Interactive marketing behavior method, system and device and storage medium

Abstract

The invention relates to an interactive marketing behavior method, system and device and a storage medium, and the method comprises the following steps: carrying out the word segmentation and part-of-speech tagging of a business demand text based on a preset marketing semantic rule, generating a demand word sequence, and extracting a marketing intention label and an interaction scene parameter according to the demand word sequence; performing semantic extension on the intention label by using a large language model, matching a verbal skill template to generate a marketing verbal skill text, and combining with components such as an interaction scene parameter binding button and a pop-up window to form an interaction element configuration table; matching user operation data through the configuration table, and analyzing a behavior path to extract user response characteristics; and finally, according to the response characteristics, performing association scoring, screening and sorting on the skill text, outputting personalized push content, and synchronously generating a behavior feedback record, thereby solving the technical problem of how to efficiently map abstract user intention into executable interaction logic and respond to user operation in real time to form closed-loop feedback.

Description

Technical Field

[0001] This invention relates to the field of interactive marketing technology, and in particular to interactive marketing methods, systems, devices and storage media. Background Technology

[0002] In recent years, with the rapid evolution of the digital marketing environment, enterprises have increasingly urgent needs to accurately reach users and improve conversion efficiency. Traditional marketing strategies rely heavily on human experience to develop scripts and interaction processes, making it difficult to dynamically adapt to diverse user behaviors and complex and ever-changing business scenarios. Especially when faced with massive amounts of user data, existing rule engines and static templates often lack sufficient semantic understanding and context awareness capabilities, resulting in severe homogenization of marketing content, low user engagement, and difficulty in achieving truly personalized interaction.

[0003] Meanwhile, despite the powerful potential of Large Language Models (LLMs) in the field of natural language processing, their application in marketing scenarios still faces numerous challenges. On the one hand, content generated by the model, if detached from specific business contexts, can easily become vague or deviate from marketing objectives. On the other hand, how to efficiently map abstract user intentions into executable interactive logic and respond to user actions in real time to form a closed-loop feedback loop remains a weak link in current technical solutions. Existing systems often fragment the three key stages of "semantic understanding—content generation—behavior tracking," creating a disconnect between marketing actions and users' actual needs. Summary of the Invention

[0004] The main technical problem addressed in this application is to provide an interactive marketing method, system, device, and storage medium, which solves the technical problem of how to efficiently map abstract user intent into executable interactive logic and respond to user operations in real time to form a closed-loop feedback.

[0005] To address the aforementioned technical problems, this application employs an interactive marketing approach, comprising the following steps: Based on preset marketing semantic rules, the acquired business requirement text is segmented and part-of-speech tagging is performed to obtain a requirement word sequence. The requirement word sequence is then classified for intent and extracted for scene elements to obtain marketing intent tags and interaction scene parameters. The marketing intent tags are semantically expanded and matched with the script templates using a large language model to obtain marketing script text. Interactive components are then bound to the marketing script text based on the interaction scenario parameters to obtain an interactive element configuration table. Based on the interactive element configuration table, the collected user operation data is matched with events and the behavior path is parsed to obtain user response characteristics. Based on the user response characteristics, the marketing text is filtered and sorted to obtain personalized push content and behavior feedback records.

[0006] Furthermore, the obtained business requirement text is segmented and part-of-speech tagged based on preset marketing semantic rules to obtain a sequence of requirement words, including: The business requirement text is preprocessed to remove special characters and extra spaces to obtain standardized text. The standardized text is then split into multiple sentence units according to common punctuation marks. Based on a pre-defined marketing vocabulary database, the sentence units are segmented by word matching to obtain preliminary word segmentation results. Marketing keywords and ordinary words in the preliminary word segmentation results are then labeled with parts of speech and assigned corresponding part-of-speech tags to obtain the required word sequence.

[0007] Furthermore, the process of classifying the intent and extracting scene elements from the sequence of demand words to obtain marketing intent tags and interaction scene parameters includes: Action-related and object-related words are extracted from the demand word sequence to obtain a set of intent candidate words. Based on a preset marketing intent category table, the set of intent candidate words is matched and classified to obtain marketing intent tags. Based on the marketing intent tags, the demand word sequence is used to locate scene elements to obtain element field types. Then, attribute words and numerical words corresponding to the element field types are identified from the demand word sequence to obtain interaction scene parameters.

[0008] Furthermore, the marketing intent tags are semantically expanded and matched with the script template using a large language model to obtain the marketing script text, including: The marketing intent tags are parsed and element-splitting is performed to obtain the intent type field and the marketing goal field. The intent type field and the marketing goal field are then semantically expanded using a large language model to generate an intent semantic description text that includes marketing appeal statements, user benefit descriptions, and calls to action. Based on the intent semantic description text, keyword retrieval and semantic similarity calculation are performed on the preset script template library to obtain a candidate script template set. The template structure of the candidate script template set is then parsed to obtain template slot identifiers and slot filling rules. Based on the slot filling rules, the marketing appeal, user benefit description, and call to action statements in the intent semantic description text are filled into the corresponding template slot identifier positions to obtain the script filling result. The script filling result is then checked for text coherence and formatted to obtain the marketing script text.

[0009] Furthermore, the step of binding interactive components to the marketing script text based on the interactive scenario parameters to obtain an interactive element configuration table includes: The interaction scenario parameters are identified by parameter type and mapped to components to obtain a list of component types including button component identifiers, pop-up component identifiers and form component identifiers; Based on the component type list, the marketing text is segmented and divided into regions to obtain segmented text including title text, body text and operation text. Based on the component type list, the title area text, body area text, and operation area text in the speech segment content are bound to the corresponding button component identifier, pop-up component identifier, and form component identifier, respectively, to obtain the component binding relationship. The component binding relationship is then encapsulated and formatted to obtain the interactive element configuration table.

[0010] Furthermore, the step of performing event matching and behavior path parsing on the collected user operation data based on the interaction element configuration table to obtain user response characteristics includes: The collected user operation data is subjected to operation type identification to obtain operation type identifier, and the operation type identifier is matched with the component type list in the interaction element configuration table to obtain the successfully matched component operation record; Based on the successfully matched component operation records, the timestamps of the user operation data are extracted to obtain the operation time series, and the user operation data is constructed based on the operation time series to obtain the user behavior path graph. Based on the user behavior path graph, response features are extracted from the user operation data to obtain user response features.

[0011] Furthermore, based on the user response characteristics, the marketing text is filtered and sorted to obtain personalized push content and behavioral feedback records, including: The user response features are divided into feature dimensions to obtain response feature dimensions. Based on the response feature dimensions, the marketing text is analyzed for relevance to obtain a relevance score. The relevance scores are then sorted in descending order to obtain a sorted list of texts. Based on the sorted list of sales scripts, the marketing script texts are filtered to obtain filtered sales script content. The filtered sales script content is then bound and recorded with the user response characteristics to obtain behavioral feedback records. At the same time, the filtered sales script content is used as personalized push content.

[0012] The present invention also provides an interactive marketing behavior system, comprising: The extraction module is used to perform word segmentation and part-of-speech tagging on the acquired business requirement text based on preset marketing semantic rules to obtain a requirement word sequence, and to perform intent classification and scene element extraction on the requirement word sequence to obtain marketing intent tags and interaction scene parameters. The matching module is used to semantically expand the marketing intent tags and match the script templates using a large language model to obtain marketing script text, and bind interactive components to the marketing script text based on the interaction scenario parameters to obtain an interactive element configuration table. The parsing module is used to perform event matching and behavior path parsing on the collected user operation data based on the interaction element configuration table to obtain user response features, and to filter and sort the marketing text based on the user response features to obtain personalized push content and behavior feedback records.

[0013] The present invention also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps of any of the above methods.

[0014] The present invention also provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the steps of any of the above methods.

[0015] The above solution performs word segmentation and part-of-speech tagging on the acquired business requirement text based on preset marketing semantic rules to obtain a sequence of requirement words. This sequence is then categorized by intent and scene elements extracted to obtain marketing intent tags and interaction scene parameters. A large language model is used to semantically expand the marketing intent tags and match them with script templates to obtain marketing script text. Interactive components are bound to the marketing script text based on the interaction scene parameters to obtain an interaction element configuration table. Based on the interaction element configuration table, event matching and behavior path parsing are performed on the collected user operation data to obtain user response features. These features are then used to filter and sort the marketing script text, resulting in personalized push content and behavior feedback records. This solution addresses the technical challenge of efficiently mapping abstract user intents into executable interactive logic and responding to user operations in real time to form a closed-loop feedback loop. By collecting user action data such as clicks, dwell times, and conversions during the actual outreach process in real time based on the interaction element configuration table, the system can accurately analyze user behavior paths and extract response features. These features are then used to dynamically filter and sort candidate scripts, ensuring that the pushed content continuously evolves towards higher response and conversion rates. This adaptive mechanism, based on real user feedback, effectively solves the problem that traditional static marketing content cannot be adjusted to changes in user preferences, significantly improving interaction effectiveness and resource utilization efficiency. Attached Figure Description

[0016] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0017] Figure 1 This is a schematic diagram of the steps of an interactive marketing behavior method in one embodiment of the present invention; Figure 2 This is a structural block diagram of an interactive marketing behavior system according to an embodiment of the present invention; Figure 3 This is a schematic block diagram of the structure of a computer device according to an embodiment of the present invention.

[0018] The objectives, features, and advantages of this invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0019] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of the embodiments. Based on the embodiments of this application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of this application.

[0020] Specifically, the interactive marketing method of this embodiment includes the following steps: like Figure 1 As shown, Figure 1 This invention provides an interactive marketing method, comprising the following steps: Step S1: Based on preset marketing semantic rules, the acquired business requirement text is segmented and part-of-speech tagging is performed to obtain a requirement word sequence. The requirement word sequence is then classified for intent and extracted for scene elements to obtain marketing intent tags and interaction scene parameters.

[0021] Specifically, in this method, the overarching step of "segmenting and tagging the acquired business requirement text based on preset marketing semantic rules" is achieved by loading a grammar rule library predefined by marketing experts (such as keywords containing typical intents like promotion, new customer acquisition, and repeat purchase, along with their dependency structures). The system first segments the input business requirement text (e.g., "We hope to push a 300-yuan discount promotion to existing users during the 618 shopping festival") using a Chinese word segmentation tool (such as LTP or HanLP) and simultaneously tags the part of speech of each word (e.g., ...). "Spend 300 and get 50 off" is identified as a "numerical phrase" or "discount expression," thus forming a structured sequence of demand words. Subsequently, this sequence of demand words is fed into an intent recognition module based on the fusion of rule matching and a lightweight classification model. Preset rules are used to extract explicit elements (such as "old user" corresponding to user type, "618" corresponding to time node, and "recall" corresponding to marketing intent), while the classification model assists in judging implicit intent. Finally, standardized marketing intent labels (such as "old customer recall") and interaction scenario parameters (such as {user type: old user, activity time: 618 promotion, discount form: discount for spending over a certain amount}) are output. This process is a detailed implementation of the "intent classification and scenario element extraction" in the upper-level solution, ensuring that subsequent script generation and component binding have clear semantic anchors.

[0022] Step S2: The marketing intent tag is semantically expanded and matched with the script template using a large language model to obtain the marketing script text. Based on the interaction scenario parameters, the marketing script text is bound to interactive components to obtain an interactive element configuration table.

[0023] Specifically, in the aforementioned process, the step of "semantically expanding the marketing intent tags and matching them with script templates using a large language model" involves inputting the marketing intent tags output from the previous stage (such as "recalling old customers") into the finely tuned large language model. This model has been trained on a large amount of historical marketing copy data and can generate several candidate scripts with similar semantics but diverse expressions. At the same time, the system will retrieve structurally matching templates from the preset script template library (e.g., "Dear {user nickname}, long time no see! {discount content} is waiting for you to claim!") and fill the corresponding content generated by the model into the corresponding... Placeholders are used to form the initial marketing text. Then, based on the interaction scenario parameters extracted earlier (such as user type "old user" and event time "618 promotion"), the system automatically binds specific interactive components to the key information in the text. For example, it links "{discount content}" to the coupon button for 50 RMB off for purchases over 300 RMB, or maps "Waiting for you to claim" to a click event that redirects to the product page. Finally, it integrates these into a structured interactive element configuration table. This process is a specific expansion of the "interactive component binding" in the higher-level solution, ensuring that each text is not only semantically appropriate, but also triggers the expected user behavior in the actual interface.

[0024] Step S3: Based on the interaction element configuration table, perform event matching and behavior path parsing on the collected user operation data to obtain user response features, and based on the user response features, perform content filtering and sorting on the marketing text to obtain personalized push content and behavior feedback records.

[0025] Specifically, the system first uses the previously generated interaction element configuration table (e.g., the event ID for the "300 minus 50" button is defined as evt_coupon_618) to monitor and collect user operation logs on the terminal in real time, such as clicks, swipes, or page dwell time. Then, by accurately matching these raw operations with the event identifiers in the configuration table, it identifies valid interactive behaviors and reconstructs the complete behavioral path by combining timestamps and page jump sequences—for example, user A clicks the discount button but does not place an order but returns to the homepage, while user B clicks and completes the payment. Based on such path differences, the system extracts quantifiable user response features, including click-through rate, conversion delay, and churn points. Next, these features are used to weight and score candidate marketing texts. For example, it prioritizes pushing urgency-emphasizing texts to high-click, low-conversion groups, while increasing the ranking weight of benefit-oriented content for inactive users, ultimately outputting personalized push content tailored to individual preferences.

[0026] In a specific embodiment, the step of performing word segmentation and part-of-speech tagging on the acquired business requirement text based on preset marketing semantic rules to obtain a requirement word sequence includes: The business requirement text is preprocessed to remove special characters and extra spaces to obtain standardized text. The standardized text is then split into multiple sentence units according to common punctuation marks. Based on a pre-defined marketing vocabulary database, the sentence units are segmented by word matching to obtain preliminary word segmentation results. Marketing keywords and ordinary words in the preliminary word segmentation results are then labeled with parts of speech and assigned corresponding part-of-speech tags to obtain the required word sequence.

[0027] Specifically, the system performs text preprocessing on the original input business requirement text (e.g., "We hope to push a recall campaign of 50 RMB off for every 300 RMB spent to existing users during the 618 shopping festival!"). The system uses regular expressions to remove special symbols (such as exclamation marks, quotation marks, line breaks, etc.) and consecutive spaces, retaining only Chinese characters, numbers, and standard punctuation, and outputting standardized text with a uniform format. Subsequently, the standardized text is segmented according to common sentence-ending punctuation (such as periods, semicolons, commas) to generate several sentence units (e.g., "We hope to push a recall campaign of 50 RMB off for every 300 RMB spent to existing users during the 618 shopping festival"). This is intended to reduce contextual interference in subsequent word segmentation.

[0028] Next, the system calls a pre-set marketing vocabulary database—maintained by the operations team and containing high-frequency marketing phrases such as "spend 300 and get 50 off," "old users," and "618 promotion"—and performs maximum matching or bidirectional scanning on each sentence unit, prioritizing the identification and extraction of complete marketing terms. The remaining parts are then segmented according to a general dictionary, forming preliminary word segmentation results (e.g., ["hope," "at," "618," "during," "targeting," "old users," "push," "spend 300 and get 50 off," "of," "recall," "activity"]). Based on this, items identified as marketing keywords (e.g., "old users," "spend 300 and get 50 off") are assigned specific part-of-speech tags (e.g., "USER_TYPE," "DISCOUNT_EXPR"), while ordinary words (e.g., "hope," "at") are labeled with general parts of speech (e.g., "verb," ​​"preposition"), ultimately integrating them into a structured sequence of demand words.

[0029] In a specific embodiment, the step of classifying intent and extracting scenario elements from the sequence of demand words to obtain marketing intent tags and interaction scenario parameters includes: Action-related and object-related words are extracted from the demand word sequence to obtain a set of intent candidate words. Based on a preset marketing intent category table, the set of intent candidate words is matched and classified to obtain marketing intent tags. Based on the marketing intent tags, the demand word sequence is used to locate scene elements to obtain element field types. Then, attribute words and numerical words corresponding to the element field types are identified from the demand word sequence to obtain interaction scene parameters.

[0030] Specifically, the system filters action-oriented terms (such as "recall") and object-oriented terms (such as "old users") from the generated demand word sequence (e.g., ["618", "old users", "spend 300 minus 50", "recall"), and merges them to form a set of intent candidate words. Then, this set is matched item by item with a pre-built marketing intent category table (which is summarized by the operations team based on historical activities and includes standard categories such as "new user acquisition", "repeat purchase", and "churn recall" and their corresponding keyword mappings). For example, when "recall" and "old users" appear at the same time, the system determines that it belongs to the "churn recall" category, and thus outputs a standardized marketing intent label. This step is a detailed expansion of the "intent classification" in the higher-level solution.

[0031] Next, based on the marketing intent tags, the system initiates scenario element extraction: according to the element templates associated with the tag (such as the three fields of time, user type, and discount form corresponding to "churn recall"), it reverses the process to locate matching words in the original demand word sequence; specifically, it first determines the required element field types (such as "time node", "user identity", and "discount rules"), and then identifies the corresponding attribute words (such as "old user" corresponding to user identity) and numerical words (such as "618" as time and "300 minus 50" as discount numerical expression) through part-of-speech and context constraints, and finally assembles them into interactive scenario parameters in a structured way, such as {user type: old user, activity time: 618, discount form: 300 minus 50}.

[0032] In a specific embodiment, the marketing intent tag is semantically expanded and matched with the script template using a large language model to obtain the marketing script text, including: The marketing intent tags are parsed and element-splitting is performed to obtain the intent type field and the marketing goal field. The intent type field and the marketing goal field are then semantically expanded using a large language model to generate an intent semantic description text that includes marketing appeal statements, user benefit descriptions, and calls to action. Based on the intent semantic description text, keyword retrieval and semantic similarity calculation are performed on the preset script template library to obtain a candidate script template set. The template structure of the candidate script template set is then parsed to obtain template slot identifiers and slot filling rules. Based on the slot filling rules, the marketing appeal, user benefit description, and call to action statements in the intent semantic description text are filled into the corresponding template slot identifier positions to obtain the script filling result. The script filling result is then checked for text coherence and formatted to obtain the marketing script text.

[0033] Specifically, its implementation begins with marketing intent tags (such as "recalling existing customers"), which are then structured and parsed. Based on predefined tag format specifications, the system breaks down the tag into an intent type field (e.g., "recall") and a marketing objective field (e.g., "existing customer"). These two fields together form the input anchor for semantic generation. Subsequently, this is input as a prompt to a large language model (e.g., based on LLaMA or ChatGLM architecture) that has been fine-tuned from the marketing corpus. The instruction model then generates three types of content: marketing appeal statements (e.g., "We really want you back"), descriptions of user benefits (e.g., "Exclusive discount of 50 RMB off for purchases over 300 RMB has been reserved for you"), and calls to action (e.g., "Click to claim, valid for a limited time"). These three elements combine to form a complete semantic description of the intent, which is a refinement of the higher-level "semantic expansion."

[0034] Next, the system uses the semantic description text of the intent to perform a double matching on the preset script template library (containing hundreds of structured templates that have been verified by operations, such as "{appeal}!{benefit}, {call to action}"): first, it quickly filters templates containing relevant marketing keywords through the keyword inverted index, and then calculates the sentence-level cosine similarity (based on Sentence-BERT vector) between each candidate template and the current intent semantic, retaining templates with scores higher than a threshold (such as 0.75) to form a candidate script template set; then, it performs template structure parsing on each template in the set, identifies the placeholders (such as {benefit}) as template slot identifiers, and extracts the corresponding slot filling rules (for example, the "benefit" slot only accepts phrases containing numerical values ​​or benefit descriptions).

[0035] Finally, based on these rules, the three types of statements in the aforementioned semantic description text are filled into the corresponding template slots to generate preliminary wording filling results. To avoid awkward splicing, the system then calls a lightweight language model to perform text coherence verification on the results (checking subject-verb agreement, logical connection, etc.), and formats punctuation, spaces, etc., and finally outputs marketing wording text that conforms to reading habits and is semantically complete.

[0036] In a specific embodiment, the step of binding interactive components to the marketing script text based on the interactive scenario parameters to obtain an interactive element configuration table includes: The interaction scenario parameters are identified by parameter type and mapped to components to obtain a list of component types including button component identifiers, pop-up component identifiers and form component identifiers; Based on the component type list, the marketing text is segmented and divided into regions to obtain segmented text including title text, body text and operation text. Based on the component type list, the title area text, body area text, and operation area text in the speech segment content are bound to the corresponding button component identifier, pop-up component identifier, and form component identifier, respectively, to obtain the component binding relationship. The component binding relationship is then encapsulated and formatted to obtain the interactive element configuration table.

[0037] Specifically, starting with the generated interaction scenario parameters (such as {User Type: Existing User, Activity Time: 618, Discount Type: 50 RMB off for purchases over 300 RMB}), the system performs parameter type identification for each parameter item based on a preset component mapping rule library. For example, if "Discount Type" is identified as a benefit parameter that can trigger an action, it should be associated with a button; "User Type" involves identity verification and may require a pop-up explanation; if there is a form filling requirement (such as verifying a mobile phone number before claiming), then a form component is used. This generates a list of component types containing button component identifiers (such as btn_claim_coupon), pop-up component identifiers (such as modal_user_notice), and form component identifiers (such as form_verify_phone). This step is a specific expansion of the prerequisites for the higher-level "component binding".

[0038] The system performs structured parsing on the marketing text output from the previous stage (e.g., "Long time no see! Your exclusive discount of 50 RMB off orders over 300 RMB has arrived in your account, click to claim it now!"). Based on punctuation, semantic focus, and preset layout rules, it automatically divides the text into title text (e.g., "Long time no see!"), body text (e.g., "Your exclusive discount of 50 RMB off orders over 300 RMB has arrived in your account"), and action text (e.g., "Click to claim it now!"), forming segmented content for the sales pitch. It's worth noting that this division is not simply sentence-by-sentence splitting, but rather incorporates UI display conventions—titles are typically short, emotionally evocative sentences, body text conveys the core benefits, and action areas contain explicit verb commands.

[0039] Based on the aforementioned list of component types, the system binds each segment of content to a specific component identifier: the title text is bound to the pop-up component identifier (used for display at the top of the pop-up), the body text serves as the main content of the pop-up, and the operation area text is associated with the button component identifier, serving as the text on the button; if a form exists, a form component identifier is inserted below the button with reserved fields. The resulting component binding relationships are further encapsulated into structured data objects, where each component identifier corresponds to its text content, display order, and interaction events (such as clicking btn_claim_coupon to trigger the coupon API). These are then serialized in JSON or YAML format, ultimately outputting a standard interactive element configuration table.

[0040] In a specific embodiment, the step of performing event matching and behavior path parsing on the collected user operation data based on the interaction element configuration table to obtain user response characteristics includes: The collected user operation data is subjected to operation type identification to obtain operation type identifier, and the operation type identifier is matched with the component type list in the interaction element configuration table to obtain the successfully matched component operation record; Based on the successfully matched component operation records, the timestamps of the user operation data are extracted to obtain the operation time series, and the user operation data is constructed based on the operation time series to obtain the user behavior path graph. Based on the user behavior path graph, response features are extracted from the user operation data to obtain user response features.

[0041] Specifically, starting with the raw logs: The system receives user operation data (such as clicks, swipes, page dwells, etc.) reported by the front end, identifies the operation type through preset operation classification rules (e.g., "click" maps to clicks, "view" maps to exposures), and outputs standardized operation type identifiers; then, these identifiers are compared item by item with the component type list (such as btn_claim_coupon, modal_user_notice, etc.) in the interaction element configuration table generated earlier. If the target ID of a click event is exactly equal to btn_claim_coupon, it is determined to be a successful match and recorded as a successfully matched component operation record - this step is essentially a detailed implementation of the higher-level "event matching".

[0042] Based on this, the system extracts the timestamp (accurate to milliseconds) from each successfully matched component operation record, sorts them chronologically to form an operation time sequence. Then, using this sequence as a basis, it sequentially strings together all valid operations triggered by the user in a single session to construct a complete user behavior path graph. For example, if a user first triggers `modal_user_notice` (a pop-up notification), clicks `btn_claim_coupon` (a coupon claim button) 1.2 seconds later, but does not enter the payment page, this path is recorded as [modal → btn → exit]. It is worth noting that path construction not only depends on the chronological order but also needs to consider the page context to prevent cross-session data contamination.

[0043] Finally, based on this user behavior path graph, the system calls a preset feature extraction rule set to extract quantifiable metrics from the path structure as user response features, such as "whether the button click was completed," "the delay between the pop-up and the click," and "whether there was an interruption." These features directly reflect the user's true reaction to the current marketing content.

[0044] In a specific embodiment, the marketing text is filtered and sorted based on the user response characteristics to obtain personalized push content and behavioral feedback records, including: The user response features are divided into feature dimensions to obtain response feature dimensions. Based on the response feature dimensions, the marketing text is analyzed for relevance to obtain a relevance score. The relevance scores are then sorted in descending order to obtain a sorted list of texts. Based on the sorted list of sales scripts, the marketing script texts are filtered to obtain filtered sales script content. The filtered sales script content is then bound and recorded with the user response characteristics to obtain behavioral feedback records. At the same time, the filtered sales script content is used as personalized push content.

[0045] Specifically, the user response characteristics output in the previous stage (such as "clicked the button but did not pay", "pop-up window stay time > 3 seconds", "no follow-up action") are structured: the system classifies these raw characteristics by dimension according to the preset feature classification system, such as "interaction depth", "conversion intention" and "churn risk", forming clear response characteristic dimensions; then, for each dimension, the system calculates the matching strength between each marketing text and that dimension - for example, if a message emphasizes "limited-time offer", if the user response characteristics show a high churn risk but a brief stay, it is given a higher relevance weight, thereby generating a message relevance score for each message; all scores are sorted in descending order of numerical value, producing a sorted message list, which is a detailed expansion of the previous "content filtering and sorting".

[0046] The system sets thresholds (e.g., only selecting the top 30% of responses) or rules (e.g., requiring a clear call to action). It then selects entries from the sorted response list that meet the strategy requirements, forming the filtered response content. Simultaneously, this content is structurally linked to the user's response characteristics that triggered the filtering. For example, it records "User A was pushed response ID #205 due to 'high churn + short stay' characteristics," and stores this information along with a timestamp and channel information in the log, creating a traceable behavioral feedback record. Finally, this filtered response content is directly sent to the user's terminal as personalized push content.

[0047] Please see Figure 2 , Figure 2 This is a schematic diagram of the framework of an embodiment of the interactive marketing behavior system of this application. Figure 2 As shown, the interactive marketing behavior system includes an extraction module 1, which performs word segmentation and part-of-speech tagging on the acquired business requirement text based on preset marketing semantic rules to obtain a requirement word sequence, and performs intent classification and scene element extraction on the requirement word sequence to obtain marketing intent tags and interaction scene parameters; a matching module 2, which performs semantic expansion and script template matching on the marketing intent tags through a large language model to obtain marketing script text, and binds interactive components to the marketing script text based on the interaction scene parameters to obtain an interaction element configuration table; and a parsing module 3, which performs event matching and behavior path parsing on the collected user operation data based on the interaction element configuration table to obtain user response features, and performs content filtering and sorting on the marketing script text based on the user response features to obtain personalized push content and behavior feedback records.

[0048] Reference Figure 3 This invention also provides a computer device whose internal structure can be as follows: Figure 3 As shown, the computer device includes a processor, memory, display screen, input device, network interface, and database connected via a system bus. The processor provides computing and control capabilities. The memory includes a non-volatile storage medium and internal memory. The non-volatile storage medium stores the operating system, computer programs, and database. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage medium. The database stores the data corresponding to this embodiment. The network interface is used to communicate with external terminals via a network connection. When the computer program is executed by the processor, it implements the above-described method.

[0049] Those skilled in the art will understand that Figure 3The structures shown are merely block diagrams of some structures related to the present invention and do not constitute a limitation on the computer devices on which the present invention is applied.

[0050] An embodiment of the present invention also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the above-described method. It is understood that the computer-readable storage medium in this embodiment can be a volatile readable storage medium or a non-volatile readable storage medium.

[0051] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in the present invention and embodiments can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual-rate SDRAM (SSRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), Rambus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM, etc.

[0052] In some embodiments, the functions or modules of the apparatus provided in this disclosure can be used to perform the methods described in the above method embodiments. The specific implementation can be referred to the description of the above method embodiments, and for the sake of brevity, it will not be repeated here.

[0053] The description of the various embodiments above tends to emphasize the differences between the various embodiments. The similarities or similarities between them can be referred to, and for the sake of brevity, they will not be repeated here.

[0054] In the several embodiments provided in this application, it should be understood that the disclosed methods and apparatus can be implemented in other ways. For example, the apparatus implementations described above are merely illustrative. For instance, the division of modules or 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 mutual coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection of devices or units may be electrical, mechanical, or other forms.

[0055] 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; 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, depending on actual needs.

[0056] 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.

[0057] 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 medium. 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 storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) or processor to execute all or part of the steps of the methods of various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0058] If the technical solution of this application involves personal information, the product using this technical solution has clearly informed the user of the personal information processing rules and obtained the user's voluntary consent before processing the personal information. If the technical solution of this application involves sensitive personal information, the product using this technical solution has obtained the user's separate consent before processing the sensitive personal information, and also meets the requirement of "express consent". For example, at personal information collection devices such as cameras, clear and prominent signs are set up to inform users that they have entered the scope of personal information collection and that personal information will be collected. If an individual voluntarily enters the collection scope, it is deemed that they have agreed to the collection of their personal information; or on the personal information processing device, with clear signs / information informing users of the personal information processing rules, authorization is obtained from the individual through pop-up information or by asking the individual to upload their personal information; wherein, the personal information processing rules may include information such as the personal information processor, the purpose of personal information processing, the processing method, and the types of personal information processed.

Claims

1. An interactive marketing method, characterized in that, Includes the following steps: Based on preset marketing semantic rules, the acquired business requirement text is segmented and part-of-speech tagging is performed to obtain a requirement word sequence. The requirement word sequence is then classified for intent and extracted for scene elements to obtain marketing intent tags and interaction scene parameters. The marketing intent tags are semantically expanded and matched with the script templates using a large language model to obtain marketing script text. Interactive components are then bound to the marketing script text based on the interaction scenario parameters to obtain an interactive element configuration table. Based on the interaction element configuration table, the collected user operation data is matched for events and the behavior path is parsed to obtain user response characteristics. Based on the user response characteristics, the marketing text is filtered and sorted to obtain personalized push content and behavior feedback records.

2. The interactive marketing method according to claim 1, characterized in that, The process of segmenting and tagging the acquired business requirement text based on preset marketing semantic rules yields a sequence of requirement words, including: The business requirement text is preprocessed to remove special characters and extra spaces to obtain standardized text. The standardized text is then split into multiple sentence units according to common punctuation marks. Based on a pre-defined marketing vocabulary database, the sentence units are segmented by word matching to obtain preliminary word segmentation results. Marketing keywords and ordinary words in the preliminary word segmentation results are then labeled with parts of speech and assigned corresponding part-of-speech tags to obtain the required word sequence.

3. The interactive marketing method according to claim 1, characterized in that, The process of classifying intent and extracting scenario elements from the sequence of demand terms to obtain marketing intent tags and interaction scenario parameters includes: Action-related and object-related words are extracted from the demand word sequence to obtain a set of intent candidate words. Based on a preset marketing intent category table, the set of intent candidate words is matched and classified to obtain marketing intent tags. Based on the marketing intent tags, the demand word sequence is used to locate scene elements to obtain element field types. Then, attribute words and numerical words corresponding to the element field types are identified from the demand word sequence to obtain interaction scene parameters.

4. The interactive marketing method according to claim 1, characterized in that, The marketing intent tags are semantically expanded and matched with the script template using a large language model to obtain the marketing script text, including: The marketing intent tags are parsed and element-splitting is performed to obtain the intent type field and the marketing goal field. The intent type field and the marketing goal field are then semantically expanded using a large language model to generate an intent semantic description text that includes marketing appeal statements, user benefit descriptions, and calls to action. Based on the intent semantic description text, keyword retrieval and semantic similarity calculation are performed on the preset script template library to obtain a candidate script template set. The template structure of the candidate script template set is then parsed to obtain template slot identifiers and slot filling rules. Based on the slot filling rules, the marketing appeal, user benefit description, and call to action statements in the intent semantic description text are filled into the corresponding template slot identifier positions to obtain the script filling result. The script filling result is then checked for text coherence and formatted to obtain the marketing script text.

5. The interactive marketing method according to claim 1, characterized in that, The step of binding interactive components to the marketing script text based on the interactive scenario parameters to obtain an interactive element configuration table includes: The interaction scenario parameters are identified by parameter type and mapped to components to obtain a list of component types including button component identifiers, pop-up component identifiers and form component identifiers; Based on the component type list, the marketing text is segmented and divided into regions to obtain segmented text including title text, body text and operation text. Based on the component type list, the title area text, body area text, and operation area text in the speech segment content are bound to the corresponding button component identifier, pop-up component identifier, and form component identifier, respectively, to obtain the component binding relationship. The component binding relationship is then encapsulated and formatted to obtain the interactive element configuration table.

6. The interactive marketing method according to claim 1, characterized in that, The process of performing event matching and behavior path parsing on the collected user operation data based on the interaction element configuration table to obtain user response characteristics includes: The collected user operation data is subjected to operation type identification to obtain operation type identifier, and the operation type identifier is matched with the component type list in the interaction element configuration table to obtain the successfully matched component operation record; Based on the successfully matched component operation records, the timestamps of the user operation data are extracted to obtain the operation time series, and the user operation data is constructed based on the operation time series to obtain the user behavior path graph. Based on the user behavior path graph, response features are extracted from the collected user operation data to obtain user response features.

7. The interactive marketing method according to claim 1, characterized in that, Based on the user response characteristics, the marketing text is filtered and sorted to obtain personalized push content and behavioral feedback records, including: The user response features are divided into feature dimensions to obtain response feature dimensions. Based on the response feature dimensions, the marketing text is analyzed for relevance to obtain a relevance score. The relevance scores are then sorted in descending order to obtain a sorted list of texts. Based on the sorted list of sales scripts, the marketing script texts are filtered to obtain filtered sales script content. The filtered sales script content is then bound and recorded with the user response characteristics to obtain behavioral feedback records. At the same time, the filtered sales script content is used as personalized push content.

8. An interactive marketing behavior system, characterized in that, A method for performing interactive marketing activities according to any one of claims 1 to 7, comprising: The extraction module is used to perform word segmentation and part-of-speech tagging on the acquired business requirement text based on preset marketing semantic rules to obtain a requirement word sequence, and to perform intent classification and scene element extraction on the requirement word sequence to obtain marketing intent tags and interaction scene parameters. The matching module is used to semantically expand the marketing intent tags and match the script templates using a large language model to obtain marketing script text, and bind interactive components to the marketing script text based on the interaction scenario parameters to obtain an interactive element configuration table. The parsing module is used to perform event matching and behavior path parsing on the collected user operation data based on the interaction element configuration table to obtain user response features, and to filter and sort the marketing text based on the user response features to obtain personalized push content and behavior feedback records.

9. A computer device, characterized in that, The method includes a memory and a processor coupled to each other, the memory storing program instructions, and the processor executing the program instructions to implement the interactive marketing behavior method according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The system stores program instructions that can be executed by a processor, the program instructions being implemented to implement the interactive marketing behavior method according to any one of claims 1 to 7.

Images