A method and system for personalized video advertisement presentation and content unlocking

Abstract

The application provides a personalized video advertisement display and content unlocking method and system, and relates to the technical field of data processing. The method comprises the following steps: step 1, obtaining short drama content from multiple short drama operators and storing the short drama content in a content library according to themes; calculating a content value index of each short drama according to production quality, interactive heat value and commercial value weight; step 2, collecting historical behavior data of users in the content library, including short drama viewing time, theme selection frequency and content unlocking records, and calculating the initial preference intensity of users for each theme based on the historical behavior data. The application dynamically corrects theme preferences and generates a personalized recommendation list by fusing the content value index and the historical behavior data of users, combines an incentive video advertisement unlocking mechanism with strategy closed-loop optimization, and realizes the synergistic improvement of accurate recommendation, efficient advertisement monetization and user experience.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of data processing, in particular to a personalized video advertisement display and content unlocking method and system. BACKGROUND

[0002] With the large-scale development of mobile Internet and short video industry, video content platforms face the core technical challenge of insufficient coordination between recommendation algorithm and content commercial value in the process of realizing commercial monetization through personalized recommendation. The existing technology has two limitations:

[0003] Most of the models are based on historical behavior data such as user viewing time, click records, etc., lacking quantitative evaluation of content production quality (such as resolution, frame rate) and commercial attributes (advertiser investment value), which leads to over-recommendation of high user preference but low commercial value content, or insufficient exposure of high commercial value content due to mismatching user interest, affecting platform advertisement monetization efficiency.

[0004] Mainly in the form of "direct payment" or "mandatory advertisement", users passively accept the push, with low advertisement viewing completion rate, which easily leads to resistance, at the same time, the traditional mechanism lacks dynamic feedback of user behavior, making it difficult to identify abnormal operations such as short-time high-frequency unlocking, which may cause waste of system resources or damage to the interests of advertisers. SUMMARY

[0005] The technical problem to be solved by the present application is to provide a personalized video advertisement display and content unlocking method and system, realizing the technical closed loop of "accurate recommendation-high efficiency monetization-experience optimization".

[0006] To solve the above technical problems, the technical scheme of the present application is as follows:

[0007] In a first aspect, a personalized video advertisement display and content unlocking method, the method comprising:

[0008] Step 1: Obtain short drama content from multiple short drama operators and store it in the content library according to the theme; calculate the content value index of each short drama according to the production quality, interactive heat value and commercial value weight;

[0009] Step 2: Collect historical behavior data of users in the content library, including short drama viewing time, theme selection frequency and content unlocking record, and calculate the initial preference intensity of users for each theme based on the historical behavior data;

[0010] Step 3: Use the content value index as a correction parameter to dynamically weight the initial preference intensity to generate a corrected preference value for each theme;

[0011] Step 4, based on the modified preference value of each theme, screen the short drama content matching the theme from the content library, generate a candidate recommendation set, and generate a personalized recommendation list in the order of the modified preference value of each theme from high to low;

[0012] Step 5, when selecting the target short drama in the recommendation list, trigger the display of the incentive video advertisement, and after the user watches the incentive video advertisement, automatically unlock the target short drama content, and generate an unlocking behavior log;

[0013] Step 6, real-time analysis of the unlocking behavior log, identify abnormal operation mode, generate risk indicators; take the risk indicators as control instructions, dynamically adjust the configuration parameters of the business value weight, and recalculate the content value index, realize the strategy closed-loop optimization.

[0014] Further, obtain short drama content from multiple short drama operators, and store it in the content library according to the theme classification; calculate the content value index of each short drama according to the production quality, interactive heat value and business value weight, including:

[0015] Receive the short drama metadata package transmitted by the multiple short drama operators, analyze the theme label code, and store it in the content library according to the label code;

[0016] Read the production parameter set in the short drama metadata package, including resolution value, audio sampling rate and frame rate sequence, generate production quality score; obtain the user behavior record table of each short drama in the content library, extract the like counter value, comment counter value, share counter value and complete play rate percentage, and generate interactive heat value;

[0017] According to the advertiser configuration table, extract the advertiser's bid unit price value, ad position quantity value and historical conversion rate percentage corresponding to the short drama, and output the business value weight value;

[0018] Linearly superimpose the production quality score, interactive heat value and business value weight value, and output the content value index.

[0019] Further, collect the historical behavior data of the user in the content library, including short drama viewing time, theme selection frequency and content unlocking record, and calculate the initial preference intensity of the user for each theme based on the historical behavior data, including:

[0020] Monitor the operation instruction flow of the user terminal in the content library, capture and analyze three kinds of logs, including short drama playing log, theme clicking log and unlocking request log;

[0021] Group the operation logs according to the theme label code, generate the theme total time value, theme click times and unlocking success times;

[0022] The output duration proportion coefficient, the selection frequency coefficient and the unlocking conversion coefficient are determined according to the theme total duration value, the theme click number and the unlocking success number.

[0023] The initial preference intensity vector is determined according to the duration proportion coefficient, the selection frequency coefficient and the unlocking conversion coefficient.

[0024] Further, the theme total duration value, the theme click number and the unlocking success number are generated by grouping the operation log according to the theme label code, including:

[0025] The theme total duration value is output by performing a theme label aggregation operation on the short video play log and accumulating the total play duration of all short videos under the same theme label.

[0026] The theme click number is output by performing a frequency calculation algorithm on the theme click log to count the number of active clicks of each theme label by the user.

[0027] The unlocking success number is output by performing a state filtering operation on the unlocking request log to extract unlocking records with a successful verification state and counting the number of successful unlockings under each theme label.

[0028] Further, the content value index is used as a correction parameter to dynamically weight the initial preference intensity to generate a corrected preference value for each theme, including:

[0029] The content value index of each short video theme in the content library is extracted to generate a theme value vector.

[0030] The initial preference intensity vector and the theme value vector are fused to obtain a weighted preference vector.

[0031] The weighted preference vector is normalized to output the corrected preference value of each theme.

[0032] Further, based on the corrected preference value of each theme, short video content matching the theme is selected from the content library to generate a candidate recommendation set, and a personalized recommendation list is generated in order of the corrected preference value of each theme from high to low, including:

[0033] The top N themes are selected as target themes according to the corrected preference value of each theme.

[0034] Short videos belonging to the target theme are selected from the content library and sorted in order of the content value index from high to low, and the top M short videos are selected to generate a candidate recommendation set.

[0035] The candidate recommendation set is arranged in order of the corrected preference value of each theme from high to low to generate a personalized recommendation list.

[0036] Further, when the target short drama in the recommendation list is selected, the display of the incentive video advertisement is triggered, and after the user watches the incentive video advertisement, the target short drama content is automatically unlocked, and a unlocking behavior log is generated, including:

[0037] The selection operation of the user on the target short drama in the personalized recommendation list is monitored, and the incentive video advertisement is requested from the advertisement placement party according to the user portrait feature;

[0038] The matched incentive video advertisement is obtained and displayed on the user interface, and the advertisement playing progress is tracked in real time. When it is detected that the advertisement is completely played and ended, the content access restriction of the target short drama is released, and a data entry containing a timestamp, a user identifier, a short drama identifier, an advertisement identifier and an unlocking state is recorded. The data entry is written into the unlocking behavior log.

[0039] Further, the unlocking behavior log is analyzed in real time, an abnormal operation mode is identified, and a risk index is generated. The risk index is used as a regulation instruction to dynamically adjust the configuration parameters of the business value weight, and the content value index is recalculated to realize strategy closed-loop optimization, including:

[0040] The unlocking behavior log is scanned according to a preset time window, and the short drama unlocking frequency of the same user identifier and the advertisement triggering density of the same device are counted;

[0041] When the short drama unlocking frequency exceeds the dynamic threshold or the advertisement triggering density deviates from the reference range, a risk index containing an abnormal code is generated, and the advertisement conversion rate parameter and the delivery unit price parameter are adjusted according to the risk index level;

[0042] The business value weight of the affected short drama is recalculated using the adjusted advertisement conversion rate parameter and delivery unit price parameter to obtain an updated business value weight;

[0043] The content value index is recalculated based on the updated business value weight, and the recalculated value index is fed back to the recommendation list generation process to complete the strategy closed-loop optimization.

[0044] In a second aspect, a system for personalized video advertisement display and content unlocking includes:

[0045] A value evaluation module is configured to obtain short drama content from a plurality of short drama operators, store the short drama content in a content library according to themes, and calculate a content value index of each short drama according to production quality, interactive heat value and business value weight;

[0046] A user behavior analysis module is configured to collect historical behavior data of users in the content library, such as short drama viewing time, theme selection frequency and content unlocking record, and calculate the initial preference intensity of the user for each theme;

[0047] The preference correction module is used for dynamically weighting the initial preference intensity by taking the content value index as a correction parameter, and generating a corrected preference value of each theme which is more in line with actual demand.

[0048] The personalized recommendation module is used for screening short drama content matched in theme from the content library based on the corrected preference value of each theme, generating a candidate recommendation set, and generating a personalized recommendation list in the order from high to low of the corrected preference value.

[0049] The incentive advertisement unlocking module is used for triggering the display of an incentive video advertisement when the user selects a target short drama in the recommendation list, automatically unlocking the target short drama content after the user watches the advertisement, and generating an unlocking behavior log.

[0050] The strategy closed-loop optimization module is used for analyzing the unlocking behavior log in real time, identifying abnormal operation modes and generating a risk index, taking the risk index as a regulation and control instruction, dynamically adjusting the configuration parameter of the commercial value weight, recalculating the content value index, and realizing the closed-loop optimization of the recommendation strategy.

[0051] In a third aspect, a computer readable storage medium is provided, and the computer readable storage medium stores a program, and the program is executed by a processor to implement the method.

[0052] The above scheme of the present application at least has the following beneficial effects:

[0053] The content value index is calculated by fusing the production quality, the interactive heat and the commercial value weight, the limitation of the traditional recommendation which only relies on user behavior data is changed, and the platform advertisement realization efficiency is improved. The user preference intensity is corrected by the content value index, the three-dimensional recommendation logic of "theme matching + content quality + commercial value" is formed, and the recommendation list synchronously meets the user interest and the platform commercialization target. The incentive video advertisement is used to replace the forced push mode, the user unlocks the content by actively watching the advertisement, the advertisement watching completion rate is improved, the user resistance emotion is effectively reduced, and the platform user retention is improved. The user interaction data is recorded in real time by the unlocking behavior log, precise feedback is provided for the advertisement placement party, the matching degree of the advertisement and the user is optimized, and a virtuous cycle of "user voluntarily watching - high efficient conversion of advertisement" is formed.

[0054] The unlocking frequency and the advertisement triggering density are monitored in real time, the abnormal operation (such as the behavior of brushing the advertisement to unlock) is dynamically identified, the strategy is self-optimized by adjusting the commercial value weight parameter, and the system risk resistance ability is enhanced. The content value index is updated in real time according to market dynamics (such as the advertisement principal budget adjustment and user preference migration), the accuracy of the recommendation strategy is ensured for a long time, and the lag problem of the traditional offline preset model is avoided. BRIEF DESCRIPTION OF DRAWINGS

[0055] Figure 1is a flowchart of a personalized video advertisement display and content unlocking method provided by an embodiment of the present application.

[0056] Figure 2 is a schematic diagram of a personalized video advertisement display and content unlocking system provided by an embodiment of the present application. DETAILED DESCRIPTION

[0057] Exemplary embodiments of the present disclosure will be described in greater detail below with reference to the accompanying drawings. Although exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure can be implemented in various forms and should not be limited by the embodiments set forth herein. Rather, these embodiments are provided so that the present disclosure can be more thoroughly understood and the scope of the present disclosure can be accurately conveyed to those skilled in the art.

[0058] As shown in Figure 1 An embodiment of the present application proposes a personalized video advertisement display and content unlocking method, which comprises the following steps:

[0059] Step 1: Obtain short drama content from multiple short drama operators and store it in a content library according to the theme; calculate the content value index of each short drama according to the production quality, interactive heat value and commercial value weight;

[0060] Step 2: Collect historical behavior data of users in the content library, including short drama viewing time, theme selection frequency and content unlocking record, and calculate the initial preference intensity of users for each theme based on the historical behavior data;

[0061] Step 3: Use the content value index as a correction parameter to dynamically weight the initial preference intensity to generate a corrected preference value for each theme;

[0062] Step 4: Based on the corrected preference value of each theme, filter short drama content matching the theme from the content library to generate a candidate recommendation set, and generate a personalized recommendation list in order of the corrected preference value of each theme from high to low;

[0063] Step 5: When a target short drama in the recommendation list is selected, trigger the display of a rewarded video advertisement, and after the user watches the rewarded video advertisement, automatically unlock the target short drama content to generate an unlocking behavior log;

[0064] Step 6: Real-time analysis of the unlocking behavior log to identify abnormal operation patterns and generate a risk indicator; use the risk indicator as a control instruction to dynamically adjust the configuration parameters of the commercial value weight and recalculate the content value index to achieve strategy closed-loop optimization.

[0065] In the embodiment of the present application, the multi-dimensional integrated short drama production quality, user interaction data and commercial value weight are combined to form a comprehensive content value evaluation system, avoiding the imbalance between content quality and commercial realization caused by single-dimensional recommendation, laying a data foundation for accurate recommendation. Based on the behavior data modeling of real viewing time and theme selection frequency, the initial interest tendency of users for different themes is accurately captured, ensuring that the recommendation logic fits the actual preferences of users and improving the acceptance of recommended content. As a weighted parameter, the content value index deeply integrates "user interest" and "content commercial value", avoiding the problems of "high preference low value" or "high value low exposure" in traditional recommendation, and realizing the bidirectional optimization of recommendation strategy.

[0066] According to the revised preference value, the short drama matched with the theme is screened, and the recommendation list considering user interest and commercial value is generated by sorting according to the content value index, which not only improves the user's viewing willingness, but also maximizes the platform's advertising realization potential. The incentive mode of "watching advertisements-unlocking content" is adopted to change passive advertisement pushing into user active participation, reduce the sense of advertisement interference, and record user interaction data through unlocking log. Real-time analysis of unlocking behavior log, dynamic identification of abnormal operation and adjustment of commercial value weight parameter make the content value evaluation and recommendation strategy self-adaptive iteration with user behavior and market dynamics, enhancing the system's anti-risk ability and long-term accuracy.

[0067] In a preferred embodiment of the present application, step 1, short drama content is obtained from multiple short drama operators and stored in a content library according to theme classification; the content value index of each short drama is calculated according to production quality, interaction heat value and commercial value weight, which can include:

[0068] Step 100, receiving short drama metadata packets transmitted by multiple short drama operators, analyzing theme label codes, and storing them in a content library according to label codes;

[0069] Step 101, reading the production parameter set in the short drama metadata packet, including resolution value, audio sampling rate and frame rate sequence, generating production quality score; obtaining user behavior record table of each short drama in the content library, extracting like counter value, comment counter value, share counter value and complete playback rate percentage, generating interaction heat value;

[0070] Step 102, according to the advertiser configuration table, extracting the bid unit price value, ad position value and historical conversion rate percentage of the corresponding short drama of the advertiser, outputting the commercial value weight value;

[0071] Step 103, linearly superimposing production quality score, interaction heat value and commercial value weight value, outputting content value index.

[0072] In the embodiment of the present application, the metadata package uploaded by multiple short drama operators is received through an API interface or a file transfer protocol. The metadata package follows a unified format and contains short drama basic information (such as title, ID), technical parameters (resolution, frame rate, etc.), theme label (such as standardized codes of “love” and “suspense”), and user behavior data interface address. The parsing module automatically identifies the theme label code (for example, “T001” corresponds to the “science fiction” theme), performs standardized processing through a regular expression or a label mapping table, and ensures the uniformity of labels of different operators (for example, “urban emotion” and “modern love” are classified as “love” theme codes). Based on the parsed theme label code, the short drama content is stored in the corresponding folder of the distributed content library (such as “ / theme library / love / T001 / ”), and an index table is established simultaneously during storage, including the association of theme labels, production quality parameters, and commercial value parameters.

[0073] Step 101, read the resolution (such as 1080p, 720p), audio sampling rate (such as 44.1 kHz, 48 kHz), and frame rate sequence (such as 25 fps, 30 fps) in the metadata package, and map them to a preset score interval. For example: 1080p corresponds to 80-100 points, 720p corresponds to 60-80 points; 48 kHz sampling rate corresponds to 90 points, and 44.1 kHz corresponds to 80 points. According to industry standards or platform strategies, set weights (such as resolution accounts for 40%, audio accounts for 30%, and frame rate accounts for 30%), calculate the weighted average score, and generate the production quality score (range 0-100).

[0074] From the user behavior record table of the content library, real-time pull the number of likes, the number of comments, the number of shares, and the completion rate of each short drama. For example, a short drama has 5000 likes, 1000 comments, 2000 shares, and an 85% completion rate. Normalize each indicator according to the maximum value (for example, the maximum value of likes is 10000, so 5000 corresponds to 0.5), and set weights according to the interaction intensity (for example, completion rate accounts for 50%, likes accounts for 20%, comments accounts for 20%, and shares accounts for 10%), and calculate the interaction heat value (range 0-1) by weighted calculation.

[0075] Step 102, real-time access to the advertiser configuration table through a structured database or an API interface, and retrieve the corresponding commercialization parameters according to the unique ID of the short drama:

[0076] Delivery unit price: get the price set by the advertiser for each thousand displays of the short drama (such as 50 yuan / thousand), and convert it into a standardized value according to the platform's unified standard (such as mapping “yuan / thousand” to a basic score value);

[0077] Number of ad slots: count the total number of implantable ad slots in the short drama (such as 3 implantation positions in natural scenes, character props, etc.);

[0078] Historical conversion rate: Extract the historical average conversion data of ads in the same category on the platform (for example, the user click conversion rate of ads in this category is 3%, which is converted into a standardized value of 0.03).

[0079] If the bid price is 0 or negative, it is automatically marked as abnormal and corrected from the average value of the same type of ads in the last 30 days. If the historical conversion rate exceeds the industry benchmark threshold (for example, set the threshold to 15%), it is forcibly corrected to the upper limit of the threshold to avoid extreme data interference with the evaluation logic. The commercial value weight is calculated using a linear combination formula, combined with dimension standardization processing:

[0080] First, convert the bid price into the base value per unit of ad space (for example, 50 yuan / thousand times ÷ 3 ad spaces = 16.67 yuan / thousand times·ad space), and then multiply it by the standardized historical conversion rate to form the commercial value weight: Commercial value weight = (Bid price ÷ Number of ad spaces) × Historical conversion rate.

[0081] Step 103, set fixed weights for production quality score, interaction heat value, and commercial value weight (for example, production quality accounts for 40%, interaction heat accounts for 30%, and commercial value accounts for 30%), or dynamically adjust according to real-time market demand (for example, increase the commercial value weight to 40% in peak season). Linearly superimpose the scores of the three dimensions according to the weights, for example: Content value index = Production quality score × 0.4 + Interaction heat value × 100 × 0.3 + Commercial value weight × 0.3 (since the interaction heat value ranges from 0 to 1, it needs to be multiplied by 100 to convert it to the 0-100 interval to match other dimensions), and finally output the content value index of each short drama (range 0-100).

[0082] Through three-dimensional quantification of production quality, interaction heat, and commercial value, the traditional recommendation relying only on user behavior or a single indicator is avoided, making the content value evaluation more comprehensive and reducing the problem of "high traffic and low monetization" or "high commercial value and low exposure". Unified analysis of metadata packages and theme labels enables standardized storage and rapid retrieval of multi-source content. Combined with ad placement data and historical conversion results, the commercial monetization potential of short dramas is quantified, allowing the platform to balance user experience and ad revenue when recommending, thereby improving overall commercial efficiency. The linear superposition mechanism of the content value index supports dynamic adjustment of weights, allowing real-time optimization of evaluation logic according to market demand (such as holiday marketing and fluctuations in advertiser budgets), enhancing system adaptability.

[0083] In a preferred embodiment of the present application, the above-mentioned step 2, collecting historical behavior data of users in the content library, including short drama viewing time, theme selection frequency, and content unlocking records, and calculating the initial preference intensity of users for each theme based on historical behavior data, can include:

[0084] Step 200, monitoring the operation instruction stream of the user terminal in the content library, capturing and analyzing three kinds of logs, including short play log, theme click log and unlocking request log;

[0085] Step 201, grouping the operation log according to the theme label code, generating theme total time value, theme click times and unlocking success times;

[0086] Step 202, determining the output time length ratio coefficient, selection frequency coefficient and unlocking conversion coefficient according to the theme total time value, theme click times and unlocking success times, specifically including:

[0087] Step 2020, performing theme label aggregation operation on the short play log, accumulating the total play time of all short plays under the same theme label, and outputting the theme total time value;

[0088] Step 2021, performing frequency calculation algorithm on the theme click log, counting the number of active clicks of users on each theme label, and outputting the theme click times;

[0089] Step 2022, performing state filtering operation on the unlocking request log, extracting unlocking records with successful verification state, counting the number of successful unlocking under each theme label, and outputting the unlocking success times;

[0090] Step 203, determining the initial preference intensity vector according to the time length ratio coefficient, selection frequency coefficient and unlocking conversion coefficient.

[0091] In the embodiment of the application, a lightweight monitoring plug-in is deployed on the user terminal (such as a mobile phone, a tablet computer), which listens to the content library operation instruction stream in real time, including but not limited to short play, theme page click, unlocking request and other behaviors.

[0092] Three kinds of log files are generated according to operation type classification:

[0093] Short play log: records the play start time, end time, actual viewing time (excluding invalid time such as fast forward, pause, etc.) and exit reason of each short play;

[0094] Theme click log: captures the active click behavior of users on the theme classification page, including the clicked theme label, stay time and jump path;

[0095] Unlocking request log: records the time when the user triggers the unlocking operation, target short play ID, request state (success / failure) and verification information.

[0096] Log analysis process:

[0097] Regular expressions and structured parsers are used to extract key fields from raw logs. For example, from the play log, parse out "short drama ID = D001, viewing duration = 5 minutes 30 seconds, theme label = T002 (suspense)", and convert unstructured text into JSON or table format storage.

[0098] Step 201, based on the theme label code parsed in step 100 (such as T001 = love, T002 = suspense), perform grouping aggregation on three types of logs:

[0099] Traverse all short drama play logs, accumulate the viewing duration of the same theme label, and generate "theme total duration value" (such as suspense theme total viewing duration = 120 minutes);

[0100] Scan the theme click log, count the number of user active clicks by theme label, and generate "theme click count" (such as love theme clicked 35 times);

[0101] Filter the unlock request log, only keep the records with status "success", count the number of unlocks by theme label, and generate "unlock success count" (such as science fiction theme successfully unlocked 8 times).

[0102] Mark and remove obvious outliers (such as single theme daily click count exceeding 10 times the user's daily operation volume, viewing duration being negative), to ensure the authenticity of the statistical results.

[0103] Step 202, calculate the proportion of single theme total duration to user's total viewing duration. For example, if the user's total viewing duration is 300 minutes and the suspense theme is 120 minutes, then the duration proportion coefficient = 120 ÷ 300 = 0.4. Standardize the theme click count to a relative value, assuming the user's total click count is 100 times, and the love theme clicks 35 times, then the selection frequency coefficient = 35 ÷ 100 = 0.35. Calculate the conversion rate of theme unlock success count and click count, for example, the science fiction theme clicks 20 times, successfully unlocks 8 times, then the unlock conversion coefficient = 8 ÷ 20 = 0.4 (indicating that every 10 clicks, 4 times are converted into unlock behavior).

[0104] Step 203, construct a three-dimensional initial preference intensity vector P = (a, b, c), where:

[0105] a = duration proportion coefficient (reflecting the user's immersion in the theme);

[0106] b = selection frequency coefficient (reflecting the user's willingness to actively explore the theme);

[0107] c = unlock conversion coefficient (reflecting the user's willingness to pay / engage in the theme).

[0108] Mapping the values of each dimension of the vector to the interval [0, 1] (such as by Min-Max standardization) ensures that coefficients of different dimensions are comparable, and the finally generated vector can intuitively reflect the user's interest priority for each theme (such as the suspense theme vector P=(0.4, 0.35, 0.4) representing a medium preference intensity).

[0109] Through cross analysis of the three types of behavior data of viewing duration, click frequency and unlocking conversion, the preference misjudgment caused by a single indicator (such as only viewing duration) is avoided, and the initial preference model is closer to the user's real interest (for example, a user frequently clicks on science fiction themes but rarely watches them, and the unlocking conversion coefficient is low, so the system can identify it as a "potential interest but not deep conversion" group). The three-dimensional coefficient clearly maps the association between user behavior and preference (duration proportion reflects "immersion", click frequency reflects "exploration", and unlocking conversion reflects "willingness to pay"), providing a traceable behavior basis for the recommendation result, and facilitating platform optimization operation strategy (such as optimizing content design for themes with low unlocking conversion).

[0110] In a preferred embodiment of the present application, step 3 above uses the content value index as a correction parameter to dynamically weight the initial preference intensity to generate a corrected preference value for each theme, which can include:

[0111] Step 300: Extract the content value index of each short drama theme in the content library to generate a theme value vector;

[0112] Step 301: Fuse the initial preference intensity vector and the theme value vector to obtain a weighted preference vector;

[0113] Step 302: Normalize the weighted preference vector to output the corrected preference value of each theme.

[0114] In an embodiment of the present application, the content value index of each short drama is extracted from the short drama metadata of the content library according to the theme label grouping (calculated by step 103, containing the linear superposition value of production quality, interaction heat, and commercial value weight). The content value index of all short dramas under the same theme is statistically processed, and the average value (such as weighted by the number of plays) is used to generate a comprehensive value index of the theme. For example, the "suspense" theme includes 10 short dramas, and the content value indexes are 85, 78, 92,..., and the average value is 86, which is taken as the value index of the theme. The comprehensive value indexes of each theme are arranged in the order of theme labels to form a theme value vector V=(v1, v2, v3, …, vn), where vi represents the content value index of the i-th theme, and the vector dimension is consistent with the number of platform theme classifications (such as 20 themes on the platform, then the vector is 20-dimensional).

[0115] Step 301, extract the initial preference intensity vector P = (p1, p2, p3, …, pn) (pi represents the initial preference coefficient of the user for the ith theme) generated in step 203, that is, obtained by weighting or aggregating the a, b, c coefficients of the theme in step 203; and the theme value vector V is multiplied by the corresponding elements: each element wi of the weighted preference vector W is pi x vi, which represents the "corrected preference intensity" of the user for the ith theme; if the user's preference for a certain theme is high but the content value index is low (such as poor production quality and low commercial value), the weighted preference intensity will be weakened; on the contrary, if the content value of a certain theme is high but the user's initial preference is low, the weighted preference can improve its recommendation priority, and the balance of "interest-value" is realized.

[0116] Step 302, Min-Max normalization is performed on the weighted preference vector W to map each element to the [0, 1] interval, ensuring that the corrected preference values of different themes are comparable, and generating a corrected preference value vector R = (r1, r2, r3, …, rn), where ri represents the final preference intensity (after normalization) of the user for the ith theme.

[0117] By weighting and correcting the user's preference with the content value index, the "information cocoon" caused by the traditional recommendation relying only on behavior data is avoided, so that the recommendation list not only meets the user's interest, but also takes into account the content production quality and commercial realization value (for example, a theme with high user preference but low commercial value will be appropriately down-weighted, and high-value content will get more exposure). The corrected mechanism of the commercial value weight makes the platform actively regulate the recommendation priority and push the theme / short drama with high advertising investment value to potential users. Dynamic weighting avoids user resistance caused by excessive recommendation of low-quality content, and at the same time, high-quality content is selected through the content value index to improve the user's viewing experience (such as preferentially recommending short dramas with high production quality and high interaction), forming a virtuous cycle of "high-quality content-high interaction-high commercial value". Normalization processing ensures that indicators of different dimensions and different units (such as preference coefficients and content value indexes) can be directly operated, so that the corrected preference value can accurately reflect the real priority of "user interest intensity x content comprehensive value", and provide reliable sorting basis for the recommendation algorithm.

[0118] In a preferred embodiment of the present application, step 4 above, based on the corrected preference value of each theme, short drama content matching the theme is selected from the content library to generate a candidate recommendation set, and a personalized recommendation list is generated in the order of the corrected preference value of each theme from high to low, which can include:

[0119] Step 400, select the top N themes as the target themes according to the corrected preference value of each theme;

[0120] Step 401, filter short dramas belonging to the theme subject from the content library, and sort them in descending order of content value index, select the top M short dramas to generate a candidate recommendation set;

[0121] Step 402, arrange the candidate recommendation set in descending order of the modified preference value of each theme, and generate a personalized recommendation list.

[0122] In an embodiment of the present application, the modified preference values (generated by step 302, range [0, 1]) of all themes are sorted in descending order to determine the user interest priority. For example, the modified preference value of a certain user is sorted as: Suspense (0.85) > Love (0.72) > Science fiction (0.61) > Comedy (0.55)…

[0123] Top N theme selection:

[0124] Set the threshold value N (such as N=3 or 5) according to business needs, and select the top N themes with the highest modified preference value as the theme subject. For example, when N=3, select Suspense, Love, and Science fiction as the theme subject, and ignore the subsequent low-priority themes.

[0125] Step 401, traverse the content library and filter all short dramas belonging to the theme subject. For example, the theme subjects are "Suspense", "Love", and "Science fiction", so only short dramas of these three themes are retained, and other themes are excluded.

[0126] For each short drama under each theme subject, sort them in descending order of content value index (calculated by step 103, which includes a comprehensive value of production quality, interactive heat, and commercial value). For example, there are 10 short dramas under the "Suspense" theme, with content value indexes of 90, 85, 80, etc. The top M (such as M=5) are selected after sorting. Combine the top M short dramas of each theme subject to form a candidate recommendation set. For example, when N=3 and M=5, the candidate set contains 5 Suspense, 5 Love, and 5 Science fiction, a total of 15 short dramas.

[0127] Step 402, group the short dramas in the candidate recommendation set by their theme, such as "Suspense group", "Love group", and "Science fiction group", and sort the short dramas in each group by content value index. Arrange the order of each group in descending order of the modified preference value of the theme subject. For example, the modified preference value is sorted as Suspense (0.85) > Love (0.72) > Science fiction (0.61), so the recommendation list first displays the 5 short dramas of the Suspense group, and then displays the short dramas of the Love group and the Science fiction group in turn. Combine the short dramas of each group to form the final personalized recommendation list, and the order of each short drama in the list satisfies both the "theme preference priority" and the "content value priority within the group". For example, the top 5 in the list are the top 5 in terms of content value in the Suspense theme, followed by the top 5 in terms of content value in the Love theme, and so on.

[0128] Through the double-layer mechanism of "amending the preference value to select the theme + content value index to select the short drama", the recommendation list is ensured to be consistent with the recent interests of the user (such as preferentially displaying high-preference themes), and the over-recommendation of low-quality content in the same theme is avoided, and the content diversity and user experience are improved. The content value index includes a commercial value weight, so that the recommendation list preferentially displays short dramas with high advertisement monetization potential while meeting the user's interests. According to industry simulation data, the strategy of filtering by theme first and then sorting by value reduces the invalid calculation amount (such as directly excluding low-preference themes), which is suitable for real-time recommendation requirements in a massive content scenario. The recommendation list is arranged in groups according to the theme preference intensity, which is consistent with the user's cognitive logic (such as viewing the most interesting theme first), reducing the sense of information overload during browsing; at the same time, the content value index ensures the production quality of the recommended content, reducing the user churn rate caused by poor content.

[0129] In a preferred embodiment of the present application, when selecting the target short drama in the recommendation list, the display of the incentive video advertisement is triggered in step 5, and after the user watches the incentive video advertisement, the target short drama content is automatically unlocked, and the unlocking behavior log can include:

[0130] Step 500, monitoring the user's selection operation on the target short drama in the personalized recommendation list, requesting an incentive video advertisement from an advertisement placement party according to the user portrait features;

[0131] Step 501, obtaining the matched incentive video advertisement and displaying it on the user interface, and tracking the advertisement playing progress in real time, when detecting that the advertisement is completely played and ended, unlocking the content access restriction of the target short drama, recording a data entry containing a timestamp, a user identifier, a short drama identifier, an advertisement identifier and an unlocking state; writing the data entry into the unlocking behavior log.

[0132] In an embodiment of the present application, an event listening mechanism is deployed in the recommendation list interface, when the user clicks the "watch" "unlock" and other interactive buttons of the target short drama, the selection operation is captured in real time, and the ID of the target short drama, the theme it belongs to and the current login state of the user are analyzed. For example, the user clicks the suspense theme short drama with ID D007 in the recommendation list, and the system immediately identifies the theme label T002 (suspense) of the short drama, the content value index 86 and the user ID U1024.

[0133] The historical behavior labels of the user are called from the user portrait database, including but not limited to:

[0134] Basic attributes: age, region, device type (such as iOS mobile phone);

[0135] Interest labels: recent high-frequency watched themes (such as suspense, science fiction), watching time (8-10 pm);

[0136] Consumption habits: number of unlocks in the past 30 days, average ad watch duration (e.g., average watch 45 seconds).

[0137] Incentivized ad request generation:

[0138] Pack user profile features and target short series attributes (genre, content value index) as request parameters and send them to the ad placement party (e.g., AdX platform) through API interface to request matching incentivized video ads. For example:

[0139] Request parameters include: user ID = U1024, interest tags = [mystery, science fiction], target short series genre = mystery, content value index = 86;

[0140] The ad placement party filters the ad library according to the parameters and returns ads that match the user's interests and are suitable for the short series genre (e.g., mystery product ads) first.

[0141] Step 501, receive the incentivized video ad returned by the ad placement party (e.g., MP4 format, duration 30-60 seconds), display it in full screen or pop-up form on the user interface, and start the play timer; when the user clicks "skip" or exits the ad page, pause the timer and mark it as "incomplete play"; when the ad play progress reaches 100% (i.e., complete play ends), trigger the unlock logic.

[0142] Content unlock automation execution:

[0143] After detecting the complete play of the ad, the system automatically removes the access restrictions of the target short series (e.g., decrypts the video file, unlocks the paid chapters), and returns a "unlock success" prompt to the user. For example, the encrypted video stream key of D007 short series is released, and the user can directly watch. Record standardized data entries, including the following fields:

[0144] Timestamp: accurate to milliseconds (e.g., 2025-06-26 14:30:25.003);

[0145] User identification: U1024 (encrypted unique ID);

[0146] Short series identification: D007, genre T002, content value index 86;

[0147] Ad identification: AdID = A12345, ad type = mystery, placement party = AdX;

[0148] Unlock status: success / failure (record play duration = 45 seconds when successful, record reason = user exits in the middle when failed).

[0149] Write data entries in chronological order to a distributed log system (e.g., HDFS or Elasticsearch).

[0150] Incentive advertising changes "forced push" to "active participation", users voluntarily watch ads to unlock content, while reducing user churn caused by ad interference. The ad request mechanism precisely matches the user portrait and the theme of the short drama, making the ad content more consistent with the user's interest (such as suspense short drama matching suspense product ads), and the user is more likely to continue consuming similar content after unlocking, forming a closed loop of "watching ads-unlocking content-more advertising value". The unlocking log records the interaction behavior of users and ads (such as playing time and exit reason), which provides precise optimization basis for advertisers (such as adjusting ad length and content creativity). The ad identifier and unlocking state recorded in the log can be used to identify abnormal operations (such as the same user unlocking the same theme at a high frequency in a short time), effectively preventing "brushing ads unlocking" and other cheating behaviors, and protecting the interests of advertisers and the platform.

[0151] In a preferred embodiment of the present application, the above step 6, real-time analysis of unlocking behavior log, identification of abnormal operation mode, generation of risk index; the risk index is used as a control instruction to dynamically adjust the configuration parameters of the business value weight, and the content value index is recalculated to realize the closed-loop optimization of the strategy, which can include:

[0152] Step 600, scanning the unlocking behavior log according to the preset time window, and counting the short drama unlocking frequency of the same user identifier and the ad triggering density of the same device;

[0153] Step 601, when the short drama unlocking frequency exceeds the dynamic threshold or the ad triggering density deviates from the reference range, a risk index containing an abnormal code is generated, and the ad conversion rate parameter and the delivery unit price parameter are adjusted according to the risk index level;

[0154] Step 602, using the adjusted ad conversion rate parameter and delivery unit price parameter to recalculate the business value weight of the affected short drama, and obtaining the updated business value weight;

[0155] Step 603, based on the updated business value weight, the content value index is recalculated, and the recalculated value index is fed back to the recommendation list generation process to complete the closed-loop optimization of the strategy.

[0156] In the embodiments of the present application, the unlocking behavior log is scanned in a preset time window (such as 5 minutes, 1 hour or 24 hours) in a cycle, which can be configured as a sliding window or a fixed window mode. For example, a 1-hour window is set, and the log data in the last 1 hour is scanned every 10 minutes. For each user identification (UserID), the number of short video unlocking times (i.e. unlocking frequency) in the window is counted; for each device identification (DeviceID), the number of advertisement triggering times (i.e. advertisement triggering density) is counted. For example, user U1024 unlocks short videos 5 times in 1 hour, and device D001 triggers advertisements 30 times in 24 hours. The data across the time window is smoothed (such as moving average) to avoid accidental peak interference with the statistical results. For example, the average unlocking frequency of the user in the last 3 time windows is calculated to determine whether the trend is abnormal.

[0157] In step 601, the threshold value is set according to the historical data, such as the average unlocking frequency of the user μ+2σ (μ is the mean value and σ is the standard deviation), or the normal range of the device advertisement triggering density is set according to the theme and user level, such as [10 times / 24 hours, 20 times / 24 hours], and the deviation from the range is regarded as abnormal.

[0158] Abnormality determination and risk index generation:

[0159] If the user unlocking frequency exceeds the dynamic threshold value (such as 3 times for an ordinary user in 1 hour), or the device advertisement triggering density is lower / higher than the reference range (such as ≤5 times or ≥40 times in 24 hours), it is determined as an abnormal operation mode.

[0160] The risk index is generated, including an abnormal code (such as “U001” representing high-frequency unlocking of the user), a risk level (low / medium / high) and an abnormal description (such as “user U1024 unlocks 5 times in 1 hour, exceeding the threshold value of 3 times”).

[0161] Dynamic adjustment of parameters:

[0162] According to the risk level, the configuration parameters of the commercial value weight are adjusted:

[0163] Low risk: advertisement conversion rate parameter is reduced by 5%-10%;

[0164] Medium risk: bid price parameter is reduced by 10%-20%, and conversion rate parameter is reduced by 15%;

[0165] High risk: bid price parameter is reduced by 20%-30%, and high-value advertisement delivery of the user / device is suspended.

[0166] Step 602, identify the theme or short drama related to abnormal operation. For example, the user frequently unlocks the "suspense" theme short drama, or the device frequently triggers the specific short drama ID of the advertisement. Using the adjusted advertisement conversion rate parameter and the delivery unit price parameter, recalculate the commercial value weight of the affected short drama. For example, the original delivery unit price is 50 yuan / thousand times, which is adjusted to 40 yuan / thousand times, and the original conversion rate is 3%, which is adjusted to 2.5%. The new commercial value weight = 40 x 3 x 0.025 = 3 (the original weight is 50 x 3 x 0.03 = 4.5). According to the linear superposition logic of step 103, use the updated commercial value weight to recalculate the content value index of the short drama.

[0167] Step 603, synchronize the recalculated content value index to the content library in real time, and update the sorting priority of the short drama. For example, the content value index of a certain short drama increases from 54.35 to 54.9, and the ranking in the recommendation list may rise. The recommendation list generation process of step 402 is reordered based on the new content value index, realizing the closed loop of "abnormality identification-parameter adjustment-recommendation optimization". For example, the short drama of high-risk theme reduces the priority in the recommendation list due to the decrease of content value index, and reduces the exposure.

[0168] By monitoring the unlocking frequency and the advertisement triggering density through dynamic threshold and benchmark range, abnormal behaviors such as "advertisement unlocking" and "device cheating" are found in time, and system resource waste and damage to the interests of advertisers are avoided. According to the risk indicators, the advertisement conversion rate and delivery unit price parameters are dynamically optimized, so that the commercial value weight is more consistent with the actual realization ability. For example, when the advertisement conversion rate of a certain theme decreases due to user resistance, the system automatically reduces its commercial value weight and reduces the inefficient recommendation. The closed loop feedback mechanism makes the content value index and the recommendation list updated in real time with market dynamics (user preference migration, advertiser budget adjustment), and the recommendation accuracy remains stable for a long time, avoiding the lag of traditional offline models. When the high-frequency unlocking is abnormal, the recommendation priority of the corresponding theme is reduced, and the user's resistance caused by excessive advertisement interference is reduced. At the same time, high-value advertisements are maintained for low-risk users, balancing platform revenue and user experience.

[0169] As shown in Figure 2 The embodiment of the present application also provides a system for personalized video advertisement display and content unlocking, comprising:

[0170] A value evaluation module is configured to obtain short drama content from a plurality of short drama operators, store the short drama content in the content library according to themes, and calculate a content value index of each short drama according to production quality, interactive heat value and commercial value weight;

[0171] A user behavior analysis module is configured to collect historical behavior data of users in the content library, such as short drama viewing time, theme selection frequency and content unlocking record, and calculate the initial preference intensity of the user for each theme;

[0172] The preference correction module is used for dynamically weighting the initial preference intensity by taking the content value index as a correction parameter, and generating a corrected preference value of each theme which is more in line with actual demand;

[0173] The personalized recommendation module is used for screening short drama content matched with the theme from the content library based on the corrected preference value of each theme, generating a candidate recommendation set, and generating a personalized recommendation list in the order from high to low of the corrected preference value;

[0174] The incentive advertisement unlocking module is used for triggering the display of an incentive video advertisement when the user selects a target short drama in the recommendation list, automatically unlocking the target short drama content after the user watches the advertisement, and generating an unlocking behavior log;

[0175] The strategy closed-loop optimization module is used for analyzing the unlocking behavior log in real time, identifying an abnormal operation mode and generating a risk index, taking the risk index as a regulation and control instruction, dynamically adjusting the configuration parameter of the business value weight, recalculating the content value index, and realizing the closed-loop optimization of the recommendation strategy.

[0176] Embodiments of the present application also provide a computer readable storage medium storing instructions which, when executed on a computer, cause the computer to perform the method described above. All implementation manners in the above method embodiments are applicable to this embodiment and can also achieve the same technical effects.

[0177] The above is the preferred embodiment of the present application. It should be noted that for those skilled in the art, without departing from the principles of the present application, a number of improvements and refinements can be made, which should also be considered within the scope of protection of the present application.

Claims

1. A method for displaying personalized video advertisements and unlocking content, characterized in that, The method includes: Step 1: Obtain short drama content from multiple short drama operators and store it in the content library according to genre. Calculate the content value index for each short drama based on production quality, interaction popularity, and commercial value weight. This includes: receiving short drama metadata from multiple operators, parsing genre tag codes, and storing them in the content library according to tag codes; reading the production parameter set from the short drama metadata, including resolution values, audio sampling rate, and frame rate sequence, to generate a production quality score; obtaining the user behavior record table for each short drama in the content library, extracting like counter values, comment counter values, share counter values, and completion rate percentage to generate an interaction popularity value; extracting the advertiser's unit price for the corresponding short drama, the number of ad slots, and historical conversion rate percentage based on the advertiser configuration table, and outputting the commercial value weight value; and linearly superimposing the production quality score, interaction popularity value, and commercial value weight value to output the content value index. Step 2: Collect users' historical behavior data from the content library, including short drama viewing time, theme selection frequency, and content unlocking records, and calculate the initial preference intensity of users for each theme based on the historical behavior data; Step 3: Use the content value index as a correction parameter to dynamically weight the initial preference intensity and generate the corrected preference value for each subject. Step 4: Based on the modified preference values ​​of each genre, select short drama content that matches the genre from the content library, generate a candidate recommendation set, and generate a personalized recommendation list in descending order of the modified preference values ​​of each genre. Step 5: When a target short drama is selected from the recommended list, a rewarded video ad is triggered. After the user watches the rewarded video ad, the target short drama content is automatically unlocked, and an unlocking behavior log is generated. This includes: monitoring the user's selection of a target short drama from the personalized recommended list; requesting a rewarded video ad from the advertiser based on the user profile characteristics; obtaining the matching rewarded video ad and displaying it on the user interface, and tracking the ad playback progress in real time; when the ad is detected to have finished playing completely, removing the content access restriction of the target short drama and recording data entries including timestamp, user ID, short drama ID, ad ID, and unlocking status; and writing the data entries to the unlocking behavior log. Step 6: Analyze unlocking behavior logs in real time, identify abnormal operation patterns, and generate risk indicators. Use these risk indicators as control commands to dynamically adjust the configuration parameters of the commercial value weight and recalculate the content value index to achieve closed-loop strategy optimization. This includes: scanning unlocking behavior logs according to a preset time window, statistically analyzing the unlocking frequency of short dramas for the same user identifier and the ad trigger density on the same device; when the unlocking frequency of short dramas exceeds a dynamic threshold or the ad trigger density deviates from the benchmark range, generating risk indicators containing abnormal codes, and adjusting ad conversion rate parameters and ad placement unit price parameters according to the risk indicator level; recalculating the commercial value weight of the affected short dramas using the adjusted ad conversion rate parameters and ad placement unit price parameters to obtain updated commercial value weights; recalculating the content value index based on the updated commercial value weights, and feeding the recalculated value index back to the recommendation list generation process to complete the closed-loop strategy optimization.

2. The method for personalized video ad display and content unlocking according to claim 1, characterized in that, Collect users' historical behavioral data from the content library, including short drama viewing time, frequency of genre selection, and content unlocking records. Calculate the initial preference strength of users for each genre based on this historical behavioral data, including: Monitor the user terminal's operation command flow in the content library, capture and parse three types of logs, including short drama playback logs, theme click logs and unlock request logs; The operation logs are grouped by topic tag code to generate the total topic duration value, topic click count, and number of successful unlocks; The output duration ratio coefficient, selection frequency coefficient, and unlock conversion coefficient are determined based on the total duration of the subject matter, the number of clicks on the subject matter, and the number of successful unlocks. The initial preference intensity vector is determined based on the duration ratio coefficient, selection frequency coefficient, and unlock conversion coefficient.

3. The method for personalized video ad display and content unlocking according to claim 2, characterized in that, Operation logs are grouped by topic tag code, generating total topic duration, topic click count, and number of successful unlocks, including: Perform a theme tag aggregation operation on the short drama playback logs, accumulate the total playback time of all short dramas under the same theme tag, and output the total theme duration value; Perform a frequency counting algorithm on the topic click logs to count the number of times users actively click on each topic tag and output the topic click count; Perform a status filtering operation on the unlock request log, extract unlock records with a successful verification status, count the number of successful unlocks under each theme tag, and output the number of successful unlocks.

4. The method for personalized video ad display and content unlocking according to claim 3, characterized in that, Using the content value index as a correction parameter, the initial preference intensity is dynamically weighted to generate corrected preference values ​​for each subject matter, including: Extract the content value index of each short drama genre from the content library and generate a genre value vector; The initial preference intensity vector and the subject value vector are fused to obtain a weighted preference vector; The weighted preference vector is normalized, and the corrected preference values ​​for each subject are output.

5. The method for personalized video ad display and content unlocking according to claim 4, characterized in that, Based on the modified preference values ​​for each genre, short drama content matching the genre is selected from the content library to generate a candidate recommendation set. A personalized recommendation list is then generated, ordered from highest to lowest modified preference value for each genre, including: The top N themes are selected as target themes based on the modified preference values ​​of each theme. Short dramas belonging to the target theme are selected from the content library and sorted from high to low according to the content value index. The top M short dramas in the sorting are selected to generate a candidate recommendation set. The candidate recommendation set is sorted from high to low according to the modified preference value of each subject, and a personalized recommendation list is generated.

6. A system for personalized video ad display and content unlocking, the system implementing the method as described in any one of claims 1 to 5, characterized in that, include: The value assessment module is used to obtain short drama content from multiple short drama operators, store it in the content library according to the theme, and calculate the content value index of each short drama based on production quality, interaction popularity value and commercial value weight. The user behavior analysis module is used to collect users' short drama viewing time, subject selection frequency, and content unlocking history data in the content library, and calculate the initial preference intensity of users for each subject. The preference correction module is used to dynamically weight the initial preference intensity by using the content value index as a correction parameter, and generate a corrected preference value for each subject to better meet actual needs. The personalized recommendation module is used to filter short drama content that matches the theme from the content library based on the modified preference value of each theme, generate a candidate recommendation set, and generate a personalized recommendation list in descending order of modified preference value. The incentivized ad unlocking module is used to trigger the display of incentivized video ads when a user selects a target short drama from the recommended list. After the user watches the ad, the target short drama content is automatically unlocked, and an unlocking behavior log is generated. The strategy closed-loop optimization module is used to analyze unlocking behavior logs in real time, identify abnormal operation patterns and generate risk indicators. These risk indicators are then used as control instructions to dynamically adjust the configuration parameters of the commercial value weight, recalculate the content value index, and achieve closed-loop optimization of the recommendation strategy.

7. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a program that, when executed by a processor, implements the method as described in any one of claims 1 to 5.

Images