Method, system, electronic device and storage medium for dynamically optimizing an advertisement script

Abstract

The application provides a method, system, electronic device and storage medium for dynamically optimizing an advertisement script, and relates to the technical field of digital marketing. In the method, behavior data of a target user is obtained, and emotion analysis is performed on the behavior data to obtain an emotional state of the target user. Based on the emotional state, a script module is selected from a preset script material library, and a personalized advertisement script matching the emotional state is generated. The personalized advertisement script is displayed to the target user, and the attention behavior indicators of the target user are monitored in real time. Based on the attention behavior indicators and a preset intervention node, attention intervention compensation is triggered on the personalized advertisement script based on the emotional state to generate a push advertisement script. The conversion results of the push advertisement script are counted, and the personalized advertisement script, the push advertisement script, the compensation records corresponding to the push advertisement script and the conversion results are returned to data to enable automatic updating of the script module and improve the advertisement conversion rate.

Description

Technical Field

[0001] This application relates to the field of digital marketing technology, and specifically to a method, system, electronic device, and storage medium for dynamically optimizing advertising scripts. Background Technology

[0002] In the current digital advertising ecosystem, such as mobile feed ads or social media video ads, the core technical challenge facing ad delivery platforms is how to capture users' attention and effectively deliver information within a very short time window (usually 3-8 seconds) in order to improve ad click-through rates and conversion rates.

[0003] In related technologies, A / B testing or rule-based templates are commonly used to adjust ad placement. A / B testing is an offline optimization strategy. Multiple sets of static ad scripts (combinations of text, images, and videos) are manually pre-set and delivered to different user groups. After a relatively long data collection period (e.g., 24-48 hours), the conversion rate of each group is statistically analyzed, and the optimal group is selected for large-scale deployment. Rule-based templates use pre-defined rules and templates to fill in static user tags (such as region and gender) into the variables of the ad scripts, achieving basic personalization. Both rely on predefined, fixed ad script content. A / B testing involves choosing from multiple static scripts, while rule-based templates make limited parameter adjustments to static templates. This reliance on static content results in ad scripts with limited personalization dimensions, leading to low ad conversion rates. Summary of the Invention

[0004] To alleviate the problem of low conversion rates caused by the lack of personalization in static advertisements, this application provides a method, system, electronic device, and storage medium for dynamically optimizing advertisement scripts.

[0005] The technical solution of this application embodiment is as follows:

[0006] In a first aspect, embodiments of this application provide a method for dynamically optimizing advertising scripts, the method comprising:

[0007] Obtain behavioral data of the target user, perform sentiment analysis on the behavioral data, and obtain the emotional state of the target user;

[0008] Based on the emotional state, a script module is selected from a preset script material library to generate a personalized advertising script that matches the emotional state. The script material includes a variety of emotional materials and the script module corresponding to the emotional material. The emotional state matches one or more of the emotional materials.

[0009] The personalized advertising script is displayed to the target user, and the target user's attention behavior metrics are monitored in real time.

[0010] Based on the attention behavior indicators and preset intervention nodes, attention intervention compensation is triggered on the personalized advertising script based on the emotional state to generate push advertising script, so as to improve the target user's attention to the push advertising script.

[0011] The conversion results of the push advertising script are statistically analyzed, and the personalized advertising script, the push advertising script, the compensation record corresponding to the push advertising script, and the conversion results are transmitted back to the script module so as to automatically update the script module.

[0012] In the above technical solution, behavioral data of the target user is acquired, and sentiment analysis is performed on the behavioral data to obtain the target user's emotional state. By acquiring and analyzing behavioral data, preparation is made for adding emotions in the subsequent generation of personalized advertising scripts to improve the relevance of the advertisements to the user. Based on the emotional state, a script module is selected from a preset script material library to generate a personalized advertising script that matches the emotional state. The script materials include various emotional materials and corresponding script modules. The emotional state matches one or more of the emotional materials. By introducing the emotional state into the advertising script generation process, and by paying attention to the target user's emotions, advertisements that may cause user aversion are avoided. Personalized advertisements are then displayed to the target user. The system generates advertising scripts and monitors target users' attention behavior metrics in real time for dynamic adjustments. Based on these metrics and preset intervention points, it triggers attention intervention compensation for personalized advertising scripts based on emotional states, generating push advertising scripts to increase target users' attention to these scripts. Through attention intervention compensation, it achieves dynamic ad pushes, attracting user attention and increasing user engagement with the ads. The system also tracks the conversion results of push advertising scripts and feeds back data on personalized advertising scripts, push advertising scripts, corresponding compensation records, and conversion results. This allows for automatic script module updates and, through autonomous dynamic learning, provides a foundation for subsequent real-time dynamic responses.

[0013] The above process generates personalized ad scripts by incorporating the target user's emotions, ensuring the scripts align with the user's psychological needs, stimulating interest and emotional resonance, and encouraging engagement. Combined with attention monitoring and attention intervention compensation, the ad scripts are dynamically adjusted to attract the target user's attention, and through self-learning, prepare for subsequent real-time dynamic responses. This collaborative approach reduces reliance on static content, breaks away from single-dimensional personalization, and ultimately improves ad conversion rates.

[0014] In some embodiments of this application, performing sentiment analysis on the behavioral data to obtain the emotional state of the target user includes:

[0015] The behavioral data is tensorized to obtain a behavioral vector;

[0016] The behavior vector is processed using a pre-defined emotion analysis model to obtain an emotion category vector;

[0017] The emotion category vector is subjected to probability transformation to obtain a probability distribution. The probability distribution is then mapped to a preset emotion label to form an emotion vector. The emotion vector includes the emotion label and the probability value corresponding to the emotion state.

[0018] The probability value in the probability distribution is used as the confidence value, and the emotion label corresponding to the largest confidence value is selected as the emotion state.

[0019] In some embodiments of this application, the step of selecting a script module from a preset script material library based on the emotional state to generate a personalized advertising script that matches the emotional state includes:

[0020] The emotional state is matched with the emotional material in the script material library, and the script module corresponding to the matched emotional material is combined in multiple modes to generate multiple initial opening layers;

[0021] Obtain the target user's historical behavior data, analyze the historical behavior data, and obtain user preferences;

[0022] By utilizing the user preferences and emotional states, each of the initial opening layers is assembled to obtain multiple claim layers;

[0023] Click-through rate (CTR) prediction is performed on each of the claim layers, and the claim layer with a CTR greater than a preset evaluation threshold is selected as the personalized advertising script.

[0024] In some embodiments of this application, the assembly of each initial opening layer using the user preferences and the emotional state to obtain multiple claim layers includes:

[0025] The user preferences and emotional state are used as inputs, and a large language model is used to generate multiple candidate script modules. The candidate script modules are then combined with the user preferences to form an assembly set.

[0026] The candidate script modules in the assembly set are combined with the preset assembly strategy to form multiple claim documents;

[0027] The emotional state is matched with emotional words in a preset emotional expression library to determine the strategy sentence pattern corresponding to the emotional state;

[0028] The various claim statements are adjusted and encapsulated using the aforementioned strategy and sentence structure to form the claim layer.

[0029] In some embodiments of this application, the step of triggering attention intervention compensation for the personalized advertising script based on the attention behavior indicators and preset intervention nodes, and generating a push advertising script, includes:

[0030] Before the intervention node, if the attention score in the attention behavior index is less than the preset attention threshold, multiple material scripts corresponding to the emotional state are searched in the preset compensation strategy library, and the material scripts are combined and packaged to obtain compensation material.

[0031] The personalized advertising script is then subjected to attention intervention compensation using the compensation material to generate the push advertising script.

[0032] In some embodiments of this application, the step of triggering attention intervention compensation for the personalized advertising script based on the attention behavior indicators and preset intervention nodes, and generating a push advertising script, includes:

[0033] Upon reaching the intervention node, the current emotion corresponding to the intervention node is obtained, and it is determined whether the emotional state is consistent with the current emotion.

[0034] In the event of inconsistency, the current emotion is taken as the emotional state, and multiple material scripts corresponding to the current emotion are searched in the preset compensation strategy library. The material scripts are combined and packaged to obtain compensation material.

[0035] The personalized advertising script is then subjected to attention intervention compensation using the compensation material to generate the push advertising script.

[0036] In some embodiments of this application, the step of using the compensation material to perform attention intervention compensation on the personalized advertising script to generate the push advertising script includes:

[0037] The personalized advertising script is divided into video frames, and a candidate switching area is defined between the video frame corresponding to the preset turning point and the next video frame. The preset turning point is determined by the intervention node and the attention threshold.

[0038] The compensation material is switched to the candidate switching area, and a smooth animation is added to the video frame corresponding to the turning point. The volume of the compensation material is adjusted to match the volume of the personalized advertising script to generate the push advertising script.

[0039] Secondly, embodiments of this application provide a system for dynamically optimizing advertising scripts, the system comprising:

[0040] The data acquisition and preprocessing module is used to acquire behavioral data of the target user, perform sentiment analysis on the behavioral data, and obtain the emotional state of the target user.

[0041] A personalized script generation module is used to select a script module from a preset script material library based on the emotional state and generate a personalized advertising script that matches the emotional state. The script material includes a variety of emotional materials and the script module corresponding to the emotional material. The emotional state matches one or more of the emotional materials.

[0042] An attention monitoring module is used to display the personalized advertising script to the target user and monitor the target user's attention behavior metrics in real time.

[0043] The script adjustment module is used to trigger attention intervention compensation on the personalized advertising script based on the attention behavior indicators and preset intervention nodes, based on the emotional state, to generate push advertising scripts, so as to improve the target user's attention to the push advertising scripts.

[0044] The self-learning module is used to statistically analyze the conversion results of the push advertising script, and to transmit the personalized advertising script, the push advertising script, the compensation record corresponding to the push advertising script, and the conversion results back to the script module so as to automatically update the script module.

[0045] Thirdly, embodiments of this application provide an electronic device including a processor, a memory, a user interface, a communication bus, and a network interface. The processor, the memory, the user interface, and the network interface are respectively connected to the communication bus. The memory is used to store instructions. The user interface and the network interface are used to communicate with other devices. The processor is used to execute the instructions stored in the memory to cause the electronic device to perform the method described in any one of the first aspects.

[0046] Fourthly, embodiments of this application provide a computer-readable storage medium storing instructions that, when executed, perform the method described in any one of the methods provided in the first aspect above.

[0047] In summary, one or more technical solutions provided in the embodiments of this application have at least the following technical effects or advantages:

[0048] 1. By incorporating target user emotions to generate personalized ad scripts, these scripts align with users' psychological needs, stimulating their interest and emotional resonance, and encouraging their participation in ad interactions. Furthermore, by combining attention monitoring with attention intervention and compensation, the ad scripts are dynamically adjusted to attract target user attention and, through self-learning, prepare for subsequent real-time dynamic responses. This collaborative approach reduces reliance on static content, breaks away from single-dimensional personalization, and improves ad conversion rates. Therefore, it effectively solves the problems of low conversion rates caused by reliance on static content and the limited personalization of generated ad scripts in related technologies.

[0049] 2. Based on the probability values ​​in the probability distribution, by selecting the emotion tag corresponding to the probability with a higher confidence level as the emotion state, the main emotion of the target user can be generated, which makes it easier for the personalized advertising script generated based on the emotion state to be more suitable for the target user.

[0050] 3. By using a multi-layered personalized ad script generation mode and embedding user preferences within it, ads become more aligned with user needs, increasing user attention to the ads.

[0051] 4. At different nodes, attention is triggered to intervene and compensate, enabling dynamic adjustment of the advertising script. This breaks the limitations of static content and makes the advertising script richer by combining emotional states. Attached Figure Description

[0052] Figure 1 This is a flowchart illustrating a method for dynamically optimizing advertising scripts according to an embodiment of this application;

[0053] Figure 2 This is a schematic diagram of the initial opening layer formation of a method for dynamically optimizing advertising scripts provided in one embodiment of this application;

[0054] Figure 3 This is a flowchart illustrating a method for dynamically optimizing advertising scripts according to another embodiment of this application;

[0055] Figure 4 This is a flowchart illustrating a method for dynamically optimizing advertising scripts provided in another embodiment of this application;

[0056] Figure 5 This is a schematic diagram of the structure of a system for dynamically optimizing advertising scripts according to an embodiment of this application;

[0057] Figure 6This is a schematic diagram of the structure of an electronic device provided in one embodiment of this application. Detailed Implementation

[0058] To enable those skilled in the art to better understand the technical solutions in this specification, the technical solutions in the embodiments of this specification will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments.

[0059] In the description of the embodiments of this application, the words "for example" or "for instance" are used to indicate examples, illustrations, or explanations. Any embodiment or design that is described as "for example" or "for instance" in the embodiments of this application should not be construed as being more preferred or advantageous than other embodiments or design options. Rather, the use of the words "for example" or "for instance" is intended to present the relevant concepts in a specific manner.

[0060] In the description of the embodiments of this application, the term "multiple" means two or more. For example, multiple systems means two or more systems, and multiple screen terminals means two or more screen terminals. Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the indicated technical features. Thus, a feature defined with "first" or "second" may explicitly or implicitly include one or more of that feature. The terms "comprising," "including," "having," and variations thereof all mean "including but not limited to," unless otherwise specifically emphasized.

[0061] Among related technologies, A / B testing has high latency, requiring 24-48 hours to converge, thus missing the window of trending sentiment. While rule-based template solutions merely replace fields, they cannot deeply match user needs. Simply modifying these solutions for remarketing results in significant attention loss and a broken closed-loop mechanism, failing to match advertising content with real-time user needs.

[0062] Based on this, embodiments of this application provide a method, system, electronic device, and readable storage medium for dynamically optimizing advertising scripts. The method first acquires behavioral data of the target user, performs sentiment analysis on the behavioral data to obtain the target user's emotional state, and prepares for adding emotions to the subsequent personalized advertising script generation to improve the relevance of the advertisement to the user. Based on the emotional state, a script module is selected from a preset script material library to generate a personalized advertising script that matches the emotional state. The script materials include various emotional materials and corresponding script modules. The emotional state matches one or more of the emotional materials. By introducing the emotional state into the advertising script generation process, and by focusing on the target user's emotions, the method avoids... This approach avoids pushing ads that might annoy users; it displays personalized ad scripts to target users and monitors their attention behavior metrics in real time for dynamic adjustments; based on these metrics and preset intervention points, it triggers attention intervention compensation for personalized ad scripts based on emotional states, generating push ad scripts to increase target users' attention to these scripts. Through this attention intervention compensation, it achieves dynamic ad delivery, attracting user attention and increasing user engagement with the ads; it statistically analyzes the conversion results of push ad scripts, transmitting data from personalized ad scripts, push ad scripts, corresponding compensation records, and conversion results back to the system for automatic script module updates. Through autonomous dynamic learning, it provides a foundation for subsequent real-time dynamic responses.

[0063] It should be noted that this method of dynamically optimizing ad scripts is used for digital advertising, including embedded ads in news feeds, social media video ads, and embedded ads displayed on web pages. By adding target user emotions to generate personalized ad scripts, these scripts align with users' psychological needs, stimulating their interest and emotional resonance, and encouraging them to engage with the ads. Combined with attention monitoring and attention intervention compensation, the ad scripts are dynamically adjusted to attract target user attention and, through self-learning, prepare for subsequent real-time dynamic responses. Based on the collaborative processes described above, this dynamic adjustment method reduces reliance on static content, breaks away from single-dimensional personalized generation, and improves ad conversion rates.

[0064] The technical solutions provided in the embodiments of this application will be further described below with reference to the accompanying drawings.

[0065] Reference Figure 1 , Figure 1This is a flowchart illustrating a method for dynamically optimizing advertising scripts provided in an embodiment of this application. The method for dynamically optimizing advertising scripts is applied to a system for dynamically optimizing advertising scripts. The method is executed by a processor in an electronic device or a readable storage medium. The method includes steps S100, S200, S300, S400, and S500.

[0066] Step S100: Obtain the target user's behavioral data, perform sentiment analysis on the behavioral data, and obtain the target user's emotional state.

[0067] In one embodiment, the target users are multiple users who can view the displayed advertisements. Behavioral data includes the target users' search behavior on web pages, their behavior while browsing software programs, and their commenting behavior on displayed content. For example, behavioral data may include browsing text content: the title and summary of the page currently being viewed (parsed via DOM); search behavior: search keywords within the last 3 minutes; and interaction behavior: liking, commenting, and sharing specific types of content. By deploying a client SDK (embedded in a web page or application) on the user's terminal, the above-mentioned behavioral data of the target users is collected and saved, packaged into a JSON object for easy subsequent data retrieval. This JSON object contains the user ID, timestamp, behavior type, and behavioral data (including text content and keywords). A preset file reading function is used to obtain the target users' behavioral data, providing data support for subsequent behavioral data analysis. The file reading function can be a read() function, an open() function, etc. Since static user tags (such as age and geolocation) are low-frequency and coarse-grained, they cannot reflect the user's true intentions and emotional state at a specific moment. By acquiring the behavioral data of target users, we can monitor their real-time behavior and provide timely data input for all subsequent personalized decisions.

[0068] In one embodiment, sentiment analysis is performed on behavioral data to obtain the emotional state of the target user, including but not limited to the following steps:

[0069] Step S110: Tensor quantization is performed on the behavior data to obtain the behavior vector.

[0070] In one embodiment, based on the JSON format of the aforementioned behavioral data, the target user ID and behavioral data are extracted. Then, the text content and keywords (including search terms, summaries, and titles) are extracted from the behavioral data. The method for extracting these fields is not detailed here. The extracted target user ID and behavioral data are then vectorized. The transformed vectors are concatenated; specifically, based on the target user ID, the text content and keyword content in the behavioral data are concatenated to achieve tensor quantization, forming the input data format and obtaining the behavioral vector.

[0071] It should be noted that the spliced ​​tensor data can also undergo data preprocessing, including data normalization, data imputation, and data cleaning, to obtain behavior vectors. Data preprocessing improves data quality and further enhances the accuracy of subsequent sentiment analysis.

[0072] Step S120: Use a preset emotion analysis model to perform feature processing on the behavior vector to obtain an emotion category vector.

[0073] In one embodiment, the preset emotion analysis model can be a fine-tuned RoBERTa model, which is adapted to the format of the aforementioned behavior vectors. During feature processing of the behavior vectors, the raw, unactivated output vector of the classification layer is extracted from the output of the RoBERTa model. This vector is N, where N represents predefined emotion category data, exemplarily including (anxiety, curiosity, excitement, nature, etc.). The resulting emotion category vector is used to subsequently determine the emotional state.

[0074] Step S130: Perform probability transformation on the emotion category vector to obtain a probability distribution, and map the probability distribution to preset emotion labels to form an emotion vector. The emotion vector includes the probability value corresponding to the emotion label and the emotion state.

[0075] In one embodiment, the emotion category vector is input into a softmax function to convert it into a probability distribution. This results in a probability distribution vector corresponding to the emotion category vector. The probability distribution is then mapped to emotion labels to form an emotion vector. Preset emotion labels include anxiety, curiosity, excitement, and naturalness. Since the probability values ​​in the probability distribution correspond to the emotion categories in the emotion category vector, and the emotion category vector corresponds to the emotion labels, each emotion label corresponds to a probability value through mapping. The emotion vector includes the probability values ​​corresponding to the emotion labels and the emotion states. The emotion vector is represented in key-value pairs so that the emotion state can be determined subsequently based on these key-value pairs. For example, the emotion vector can be represented as {"anxiety": 0.85, "curiosity": 0.1}.

[0076] Step S140: Use the probability values ​​in the probability distribution as confidence values, and select the emotion label corresponding to the maximum confidence value as the emotion state.

[0077] In one embodiment, each probability value in the probability distribution represents the likelihood of being classified as a certain emotion category. The higher the probability value, the greater the likelihood of being classified as that emotion. The probability values ​​in the probability distribution are used as confidence values. These confidence values ​​can be sorted using methods such as bubble sort or quick sort. The emotion label corresponding to the highest confidence value is selected as the emotion state, indicating a higher degree of confidence in classifying the emotion category, thus reflecting the target user's current emotional state. For example, the emotion vector can be represented as {"anxiety": 0.85, "curiosity": 0.1}. Using 0.85 and 0.1 as confidence values, selecting the emotion label corresponding to the highest value of 0.85 yields the target user's emotional state as anxiety.

[0078] Step S200: Based on the emotional state, select a script module from the preset script material library to generate a personalized advertising script that matches the emotional state. The script material includes a variety of emotional materials and script modules corresponding to the emotional materials. The emotional state matches one or more of the emotional materials.

[0079] In one embodiment, the script materials include various emotion materials and corresponding script modules. Emotional states match one or more of the emotion materials. Both the emotion materials and script modules are predefined templates. Emotional materials indicate different emotion categories, such as anger, anxiety, and curiosity. Each emotion corresponds to one emotion material, which includes an emotion field consistent with the emotional state, as well as related elements (visual, text, sound effects, etc.) to facilitate matching. The script module generates an advertising script for each emotion material. For example, for anxiety, the corresponding advertising script could be "Don't let the uncertainty of the future affect your quality of life today," or it could be a motivational script, such as "Smile, no worries." The script material library provides a rich set of modules, enabling the rapid generation of personalized advertising scripts through matching.

[0080] In one embodiment, based on emotional state, a script module is selected from a preset script material library to generate a personalized advertising script that matches the emotional state, including but not limited to the following steps:

[0081] Step S210: Match the emotional state with the emotional material in the script material library, and combine the script modules corresponding to the matched emotional material in multiple modes to generate multiple initial opening layers.

[0082] In one embodiment, a preset similarity algorithm is used to calculate the similarity between emotional states and emotional content, obtaining a similarity value. This preset similarity algorithm can be a text similarity algorithm, a cosine similarity algorithm, etc. If the similarity value is greater than a preset similarity threshold, an emotional content has been matched. The script modules corresponding to the matched emotional content are randomly combined, or they can be combined in various different ways according to scene changes. In the scene-change combination method, each script module is assigned a scene number to form a more continuous and smooth advertising script. The results of each combination form multiple initial opening layers, allowing for subsequent adjustments based on these initial opening layers to generate personalized advertising scripts.

[0083] For example, emotional material can be represented as anger: {"Visual element": "red background", "Text feature": "dense exclamation marks / short sentences", "Sound effect suggestion": "deep drumbeat"}, and anxiety: {"Visual element": "gray ripples", "Text feature": "interrogative sentence / ellipsis", "Sound effect suggestion": "high-frequency electronic sound"}. Script modules include {"Relaxation": "Smile, no worries", "Philosophical": "Don't let the uncertainty of the future affect your quality of life today", "Action-oriented": "Take three deep breaths now and feel the power of the present moment", and "Interactive": "When was the last time you felt relaxed?"}.

[0084] like Figure 2 As shown, the initial opening layer can be, for example, matching the emotional state as anxiety. Visuals: gray wavy background; text features: interrogative sentences + ellipses; sound effects: high-frequency electronic sounds; script selection: philosophical and action-oriented combinations. The combined result is: When the hourglass time makes you feel suffocated, don't let the uncertainty of the future affect the quality of life today. Close your eyes and take a deep breath (4 seconds of inhalation - 7 seconds of holding your breath - 8 seconds of exhalation), write down three small things you can control at this moment, and tell yourself: "I have done very well."

[0085] Step S220: Obtain historical behavior data of the target user, analyze the historical behavior data, and obtain user preferences.

[0086] In one embodiment, historical behavior data includes the target user's search behavior on web pages, their behavior while browsing software programs, and their commenting behavior on displayed content, recording past actions. For example, historical behavior data may include browsing text content: the title and summary of the page currently being viewed (parsed via DOM); search behavior: search keywords within the last 3 minutes; and interaction behavior: liking, commenting, and sharing specific types of content. By deploying a client SDK (embedded in a web page or application) on the user's terminal, the aforementioned historical behavior data of the target user is collected and saved, packaged into a JSON object for easy subsequent data retrieval. This JSON object contains the user ID, timestamp, behavior type, and behavior data (including text content and keywords). A preset file reading function is used to obtain the target user's historical behavior data, providing data support for subsequent historical behavior data analysis. The file reading function can be a read() function, an open() function, etc. Data preprocessing can also be performed on the historical behavior data, including data normalization, missing value imputation, and data cleaning, to improve data quality and accuracy.

[0087] Then, a pre-defined deep learning model is used to analyze historical behavioral data to obtain user preferences. This deep learning model combines a text encoder and a temporal modeling layer. The text encoder uses the BERT model to extract text content, and then the Transformer Encoder is used as the temporal modeling layer to capture long-term dependencies in behavioral sequences, perform interest classification prediction, and output user preferences. User preferences can be represented as price sensitivity, quality pursuit, social identification, efficiency improvement, etc. This allows for subsequent script assembly based on user preferences to attract the attention of target users.

[0088] Step S230: Using user preferences and emotional states, assemble each initial opening layer to obtain multiple claim layers.

[0089] Specifically, by utilizing user preferences and emotional states, each initial opening layer is assembled to obtain multiple claim layers, including but not limited to the following steps:

[0090] Step S231: Using user preferences and emotional state as input, generate multiple candidate script modules using a large language model, and combine the candidate script modules with user preferences to form an assembly set.

[0091] In some possible embodiments of this application, user preferences and emotional states are processed into feature vectors separately. These processed feature vectors are then concatenated and fused, serving as input to a large language model. The large language model extracts features and, based on the dual influence of preferences and emotions, automatically generates multiple candidate script modules. These candidate script modules are then used to generate the advocacy copy. Each candidate script module is an advertising script that combines user preferences and emotional states and closely aligns with user habits. User preferences and candidate scripts are combined, with a corresponding relationship between user preferences and the candidate script modules output by the model. These are grouped into a record. Multiple candidate script modules can be output, forming multiple records. These records form an assembly set for subsequent matching to obtain the advocacy copy.

[0092] In other possible embodiments of this application, user preferences and emotional states are concatenated and fused to obtain fused labels. These fused labels are used as columns, and the initial opening layer as rows, forming a two-dimensional matrix. A Cartesian product is calculated on this two-dimensional matrix to obtain an assembly matrix. Each element in the assembly matrix represents a candidate script module. An assembly set is formed by combining user preferences with their corresponding candidate script modules. By constructing the assembly set, user preferences, emotional states, and script modules are combined, ensuring that the subsequently generated claim text meets user needs.

[0093] Step S232: Combine the candidate script modules in the assembly set with the preset assembly strategy to form multiple claim documents.

[0094] In some possible embodiments of this application, the preset assembly strategy library includes user preferences, assembly strategies corresponding to user preferences, and strategy descriptions. This assembly strategy is automatically generated by statistically analyzing a large amount of historical data, utilizing strategy expressions and core point summaries generated using a large language model, and combining them with user preferences. The user preferences in the assembly matrix are matched with the user preferences in the assembly strategies. Matching can be calculated using a similarity algorithm, which will not be elaborated here. If a match is successful, the candidate script module corresponding to the successfully matched user preference is selected and combined with the assembly strategy in the assembly strategy library to form multiple claim texts. This combination can be exhaustive or user preference-oriented, selecting assembly strategies that match the user's emotional state and tone, and then randomly combining them with candidate script modules. Obtaining the claim texts prepares for the subsequent formation of the claim layer.

[0095] It should be noted that identifiers can also be set for the beginning, end, and transition points of the claim text, and these can be marked in the assembly strategy library to meet the needs of rapid matching and generation.

[0096] Step S233: Match the emotional state with the emotional words in the preset emotional expression library to determine the strategy sentence pattern corresponding to the emotional state.

[0097] In some possible embodiments of this application, the emotion expression library includes emotion words, corresponding tone tendencies, and sentence examples. The emotion words can be words such as anger, anxiety, and curiosity. This emotion expression library is obtained by statistically analyzing a large amount of historical data and using a large language model to predict and generate vocabulary and sentence patterns. Emotional states are matched with emotion words, and a similarity calculation is performed. A successful match is achieved when the similarity value is greater than a preset similarity threshold. A strategy sentence pattern corresponding to the emotional state is selected from the emotion expression library to prepare for the subsequent formation of the claim layer. The similarity calculation uses a preset similarity algorithm, such as a text similarity algorithm or a cosine similarity algorithm; the preset similarity threshold is set to a value between 90% and 95%.

[0098] Step S234: Adjust and encapsulate each claim text using strategic sentence structures to form a claim layer.

[0099] In some possible embodiments of this application, strategic phrases are used to replace the beginning, end, and keyword descriptions of each claim document to adjust the claim copy. Specifically, the adjustment process involves identifying the beginning, transition points, and endings of the claim copy, which can be determined by inserting identifiers into the claim copy. Then, the strategic phrases replace the identifiers in the claim copy, resulting in the adjusted copy. This method is fast and convenient, and can better meet the real-time requirements of advertising. The adjusted copy is then formatted to form an advertising script format, resulting in the claim layer, for subsequent evaluation of each claim layer.

[0100] For example, in a state of anxiety, the interjections generated in step S233 are empathy, reassurance, and reassurance, and the strategic sentence structure is "Don't worry, say goodbye to unhappiness." The advocacy copy aims to simplify complexity, starting with the anxiety caused by inefficiency and introducing efficient and convenient solutions. Through the above process, the adjusted copy can be "Don't worry, the anxiety caused by inefficiency is simplified, and can be alleviated through efficient solutions; say goodbye to unhappiness." By making these adjustments, the target user's emotions can be soothed or empathized with, thereby attracting the user's attention to the advertising script.

[0101] Step S240: Perform click-through rate prediction and evaluation on each claim layer, and select the claim layer with a predicted click-through rate greater than the preset evaluation threshold as the personalized advertising script.

[0102] In one embodiment, a pre-defined prediction model is used to predict and evaluate the click-through rate (CTR) of each ad claim layer, resulting in a predicted CTR. The pre-defined prediction model can be a machine learning model or a deep learning model. The machine learning model can be a regression model, a factorization machine, etc., while the deep learning model can be an attention-based Transformer model, or a combination of machine learning and deep learning models. Taking a deep learning model as an example, an attention-based Transformer model is used for prediction and evaluation. Extensive web scraping algorithms are used to collect large amounts of user behavior data, analyze user preferences and emotional states, and analyze ad footers viewed and clicked by users. This data is fed into the attention-based Transformer model, and through continuous iterative training, a trained deep learning model is obtained. This model is then used as the prediction model for CTR prediction and evaluation, resulting in a predicted CTR. Using a prediction model for evaluation not only improves the accuracy of CTR evaluation but also increases script generation efficiency.

[0103] Based on the predicted click-through rate (CTR) obtained above, it is compared with a preset evaluation threshold. Since a higher CTR indicates greater user interest in the generated copy, the ad message layer corresponding to the CTR exceeding the evaluation threshold is selected as the personalized ad script. The preset evaluation threshold can be 80%. It should be noted that there may be multiple ad message layers corresponding to CTRs exceeding the preset evaluation threshold. The ad message layer with the highest CTR is selected as the personalized ad script to better align with user preferences.

[0104] It should also be noted that if the predicted click-through rate is not greater than the preset evaluation threshold, the claim layer is regenerated according to the above steps, and click-through rate prediction is performed again until there is a predicted click-through rate greater than the evaluation threshold.

[0105] A complete advertisement comprises three elements: "hook," "content," and "action." Deconstructing the script into three independently optimizable layers and dynamically combining them through algorithms exponentially increases the possibilities for creative combinations while ensuring that each part is specifically optimized. This "divide and conquer, recombine" strategy enables large-scale, deeply personalized, and highly efficient optimization.

[0106] Step S300: Display personalized advertising scripts to target users and monitor target users' attention behavior metrics in real time.

[0107] In one embodiment, a generated personalized advertising script is sent to a front-end page via an interface for user viewing. During viewing, the client's built-in camera or a recording tool can be used to capture the user on the front-end page. The captured video is then analyzed using a facial recognition model to determine the target user's head posture and gaze direction. Based on the correspondence between video frames and gaze directions, key information regions are identified, and the percentage of time spent in these regions and the head posture's deviation from the center angle are calculated. The head posture deviation from the center angle is normalized and then weighted and summed with the percentage of time spent in these regions to generate an attention score. This attention score serves as an attention behavior indicator, reflecting the user's attention deviation. Real-time monitoring of the target user's attention behavior indicators provides a basis for subsequent intervention and compensation based on these indicators.

[0108] Step S400: Based on attention behavior indicators and preset intervention nodes, trigger attention intervention compensation for personalized advertising scripts based on emotional state, and generate push advertising scripts to improve the target user's attention to push advertising scripts.

[0109] In one embodiment, users may become distracted during ad playback due to external interference or boring content. By triggering an intervention compensation mechanism through attention metrics and intervention nodes, the target user's attention to the pushed ad script can be improved.

[0110] like Figure 3 As shown, based on attention behavior indicators and preset intervention nodes, attention intervention compensation is triggered based on emotional state to generate push advertising scripts, including but not limited to the following steps:

[0111] Step S410: Before the intervention node, if the attention score in the attention behavior index is less than the preset attention threshold, search for multiple material scripts corresponding to the emotional state in the preset compensation strategy library, combine and encapsulate the material scripts to obtain compensation material.

[0112] In one embodiment, the preset intervention point is 8 seconds after the advertisement airs. Before the intervention point, it is necessary to monitor whether the target user's attention score has decreased, and to provide timely intervention compensation to improve the target user's attention to the pushed advertisement script. If the attention behavior indicator shows that the attention score is less than the preset attention threshold, it indicates that a decrease in user attention has been detected, and a compensation mechanism needs to be triggered. Multiple material scripts corresponding to the emotional state are searched in a preset compensation strategy library, and the material scripts are combined and packaged to obtain compensation material.

[0113] Specifically, the pre-set compensation strategy library includes compensation emotions and corresponding material scripts. The text in these material scripts is pre-labeled based on the emotions. The target user's emotional state is matched with the compensation emotions in the compensation strategy library using a similarity algorithm. A match is considered successful if the matching result is greater than 90%. This similarity algorithm can be a text similarity algorithm. The material scripts corresponding to the successfully matched compensation emotions are extracted from the compensation strategy library, and these scripts are combined and packaged with music, etc., to obtain compensation material. This allows for subsequent intervention and compensation based on the compensation material, thereby increasing user attention.

[0114] In one embodiment, based on attention behavior indicators and preset intervention nodes, attention intervention compensation is triggered on personalized advertising scripts based on emotional states to generate push advertising scripts, including but not limited to the following steps:

[0115] Step S420: Upon reaching the intervention node, obtain the current emotion corresponding to the intervention node and determine whether the emotional state is consistent with the current emotion.

[0116] In some possible embodiments of this application, if the attention score in the attention behavior index is greater than or equal to a preset attention threshold, it indicates that the target user's attention is being focused on the playing advertisement. To avoid fatigue with the advertisement, intervention and compensation are necessary to increase the user's attention to the advertisement. Upon reaching an intervention node, the current emotion corresponding to the intervention node is retrieved, and it is determined whether the emotional state is consistent with the current emotion.

[0117] Specifically, the video frames captured in step S300 are used to perform emotion analysis on the facial images within the video frames using an emotion analysis model. The current emotion is then analyzed from the video frame corresponding to the intervention node. Next, an analysis is performed to determine the consistency between the current emotion and the emotional state. A similarity algorithm is used to calculate the similarity between the current emotion and the emotional state; a similarity value greater than 90% indicates emotional consistency, otherwise, it indicates inconsistency. The similarity algorithm used may include text similarity algorithms, cosine similarity algorithms, etc.

[0118] Step S430: In case of inconsistency, the current emotion is taken as the emotional state, and multiple material scripts corresponding to the current emotion are searched in the preset compensation strategy library. The material scripts are combined and packaged to obtain compensation material.

[0119] In one embodiment, in the event of inconsistency, the current emotion is taken as the emotional state, enabling real-time monitoring of the target user's emotions so that suitable materials can be selected for recommendation. Multiple material scripts corresponding to the current emotion are searched in a preset compensation strategy library, and these scripts are combined and encapsulated to obtain compensation materials. The specific implementation process is similar to step S410 and will not be elaborated here.

[0120] Step S440: Under the condition of consistency, multiple material scripts corresponding to the emotional state are searched in the preset compensation strategy library, and the material scripts are combined and packaged to obtain compensation material. The specific implementation process is similar to step S410, and will not be described in detail here.

[0121] Step S450: Use compensation materials to perform attention intervention compensation on the personalized advertising script to generate the push advertising script.

[0122] Specifically, attention intervention compensation is applied to personalized ad scripts using compensating materials to generate push ad scripts, including but not limited to the following steps:

[0123] Step S451: Divide the personalized advertising script into video frames and divide the video frame corresponding to the preset inflection point into a candidate switching area between the video frame and the next video frame. The preset inflection point is determined by the intervention node and the attention threshold.

[0124] In some possible embodiments of this application, the generated personalized advertising script is divided into video frames. Based on the intervention node, a corresponding video frame is generated at 8 seconds into the video, and a candidate switching zone is defined between this frame and the next video frame. At the attention threshold and the video frame where attention decline is detected, a candidate switching zone is defined between this video frame and the next video frame. The video can be directly segmented in this way. The candidate switching zones are used for subsequent video transitions to attract user attention. Since users may experience ad fatigue or inattention at the video frames corresponding to the intervention node and attention threshold, switching at these points avoids visual impact on the user and still attracts their attention.

[0125] Step S452: Switch the compensation material to the candidate switching area, add a smooth animation to the video frame corresponding to the turning point, and adjust the volume of the compensation material to match the volume of the personalized advertising script to generate the push advertising script.

[0126] In some possible embodiments of this application, the compensation material is switched to the candidate switching area. To avoid conflicts or visual discrepancies with the previous frame, a smooth animation is added to the video frame corresponding to the turning point. This smooth animation can be a fade-out halo, a floating window halo, a bubbling animation, etc. Then, the volume of the compensation material is adjusted to match the volume setting of the personalized advertising script to avoid auditory conflicts, and a push advertising script is generated. Through the above process, advertising videos are spliced ​​and merged from multiple aspects such as visual and auditory perception, improving user attention and avoiding discomfort.

[0127] Step S500: Calculate the conversion results of the push advertising script, and send back the data of the personalized advertising script, the push advertising script, the compensation record corresponding to the push advertising script, and the conversion results to enable automatic update of the script module.

[0128] like Figure 4 As shown, to facilitate the rapid generation of ad scripts that match user emotions and needs, the click-through rate, view rate, and whether compensation was triggered for the push ad scripts are statistically analyzed, and the statistical results are recorded as conversion results. Personalized ad scripts, push ad scripts, corresponding compensation records for push ad scripts, and conversion results are stored in JSON format, and the data is saved and transmitted back to the system. This allows the system to adjust the weights of each script module, and through a self-learning process, it can better meet user needs in subsequent matching processes, providing a foundation for real-time dynamic responses.

[0129] like Figure 5 As shown in the figure, this application embodiment provides a system 100 for dynamically optimizing advertising scripts. This system 100 acquires target user behavior data through a data acquisition and preprocessing module 110, performs sentiment analysis on the behavior data to obtain the target user's emotional state; utilizes a personalized script generation module 120 to select script modules from a preset script material library based on the emotional state, generating a personalized advertising script that matches the emotional state. The script materials include various emotional materials and corresponding script modules, with the emotional state matching one or more of the emotional materials; an attention monitoring module 130 displays the personalized advertising script to the target user and monitors the target user's attention behavior indicators in real time; then, a script adjustment module 140, based on the attention behavior indicators and preset intervention nodes, triggers attention intervention compensation for the personalized advertising script based on the emotional state, generating a push advertising script to enhance the target user's attention to the push advertising script; finally, a self-learning module 150 statistically analyzes the conversion results of the push advertising script, and transmits the personalized advertising script, the push advertising script, the compensation records corresponding to the push advertising script, and the conversion results back to the system to automatically update the script modules.

[0130] It should be noted that the data acquisition and preprocessing module 110 is connected to the personalized script generation module 120, the personalized script generation module 120 is connected to the attention monitoring module 130, the attention monitoring module 130 is connected to the script adjustment module 140, and the script adjustment module 140 is connected to the self-learning module 150. The above-mentioned method for dynamically optimizing advertising scripts is applied to the system 100 for dynamically optimizing advertising scripts. The system 100 generates personalized advertising scripts by adding target user emotions, ensuring that the advertising script meets the user's psychological needs, stimulates user interest and emotional resonance, and encourages participation in advertising interaction. Combined with attention monitoring, the system dynamically adjusts the advertising script through attention intervention and compensation, attracting the target user's attention to the advertising script. Through self-learning, it prepares for subsequent real-time dynamic responses. Based on the collaboration between the above steps, the dynamic adjustment method reduces reliance on static content, breaks away from single-dimensional personalized generation, and improves advertising conversion rates.

[0131] It should also be noted that the apparatus provided in the above embodiments is only illustrated by the division of the above functional modules. In actual applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the device can be divided into different functional modules to complete all or part of the functions described above. In addition, the apparatus and method embodiments provided in the above embodiments belong to the same concept, and their specific implementation process can be found in the method embodiments, which will not be repeated here.

[0132] This application also discloses an electronic device. (See reference...) Figure 6 , Figure 6 This is a schematic diagram of the structure of an electronic device according to an embodiment of this application. The electronic device 500 may include: at least one processor 501, at least one network interface 504, a user interface 503, a memory 505, and at least one communication bus 502.

[0133] The communication bus 502 is used to enable communication between these components.

[0134] The user interface 503 may include a display screen and a camera. Optionally, the user interface 503 may also include a standard wired interface and a wireless interface.

[0135] The network interface 504 may optionally include a standard wired interface or a wireless interface (such as a Wi-Fi interface).

[0136] The processor 501 may include one or more processing cores. The processor 501 connects to various parts of the server using various interfaces and lines, and performs various server functions and processes data by running or executing instructions, programs, code sets, or instruction sets stored in memory 505, and by calling data stored in memory 505. Optionally, the processor 501 may be implemented using at least one hardware form of Digital Signal Processing (DSP), Field-Programmable Gate Array (FPGA), or Programmable Logic Array. The processor 501 may integrate one or a combination of several of the following: Central Processing Unit (CPU), Graphics Processing Unit (GPU), and Modem. The CPU primarily handles the operating system, user interface, and applications; the GPU is responsible for rendering and drawing the content required for display; and the modem handles wireless communication. It is understood that the modem may also be implemented as a separate chip without being integrated into the processor 501.

[0137] The memory 505 may include random access memory (RAM) or read-only memory. Optionally, the memory 505 may include a non-transitory computer-readable storage medium. The memory 505 may be used to store instructions, programs, code, code sets, or instruction sets. The memory 505 may include a program storage area and a data storage area, wherein the program storage area may store instructions for implementing an operating system, instructions for at least one function (such as touch function, sound playback function, image playback function, etc.), instructions for implementing the above-described method embodiments, etc.; the data storage area may store data involved in the above-described method embodiments, etc. Optionally, the memory 505 may also be at least one storage device located remotely from the aforementioned processor 501. (Refer to...) Figure 6 The memory 505, which serves as a computer storage medium, may include an operating system, a network communication module, a user interface module, and an application program for a method of dynamically optimizing advertising scripts.

[0138] exist Figure 6In the illustrated electronic device 500, the user interface 503 is mainly used to provide an input interface for the user and to acquire user input data; while the processor 501 can be used to call an application program stored in the memory 505 that stores a method for dynamically optimizing advertising scripts. When executed by one or more processors 501, the electronic device 500 performs one or more methods as described in the above embodiments. It should be noted that, for the foregoing method embodiments, for the sake of simplicity, they are all described as a series of actions. However, those skilled in the art should understand that this application is not limited to the described order of actions, because according to this application, some steps can be performed in other orders or simultaneously. Secondly, those skilled in the art should also understand that the embodiments described in the specification are all preferred embodiments, and the actions and modules involved are not necessarily essential to this application.

[0139] In the above embodiments, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions in other embodiments.

[0140] In the various embodiments provided in this application, it should be understood that the disclosed apparatus can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some service interface; the indirect coupling or communication connection between apparatuses or units may be electrical or other forms.

[0141] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0142] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0143] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage device (CMD). Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a memory and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of this application. The aforementioned memory includes various media capable of storing program code, such as USB flash drives, portable hard drives, magnetic disks, or optical disks.

[0144] The above are merely exemplary embodiments of this disclosure and should not be construed as limiting the scope of this disclosure. Any equivalent changes and modifications made in accordance with the teachings of this disclosure shall still fall within the scope of this disclosure. Other embodiments of this disclosure will readily conceive of those skilled in the art upon consideration of the specification and the disclosure of practical truths.

[0145] This application is intended to cover any variations, uses, or adaptations of this disclosure that follow the general principles of this disclosure and include common knowledge or customary techniques in the art not described in this disclosure. The specification and embodiments are to be considered exemplary only, and the scope and spirit of this disclosure are defined by the claims.

Claims

1. A method for dynamically optimizing advertising scripts, characterized in that, The method includes: Obtain behavioral data of the target user, perform sentiment analysis on the behavioral data, and obtain the emotional state of the target user; Based on the emotional state, a script module is selected from a preset script material library to generate a personalized advertising script that matches the emotional state. The script material library includes a variety of emotional materials and the script modules corresponding to the emotional materials. The emotional state matches one or more of the emotional materials. The personalized advertising script is displayed to the target user, and the target user's attention behavior metrics are monitored in real time. Based on the attention behavior indicators and preset intervention nodes, attention intervention compensation is triggered on the personalized advertising script based on the emotional state to generate push advertising script, so as to improve the target user's attention to the push advertising script. The conversion results of the push advertising script are statistically analyzed, and the personalized advertising script, the push advertising script, the compensation record corresponding to the push advertising script, and the conversion results are transmitted back to the script module so as to automatically update the script module. The step of triggering attention intervention compensation for the personalized advertising script based on the attention behavior indicators and preset intervention nodes, and generating a push advertising script based on the emotional state, includes: Before the intervention node, if the attention score in the attention behavior index is less than the preset attention threshold, multiple material scripts corresponding to the emotional state are searched in the preset compensation strategy library, and the material scripts are combined and packaged to obtain compensation material. The personalized advertising script is subjected to attention intervention compensation using the compensation material to generate the push advertising script; Upon reaching the intervention node, the current emotion corresponding to the intervention node is obtained, and it is determined whether the emotional state is consistent with the current emotion. In the event of inconsistency, the current emotion is taken as the emotional state, and multiple material scripts corresponding to the current emotion are searched in the preset compensation strategy library. The material scripts are combined and packaged to obtain compensation material. The personalized advertising script is subjected to attention intervention compensation using the compensation material to generate the push advertising script; The step of using the compensation material to perform attention intervention compensation on the personalized advertising script to generate the push advertising script includes: The personalized advertising script is divided into video frames, and a candidate switching area is defined between the video frame corresponding to the preset turning point and the next video frame. The preset turning point is determined by the intervention node and the attention threshold. The candidate switching area is used to switch the video transition to attract the user's attention. The compensation material is switched to the candidate switching area, and a smooth animation is added to the video frame corresponding to the turning point. The volume of the compensation material is adjusted to match the volume of the personalized advertising script to generate the push advertising script.

2. The method according to claim 1, characterized in that, The step of performing sentiment analysis on the behavioral data to obtain the emotional state of the target user includes: The behavioral data is tensorized to obtain a behavioral vector; The behavior vector is processed using a pre-defined emotion analysis model to obtain an emotion category vector; The emotion category vector is subjected to probability transformation to obtain a probability distribution. The probability distribution is then mapped to a preset emotion label to form an emotion vector. The emotion vector includes the emotion label and the probability value corresponding to the emotion state. The probability value in the probability distribution is used as the confidence value, and the emotion label corresponding to the largest confidence value is selected as the emotion state.

3. The method according to claim 2, characterized in that, The step of selecting a script module from a preset script material library based on the emotional state to generate a personalized advertising script that matches the emotional state includes: The emotional state is matched with the emotional material in the script material library, and the script module corresponding to the matched emotional material is combined in multiple modes to generate multiple initial opening layers; Obtain the target user's historical behavior data, analyze the historical behavior data, and obtain user preferences; By utilizing the user preferences and emotional states, each of the initial opening layers is assembled to obtain multiple claim layers; Click-through rate (CTR) prediction is performed on each of the claim layers, and the claim layer with a CTR greater than a preset evaluation threshold is selected as the personalized advertising script.

4. The method according to claim 3, characterized in that, The method utilizes the user preferences and emotional states to assemble each of the initial opening layers, resulting in multiple claim layers, including: The user preferences and emotional state are used as inputs, and a large language model is used to generate multiple candidate script modules. The candidate script modules are then combined with the user preferences to form an assembly set. The candidate script modules in the assembly set are combined with the preset assembly strategy to form multiple claim documents; The emotional state is matched with emotional words in a preset emotional expression library to determine the strategy sentence pattern corresponding to the emotional state; The various claim statements are adjusted and encapsulated using the aforementioned strategy and sentence structure to form the claim layer.

5. A system for dynamically optimizing advertising scripts, characterized in that, The system for executing the dynamically optimized advertising script as described in claim 1, the system comprising: The data acquisition and preprocessing module is used to acquire behavioral data of the target user, perform sentiment analysis on the behavioral data, and obtain the emotional state of the target user. A personalized script generation module is used to select a script module from a preset script material library based on the emotional state and generate a personalized advertising script that matches the emotional state. The script material includes a variety of emotional materials and the script module corresponding to the emotional material. The emotional state matches one or more of the emotional materials. An attention monitoring module is used to display the personalized advertising script to the target user and monitor the target user's attention behavior metrics in real time. The script adjustment module is used to trigger attention intervention compensation on the personalized advertising script based on the attention behavior indicators and preset intervention nodes, based on the emotional state, to generate push advertising scripts, so as to improve the target user's attention to the push advertising scripts. The self-learning module is used to statistically analyze the conversion results of the push advertising script, and to transmit the personalized advertising script, the push advertising script, the compensation record corresponding to the push advertising script, and the conversion results back to the script module so as to automatically update the script module.

6. An electronic device, characterized in that, The device includes a processor, a memory, a user interface, a communication bus, and a network interface. The processor, the memory, the user interface, and the network interface are respectively connected to the communication bus. The memory is used to store instructions. The user interface and the network interface are used to communicate with other devices. The processor is used to execute the instructions stored in the memory to cause the electronic device to perform the method as described in any one of claims 1-4.

7. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores instructions that, when executed, perform the method as described in any one of claims 1-4.

Images