjusted threshold value, the estimated click value of the historical advertising position when the estimated threshold value is not adopted can also be calculated, thereby comparing the actual click value and the estimated click value of the advertising position.

[0063] Among them, the estimated click value is calculated through the estimated click-through rate of the position. At this time, it is necessary to ensure that the difference between the estimated click-through rate of the position of the ad position and the click-through rate of the natural result is within a certain range, that is, the estimated click-through rate of the position meets the preset threshold. For example, if the ratio of the advertisement click-through rate to the natural result click-through rate is σ, then the decrease in the estimated click-through rate of the position should not be less than 1-σ. For example, if the advertisement click-through rate is above 85% of the natural result click-through rate, then relative to the natural result, the difference between the advertisement click-through rate and the natural result click-through rate is (1-85%)=15%, that is, the estimated click-through rate of the position should be between 85% and 100%. The estimated click value of the historical ad position is determined based on the estimated click-through rate of the position and the cost per click of the advertisement.

[0064] At this point, the cumulative value corresponding to the historical ad slots can be calculated as follows:

[0065]

[0066] Among them, ΣClickValue is the sum of the actual click values ​​corresponding to the historical ad positions, ∑pCtr pos*σ / σ′ is the sum of the estimated click values ​​corresponding to the historical ad positions, Σimpression represents the number of ads actually exposed in the historical ad recall list, and ΣPV represents the number of ads that can be exposed in the historical ad recall list. The cumulative value corresponding to the historical ad positions is obtained by the above formula (3).

[0067] In the present disclosure, the above formula (3) can be used as a reward function for deep learning in the threshold coefficient model training process. At this time, the state value in the deep learning process may include cityid (city id), cateid (category id), query (user search term), hour (time period of request), maxCtr (maximum click-through rate of the advertisement recall list), minCtr (minimum click-through rate of the advertisement recall list), avgCtr (average click-through rate of the advertisement recall list), maxCvr (maximum conversion rate of the advertisement recall list), minCvr (minimum conversion rate of the advertisement recall list), avgCvr (average conversion rate of the advertisement recall list), maxPrice (maximum conversion rate of the advertisement recall list), minCvr (minimum conversion rate of the advertisement recall list), avgCvr (average conversion rate of the advertisement recall list), maxPrice (maximum conversion rate of the advertisement recall list), minCvr (minimum conversion rate of the advertisement recall list), avgCvr (average conversion rate of the advertisement recall list), maxPrice (maximum conversion rate of the advertisement recall list), minCvr (minimum conversion rate of the advertisement recall list), avgCvr (average conversion rate of the advertisement recall list), maxPrice (maximum conversion rate of the advertisement recall list), minCvr (minimum conversion rate of the advertisement recall list), avgCvr (average conversion rate of the advertisement recall list), maxCvr ... The fine-grained feature data related to user requests may include maximum billing price of the ad recall list, minPrice (minimum billing price of the ad recall list), avgPrice (average billing price of the ad recall list), ctr list (ad click-through rate list), cvr list (ad conversion rate list), bid list (ad bid list), price list (ad billing list), ctr*bid list, ctr*price list, etc. The action may be the threshold coefficient k value, where the product of the threshold coefficient and the estimated click-through rate posCtr of the historical ad slot posCtr*k can be used as the threshold value of the historical ad slot, and the experience may be the user requests of each user within a period of time.

[0068] In an optional embodiment of the present disclosure, the reward function further includes a first penalty item and a second penalty item corresponding to the cumulative value; the first penalty item includes a first penalty coefficient and a first penalty factor; the second penalty item includes a second penalty coefficient and a second penalty factor;

[0069] Among them, the first penalty factor is the maximum value of the historical advertisement estimated click-through rates corresponding to each advertisement in the historical advertisement recall list, and the threshold value is greater than the maximum value; the second penalty factor is the difference in average quality of advertisements displayed before and after the historical advertisement position adopts the estimated threshold value.

[0070] In the disclosed embodiment, in order to avoid the threshold coefficient output by the threshold coefficient model obtained through training being too high, resulting in the threshold value being higher than the maximum value of the estimated click-through rate in the advertisement recall list, thereby resulting in the situation where no advertisement is displayed in the advertisement position, a first penalty item and a second penalty item corresponding to the cumulative value may be added to the reward function, wherein the first penalty factor of the first penalty item is the maximum value of the estimated click-through rates of the historical advertisements corresponding to the advertisements in the historical advertisement recall list, and the threshold value is greater than the maximum value, thereby avoiding the problem of no advertisement being displayed due to the output threshold coefficient being too high, and the second penalty factor of the second penalty item is the difference in the average quality of the advertisements displayed before and after the estimated threshold value is adopted in the historical advertisement position, thereby ensuring that the average quality of the advertisement is improved after the threshold value is added, and the reward function finally obtained is as follows:

[0071] reward=R+alpha*PenaltyCtr+beta*PenaltyScoresDiff (4)

[0072] Where R is the above formula (3), penaltyCtr represents the penalty for a threshold value greater than the maximum click rate of the list; PenaltyScoresDiff represents the difference in average ad quality before and after the threshold value is added; alpha and beta are parameters found through parameter adjustment.

[0073] In actual applications, if after the above adjustments, the threshold value of the ad slot calculated by the threshold coefficient output by the threshold coefficient model is still greater than the maximum click-through rate of the ad recall list, optionally, a preset number of ads in the ad recall list whose click-through rates are arranged from high to low can be used as candidate ads for the ad slot to avoid the problem of having no ads to display.

[0074] In the disclosed embodiments, historical data is collected, and deep learning training of the threshold coefficient model is performed using the above-mentioned reward function. When online predictions are performed using the obtained threshold coefficient model, the threshold coefficient model can also be updated in real time or offline based on real-time user requests, advertisement recall lists, threshold coefficients, candidate advertisements, advertisement click-through rates and other data, thereby ensuring the accuracy of the threshold coefficient model output.

[0075] In an optional embodiment of the present disclosure, after obtaining the trained threshold coefficient model, the method further includes:

[0076] Step S26, inputting the estimated threshold coefficient and the historical characteristic data into a value model, and outputting an estimated threshold value corresponding to the estimated threshold coefficient through the value model;

[0077] Step S26: adjusting the model parameters of the threshold coefficient model according to the difference between the estimated threshold value and the cumulative value.

[0078] In the disclosed embodiment, whether the estimated threshold coefficient output by the threshold coefficient model can achieve the greatest value balance between user experience and platform revenue can be evaluated by a value model. The value model receives the estimated threshold coefficient output by the threshold coefficient model, and outputs the historical feature data on which the estimated threshold coefficient is based. The value model can output an estimated threshold value corresponding to the estimated threshold coefficient. The estimated threshold value represents the estimated value obtained after the threshold value calculated by the threshold coefficient is used for the ad slot. The model parameters of the threshold coefficient model are adjusted through the deviation between the estimated threshold value and the actual cumulative value of the historical ad slot, thereby ensuring that the prediction result of the threshold coefficient model is closer to the actual threshold coefficient, making the prediction result more accurate.

[0079] In addition, optionally, the disclosed embodiment may also iteratively update the value model based on the estimated threshold coefficient, historical feature data and evaluation results, while evaluating the authenticity of the threshold coefficient output by the threshold coefficient model, so as to ensure the accuracy of the model evaluation.

[0080] In summary, the embodiment of the present disclosure provides an advertisement display method, the method comprising: determining the threshold value of the advertisement position corresponding to the user request according to the user request and the advertisement recall list corresponding to the user request, estimating the click rate of each advertisement in the advertisement recall list, obtaining the estimated click rate of each advertisement, taking the advertisement whose estimated click rate exceeds the threshold value as the candidate advertisement, and sorting and displaying the candidate advertisement. In the embodiment of the present disclosure, since the threshold value of different advertisement positions is determined for each user, and each advertisement is screened according to the threshold value and the click rate of each advertisement, the problem of the same designated advertisement access threshold in the prior art causing the delivery of high-priced and low-click rate advertisements, which affects the user experience, is avoided, and the accuracy of advertisement display is improved at the user granularity, and the present disclosure can adjust the threshold coefficient according to the real-time feedback of the user, and can make different degrees of adjustment and exchange according to the granularity of the feature data, and the advertisement display is flexible and targeted. Finally, the present disclosure screens advertisements at the advertisement access stage, achieving the direct and accurate exchange of user experience and platform revenue, which can better optimize the accuracy of advertisement recommendation and improve user experience.

[0081] Embodiment 2

[0082] Reference Figure 2 , which shows a structural diagram of an advertising display device 200 in an embodiment of the present disclosure, as follows.

[0083] A threshold determination module 201, configured to determine a threshold value of an advertisement position corresponding to a user request according to a user request and an advertisement recall list corresponding to the user request;

[0084] Advertisement screening module 202, used to estimate the click rate of each advertisement in the advertisement recall list to obtain the estimated click rate of each advertisement;

[0085] The sorting and displaying module 203 is used to select the advertisements whose estimated click-through rates obtained by the advertisement screening module exceed the threshold value as candidate advertisements, and sort and display the candidate advertisements.

[0086] Optionally, the threshold determination module 201 includes:

[0087] A feature determination submodule, configured to determine feature data related to the user request according to the user request and the advertisement recall list;

[0088] A coefficient determination submodule, used for inputting the feature data determined by the feature determination submodule into a threshold coefficient model, and outputting a threshold coefficient of the advertisement position corresponding to the user request through the threshold coefficient model;

[0089] A click rate estimation submodule, used to estimate the click rate of the advertisement position corresponding to the user request, so as to obtain the estimated click rate of the position of the advertisement position;

[0090] The threshold determination submodule is used to determine the threshold value of the advertisement position according to the threshold coefficient of the advertisement position output by the coefficient determination submodule and the estimated click rate of the position of the advertisement position.

[0091] Optionally, the characteristic data includes at least any one of the following: an estimated click rate, an estimated conversion rate, an estimated transaction amount corresponding to each advertisement in the advertisement recall list, and a search term, a city, and a category corresponding to the user request.

[0092] Optionally, the device further comprises: a model training module, used for training the threshold coefficient model; the model training module comprises:

[0093] A data acquisition submodule, configured to determine historical feature data related to historical user requests based on historical user requests and a historical advertisement recall list corresponding to the historical user requests;

[0094] A data prediction submodule, used to use the historical feature data determined by the data acquisition submodule as an input of an initial threshold coefficient model, and output an estimated threshold coefficient of a historical advertising position corresponding to the historical user request through the initial threshold coefficient model;

[0095] A first determination submodule, configured to determine an estimated threshold value of the advertisement position according to the estimated threshold coefficient of the historical advertisement position output by the data prediction submodule and the estimated click rate of the historical position of the historical advertisement position;

[0096] The second determination submodule is used to determine the cumulative value corresponding to the historical advertising position according to the actual click value after the historical advertising position adopts the estimated threshold value, the estimated click value before the historical advertising position adopts the estimated threshold value, the number of advertisements that can be exposed in the historical advertising recall list, and the number of advertisements that are actually exposed in the historical advertising recall list;

[0097] A parameter adjustment submodule is used to use the accumulated value determined by the second determination submodule as the reward function of the initial threshold coefficient model, train the initial threshold coefficient model and adjust the model parameters of the initial threshold coefficient model to obtain a trained threshold coefficient model.

[0098] Optionally, the actual click value is calculated based on a preset click value of a natural result, a single click cost of an advertisement, and a conversion coefficient between advertisement revenue and advertisement click volume;

[0099] The estimated click value is determined based on the estimated click rate of the historical advertising position, and the estimated click rate of the historical advertising position meets a preset threshold.

[0100] Optionally, the reward function further includes a first penalty item and a second penalty item corresponding to the accumulated value; the first penalty item includes a first penalty coefficient and a first penalty factor; the second penalty item includes a second penalty coefficient and a second penalty factor;

[0101] Among them, the first penalty factor is the maximum value of the historical advertisement estimated click-through rates corresponding to each advertisement in the historical advertisement recall list, and the threshold value is greater than the maximum value; the second penalty factor is the difference in average quality of advertisements displayed before and after the historical advertisement position adopts the estimated threshold value.

[0102] Optionally, the device further comprises:

[0103] A value estimation module, used for inputting the estimated threshold coefficient determined by the coefficient determination submodule and the historical characteristic data into a value model, and outputting an estimated threshold value corresponding to the estimated threshold coefficient through the value model;

[0104] The model optimization module is used to adjust the model parameters of the threshold coefficient model according to the difference between the estimated threshold value output by the value estimation module and the cumulative value.

[0105] In summary, an embodiment of the present disclosure provides an advertising display device, which includes: a threshold determination module, which is used to determine the threshold value of the advertising position corresponding to the user request based on the user request and the advertising recall list corresponding to the user request; an advertising screening module, which is used to estimate the click rate of each advertisement in the advertising recall list to obtain the estimated click rate of each advertisement; a sorting and display module, which is used to select advertisements whose estimated click rate exceeds the threshold value as candidate advertisements, and to sort and display the candidate advertisements. It is able to determine the threshold value of different advertising positions for each user, and screen each advertisement according to the threshold value and the click-through rate of each advertisement. Therefore, it avoids the problem of the same designated advertisement access threshold in the prior art leading to the delivery of high-priced, low-click-rate advertisements, which affects the user experience, and improves the accuracy of advertisement display at the user granularity. In addition, the present invention can adjust the threshold coefficient according to the real-time feedback of the user, and can make different degrees of adjustment and exchange according to the granularity of the feature data. The advertisement display is flexible and highly targeted. Finally, the present invention screens the advertisements at the advertisement access stage, achieves a direct and accurate exchange of user experience and platform revenue, and can better optimize the accuracy of advertisement recommendations and enhance user experience.

[0106] The embodiment of the present disclosure also provides an electronic device, see Figure 3 , including: a processor 301, a memory 302, and a computer program 3021 stored in the memory and executable on the processor, wherein the processor implements the advertisement display method of the aforementioned embodiment when executing the program.

[0107] An embodiment of the present disclosure further provides a readable storage medium. When instructions in the storage medium are executed by a processor of an electronic device, the electronic device is enabled to execute the advertisement display method of the aforementioned embodiment.

[0108] As for the device embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment.

[0109] The algorithms and displays provided herein are not inherently related to any particular computer, virtual system or other device. Various general purpose systems may also be used together with the teachings based thereon. According to the above description, it is apparent that the structure required for constructing such systems is. In addition, the embodiments of the present disclosure are not directed to any particular programming language either. It should be understood that the contents of the embodiments of the present disclosure described herein may be realized using various programming languages, and the description of the specific languages ​​above is intended to disclose the best mode of implementation of the embodiments of the present disclosure.

[0110] In the description provided herein, a large number of specific details are described. However, it is understood that the embodiments of the embodiments of the present disclosure can be practiced without these specific details. In some instances, well-known methods, structures and techniques are not shown in detail so as not to obscure the understanding of this description.

[0111] Similarly, it should be understood that in order to streamline the present disclosure and aid in understanding one or more of the various inventive aspects, in the above description of the exemplary embodiments of the present disclosure, the various features of the embodiments of the present disclosure are sometimes grouped together into a single embodiment, figure, or description thereof. However, the disclosed method should not be interpreted as reflecting the following intention: that the claimed embodiments of the present disclosure require more features than the features explicitly recited in each claim. More specifically, as reflected in the claims below, the inventive aspects lie in less than all the features of the single embodiment disclosed above. Therefore, the claims that follow the specific embodiment are hereby expressly incorporated into the specific embodiment, with each claim itself as a separate embodiment of the embodiments of the present disclosure.

[0112] Those skilled in the art will appreciate that the modules in the devices in the embodiments may be adaptively changed and arranged in one or more devices different from the embodiments. The modules or units or components in the embodiments may be combined into one module or unit or component, and in addition they may be divided into a plurality of submodules or subunits or subcomponents. Except that at least some of such features and / or processes or units are mutually exclusive, all features disclosed in this specification (including the accompanying claims, abstracts and drawings) and all processes or units of any method or device disclosed in this manner may be combined in any combination. Unless otherwise expressly stated, each feature disclosed in this specification (including the accompanying claims, abstracts and drawings) may be replaced by an alternative feature providing the same, equivalent or similar purpose.

[0113] The various component embodiments of the embodiments of the present disclosure may be implemented in hardware, or in software modules running on one or more processors, or in a combination thereof. It should be understood by those skilled in the art that a microprocessor or a digital signal processor (DSP) may be used in practice to implement some or all of the functions of some or all of the components in the sorting device according to the embodiments of the present disclosure. The embodiments of the present disclosure may also be implemented as a device or apparatus program for executing part or all of the methods described herein. Such a program implementing an embodiment of the present disclosure may be stored on a computer-readable medium, or may be in the form of one or more signals. Such a signal may be downloaded from an Internet website, or provided on a carrier signal, or provided in any other form.

[0114] It should be noted that the above embodiments illustrate rather than limit the embodiments of the present disclosure, and that those skilled in the art may design alternative embodiments without departing from the scope of the appended claims. In the claims, any reference symbol between brackets should not be constructed as a limitation on the claims. The word "comprising" does not exclude the presence of elements or steps not listed in the claims. The word "one" or "an" preceding an element does not exclude the presence of multiple such elements. The embodiments of the present disclosure may be implemented by means of hardware including several different elements and by means of appropriately programmed computers. In a unit claim that lists several devices, several of these devices may be embodied by the same hardware item. The use of the words first, second, and third, etc. does not indicate any order. These words may be interpreted as names.

[0115] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0116] The above description is only a preferred embodiment of the embodiments of the present disclosure and is not intended to limit the embodiments of the present disclosure. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the embodiments of the present disclosure should be included in the protection scope of the embodiments of the present disclosure.

[0117] The above is only a specific implementation of the embodiment of the present disclosure, but the protection scope of the embodiment of the present disclosure is not limited thereto. Any technician familiar with the technical field can easily think of changes or replacements within the technical scope disclosed by the embodiment of the present disclosure, which should be included in the protection scope of the embodiment of the present disclosure. Therefore, the protection scope of the embodiment of the present disclosure should be based on the protection scope of the claims.

Claims

1. An advertisement display method, It is characterized in that The method comprises: Determining a threshold value of an advertisement slot corresponding to the user request according to the user request and an advertisement recall list corresponding to the user request; Estimating the click rate of each advertisement in the advertisement recall list to obtain an estimated click rate of each advertisement; Advertisements with estimated click-through rates exceeding the threshold are considered candidate ads, and the candidate ads are ranked and displayed; The determining, according to the user request and the advertisement recall list corresponding to the user request, a threshold value of the advertisement position corresponding to the user request includes: Determining characteristic data related to the user request according to the user request and the advertisement recall list; Inputting the characteristic data into a threshold coefficient model, and outputting a threshold coefficient of the advertisement position corresponding to the user request through the threshold coefficient model; Estimating the click rate of the advertisement position corresponding to the user request to obtain an estimated click rate of the advertisement position; Determining a threshold value of the advertisement position according to a threshold coefficient of the advertisement position and an estimated click rate of the position of the advertisement position; The threshold coefficient model is trained by the following steps: Determining historical feature data related to the historical user request according to the historical user request and the historical advertisement recall list corresponding to the historical user request; Using the historical feature data as input to an initial threshold coefficient model, and outputting an estimated threshold coefficient of a historical advertisement position corresponding to the historical user request through the initial threshold coefficient model; Determine the estimated threshold value of the historical advertisement position according to the estimated threshold coefficient of the historical advertisement position and the estimated click rate of the historical position of the historical advertisement position; Determine the cumulative value corresponding to the historical advertising position according to the actual click value after the historical advertising position adopts the estimated threshold value, the estimated click value before the historical advertising position adopts the estimated threshold value, the number of advertisements that can be exposed in the historical advertising recall list, and the number of advertisements that are actually exposed in the historical advertising recall list; Using the accumulated value as a reward function of the initial threshold coefficient model, training the initial threshold coefficient model, and adjusting model parameters of the initial threshold coefficient model to obtain a trained threshold coefficient model; The actual click value is calculated based on the preset click value of the natural result, the single click cost of the advertisement, and the conversion coefficient between the advertisement revenue and the advertisement click volume; The estimated click value is determined based on the estimated click rate of the historical advertising position, and the estimated click rate of the historical advertising position meets a preset threshold.

2. The method according to claim 1, It is characterized in that The characteristic data includes at least any one of the following: the estimated click rate, estimated conversion rate, estimated transaction amount corresponding to each advertisement in the advertisement recall list, and the search term, city, and category corresponding to the user request.

3. The method according to claim 1, It is characterized in that The reward function also includes a first penalty item and a second penalty item corresponding to the accumulated value; the first penalty item includes a first penalty coefficient and a first penalty factor; the second penalty item includes a second penalty coefficient and a second penalty factor; Among them, the first penalty factor is the maximum value of the historical advertisement estimated click-through rates corresponding to each advertisement in the historical advertisement recall list, and the threshold value is greater than the maximum value; the second penalty factor is the difference in average quality of advertisements displayed before and after the historical advertisement position adopts the estimated threshold value.

4. The method according to claim 1, It is characterized in that After obtaining the trained threshold coefficient model, the method further includes: Inputting the estimated threshold coefficient and the historical characteristic data into a value model, and outputting an estimated threshold value corresponding to the estimated threshold coefficient through the value model; According to the difference between the estimated threshold value and the accumulated value, the model parameters of the threshold coefficient model are adjusted.

5. An advertising display device, It is characterized in that The device comprises: A threshold determination module, configured to determine a threshold value of an advertisement slot corresponding to a user request according to a user request and an advertisement recall list corresponding to the user request; An advertisement screening module, used to estimate the click rate of each advertisement in the advertisement recall list to obtain an estimated click rate of each advertisement; A sorting and displaying module, used to select advertisements whose estimated click-through rates obtained by the advertisement screening module exceed the threshold value as candidate advertisements, and to sort and display the candidate advertisements; The threshold determination module comprises: A feature determination submodule, configured to determine feature data related to the user request according to the user request and the advertisement recall list; A coefficient determination submodule, used for inputting the feature data determined by the feature determination submodule into a threshold coefficient model, and outputting a threshold coefficient of the advertisement position corresponding to the user request through the threshold coefficient model; A click rate estimation submodule, used to estimate the click rate of the advertisement position corresponding to the user request, so as to obtain the estimated click rate of the position of the advertisement position; A threshold determination submodule, configured to determine a threshold value of the advertisement position according to the threshold coefficient of the advertisement position output by the coefficient determination submodule and the estimated click rate of the position of the advertisement position; The device further includes: a model training module for training the threshold coefficient model; the model training module includes: A data acquisition submodule, configured to determine historical feature data related to historical user requests based on historical user requests and a historical advertisement recall list corresponding to the historical user requests; A data prediction submodule, used to use the historical feature data determined by the data acquisition submodule as an input of an initial threshold coefficient model, and output an estimated threshold coefficient of a historical advertising position corresponding to the historical user request through the initial threshold coefficient model; A first determination submodule, configured to determine an estimated threshold value of the advertisement position according to the estimated threshold coefficient of the historical advertisement position output by the data prediction submodule and the estimated click rate of the historical position of the historical advertisement position; The second determination submodule is used to determine the cumulative value corresponding to the historical advertising position according to the actual click value after the historical advertising position adopts the estimated threshold value, the estimated click value before the historical advertising position adopts the estimated threshold value, the number of advertisements that can be exposed in the historical advertising recall list, and the number of advertisements that are actually exposed in the historical advertising recall list; a parameter adjustment submodule, configured to use the accumulated value determined by the second determination submodule as a reward function of the initial threshold coefficient model, train the initial threshold coefficient model, and adjust model parameters of the initial threshold coefficient model to obtain a trained threshold coefficient model; The actual click value is calculated based on the preset click value of the natural result, the single click cost of the advertisement, and the conversion coefficient between the advertisement revenue and the advertisement click volume; The estimated click value is determined based on the estimated click rate of the historical advertising position, and the estimated click rate of the historical advertising position meets a preset threshold.

6. The device according to claim 5, It is characterized in that The characteristic data includes at least any one of the following: the estimated click rate, estimated conversion rate, estimated transaction amount corresponding to each advertisement in the advertisement recall list, and the search term, city, and category corresponding to the user request.

7. The device according to claim 5, It is characterized in that The reward function also includes a first penalty item and a second penalty item corresponding to the accumulated value; the first penalty item includes a first penalty coefficient and a first penalty factor; the second penalty item includes a second penalty coefficient and a second penalty factor; Among them, the first penalty factor is the maximum value of the historical advertisement estimated click-through rates corresponding to each advertisement in the historical advertisement recall list, and the threshold value is greater than the maximum value; the second penalty factor is the difference in average quality of advertisements displayed before and after the historical advertisement position adopts the estimated threshold value.

8. The device according to claim 5, It is characterized in that The device also includes: A value estimation module, used for inputting the estimated threshold coefficient determined by the coefficient determination submodule and the historical characteristic data into a value model, and outputting an estimated threshold value corresponding to the estimated threshold coefficient through the value model; The model optimization module is used to adjust the model parameters of the threshold coefficient model according to the difference between the estimated threshold value output by the value estimation module and the cumulative value.

9. An electronic device, It is characterized in that include: A processor, a memory, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the advertisement display method as described in one or more of claims 1-4 is implemented.

10. A readable storage medium, It is characterized in that When the instructions in the storage medium are executed by a processor of an electronic device, the electronic device is enabled to execute the advertisement display method as described in one or more of method claims 1-4.

Images