Marketing knowledge base construction method, device and equipment based on artificial intelligence

Abstract

The invention relates to the technical field of artificial intelligence, and discloses a marketing knowledge base construction method, device and equipment based on artificial intelligence, and the method comprises the steps: obtaining basic information and session information, and converting the session information into text data according to the basic information; analyzing implicit relation data among the employees, the products and the clients based on the basic information, and extracting marketing knowledge from the text data; and carrying out hierarchical classified storage on the implicit relation data and the marketing knowledge, and pasting multi-dimensional labels on different knowledge to form a marketing knowledge base. According to the invention, through integration and structured processing of multi-source data, potential association among employees, products and clients is mined, and actual marketing knowledge is accurately extracted; hierarchical storage enables a knowledge system to be clearer, and multi-dimensional labels improve retrieval efficiency; the dynamic updating mechanism ensures the timeliness of knowledge, provides accurate knowledge support for a marketing scene, assists employees to quickly match verbal skills and schemes, and improves the communication efficiency and the transaction rate.

Description

Technical Field

[0001] This invention relates to the field of artificial intelligence technology, specifically to a method, apparatus, and equipment for constructing a marketing knowledge base based on artificial intelligence. Background Technology

[0002] In today's highly competitive and rapidly changing market environment, the success of corporate marketing increasingly depends on efficient and personalized communication between frontline marketers and customers. However, traditional methods of processing and applying marketing knowledge have several significant problems in supporting the practical application of human marketing personnel: traditional marketing script libraries do not consider the differences among marketing staff, and uniform scripts fail to allow each employee to fully leverage their strengths; the tacit and experiential knowledge within the marketing team lacks effective recording and organization, resulting in unstructured and contextualized instructions that cannot be absorbed by marketers, leading to slow improvement in the overall practical skills of the marketing team; when communicating with customers, marketers struggle to quickly identify customer pain points and respond effectively with product knowledge, causing product knowledge to fail to play its due supporting role in marketing communication, resulting in low marketing conversion efficiency.

[0003] While some AI-based dialogue systems (such as intelligent customer service and digital humans) can engage in question-and-answer interactions or push information, they are designed to replace or partially replace humans, rather than enhance or assist human marketers in making decisions and expressing themselves in complex and ever-changing real-world communication scenarios. Existing marketing content generation tools (such as those based on large language models) can automatically generate text, but they typically lack deep integration with the company's internal practical knowledge base, employee profiles, and real-time communication contexts, making it difficult to provide real-time, personalized, and scenario-specific auxiliary suggestions. Summary of the Invention

[0004] In view of this, the present invention provides a marketing knowledge base construction method, apparatus and equipment based on artificial intelligence to solve the problem of difficulty in providing real-time, personalized and specific communication scenario-adaptive auxiliary suggestions.

[0005] In a first aspect, the present invention provides a method for constructing a marketing knowledge base based on artificial intelligence, the method comprising:

[0006] Obtain basic information and session information, and convert the session information into text data according to the basic information;

[0007] Based on basic information analysis, implicit relationship data between employees, products, and customers is used to extract marketing knowledge from text data.

[0008] Implicit relationship data and marketing knowledge are hierarchically categorized and stored, and different knowledge is labeled with multi-dimensional tags to form a marketing knowledge base.

[0009] The marketing knowledge base construction method based on artificial intelligence provided by this invention integrates and structures multi-source data to mine potential connections between employees, products, and customers, and accurately extract practical marketing knowledge; hierarchical storage makes the knowledge system clearer, multi-dimensional tags improve retrieval efficiency; dynamic update mechanism ensures the timeliness of knowledge, provides accurate knowledge support for marketing scenarios, helps employees quickly match scripts and solutions, and improves communication efficiency and conversion rate.

[0010] In one alternative implementation, the method further includes:

[0011] Acquire real-time conversations and use intelligent recommendation models to retrieve multiple relevant knowledge fragments from the marketing knowledge base and recommend them to the conversation employees;

[0012] Obtain the conversation content of employees and the feedback content of customers, evaluate the effectiveness of knowledge fragment recommendations, and optimize the marketing knowledge base based on the effectiveness evaluation results.

[0013] This invention provides an AI-based marketing knowledge base construction method that enables real-time dynamic recommendation of marketing knowledge, helping employees respond quickly to customer needs and improve communication efficiency. By forming a data loop through customer feedback and employee usage behavior, the effectiveness scoring and recommendation model of knowledge fragments are precisely optimized, making the knowledge base more aligned with real-world scenarios. A multi-dimensional evaluation mechanism ensures that recommended content is highly compatible with customer intent and employee style, enhancing the targeting and success rate of marketing communication. Simultaneously, through data loop optimization, the practicality and intelligence of the knowledge base are continuously enhanced, providing efficient support for enterprise marketing decisions.

[0014] In one optional implementation, basic information and session information are obtained, and the session information is converted into text data according to the basic information, including:

[0015] Raw attribute information is collected from the customer management system, employee management system, and product system. The raw attribute information is preprocessed into text to obtain basic information, which includes customer information, employee information, and product information.

[0016] Raw conversation data is collected from voice communication scenarios, and the raw conversation data is subjected to structured cleaning, speech-to-text conversion, and text preprocessing to obtain conversation information.

[0017] Based on customer and employee information, the conversation information is divided into text data from different speakers.

[0018] The marketing knowledge base construction method based on artificial intelligence provided by this invention collects and preprocesses raw attribute information from multiple sources to integrate and structure customer, employee, and product data, thereby improving information integrity and usability. It cleans, converts speech to text, and preprocesses text data to transform unstructured speech into analyzable text data, laying the foundation for knowledge extraction. Based on customer and employee information, it segments speaker text data, accurately distinguishing the subjects of the dialogue (such as employees and customers), facilitating subsequent analysis of the communication patterns, intentions, and verbal characteristics of different roles.

[0019] In one alternative implementation, the implicit relationship data between employees, products, and customers is analyzed based on basic information, including:

[0020] Construct an employee-product-customer relationship graph, where nodes include employee profile features, product feature tags, and customer attribute tags, and edges represent interaction relationships and weights;

[0021] Using graph neural networks or collaborative filtering algorithms, we can analyze the implicit relationships between nodes in a relational graph.

[0022] In one alternative implementation, a graph neural network or collaborative filtering algorithm is used to analyze the implicit relationships between nodes in the association graph, including:

[0023] Based on employees' historical communication records and transaction data, a regression model was established to link employee style with the effectiveness of their communication techniques.

[0024] Identify the matching patterns between product features and customer needs by analyzing customer purchase behavior sequences;

[0025] By combining the characteristics of the communication context, we can predict the probability increase of a specific employee using specific language to close a deal with a specific customer.

[0026] The marketing knowledge base construction method based on artificial intelligence provided by this invention constructs a dual-precise profile by deeply analyzing multi-dimensional data of human employees and customers. Combined with real-time communication context, the artificial intelligence system intelligently judges and recommends highly personalized operational scripts, product knowledge, solutions, etc., that are most suitable for the current scenario and the individual style of human marketers. This significantly improves the adaptability and effectiveness of marketing communication, helps employees with different styles to conduct marketing more effectively, reduces manual maintenance costs, and ensures the timeliness and effectiveness of knowledge.

[0027] In one alternative implementation, the text data includes conversational text data and meeting text data, and the extraction of marketing knowledge from the text data includes:

[0028] Using sequence labeling models, functional dialogue fragments in dialogue text data are identified. These functional dialogue fragments include: opening remarks, product introduction sentences, objection handling sentences, and sales-closing sentences.

[0029] Using event extraction and relation extraction models, we analyze entity relationships and event sequences in dialogue text data. The event sequences include: customer questioning, objection generation, objection resolution, and transaction completion.

[0030] Acquire dialogue result data and filter out dialogue patterns and speech fragments that are strongly correlated with positive results from the dialogue result data;

[0031] The core topics in the meeting text data are identified using a topic recognition model, the meeting minutes are generated using a text summarization model, and the core topics are combined to determine strategy adjustments, market insights, or customer feedback summaries.

[0032] Based on dialogue patterns and corresponding script segments, strategy adjustments, market insights, or customer feedback, marketing knowledge for different scenarios is summarized.

[0033] The marketing knowledge base construction method based on artificial intelligence provided by this invention includes unstructured voice data generated in the daily marketing operations of enterprises (such as internal communications, meetings, and sales calls with customers) into the core knowledge collection scope. It uses advanced artificial intelligence technology to deeply mine and extract the valuable practical experience, implicit skills, and contextualized cases contained therein, and performs systematic classification and tag management. This transforms past operational experience into reusable structured knowledge assets, directly providing marketing personnel with auxiliary suggestions based on practical verification, and effectively improving the overall practical level of the team.

[0034] In one alternative implementation, a real-time conversation is acquired, and a smart recommendation model is used to retrieve multiple relevant knowledge fragments from a marketing knowledge base and recommend them to the conversation staff, including:

[0035] Real-time acquisition of text streams or speech-to-text streams of conversations between employees and customers; use low-latency natural language processing models for word segmentation, entity recognition, sentiment analysis, and intent recognition to determine customer sentiment and intent.

[0036] Based on the remote dictionary service cache, dynamically update customer profiles containing customers' current intentions and emotions, and load employee profiles;

[0037] When a customer finishes speaking or an employee enters text, context recognition is triggered to determine the fine-grained communication scenario, and matching knowledge fragments are retrieved from the marketing knowledge base in combination with customer and employee profiles.

[0038] The system calculates the fit score between knowledge fragments and the current "customer-employee-context" combination using a deep ranking model or rule engine, sorts the knowledge fragments based on the fit score, pushes the most relevant knowledge fragments to the communication interface, and updates the recommendation list in real time as the conversation context changes.

[0039] The marketing knowledge base construction method based on artificial intelligence provided by this invention, based on customer profiles and a comprehensive marketing knowledge base, intelligently matches product highlights, features, use cases or solutions most relevant to the customer's current focus and needs during real-time communication between human marketers and customers. It then pushes these to employees in a way that is easy for human marketers to understand and present to customers, helping employees to give accurate and attractive product introductions, reducing customer understanding costs, and enhancing product marketing appeal and conversion rates.

[0040] Secondly, the present invention provides an artificial intelligence-based marketing knowledge base construction device, the device comprising:

[0041] The data acquisition module is used to acquire basic information and session information, and convert the session information into text data according to the basic information.

[0042] The knowledge extraction module is used to analyze implicit relationship data between employees, products, and customers based on basic information, and to extract marketing knowledge from text data.

[0043] The knowledge base building module is used to hierarchically classify and store implicit relational data and marketing knowledge, and to attach multi-dimensional tags to different knowledge to form a marketing knowledge base.

[0044] Thirdly, the present invention provides a computer device, comprising: a memory and a processor, the memory and the processor being communicatively connected to each other, the memory storing computer instructions, and the processor executing the computer instructions to perform the method described in the first aspect or any corresponding embodiment thereof.

[0045] Fourthly, the present invention provides a computer-readable storage medium storing computer instructions for causing a computer to perform the method described in the first aspect or any corresponding embodiment thereof. Attached Figure Description

[0046] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0047] Figure 1 This is a flowchart illustrating a marketing knowledge base construction method based on artificial intelligence according to an embodiment of the present invention;

[0048] Figure 2This is a schematic diagram of relevant data and data flow in the marketing knowledge base construction method based on artificial intelligence according to an embodiment of the present invention;

[0049] Figure 3 This is a flowchart illustrating another method for constructing a marketing knowledge base based on artificial intelligence according to an embodiment of the present invention;

[0050] Figure 4 This is a flowchart illustrating another method for constructing a marketing knowledge base based on artificial intelligence according to an embodiment of the present invention;

[0051] Figure 5 This is a structural block diagram of an artificial intelligence-based marketing knowledge base construction device according to an embodiment of the present invention;

[0052] Figure 6 This is a schematic diagram of the hardware structure of a computer device according to an embodiment of the present invention. Detailed Implementation

[0053] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0054] This invention provides a method for constructing a marketing knowledge base based on artificial intelligence. By integrating and structuring multi-source data, a marketing knowledge base is constructed to provide accurate knowledge support for marketing scenarios and improve communication efficiency and conversion rates.

[0055] According to an embodiment of the present invention, an embodiment of a marketing knowledge base construction method based on artificial intelligence is provided. It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions. Furthermore, although a logical order is shown in the flowchart, in some cases, the steps shown or described may be executed in a different order than that shown here.

[0056] This embodiment provides a method for constructing a marketing knowledge base based on artificial intelligence, which can be used in the aforementioned computer system. Figure 1 This is a flowchart of a marketing knowledge base construction method based on artificial intelligence according to an embodiment of the present invention, such as... Figure 1 As shown, the process includes the following steps:

[0057] Step S101: Obtain basic information and session information, and convert the session information into text data according to the basic information.

[0058] Specifically, this involves acquiring various basic and conversational information related to marketing. Basic information includes, but is not limited to: customer information (basic information, purchase history, browsing behavior, and communication records with customer service or sales (text and voice)) to analyze customer characteristics and needs; employee information (personal information, sales performance, historical communication data (text and voice), and data on communication styles and skills assessed by Artificial Intelligence (AI) or manually inputted data) to provide a basis for building employee profiles; and detailed product information (features, technical parameters, use cases, market positioning, customer reviews, etc.) to facilitate AI's deeper understanding of product knowledge. Conversational information includes, but is not limited to: daily communication and meeting voice recordings. This data collection function is integrated or deployed in the communication and collaboration tools or meeting systems used by the company's internal marketing team (such as instant messaging, audio / video conferencing, and telephone systems). Subject to compliance with national laws and regulations, internal company policies, and authorization from relevant personnel, unstructured voice data such as daily internal communications related to marketing, marketing strategy meetings, and actual communications (calls or meetings) between sales personnel and customers are collected as key and unique sources of material for mining practical experience and knowledge.

[0059] For meeting information, the format is usually voice data. To analyze it, it needs to be converted into text data. The specific conversion methods are mature existing technologies and will not be elaborated here. The text data includes the text after voice conversion and the original text record. Natural Language Processing (NLP) libraries are used for preprocessing such as word segmentation, part-of-speech tagging, named entity recognition (person names, place names, company names, product names), sentiment analysis (judging the mood of sentences), intent recognition (judging whether sentences are questions, statements, complaints, etc.), and pronoun resolution, and stop words and punctuation are removed.

[0060] Step S102: Analyze the implicit relationship data between employees, products, and customers based on basic information, and extract marketing knowledge from the text data.

[0061] Specifically, it is responsible for automatically identifying, extracting, associating, and learning marketing knowledge from preprocessed multi-source, multi-modal data to build a knowledge base with depth and breadth.

[0062] By leveraging machine learning and data mining techniques from customer, employee, and product data, we can learn and discover implicit connections and patterns. Examples include: the core needs of specific customer groups; which types of customers are more likely to be persuaded by a particular employee's communication style; which industry pain points are best addressed by specific product features; and the predictive effectiveness of different marketing phrases in different communication contexts. These patterns and predictive capabilities also serve as a supplement and support to the knowledge base.

[0063] By using NLP techniques (information extraction, relation extraction, event extraction, etc.) to automatically identify and extract practical experience knowledge fragments from daily communication or meeting speech-to-text, such as: successful opening remarks, effective wording for dealing with specific customer issues, techniques for closing deals, product highlights emphasized in practical applications, actual customer feedback on products / services and corresponding solutions, and strategic consensus formed in meetings.

[0064] Step S103: Hidden relationship data and marketing knowledge are hierarchically classified and stored, and different knowledge is labeled with multi-dimensional tags to form a marketing knowledge base.

[0065] Specifically, the extracted marketing knowledge is organized and managed in a structured, searchable, and applicable manner. Following a pre-defined, flexible, and scalable marketing knowledge classification system (e.g., by knowledge type: sales scripts, products, solutions; by business scenario: sales, customer service, marketing; by product line; by customer lifecycle, etc.), the extracted knowledge is hierarchically categorized and stored. A flexible tree-like or graph-like knowledge classification system is established in the system backend. For example: root node -> [sales script knowledge, product knowledge, marketing solutions, customer cases, market insights, industry knowledge]; sales script knowledge -> [opening remarks, exploring needs, product introduction, objection handling, closing, farewell remarks]; product knowledge -> [product A series, product B series]; product A series -> [product A1 function, product A1 parameter, product A1 application scenario], etc. The classification system supports multi-level nesting and cross-category links.

[0066] like Figure 2 The diagram shown illustrates the relevant data and data flow of the AI-based marketing knowledge base construction method provided in this embodiment. It adds a detailed, multi-dimensional tagging system to various knowledge fragments. This tagging system supports complex combined queries and multi-dimensional filtering, forming the foundation for accurate recommendations. This includes, but is not limited to:

[0067] (1) Content attribute tags: Product name, function, parameters, applicable industries, pain points solved, objection type (price, quality, brand), marketing campaign name, and solution type;

[0068] (2) Situational labels: communication stage, customer life cycle stage, communication channel (telephone, WeChat, face-to-face meeting), customer emotions, objection type, applicable employee communication style;

[0069] (3) Target attribute tags: target customer type, marketing stage, customer pain points, customer profile characteristics (age group, region, consumption preference, decision-making style), employee profile characteristics (communication style, skills and expertise, department);

[0070] (4) Effectiveness / Source Tags: Practical verification effectiveness score (e.g., 1-5 points, dynamically adjusted based on feedback data), knowledge source (meeting time, meeting minutes, contributions from outstanding employees, AI extraction), update time, and relevant risk warnings.

[0071] The system utilizes NLP techniques (such as keyword extraction, text classification, and sequence labeling) and machine learning models to automatically recommend tags for new knowledge fragments. The administrator interface provides tag editing, batch management, synonym management, and tag system optimization functions.

[0072] Based on the knowledge content and AI-generated results, the system automatically recommends the categories and tags to which the knowledge belongs. It uses text classification algorithms for automatic category prediction and keyword extraction and topic modeling algorithms to automatically generate tag suggestions. The administrator interface provides visual tools that allow manual review, modification, and improvement of the automatic categories and tags, maintaining the standardization of the classification system and tag dictionary.

[0073] The AI-based marketing knowledge base construction method provided in this embodiment integrates and structures multi-source data to uncover potential connections between employees, products, and customers, and accurately extracts practical marketing knowledge. Hierarchical storage makes the knowledge system clearer, and multi-dimensional tags improve retrieval efficiency. The dynamic update mechanism ensures the timeliness of knowledge, provides accurate knowledge support for marketing scenarios, helps employees quickly match scripts and solutions, and improves communication efficiency and conversion rate.

[0074] In some alternative implementations, such as Figure 3 As shown, the method also includes:

[0075] Step S104: Obtain the real-time conversation and use the intelligent recommendation model to retrieve multiple relevant knowledge fragments from the marketing knowledge base and recommend them to the conversation employees.

[0076] Specifically, such as Figure 2 As shown, an intelligent recommendation model is used to retrieve multiple relevant knowledge fragments from a marketing knowledge base, providing real-time, contextualized, and personalized marketing assistance during communication between marketers and customers. The feature vectors of customer and employee profiles are continuously updated. Customer profiles include interests, preferences, intentions, and emotions based on behavior and communication history, while employee profiles include styles, skills, and expertise based on performance and communication analysis.

[0077] A lightweight, high-efficiency NLP model is used to quickly analyze real-time input dialogue text, identifying the current communication stage, the customer's latest intent, and emotions. Based on the perceived context and customer intent, a rapid, multi-dimensional retrieval is performed in the marketing knowledge base to obtain a batch of relevant knowledge fragments. The retrieval can employ keyword matching, semantic vector similarity search, or a combination of tag filtering. The AI ​​recommendation algorithm receives the retrieved knowledge fragments, current customer profiles, employee profiles, and real-time contextual features. Through a pre-trained ranking model, the predicted fit or recommendation score of each knowledge fragment with the current "customer-employee-context" combination is calculated. The factors considered by this model include, but are not limited to: the relevance of knowledge to customer intent, the matching degree of knowledge to employee style, the practical effectiveness score of the knowledge itself, the timeliness of the knowledge, and the credibility of the knowledge source.

[0078] Step S105: Obtain the conversation content of the conversational staff and the feedback content of the conversational customers, evaluate the effectiveness of the knowledge fragment recommendation, and optimize the marketing knowledge base based on the effectiveness evaluation results.

[0079] Specifically, continuously monitor various new data (new customers, new performance, new product information, etc.) in the company's marketing activities. Automatously record employee adoption of system-recommended content (e.g., whether they clicked, copied, or adopted the recommended scripts, or adopted the recommended suggestions), as well as customer reactions to employees' use of the recommended content (e.g., whether the customer's mood was positive, whether they continued to communicate further, or whether a sale was ultimately made). Construct new training data samples using "context + employee profile + recommended knowledge + employee behavior + customer feedback / results," and periodically, or when sufficient data has been accumulated, use these samples to retrain the AI ​​knowledge extraction model, AI recommendation model, and knowledge effectiveness evaluation model to optimize the marketing knowledge base. For example, using reinforcement learning, positive customer feedback can be used as a reward signal to optimize recommendation strategies.

[0080] When new data is detected or feedback is collected, the system automatically initiates the knowledge update process, inputting the newly collected data into the AI ​​knowledge extraction and learning module to extract new knowledge or update existing knowledge content.

[0081] By utilizing collected feedback data based on real-world applications, the effectiveness of knowledge in the marketing knowledge base is evaluated and its weights adjusted. For example, if a recommendation is frequently adopted by employees in a specific context, and subsequent customer feedback is generally positive, ultimately leading to more transactions, then the effectiveness score of that recommendation is increased; conversely, it is decreased or marked as ineffective.

[0082] By leveraging this feedback-rich human-computer collaboration interaction data and customer response data, the AI ​​intelligent recommendation algorithm model can be retrained and optimized (e.g., using reinforcement learning techniques) so that its recommendation results can more accurately predict the actual application effects and more effectively guide employees toward success.

[0083] This embodiment also provides a user-friendly administrator interface, supporting manual review of AI-extracted knowledge, manual addition or editing of knowledge, adjustment of knowledge categories and tags, correction of employee styles or customer profiles evaluated by AI, optimization of recommendation rules, and viewing of knowledge base usage and recommendation performance statistics. Examples include deleting outdated product knowledge, adding new marketing campaign scripts, adjusting the tags of specific knowledge items, and setting recommendation rule priorities for specific scenarios. This ensures the quality and credibility of the knowledge base and allows for human guidance of the system.

[0084] The AI-based marketing knowledge base construction method provided in this embodiment enables real-time dynamic recommendation of marketing knowledge, helping employees respond quickly to customer needs and improve communication efficiency. By forming a data loop through customer feedback and employee usage behavior, the effectiveness scoring and recommendation model of knowledge fragments are precisely optimized, making the knowledge base more aligned with real-world scenarios. A multi-dimensional evaluation mechanism ensures that recommended content is highly compatible with customer intent and employee style, enhancing the targeting and success rate of marketing communication. Simultaneously, continuous optimization through the data loop enhances the practicality and intelligence of the knowledge base, providing efficient support for enterprise marketing decisions.

[0085] In an optional implementation, step S104 includes:

[0086] Step S1041: Real-time acquisition of the dialogue text stream or speech-to-text stream between employees and customers; use a low-latency natural language processing model for word segmentation, entity recognition, sentiment analysis and intent recognition to determine the customer's sentiment and intent.

[0087] Step S1042: Based on the remote dictionary service cache, dynamically update the customer profile containing the customer's current intent and emotions, and load the employee profile.

[0088] Specifically, based on data provided by the multi-source data acquisition and preprocessing module, and combined with the learning results of the AI ​​knowledge extraction and learning module, accurate, comprehensive, and continuously dynamically updated customer and employee profiles are constructed. Customer profiles include, but are not limited to, static information such as basic attributes and purchase history, but also include dynamic and in-depth features such as customer interests, consumption habits, decision-making styles, focus areas, historical communication feedback analysis (emotional tendencies, high-frequency vocabulary, questioning patterns), and AI-recognized intent, emotions, and current focus during real-time communication. Employee profiles include, but are not limited to, basic personal information and sales performance, but also include individual characteristics such as communication style assessed by AI (e.g., rigorous, enthusiastic, concise), types of customers they are good at handling, expertise in specific products or business areas, skills demonstrated in past success stories, and their communication performance and status on the day of the interview.

[0089] It receives and analyzes in real time text or voice-to-text data from communication between marketers and customers, and uses NLP technology to accurately perceive the current communication stage (e.g., opening, exploring needs, product introduction, handling objections, closing the deal), the specific questions raised by the customer, the expressed needs, potential pain points, explicit or implicit objections, and the customer's emotional state.

[0090] Step S1043: When the customer finishes speaking or the employee enters text, context recognition is triggered to determine the fine-grained communication scenario, and matching knowledge fragments are retrieved from the marketing knowledge base in combination with customer and employee profiles.

[0091] Specifically, during real-time communication between marketers and customers (e.g., after an employee enters text or speaks, or after a customer finishes speaking), the system immediately activates the AI ​​recommendation algorithm, taking into account: the current communication context (obtained through the real-time perception module), the real-time updated customer profile, and the current employee profile, and preliminarily retrieves relevant knowledge fragments (scripts, product information, cases, solutions, etc.) from the marketing knowledge base based on context, customer needs, etc.

[0092] Step S1044: Calculate the fit score between knowledge fragments and the current "customer-employee-context" combination using a deep ranking model or rule engine, sort the knowledge fragments based on the fit score, push the knowledge fragments with the highest fit to the communication interface, and update the recommendation list in real time as the dialogue context changes.

[0093] Specifically, a multi-factor matching and ranking model is used. This model comprehensively evaluates the fit or effectiveness prediction score of knowledge fragments with the current "customer-employee-situation" combination, and pushes the top-ranked, most suitable knowledge fragments (such as a suggested sales pitch, a product highlight description, or a relevant case link) to the marketing personnel's communication interface in real time (e.g., sidebar, suggested reply list). For example, the model considers: Does this sales pitch solve the customer's current problem? Does it align with the customer's decision-making preferences? Does it match the employee's communication style? What is the historical success rate of this sales pitch when used by employees with similar styles in similar scenarios? This is just an example, but not a limitation.

[0094] AI recommendation models (such as a trained deep ranking model or a complex rule engine + machine learning model) receive a list of relevant knowledge, real-time context features, customer profile features, and employee profile features. They calculate a prediction fit score for each knowledge fragment with the current "customer-employee-context" combination. The scoring considers factors including, but not limited to: relevance - the direct relevance to the customer's current problem; effectiveness - the practical effectiveness score based on historical feedback data; fit - the degree of matching with the customer's decision-making style and the employee's communication style; timeliness - the newness of the knowledge; priority - certain key knowledge (such as legal compliance requirements, important promotional information) may have high priority.

[0095] The N knowledge fragments with the highest predicted suitability (such as the 3 most suitable objection handling scripts, or 1 script + 1 product value point + 1 success case link) are pushed to the employee's communication interface in real time with low latency, usually in the form of sidebar suggestions, quick reply list, or smart pop-up.

[0096] The system continuously monitors the progress of conversations and real-time customer feedback. Once it detects new customer input that changes the communication context (e.g., a change in customer intent, the raising of new objections, or a change in mood), the system immediately re-performs context awareness, knowledge retrieval, and fit calculation, dynamically updating the recommendation list to ensure that the auxiliary information provided to employees is always highly relevant and effective to the latest state of the conversation.

[0097] The AI-based marketing knowledge base construction method provided in this embodiment, based on customer profiles and a comprehensive marketing knowledge base, intelligently matches product highlights, features, use cases, or solutions most relevant to the customer's current focus and needs during real-time communication between human marketers and customers. This information is then pushed to employees in a way that is easy for human marketers to understand and present to customers, helping them to provide accurate and attractive product introductions, reducing customer comprehension costs, and enhancing product marketing appeal and conversion rates.

[0098] This embodiment provides a method for constructing a marketing knowledge base based on artificial intelligence, which can be used in the aforementioned computer system. Figure 4 This is a flowchart of a marketing knowledge base construction method based on artificial intelligence according to an embodiment of the present invention, such as... Figure 4 As shown, the process includes the following steps:

[0099] Step S201: Obtain basic information and session information, and convert the session information into text data according to the basic information.

[0100] Specifically, step S201 includes:

[0101] Step S2011: Collect raw attribute information from the customer management system, employee management system, and product system; perform text preprocessing on the raw attribute information to obtain basic information, which includes customer information, employee information, and product information.

[0102] Specifically, for customer information, standard application programming interfaces (APIs) are used to interface with the enterprise's existing customer relationship management (CRM) system, e-commerce platform backend, marketing automation platform, etc., to collect key customer behavior data according to a preset data synchronization frequency (e.g., full / incremental synchronization every morning, with key behavior data such as browsing / clicking / adding to cart synchronized in real-time / near real-time). Collected fields include, but are not limited to: user ID, name, contact information, region, demographic information, registration time, most recent login time, historical order details (product, amount, time, status), viewed product / page ID, page dwell time, search keywords, click behavior (buttons, links), marketing campaign participation records, points / membership level.

[0103] For employee information, the system retrieves basic employee information (employee ID, name, department, position, and start date) from the enterprise HR system via API or file import, sales performance (sales revenue, number of orders closed, average order value, sales cycle, customer development volume, and win rate) from the sales force automation (SFA) system, and customer satisfaction ratings from the after-sales feedback system. To build more accurate employee profiles, the system also collects employees' historical communication records (including recorded and transcribed calls with customers, internal chats, emails, etc.) and performs AI analysis on them. An interface is provided allowing employees to self-assess their communication style, or allowing team leaders and internal trainers to evaluate and input employee communication styles and professional skills.

[0104] For product information, product information is synchronized from the Product Information Management (PIM) system or a structured product database. Fields include: product ID, name, model, category, minimum stock keeping unit (SKU), detailed functional description, technical parameters (in tabular form), applicable scenarios, target customer groups, selling points, frequently asked questions, user manual, product image / video links, market positioning, and competitor comparison information. Updates to the product information database are monitored regularly, and the product knowledge update process is automatically triggered.

[0105] The collected structured and unstructured data are cleaned, denoised, converted into formats, and standardized to ensure that the data is accurate and usable. This process is a mature existing technology and will not be described in detail here.

[0106] Step S2012: Collect raw conversation data from the voice communication scenario, and perform structured cleaning, speech-to-text conversion and text preprocessing on the raw conversation data to obtain conversation information.

[0107] Specifically, for communication records, text chat logs between customers and customer service or sales personnel are collected by integrating with or monitoring customer service systems, sales call systems, and online chat tools (such as WeChat Work, DingTalk, and Web Chat), and the time and participants of the call / chat are recorded. For voice calls, if legal, compliant, and authorized, the call recording file is also uploaded.

[0108] In widely used audio and video conferencing systems within enterprises (such as Zoom, Lark Meeting, DingTalk Meeting, etc.), develop or integrate meeting recording plugins. The meeting initiator or designated personnel can enable the recording function, and the recorded files will be automatically uploaded to the system after the meeting ends.

[0109] For enterprises using softphones, call metadata (caller / called number, call time, call duration) can be obtained through API technology, and call recordings can be obtained provided compliance and authorization requirements are met. Before implementing voice data collection, enterprises must develop detailed data security and privacy policies, clearly inform employees and relevant customers (e.g., in the service agreement or opening line of the call, informing them that recordings may be used for service improvement), and ensure compliance with relevant Chinese data security and personal information protection laws and regulations. Collected voice data must be encrypted, stored, and access controlled.

[0110] The collected voice data is automatically converted into text data through speech recognition.

[0111] Step S2013: Based on customer information and employee information, divide the conversation information into text data of different speakers.

[0112] Specifically, based on customer and employee information, the conversation information is processed by speaker separation, noise reduction, and sentence segmentation to provide high-quality input for subsequent AI model processing. Speaker Diarization toolkits (such as PyAudioAnalysis, Kaldi) can be used to separate and label different speakers in the same recording (e.g., [Employee]: "Hello", [Customer]: "Hello"). This is just an example, but not a limitation.

[0113] The AI-based marketing knowledge base construction method provided in this embodiment collects and preprocesses raw attribute information from multiple sources to integrate and structure customer, employee, and product data, improving information integrity and usability. It cleans, converts speech to text, and preprocesses text data to transform unstructured speech into analyzable text data, laying the foundation for knowledge extraction. By segmenting speaker text data based on customer and employee information, it can accurately distinguish the subjects of the conversation (such as employees and customers), facilitating subsequent analysis of the communication patterns, intentions, and verbal characteristics of different roles.

[0114] Step S202: Analyze the implicit relationship data between employees, products, and customers based on basic information, and extract marketing knowledge from the text data.

[0115] Specifically, step S202 above, which analyzes implicit relationship data between employees, products, and customers based on basic information, includes:

[0116] Step S2021: Construct an employee-product-customer relationship graph, where nodes include employee profile features, product feature tags, and customer attribute tags, and edges represent interaction relationships and weights.

[0117] Specifically, by using association graph analysis to examine customer behavior and purchase history, we can identify which product features or solutions customers are interested in. We then construct a classification or regression model, taking into input employee profile features, script text features, customer profile features, and communication context features. The model outputs the predicted effect of the script being used by that employee on the customer in the current context (e.g., increased probability of positive customer feedback, increased probability of closing the deal). The model's training data comes from historical communication records and results.

[0118] Step S2022: Analyze the implicit relationships between nodes in the association graph using graph neural networks or collaborative filtering algorithms.

[0119] Specifically, machine learning algorithms (such as decision trees, support vector machines, and neural networks) are used to analyze structured data and extract knowledge to build predictive models. For example, these models can predict a customer's level of interest in a specific product feature or the effectiveness of a particular sales pitch for a specific customer. Furthermore, they can learn to construct association rules, such as identifying which product features customers with specific behavioral patterns frequently focus on.

[0120] In some optional implementations, step S2022 above includes:

[0121] Step a1: Based on employees' historical communication records and transaction data, establish a regression model of employee style and the effectiveness of their communication skills.

[0122] Specifically, historical communication records and transaction data of employees are collected. The communication records are preprocessed to extract features such as keywords and semantic information; the transaction data is structured. Then, an appropriate regression model is selected, such as linear regression, logistic regression, or decision tree regression, with employee communication style (such as tone of voice, expression habits, etc.), content of the script, and customer attributes as independent variables, and the effectiveness of the script (such as conversion rate, customer satisfaction, etc.) as the dependent variable. The regression model is trained, and the model is optimized by continuously adjusting the parameters to analyze the degree of influence of each factor on the effectiveness of the script.

[0123] Step a2: Identify the matching patterns between product features and customer needs through customer purchase behavior sequences.

[0124] Specifically, customer purchase behavior sequence data is cleaned and processed to remove noise and outliers. Sequence pattern mining algorithms, such as PrefixSpan and GSP, are used to analyze the order and time intervals of customer product purchases to identify statistically significant purchase patterns. These patterns are then linked to product characteristics by combining them with product feature tags, identifying the preferred combinations of product features for different customer groups in various scenarios. For example, it was found that younger customers tend to prefer product combinations that offer high cost-performance and stylish appearance when purchasing electronic products, thus building a matching pattern library between product features and customer needs.

[0125] Step a3: Based on the characteristics of the communication context, predict the probability increase of a specific employee using specific language to close a deal with a specific customer.

[0126] Specifically, a multi-dimensional feature vector is constructed by integrating historical employee communication data, customer information, and communication context data. Communication context features include communication time, communication channel, and the urgency of the customer's current needs. Using classification or regression models in machine learning, such as random forests and neural networks, historical transaction data is used as training samples. Employee style, script content, customer attributes, and communication context features are taken as input, and the transaction result is used as output for model training. After training, by inputting feature information of a specific employee, a specific script, a specific customer, and their communication context, the model can predict the probability increase of closing a deal with that customer using that script, providing data support for employees to optimize their script strategies.

[0127] The AI-based marketing knowledge base construction method provided in this embodiment constructs a dual-precise profile by deeply analyzing multi-dimensional data of human employees and customers. Combined with real-time communication context, the AI ​​system intelligently judges and recommends highly personalized operational scripts, product knowledge, solutions, etc., that are most suitable for the current scenario and the individual's style to human marketers. This significantly improves the adaptability and effectiveness of marketing communication, helps employees with different styles to conduct marketing more effectively, reduces manual maintenance costs, and ensures the timeliness and effectiveness of knowledge.

[0128] In some optional implementations, the text data includes conversation text data and meeting text data. Step S202 above, which involves extracting marketing knowledge from the text data, includes:

[0129] Step S2023: Using a sequence labeling model, identify functional dialogue fragments in the dialogue text data. Functional dialogue fragments include: opening remarks, product introduction sentences, objection handling sentences, and sales promotion sentences.

[0130] Specifically, for speech-to-text data, deep learning-based sequence labeling models are used to identify key dialogue segments such as utterances, questions, and answers. Graph neural networks or Transformer models are used to analyze the dialogue structure and context, extracting key events and relationships, such as identifying a successful objection handling process or a clever product analogy. Deep learning-based sequence labeling models (such as BiLSTM-CRF and Transformer) are used to identify functionally specific segments in the dialogue text, such as opening remarks, product introductions, objection handling sentences, and sales-driving phrases.

[0131] Step S2024: Using event extraction and relation extraction models, analyze the entity relationships and event sequences in the dialogue text data. The event sequences include: customer questions, objections, objection resolution, and transaction events.

[0132] Specifically, using event extraction and relation extraction models, we analyze the key entities (customers, products, issues) in the dialogue, the relationships between them, and the events that occur (customer questions, employee answers, objections, objection resolution, and successful closing).

[0133] Step S2025: Obtain dialogue result data and filter out dialogue patterns and speech fragments that are strongly correlated with positive results from the dialogue result data.

[0134] Specifically, by combining dialogue outcomes (such as whether a deal was ultimately closed and customer satisfaction), we identify dialogue patterns, script segments, and processing flows that are strongly correlated with positive outcomes, and treat them as "highly effective" practical experience and knowledge. For example, we identify the dialogue path and key scripts of an employee who ultimately closed a deal by emphasizing the long-term value of the product when faced with the objection of "the price is too high," and add this as a knowledge segment in the "objection handling skills - price" database.

[0135] Step S2026: Use the topic recognition model to identify the core topics in the meeting text data, use the text summarization model to generate meeting minutes, and combine the core topics to determine strategy adjustments, market insights, or customer feedback summaries.

[0136] Specifically, topic models (such as LDA and Top2Vec) are used to identify the core topics discussed in the meeting. Text summarization models (extractive or abstractive) are used to generate meeting minutes or summaries of key conclusions, identifying strategic adjustments, market insights, customer feedback summaries, etc.

[0137] Step S2027: Based on the dialogue pattern and corresponding script segments, strategy adjustments, market insights or customer feedback, summarize marketing knowledge for different scenarios.

[0138] Specifically, different marketing knowledge is developed for different scenarios. For example, when expressions such as "the price is too high" or "more expensive than competitors" appear in the conversation, the system identifies the type of customer objection through event extraction. Combined with the employee's speech fragment "emphasizing the product's lifetime warranty and energy saving" and the final transaction result, the system extracts the knowledge fragment of "cost breakdown method" - "using the formula 'daily cost = total price / years of use' to compare the short-term low price of competitors with the long-term value of this product (such as 'spending an extra 2 yuan per day to get 3 years of free maintenance and 30% energy saving')", and marks it as "highly effective". This is just an example, but not a limitation.

[0139] The AI-based marketing knowledge base construction method provided in this embodiment includes unstructured voice data generated in the daily marketing operations of enterprises (such as internal communications, meetings, and sales calls with customers) in the core knowledge collection scope. It uses advanced AI technology to deeply mine and extract the valuable practical experience, implicit skills, and contextualized cases contained therein, and performs systematic classification and tag management. This transforms past operational experience into reusable structured knowledge assets, directly providing marketing personnel with practice-verified auxiliary suggestions, effectively improving the overall practical level of the team.

[0140] Step S203 involves hierarchically classifying and storing implicit relationship data and marketing knowledge, and assigning multi-dimensional tags to different knowledge items to form a marketing knowledge base. For details, please refer to [link to details]. Figure 1 Step S103 of the illustrated embodiment will not be described again here.

[0141] The AI-based marketing knowledge base construction method provided in this embodiment integrates and structures multi-source data to uncover potential connections between employees, products, and customers, and accurately extracts practical marketing knowledge. Hierarchical storage makes the knowledge system clearer, and multi-dimensional tags improve retrieval efficiency. The dynamic update mechanism ensures the timeliness of knowledge, provides accurate knowledge support for marketing scenarios, helps employees quickly match scripts and solutions, and improves communication efficiency and conversion rate.

[0142] This embodiment also provides an artificial intelligence-based marketing knowledge base construction device, which is used to implement the above embodiments and preferred embodiments; details already described will not be repeated. As used below, the term "module" can be a combination of software and / or hardware that implements a predetermined function. Although the device described in the following embodiments is preferably implemented in software, hardware implementation, or a combination of software and hardware, is also possible and contemplated.

[0143] This embodiment provides a marketing knowledge base construction device based on artificial intelligence, such as... Figure 5 As shown, it includes:

[0144] The data acquisition module 501 is used to acquire basic information and session information, and convert the session information into text data according to the basic information.

[0145] The knowledge extraction module 502 is used to analyze the implicit relationship data between employees, products and customers based on basic information, and to extract marketing knowledge from text data.

[0146] The knowledge base construction module 503 is used to hierarchically classify and store implicit relational data and marketing knowledge, and to attach multi-dimensional tags to different knowledge to form a marketing knowledge base.

[0147] Further functional descriptions of the above modules and units are the same as those in the corresponding embodiments described above, and will not be repeated here.

[0148] In this embodiment, the marketing knowledge base construction device based on artificial intelligence is presented in the form of functional units. Here, a unit refers to an ASIC (Application Specific Integrated Circuit) circuit, a processor and memory that execute one or more software or fixed programs, and / or other devices that can provide the above functions.

[0149] This invention also provides a computer device having the above-described features. Figure 5 The device shown is an AI-based marketing knowledge base construction device.

[0150] Please see Figure 6 , Figure 6 This is a schematic diagram of the structure of a computer device provided in an optional embodiment of the present invention, such as... Figure 6 As shown, the computer device includes one or more processors 10, memory 20, and interfaces for connecting the components, including high-speed interfaces and low-speed interfaces. The components communicate with each other via different buses and can be mounted on a common motherboard or otherwise installed as needed. The processors can process instructions executed within the computer device, including instructions stored in or on memory to display graphical information of a GUI on external input / output devices (such as display devices coupled to the interfaces). In some alternative implementations, multiple processors and / or multiple buses can be used with multiple memories and multiple memory modules, if desired. Similarly, multiple computer devices can be connected, each providing some of the necessary operations (e.g., as a server array, a group of blade servers, or a multiprocessor system). Figure 6 Take a processor 10 as an example.

[0151] Processor 10 may be a central processing unit, a network processor, or a combination thereof. Processor 10 may further include a hardware chip. The hardware chip may be an application-specific integrated circuit (ASIC), a programmable logic device (PLD), or a combination thereof. The programmable logic device may be a complex programmable logic device (CAMP), a field-programmable gate array (FPGA), a general-purpose array logic (GDA), or any combination thereof.

[0152] The memory 20 stores instructions executable by at least one processor 10 to cause the at least one processor 10 to perform the method shown in the above embodiments.

[0153] The memory 20 may include a program storage area and a data storage area. The program storage area may store the operating system and applications required for at least one function; the data storage area may store data created based on the use of the computer device. Furthermore, the memory 20 may include high-speed random access memory and may also include non-transitory memory, such as at least one disk storage device, flash memory device, or other non-transitory solid-state storage device. In some alternative embodiments, the memory 20 may optionally include memory remotely located relative to the processor 10, and these remote memories may be connected to the computer device via a network. Examples of such networks include, but are not limited to, the Internet, intranets, local area networks, mobile communication networks, and combinations thereof.

[0154] The memory 20 may include volatile memory, such as random access memory; the memory may also include non-volatile memory, such as flash memory, hard disk or solid-state drive; the memory 20 may also include a combination of the above types of memory.

[0155] The computer device also includes a communication interface 30 for communicating with other devices or communication networks.

[0156] This invention also provides a computer-readable storage medium. The methods described above according to embodiments of the invention can be implemented in hardware or firmware, or implemented as computer code that can be recorded on a storage medium, or implemented as computer code downloaded via a network and originally stored on a remote storage medium or a non-transitory machine-readable storage medium and then stored on a local storage medium. Thus, the methods described herein can be processed by software stored on a storage medium using a general-purpose computer, a dedicated processor, or programmable or dedicated hardware. The storage medium can be a magnetic disk, optical disk, read-only memory, random access memory, flash memory, hard disk, or solid-state drive, etc.; further, the storage medium can also include combinations of the above types of memory. It is understood that computers, processors, microprocessor controllers, or programmable hardware include storage components capable of storing or receiving software or computer code, which, when accessed and executed by the computer, processor, or hardware, implements the methods shown in the above embodiments.

[0157] Although embodiments of the invention have been described in conjunction with the accompanying drawings, those skilled in the art can make various modifications and variations without departing from the spirit and scope of the invention, and such modifications and variations all fall within the scope defined by the appended claims.

Claims

1. A method for constructing a marketing knowledge base based on artificial intelligence, characterized in that, The method includes: Obtain basic information and session information, and convert the session information into text data according to the basic information; Based on the basic information, analyze the implicit relationship data between employees, products, and customers, and extract marketing knowledge from the text data; The implicit relationship data and marketing knowledge are hierarchically classified and stored, and different knowledge is labeled with multi-dimensional tags to form a marketing knowledge base.

2. The method according to claim 1, characterized in that, The method further includes: Acquire real-time conversations and use intelligent recommendation models to retrieve multiple relevant knowledge fragments from the marketing knowledge base and recommend them to the conversation employees; Obtain the conversation content of employees and the feedback content of customers, evaluate the effectiveness of knowledge fragment recommendations, and optimize the marketing knowledge base based on the effectiveness evaluation results.

3. The method according to claim 1, characterized in that, The process of acquiring basic information and session information, and converting the session information into text data according to the basic information, includes: Raw attribute information is collected from the customer management system, employee management system, and product system. The raw attribute information is preprocessed into text to obtain basic information, which includes customer information, employee information, and product information. Raw conversation data is collected from voice communication scenarios, and the raw conversation data is subjected to structured cleaning, speech-to-text conversion, and text preprocessing to obtain conversation information. Based on the customer information and employee information, the conversation information is divided into text data from different speakers.

4. The method according to claim 1, characterized in that, Based on the aforementioned basic information, we analyze the implicit relationship data between employees, products, and customers, including: Construct an employee-product-customer relationship graph, where nodes include employee profile features, product feature tags, and customer attribute tags, and edges represent interaction relationships and weights; The implicit relationships between nodes in the association graph are analyzed using graph neural networks or collaborative filtering algorithms.

5. The method according to claim 4, characterized in that, Using graph neural networks or collaborative filtering algorithms, the implicit relationships between nodes in the association graph are analyzed, including: Based on employees' historical communication records and transaction data, a regression model was established to link employee style with the effectiveness of their communication techniques. Identify the matching patterns between product features and customer needs by analyzing customer purchase behavior sequences; By combining the characteristics of the communication context, we can predict the probability increase of a specific employee using specific language to close a deal with a specific customer.

6. The method according to claim 1, characterized in that, The text data includes conversation text data and meeting text data. Marketing knowledge is extracted from the text data, including: Using a sequence labeling model, functional dialogue segments in dialogue text data are identified. These functional dialogue segments include: opening remarks, product introduction sentences, objection handling sentences, and sales promotion sentences. Using event extraction and relation extraction models, we analyze entity relationships and event sequences in dialogue text data. The event sequences include: customer questioning, objection generation, objection resolution, and transaction events. Acquire dialogue result data, and filter out dialogue patterns and speech fragments that are strongly correlated with positive results from the dialogue result data; The core topics in the meeting text data are identified using a topic recognition model, the meeting minutes are generated using a text summarization model, and the core topics are combined to determine strategy adjustments, market insights, or customer feedback summaries. Based on the aforementioned dialogue patterns and corresponding script segments, strategy adjustments, market insights, or customer feedback, marketing knowledge for different scenarios is summarized.

7. The method according to claim 2, characterized in that, Acquire real-time conversations and leverage intelligent recommendation models to retrieve multiple relevant knowledge fragments from the marketing knowledge base and recommend them to the conversation participants, including: Real-time acquisition of text streams or speech-to-text streams of conversations between employees and customers; use low-latency natural language processing models for word segmentation, entity recognition, sentiment analysis, and intent recognition to determine customer sentiment and intent. Based on the remote dictionary service cache, dynamically update customer profiles containing customers' current intentions and emotions, and load employee profiles; When a customer finishes speaking or an employee enters text, context recognition is triggered to determine the fine-grained communication scenario, and matching knowledge fragments are retrieved from the marketing knowledge base in combination with customer and employee profiles. The system calculates the fit score between knowledge fragments and the current "customer-employee-context" combination using a deep ranking model or rule engine, sorts the knowledge fragments based on the fit score, pushes the most relevant knowledge fragments to the communication interface, and updates the recommendation list in real time as the conversation context changes.

8. A marketing knowledge base construction device based on artificial intelligence, characterized in that, The device includes: The data acquisition module is used to acquire basic information and session information, and convert the session information into text data according to the basic information. The knowledge extraction module is used to analyze the implicit relationship data between employees, products, and customers based on the basic information, and extract marketing knowledge from the text data. The knowledge base construction module is used to hierarchically classify and store the implicit relationship data and the marketing knowledge, and to attach multi-dimensional tags to different knowledge to form a marketing knowledge base.

9. A computer device, characterized in that, include: A memory and a processor, the memory and the processor being communicatively connected to each other, the memory storing computer instructions, the processor executing the computer instructions to perform the method of any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions for causing the computer to perform the method of any one of claims 1 to 7.

Images