Marketing knowledge base construction system and method based on large model

Abstract

The invention provides a marketing knowledge base construction system and method based on a large model, and the method comprises the steps: firstly determining a knowledge construction demand in an automobile marketing field, generating a marketing knowledge demand description, inputting the marketing knowledge demand description into a pre-training large model to generate candidate knowledge contents, and then carrying out the evaluation of the candidate knowledge contents according to a preset knowledge quality evaluation standard, the method comprises the steps of generating an evaluation result, inputting marketing knowledge demand description, candidate knowledge content and the evaluation result into a large model for instruction fine tuning to obtain optimized knowledge content, and finally performing structured organization on the optimized knowledge content to generate an automobile marketing knowledge unit and adding the automobile marketing knowledge unit to a knowledge base. Therefore, the construction efficiency and quality of the automobile marketing knowledge base can be effectively improved, and a systematic, accurate and practical knowledge base is provided for automobile marketing.

Description

Technical Field

[0001] The present invention relates to the field of large model technology, and in particular to a system and method for constructing a marketing knowledge base based on a large model. Background Art

[0002] In the increasingly competitive automotive industry, automotive marketing plays a crucial role in boosting product sales and brand influence. Traditional approaches to acquiring and managing automotive marketing knowledge have numerous limitations. First, marketers often rely on their own experience or fragmented data to acquire marketing knowledge, which lacks systematicity and comprehensiveness, making it difficult to cover all aspects of automotive marketing, such as the precise positioning of different vehicle models, the development of diverse marketing strategies, and communication skills for different customer segments. Second, existing knowledge management systems, mostly based on fixed rules and templates, lack flexibility and intelligence, and are unable to timely update and optimize knowledge content based on evolving marketing needs and market dynamics. Furthermore, the lack of effective mechanisms for evaluating and optimizing the quality of generated marketing knowledge leads to the potential for inaccuracy, incompleteness, or inapplicability of this knowledge, which in turn impacts the effectiveness and efficiency of automotive marketing. Summary of the Invention

[0003] In view of the above-mentioned problems, in combination with the first aspect of the present invention, an embodiment of the present invention provides a method for constructing a marketing knowledge base based on a large model, the method comprising: Determine the knowledge construction needs in the automotive marketing field and generate a marketing knowledge needs description, which includes the marketing subject scope, knowledge content type, and knowledge application scenario limitations; Input the marketing knowledge demand description into the pre-trained big model, call the text generation interface of the big model to perform preliminary knowledge content generation processing, and obtain candidate knowledge content corresponding to the marketing knowledge demand description; Acquiring a preset knowledge quality assessment standard, performing a knowledge quality assessment process on the candidate knowledge content according to the knowledge quality assessment standard, and generating a knowledge quality assessment result; Inputting the marketing knowledge requirement description, the candidate knowledge content, and the knowledge quality assessment result into the macro model, calling the instruction fine-tuning interface of the macro model to perform knowledge content optimization processing, and obtaining optimized knowledge content that meets the knowledge quality assessment standard; The optimized knowledge content is subjected to knowledge structuring and organization processing to generate an automobile marketing knowledge unit including a knowledge subject identifier, a content hierarchical relationship, and associated index information, and the automobile marketing knowledge unit is added to an automobile marketing knowledge base.

[0004] On the other hand, an embodiment of the present invention also provides a marketing knowledge base construction system based on a large model, including a processor and a machine-readable storage medium, wherein the machine-readable storage medium is connected to the processor, the machine-readable storage medium is used to store programs, instructions or codes, and the processor is used to execute the programs, instructions or codes in the machine-readable storage medium to implement the above method.

[0005] Based on the above aspects, the embodiment of the present invention determines the knowledge construction needs in the field of automotive marketing and generates a detailed description, and uses the text generation interface of the pre-trained large model to preliminarily generate candidate knowledge content, fully leveraging the large model's powerful language understanding and generation capabilities, and can quickly obtain rich and diverse knowledge information. The preset knowledge quality assessment standards are used to evaluate the candidate knowledge content, and the quality problems of the knowledge content can be accurately identified. The marketing knowledge demand description, candidate knowledge content and knowledge quality assessment results are jointly input into the large model for instruction fine-tuning, thereby achieving targeted optimization of the knowledge content, obtaining optimized knowledge content that meets the quality standards, and improving the accuracy and practicality of the knowledge. Finally, the optimized knowledge content is structured and organized to generate automotive marketing knowledge units and add them to the knowledge base, making knowledge management more systematic and orderly, facilitating rapid retrieval and utilization by marketing personnel, and significantly improving the construction efficiency and quality of the automotive marketing knowledge base. BRIEF DESCRIPTION OF THE DRAWINGS

[0006] Figure 1 It is a schematic diagram of the execution flow of the method for constructing a marketing knowledge base based on a large model provided by an embodiment of the present invention.

[0007] Figure 2 Schematic diagram of exemplary hardware and software components of a marketing knowledge base construction system based on a large model provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0008] The present invention will be described in detail below with reference to the accompanying drawings. Figure 1 This is a flow chart of a method for constructing a marketing knowledge base based on a large model provided by an embodiment of the present invention. The method for constructing a marketing knowledge base based on a large model is introduced in detail below.

[0009] Step S110: Determine the knowledge construction requirements in the field of automobile marketing and generate a marketing knowledge requirement description, which includes the marketing subject scope, knowledge content type, and knowledge application scenario limitations.

[0010] This example focuses on the need to build knowledge related to battery life and charging facilities in the new energy vehicle marketing field. The automotive marketing department has discovered in its daily marketing activities that customers are extremely interested in the battery life of new energy vehicles and the distribution of surrounding charging facilities. Existing knowledge reserves cannot fully answer these questions, leading to the need to build a relevant knowledge base.

[0011] Step S111: obtaining a preliminary knowledge requirement document provided by the automobile marketing business department, performing text parsing on the preliminary knowledge requirement document, and extracting content type indication information therein, wherein the content type indication information includes requirement keywords, subject phrases, and scenario description paragraphs.

[0012] Obtain a preliminary knowledge requirements document from the business department. This document, in natural language, outlines the general requirements for "battery life and charging facility related knowledge." Use a rule-based text parsing method to process the document, first segmenting the text into sentences based on punctuation and then performing part-of-speech tagging on each sentence.

[0013] For the extraction of demand keywords, a preset domain vocabulary is used for matching. The domain vocabulary includes words such as "new energy vehicles", "battery life", "charging facilities", "range", "charging piles", "fast charging", and "slow charging". These demand keywords are identified and extracted from sentence units.

[0014] For the extraction of topic phrases, by analyzing the subject-predicate-object structure of the sentence, we identify phrases composed of multiple words that express complete topic meanings, such as "factors affecting battery life", "charging facility coverage", "charging strategies corresponding to different driving ranges", etc.

[0015] Scenario description paragraphs are extracted by identifying a set of consecutive sentences containing elements such as time, place, and behavior. For example, "At the new energy vehicle promotion conference for family users, it is necessary to explain in detail the daily usage scenarios corresponding to different battery lifespans and the accessibility of surrounding charging facilities" and "For customers who frequently travel long distances, it is necessary to provide battery life planning based on their frequent routes and relevant knowledge of the distribution of charging facilities along the way."

[0016] Step S112: performing word frequency statistics and semantic clustering processing on the extracted demand keywords, identifying core demand topics and secondary demand topics, and generating a demand topic hierarchical structure.

[0017] We perform a word frequency count on the extracted demand keywords. This count does not involve specific numerical values; instead, we determine their importance by comparing how often they appear in documents. We found that "battery life" and "charging facilities" appear relatively frequently.

[0018] Semantic clustering is then performed using a clustering algorithm based on semantic similarity. Each demand keyword is first converted into a vector. The dimension of the vector is determined by the semantic features of the word, with different words corresponding to different combinations of semantic features. The semantic similarity between any two keyword vectors is calculated. Semantic similarity is measured by the degree of correlation between the vectors; the higher the correlation, the closer the semantics.

[0019] Keywords are divided into different clustering groups based on semantic similarity. For example, “battery life”, “range” and “endurance” are clustered into one group, and “charging facilities”, “charging piles”, “fast charging stations” and “slow charging stations” are clustered into another group.

[0020] Based on the clustering results, the core demand theme is identified as "the correlation and application of battery life and charging facilities", and the secondary demand themes include "factors affecting battery life", "types and characteristics of charging facilities", "differences in battery life and charging needs of different user groups", etc.

[0021] When generating a hierarchical structure of demand topics, the core demand topic is used as the top-level node, and each secondary demand topic is used as a child node of the core demand topic. Each secondary demand topic can be further subdivided into the next level node according to the specific content it contains, forming a hierarchical tree structure.

[0022] Step S113: performing scene feature extraction processing on the extracted scene description paragraphs to determine scene feature parameters of knowledge application, wherein the scene feature parameters include specific business scenarios, target user groups, and expected application methods.

[0023] Scene feature extraction is performed on the extracted scene description paragraphs. First, the paragraphs are segmented and semantically understood to identify the feature information related to the scene.

[0024] Specific business scenarios are determined by extracting the activity types and occasions described in the paragraphs, such as “family user promotion meeting”, “long-distance travel customer consultation”, “4S store daily reception”, etc.

[0025] The identification of target user groups is based on the description of user characteristics in the paragraph, such as "family users", "customers who frequently travel long distances", "city commuters", etc. These user groups differ in their battery life and charging facility needs.

[0026] Determining the expected application method is to analyze how the knowledge will be used, such as "sales staff verbally explaining to customers", "text description in marketing materials", "automatic reply on online consultation platform", etc.

[0027] The extracted information is integrated into scenario feature parameters, each of which contains multiple specific description items, which together constitute the complete characteristics of the knowledge application scenario.

[0028] Step S114: Integrate the demand subject hierarchical structure, the content type indication information and the scenario characteristic parameters to generate a preliminary marketing knowledge demand description.

[0029] Integrate the required subject hierarchical structure, content type indication information and scenario feature parameters, and ensure the logical consistency between the various parts of information during the integration process.

[0030] Using the demand subject hierarchy structure as the framework, the demand keywords and subject phrases in the content type indication information are mapped to the corresponding subject nodes. Then, combined with the requirements of different scenarios in the scenario feature parameters, the knowledge content that needs to be covered in each scenario is clarified.

[0031] For example, in the "Home User Promotion Meeting" scenario, combining the core demand themes and secondary demand themes, it is clear that knowledge content needs to include the battery life requirements of home users' daily travel, the distribution of charging facilities in the community and surrounding areas, etc.

[0032] Through the above integration processing, a preliminary description of marketing knowledge needs is formed. This marketing knowledge needs description is presented in natural language and comprehensively covers the subject scope, content type and application scenario limitations of knowledge construction.

[0033] Step S115: Feedback the preliminary marketing knowledge requirement description to the automobile marketing business department for requirement confirmation, receive the requirement modification opinions returned by the business department, iteratively adjust the preliminary marketing knowledge requirement description according to the requirement modification opinions, and generate a final marketing knowledge requirement description.

[0034] The initial description of marketing knowledge needs is fed back to the automotive marketing business department, which then organizes relevant personnel to review it. During the review process, the business personnel may propose revisions, such as "In long-distance travel scenarios, the impact of different weather conditions on battery life and corresponding charging facility response strategies need to be supplemented" or "Analysis of the correlation between charging facility usage costs and battery life needs to be added."

[0035] After receiving these revision suggestions, we made adjustments to the initial marketing knowledge requirements description point by point. New content requirements were incorporated into the corresponding themes and scenarios, and inaccurate or incomplete sections were corrected and supplemented.

[0036] After multiple iterations and adjustments, the final marketing knowledge demand description is generated until the business department confirms that the description accurately reflects the knowledge construction needs.

[0037] Step S120: Input the marketing knowledge requirement description into a pre-trained large model, and call the text generation interface of the large model to perform preliminary knowledge content generation processing, so as to obtain candidate knowledge content corresponding to the marketing knowledge requirement description.

[0038] Input the finally generated marketing knowledge requirement description into a pre-trained large model, which is trained on a large amount of text data and has strong text understanding and generation capabilities. Call the text generation interface of the large model. The interface receives text-form input and outputs corresponding text content according to the preset generation logic.

[0039] When performing the preliminary knowledge content generation processing, the large model will deeply understand the marketing knowledge requirement description, combine the relevant knowledge stored in it, and generate knowledge content corresponding to the requirement description, that is, candidate knowledge content. The candidate knowledge content covers all aspects related to the association between battery life and charging facilities and is presented in the form of natural language text.

[0040] Step S121: Perform demand feature vectorization processing on the marketing knowledge requirement description, and convert the demand description in text form into a demand feature vector that conforms to the input format of the large model.

[0041] Perform demand feature vectorization processing on the marketing knowledge requirement description. First, preprocess the text, including removing stop words, performing word form reduction, etc. Stop words refer to those words that frequently appear in the text but have little impact on semantic expression, such as "de", "zai", "he", etc.

[0042] Then, use word embedding technology to convert each word in the processed text into a vector form. The dimension of each word vector is determined by the parameters of the word embedding model, and different words correspond to different vector representations.

[0043] Arrange the vectors of all words in the text in the order in which they appear in the text to form a matrix, and then process the matrix through a pooling operation to obtain a vector with a fixed dimension, that is, the demand feature vector. This demand feature vector can reflect the semantic information of the marketing knowledge requirement description and meets the requirements of the large model for the input format.

[0044] Step S122: Call the domain knowledge awakening module of the large model, associate and match the demand feature vector with the pre-trained knowledge graph in the automotive marketing domain, and activate the domain knowledge parameters in the large model related to the marketing knowledge requirement description.

[0045] The domain knowledge awakening module of the large model is called. This module stores a pre-trained knowledge graph for the automotive marketing domain. The knowledge graph consists of entity nodes and relationship edges. Entity nodes include "new energy vehicle," "battery," "charging pile," and so on. Relationship edges represent the associations between entities, such as "battery includes range attribute," and "charging pile provides charging services for new energy vehicles."

[0046] The demand feature vector is associated and matched with the entity node vector in the knowledge graph, and the correlation between the demand feature vector and each entity node vector is calculated. The correlation is reflected by the degree of semantic correlation between the vectors.

[0047] Based on the degree of correlation, the domain knowledge parameters related to the description of marketing knowledge needs in the big model are activated. These parameters correspond to the relevant entity and relationship information in the knowledge graph, so that the big model can call more knowledge related to the field when generating knowledge content.

[0048] Step S123: configuring a generation parameter set of the text generation interface, wherein the generation parameter set includes a topic relevance weight, a content depth coefficient, a professional term density threshold, and an output length range.

[0049] Configure the generation parameter set of the text generation interface. The topic relevance weight is used to control the degree of relevance between the generated content and the marketing knowledge demand description topic. The higher the weight setting, the stronger the relevance of the generated content to the topic.

[0050] The content depth coefficient determines the level of detail and depth of the generated knowledge content. The larger the coefficient, the more in-depth and detailed the content is, and it can cover more details and the principles behind it.

[0051] The professional term density threshold is used to limit the proportion of professional terms in the generated text. It is set according to different knowledge application scenarios. For example, in scenarios targeting ordinary customers, the threshold is set lower and fewer professional terms are used; in scenarios targeting professionals, the threshold can be set higher.

[0052] The output length range specifies the approximate length of the generated text, ensuring that the generated knowledge content is neither too short, resulting in incomplete information, nor too long, affecting its usability.

[0053] These parameters are set through the configuration interface of the text generation interface to form a generation parameter set and passed to the interface.

[0054] Step S124: input the demand feature vector and the generation parameter set into the text generation interface, triggering the large model to execute the knowledge content generation operation to obtain the initial knowledge text.

[0055] The requirement feature vector and the generation parameter set are input into the text generation interface. The interface parses and processes the input information and converts it into an internal format that can be understood by the large model.

[0056] After receiving this information, the large model combines the activated domain knowledge parameters and executes knowledge content generation according to the requirements of the generation parameter set. During the generation process, the large model organizes the language based on the semantic information of the required feature vector to generate content related to the topic. It also controls the level of detail by referring to the content depth coefficient, uses an appropriate number of professional terms based on the professional term density threshold, and completes text generation within the output length range, ultimately producing the initial knowledge text.

[0057] Step S125: performing redundant information filtering on the initial knowledge text, deleting repeated expressions, extended contents irrelevant to the requirements, and contradictory contents with conflicting expressions, to obtain a purified intermediate knowledge text.

[0058] The initial knowledge text is filtered for redundant information. First, the text similarity calculation method is used to identify repeated expressions. Sentences or paragraphs with highly similar semantics are determined to be repeated content, and only one of them is retained.

[0059] Then, by comparing with the demand feature vector, the extended content that is not related to the marketing knowledge demand description is identified. Although these contents may involve related fields, they are beyond the scope of this knowledge construction and are deleted.

[0060] For conflicting content, we identify it through logical analysis. For example, if there are two completely opposite statements about the same problem, we need to make a judgment based on domain knowledge and demand scenarios, and delete the content that is illogical or does not match the requirements.

[0061] After the above filtering process, a purified intermediate knowledge text is obtained, which removes redundant and conflicting content while retaining the core information.

[0062] Step S126: Check the content integrity of the intermediate knowledge text. If there are obvious missing paragraphs, the missing paragraph identifier and the original requirement feature vector are re-entered into the text generation interface for supplementary generation. The supplementary generated content is spliced ​​and merged with the intermediate knowledge text to obtain candidate knowledge content.

[0063] Conduct content integrity checks on intermediate knowledge texts, and verify one by one whether the intermediate knowledge texts cover all necessary content based on the subject scope and content type specified in the marketing knowledge requirements description.

[0064] If any paragraph is found to be obviously missing content, such as not covering the topic of “the impact of charging speeds of different charging facilities on battery life,” the missing paragraph will be marked.

[0065] The missing paragraph identifier and the original requirement feature vector are re-entered into the text generation interface. The interface triggers the large model to perform supplementary generation operations according to the previous generation parameter set requirements to generate content related to the missing paragraph.

[0066] The supplementary content is spliced ​​and fused with the intermediate knowledge text. When splicing, the content is arranged in a logical order to ensure that the fused text is smooth and coherent, and finally the candidate knowledge content is obtained.

[0067] Step S130: obtaining a preset knowledge quality assessment standard, performing a knowledge quality assessment process on the candidate knowledge content according to the knowledge quality assessment standard, and generating a knowledge quality assessment result.

[0068] Obtain the preset knowledge quality evaluation standard, which is formulated based on the characteristics and application requirements of automotive marketing knowledge and includes multiple evaluation dimensions and corresponding evaluation indicators.

[0069] A comprehensive quality assessment of candidate knowledge content is conducted according to the knowledge quality assessment standards. During the assessment process, the candidate knowledge content is checked and scored according to the indicators of each dimension. Finally, the assessment results of each dimension are summarized to generate the knowledge quality assessment results.

[0070] Step S131: Retrieve the preset knowledge quality evaluation standard from the standard database of the knowledge management system, parse the knowledge quality evaluation standard, and determine the specific evaluation indicators and indicator weights contained in the content completeness evaluation dimension, the expression accuracy evaluation dimension, and the application adaptability evaluation dimension.

[0071] The pre-set knowledge quality assessment criteria, stored as structured data, are retrieved from the standard database of the knowledge management system. These criteria are parsed to extract the content completeness assessment dimension, the expression accuracy assessment dimension, and the application adaptability assessment dimension.

[0072] The content completeness evaluation dimension includes specific evaluation indicators such as knowledge point coverage and the degree of elaboration of each knowledge point; the expression accuracy evaluation dimension includes indicators such as the standardization of professional terminology use, logical coherence, and data accuracy; the application adaptability evaluation dimension includes indicators such as scenario matching, user applicability, and operability.

[0073] Each specific evaluation indicator has a corresponding indicator weight, which reflects the importance of the indicator in the evaluation dimension to which it belongs. The sum of the indicator weights is 1.

[0074] Step S132: With respect to the content completeness assessment dimension, a knowledge point coverage test is performed on the candidate knowledge content to identify whether the candidate knowledge content contains all the core knowledge points required in the marketing knowledge demand description, and the number of missing knowledge points and the score of the adequacy of the explanation of each knowledge point are counted.

[0075] Regarding the content completeness assessment dimension, we first need to identify all the core knowledge points required in the marketing knowledge demand description. These knowledge points are the key contents determined based on the hierarchical structure of the demand topics.

[0076] Perform knowledge point coverage detection on candidate knowledge content, identify the knowledge points contained in the candidate knowledge content through text matching and semantic understanding, and compare them with the core knowledge points to determine the core knowledge points that are included and the core knowledge points that are not included.

[0077] Count the number of missing knowledge points, that is, the number of core knowledge points that are not included. At the same time, evaluate each included core knowledge point based on the breadth and depth of its explanation and give it a score for adequacy of explanation. The score reflects the degree of completeness of the knowledge point's explanation.

[0078] Step S1321: extract core knowledge points from the marketing knowledge demand description, and determine the core knowledge point set corresponding to the marketing knowledge demand description in combination with the automotive marketing field knowledge system. Each core knowledge point includes a topic name, necessary elaboration aspects, and detail level requirements.

[0079] The core knowledge points of the marketing knowledge demand description are extracted and processed, and combined with the knowledge system in the field of automobile marketing, which includes various knowledge frameworks and classifications related to new energy vehicle marketing.

[0080] Through semantic analysis of marketing knowledge demand descriptions, we identify key concepts and core contents, match these contents with knowledge units in the domain knowledge system, and determine the corresponding core knowledge points.

[0081] Each core knowledge point includes a topic name, such as "Battery Range Calculation Methods," required explanations, such as "Factors Affecting Range" and "Range Performance Under Different Road Conditions," and required levels of detail, such as the need to explain basic principles and provide real-world examples. These constitute the core knowledge point set.

[0082] Step S1322: Construct a knowledge point recognition model, which uses a bidirectional long short-term memory network combined with a conditional random field algorithm to scan the candidate knowledge content section by section to identify the knowledge points already included in the candidate knowledge content and the actual elaboration of each knowledge point.

[0083] A knowledge point recognition model was constructed using a bidirectional long short-term memory network combined with a conditional random field algorithm. The bidirectional long short-term memory network consists of forward and backward long short-term memory layers, enabling bidirectional processing of text sequences and capturing contextual information.

[0084] The conditional random field algorithm is used to perform sequence labeling on the output results of the bidirectional long short-term memory network to determine which parts of the text belong to knowledge points and the boundaries of knowledge points.

[0085] When training the model, text data in the automotive marketing field with labeled knowledge points is used as the training set, and the model parameters are adjusted through the back-propagation algorithm so that the model can accurately identify knowledge points.

[0086] When scanning candidate knowledge content paragraph by paragraph, the text is segmented into multiple paragraphs, each of which serves as the model input. The model first performs word embedding on the paragraph, converting the vocabulary into vectors. These vectors are then input into a bidirectional long short-term memory network, which outputs the hidden state vector for each vocabulary word.

[0087] The conditional random field algorithm calculates the probability of the label sequence based on these hidden state vectors and selects the label sequence with the highest probability as the output, thereby identifying the knowledge points already included in the candidate knowledge content and the actual elaboration of each knowledge point.

[0088] Step S1323: Compare and match the identified included knowledge points with the core knowledge point set, determine the core knowledge points that are not included, and count the number of missing knowledge points.

[0089] The included knowledge points identified by the knowledge point recognition model are compared and matched with the core knowledge point set. The comparison is based on the subject name and explanation of the knowledge points.

[0090] For each core knowledge point, check whether there is a knowledge point with the same topic name and related explanation among the included knowledge points. If not, the core knowledge point is determined to be an unincluded core knowledge point.

[0091] The number of all core knowledge points that are not included is counted, that is, the number of missing knowledge points is obtained, which reflects the lack of coverage of core knowledge points by the candidate knowledge content.

[0092] Step S1324: For each included core knowledge point, match its actual elaboration aspects with the necessary elaboration aspects, and calculate the elaboration aspect coverage.

[0093] For each included core knowledge point, match its actual elaboration aspects with the necessary elaboration aspects. First, list all the necessary elaboration aspects of the core knowledge point, then list the actual elaboration aspects, and then count the number of actual elaboration aspects that match the necessary elaboration aspects.

[0094] The aspect coverage is calculated by dividing the number of successfully matched aspects by the total number of required aspects. For example, if a core knowledge point has five required aspects and only three of the required aspects match, the aspect coverage is 3 divided by 5.

[0095] Step S1325: Calculate the elaboration adequacy score of each knowledge point according to the elaboration aspect coverage and the detail level of each elaboration aspect according to the preset scoring rules. The elaboration adequacy score is positively correlated with the elaboration aspect coverage and the detail level.

[0096] The adequacy score is calculated based on the coverage of each aspect and the level of detail of each aspect. The pre-set scoring rules give a certain percentage of the score to the coverage of each aspect and another percentage to the level of detail of each aspect.

[0097] The level of detail is assessed based on the depth and breadth of information contained in each practical aspect, such as whether it includes explanations of principles, case analysis, data support, etc. The higher the level of detail, the higher the corresponding score.

[0098] The score corresponding to the coverage of the explanation aspect and the score corresponding to the level of detail are added together to obtain the explanation adequacy score of each knowledge point. The higher the explanation adequacy score, the more complete and comprehensive the explanation of the knowledge point is.

[0099] Step S1326: Record the number of missing knowledge points and the score of the adequacy of the explanation of each knowledge point in the content completeness assessment sub-result as a basis for calculating the comprehensive content completeness score.

[0100] The number of missing knowledge points and the adequacy of the explanation for each knowledge point are recorded in the content completeness assessment sub-result. This sub-result is presented in a structured format, clearly showing the coverage and adequacy of each core knowledge point. By comprehensively considering the number of missing knowledge points and the adequacy of the explanation for each knowledge point, a comprehensive assessment of the candidate knowledge content's performance in terms of content completeness is conducted.

[0101] Step S133: For the dimension of expression accuracy assessment, the candidate knowledge content is checked for terminology standardization, logical coherence, and data accuracy, and incorrect professional terms, logically contradictory sentences and paragraphs, and content fragments with inaccurate data expressions are marked.

[0102] Focusing on the accuracy of expression, candidate knowledge content is examined from three perspectives: terminology, logical structure, and data information. Using specialized inspection methods and tools, non-compliant sections are identified and marked.

[0103] Step S1331: Retrieve a set of standard terms from the automotive marketing professional terminology library, wherein the set of standard terms includes term names, standard definitions, correct usage scenarios, and examples of common incorrect usage.

[0104] The standard terminology set is retrieved from the automotive marketing professional terminology database, which is the result of long-term accumulation and professional review, and contains various standardized professional terms in the automotive marketing field.

[0105] Each term in the standard term set includes the term name, such as "battery energy density" and "charging power"; the standard definition, which clarifies the exact meaning of the term; the correct usage scenario, which explains in what context the term is appropriate; and examples of common incorrect usage, which show situations in which the term is easily misused, so as to facilitate comparative checks.

[0106] Step S1332: extract terms from the candidate knowledge content, identify all professional terms appearing in the text, match and compare the extracted professional terms with the standard term set, check the correctness of term spelling, consistency of definition and adaptability to usage scenarios, and mark incorrect terms that do not meet the standards.

[0107] We extract terms from candidate knowledge content and identify all professional terms in the text using a dictionary matching and part-of-speech tagging approach. We then compare the extracted professional terms with the standard term set one by one.

[0108] Check the spelling of the term to see if it is completely consistent with the standard term name, and check for typos, omissions, or duplications. Check definition consistency to analyze whether the meaning of the term in the candidate knowledge content is consistent with the standard definition. Check the adaptability of the usage scenario to determine whether the use of the term in the current context conforms to the correct usage scenario specified in the standard term set.

[0109] For terms with spelling errors, inconsistent definitions, or inappropriate usage scenarios, mark them as incorrect terms and record the error type and specific location.

[0110] Step S1333: Perform sentence-level logical relationship analysis on the candidate knowledge content, and use dependency syntax analysis to identify the logical connection relationship between adjacent sentences, including causal relationship, progressive relationship, transition relationship and parallel relationship, check whether there is improper use of logical conjunctions or logical contradictions, and mark sentences and paragraphs with logical contradictions.

[0111] Conduct sentence-level logical relationship analysis on the candidate knowledge content, use dependency syntax analysis method to analyze the syntactic structure of each sentence, identify the subject, predicate, object, attributive, adverbial, complement and other components in the sentence, as well as the dependency relationship between words.

[0112] Based on this, analyze the logical connection between adjacent sentences to determine whether it is a causal relationship (such as "because...so..."), a progressive relationship (such as "not only...but also..."), a transitional relationship (such as "although...but..."), or a parallel relationship (such as "at the same time...in addition..."), etc.

[0113] Check whether logical connectives are used appropriately, whether connectives do not match the actual logical relationships, and whether there are logical contradictions between sentences, such as conflicting content or inverted cause and effect. Mark sentences and paragraphs with logical contradictions.

[0114] Step S1334: Extract the data representation from the candidate knowledge content, compare the extracted data representation with the preset authoritative data source for automobile marketing, check whether the data values ​​are accurate, whether the units are consistent, and whether the descriptions are objective, and mark the content segments with inaccurate data representations. The data representation includes verifiable information such as values, percentages, time, location, and event descriptions.

[0115] Extract data representations from candidate knowledge content. These data representations include verifiable information such as numerical values, percentages, time, location, and event descriptions. Compare these extracted data representations with pre-defined authoritative automotive marketing data sources, which contain verified and accurate data information.

[0116] Check whether the data values ​​are consistent with the corresponding data in the authoritative data source, whether the data units are unified, whether the description is objective and true, and whether there is any exaggeration, reduction or false description. Mark any content fragments with inaccurate data descriptions.

[0117] Step S1335: Classify and count the incorrect terminology marks, logical contradiction paragraph marks, and data error segment marks, and calculate the terminology error rate, logical contradiction occurrence rate, and data error ratio as the basis for calculating the comprehensive score of expression accuracy.

[0118] The incorrect term markers, logically contradictory paragraph markers, and data error segment markers are classified and counted according to different error types. The number of incorrect terms, the number of logically contradictory paragraphs, and the number of data error segments are counted.

[0119] Calculate the term error rate, that is, the ratio of the number of incorrect terms to the total number of professional terms in the candidate knowledge content; calculate the logical contradiction incidence rate, that is, the ratio of the number of paragraphs with logical contradictions to the total number of paragraphs; calculate the data error ratio, that is, the ratio of the number of data error fragments to the total number of data expression fragments.

[0120] These proportion data will serve as the basis for calculating the comprehensive score of expression accuracy, comprehensively reflecting the problems with the expression accuracy of the candidate's knowledge content.

[0121] Step S134: With respect to the application adaptability evaluation dimension, the candidate knowledge content is analyzed for matching with the knowledge application scenarios defined in the marketing knowledge demand description to evaluate the degree of support for specific business scenarios, the applicability to the target user group, and the operability in actual marketing activities.

[0122] Focusing on application suitability, we conduct a comprehensive match analysis between candidate knowledge content and the knowledge application scenarios defined in the marketing knowledge requirements description. We assess the knowledge content's support for the business scenario, its fit with the user group, and its feasibility in actual marketing to determine whether the candidate knowledge content meets the requirements of the application scenario.

[0123] Step S1341: parse the marketing knowledge demand description and extract scenario characteristic parameters of the knowledge application scenario, wherein the scenario characteristic parameters include business scenario type, target user group characteristics, marketing activity purpose, knowledge application method and expected effect indicators.

[0124] Analyze marketing knowledge need descriptions and extract scenario-specific parameters for knowledge application scenarios through text analysis and semantic understanding. Examples of business scenarios include the aforementioned "family user promotion meeting" and "long-distance travel customer consultation." Target user group characteristics include travel habits, spending power, and awareness of new energy vehicles. Marketing activity objectives include increasing customer purchase intent, answering customer questions, and enhancing brand image. Knowledge application methods include oral explanations, written material presentations, and online interactions. Expected performance indicators include increased customer satisfaction and increased consultation conversion rates.

[0125] Step S1342: extracting application scenario-related information from the candidate knowledge content, and identifying the applicable scenario description, recommended usage objects, suggested application methods, and expected goals implicit in the knowledge content.

[0126] Extract application scenario-related information from candidate knowledge content, and identify the information related to the application scenario implicit in the knowledge content through methods such as semantic analysis and keyword matching.

[0127] The applicable scenario description refers to the specific occasions where the knowledge content is suitable for application; the recommended users refer to the user groups to which the knowledge content is more suitable; the recommended application method refers to the recommended dissemination or display method of the knowledge content; and the expected goal refers to the effect that is hoped to be achieved through the use of the knowledge content.

[0128] The extracted information is sorted to form an application scenario-related information set of the candidate knowledge content.

[0129] Step S1343: Calculate the similarity between the applicable scenario description of the candidate knowledge content and the business scenario type in the scenario feature parameters, and evaluate the support level of the knowledge content for the specific business scenario.

[0130] The similarity between the applicable scenario description of the candidate knowledge content and the business scenario type in the scenario feature parameters is calculated. First, both are converted into vector form. The dimension of the vector is determined based on the characteristic elements of the scenario, such as activity type, participants, and environmental atmosphere.

[0131] Calculate the similarity between the two vectors. The higher the similarity, the more closely the candidate knowledge content matches the business scenario, and the more support the knowledge content provides for the specific business scenario. Based on the similarity, a corresponding support score is assigned.

[0132] Step S1344: performing a matching analysis on the recommended users of the candidate knowledge content and the target user group characteristics in the scenario characteristic parameters, and evaluating the applicability score of the knowledge content to the target user group.

[0133] Match the recommended users of the candidate knowledge content with the target user group characteristics in the scenario feature parameters. Compare the recommended user characteristics with the characteristics of the target user group, such as age, occupation, and travel needs.

[0134] The higher the degree of match, the more suitable the knowledge content is for the target user group, and the higher the applicability score. The applicability score for the knowledge content to the target user group is determined by comprehensively considering the matching of multiple features.

[0135] Step S1345: Evaluate the operability score of the knowledge content in actual marketing activities based on the recommended application method and expected goal of the candidate knowledge content and the degree of fit with the marketing activity purpose, knowledge application method and expected effect indicators in the scenario feature parameters.

[0136] Analyze whether the recommended application method of the candidate knowledge content is consistent with the knowledge application method in the scenario characteristic parameters, and whether the expected goals are consistent with the marketing activity objectives and expected effect indicators.

[0137] If the suggested application method is easy to implement in actual marketing activities and the expected goals are highly consistent with the marketing activity objectives and expected performance indicators, the knowledge content is highly actionable and receives a high actionability score. Otherwise, the score is low. These factors are combined to determine the actionability score of the knowledge content in actual marketing activities.

[0138] Step S1346: The scenario support score, user applicability score, and operability score are weighted and calculated according to preset weights to obtain a comprehensive application adaptability score, which reflects the overall matching level between the candidate knowledge content and the application scenario.

[0139] The scenario support score, user applicability score, and operability score are weighted according to the preset weights. The preset weights are determined based on the importance of each score in the application adaptability assessment. For example, the scenario support score has a weight of 0.4, the user applicability score has a weight of 0.3, and the operability score has a weight of 0.3.

[0140] The calculation method is to multiply each score by its corresponding weight and then add the products together to obtain the application adaptability comprehensive score. This application adaptability comprehensive score comprehensively reflects the overall matching level between the candidate knowledge content and the application scenario.

[0141] Step S135: Based on the specific evaluation indicator scores and indicator weights of each evaluation dimension, a weighted summation method is used to calculate the comprehensive score of content completeness, the comprehensive score of expression accuracy, and the comprehensive score of application adaptability.

[0142] For the content completeness evaluation dimension, the scores of specific evaluation indicators such as knowledge point coverage and the degree of adequacy of explanation of each knowledge point are weighted and summed according to their corresponding indicator weights to obtain a comprehensive content completeness score.

[0143] Similarly, for the expression accuracy assessment dimension, the scores of indicators such as the standardization of professional terminology use, logical coherence, and data accuracy are weighted and summed with the corresponding weights to obtain a comprehensive expression accuracy score.

[0144] For the application adaptability evaluation dimension, the comprehensive application adaptability score is calculated according to the method of step S1346 above.

[0145] The specific method of weighted summation is to multiply the score of each indicator by its weight and then add them together. The sum of the weights of each indicator is 1, ensuring that the comprehensive score can reasonably reflect the overall performance of each dimension.

[0146] Step S136: Summarize and organize the comprehensive scores of each dimension, specific evaluation indicator scores, missing knowledge point lists, error marking information and scenario matching analysis results to generate a knowledge quality evaluation result including scoring results, problem descriptions and improvement suggestions.

[0147] The comprehensive scores for content completeness, presentation accuracy, and application adaptability, as well as the scores for specific evaluation indicators in each dimension, are summarized. Error information such as lists of missing knowledge points, incorrect terminology markers, logically contradictory paragraph markers, and data error segment markers are also compiled, along with issues discovered during the scenario matching analysis.

[0148] Based on this summarized information, a problem description is made to clearly point out the deficiencies in the candidate knowledge content, and corresponding improvement suggestions are made based on the problems, such as supplementing missing knowledge points, correcting incorrect terminology, and adjusting content to improve its match with the scenario.

[0149] These contents are integrated together to generate knowledge quality assessment results, which comprehensively reflect the quality status of candidate knowledge contents and the directions that need improvement.

[0150] Step S140: input the marketing knowledge requirement description, the candidate knowledge content and the knowledge quality assessment result into the big model, call the instruction fine-tuning interface of the big model to perform knowledge content optimization processing, and obtain optimized knowledge content that meets the knowledge quality assessment standard.

[0151] The marketing knowledge demand description, candidate knowledge content and knowledge quality assessment results are input into the big model together, and the instruction fine-tuning interface of the big model is called. The instruction fine-tuning interface can receive specific instructions and data to make targeted adjustments and optimizations to the model output.

[0152] By performing knowledge content optimization processing, the big model modifies and improves the candidate knowledge content based on the problems and suggestions in the knowledge quality assessment results, and finally obtains the optimized knowledge content that meets the knowledge quality assessment standards.

[0153] Step S141: Perform instruction conversion processing on the knowledge quality assessment results, and convert the problem descriptions and improvement suggestions therein into optimization instructions that conform to the large model instruction format. The optimization instructions include supplementary instructions for content completeness, correction instructions for expression accuracy, and adjustment instructions for application adaptability.

[0154] The problem descriptions and improvement suggestions in the knowledge quality assessment results are converted into instructions. According to the instruction format that the large model can understand, the problem descriptions are converted into tasks to be solved, and the improvement suggestions are converted into specific operational requirements.

[0155] Regarding the supplementary instructions for content completeness, they clearly point out the missing knowledge points that need to be supplemented and the requirements for improving the explanation of each knowledge point; regarding the correction instructions for expression accuracy, they indicate the incorrect terms, logically contradictory paragraphs and data error fragments that need to be corrected; regarding the adjustment instructions for application adaptability, they propose how to adjust the content to improve the match with the application scenario.

[0156] Step S142: construct an optimization prompt word template, fill the marketing knowledge demand description, the candidate knowledge content and the optimization instruction into the optimization prompt word template according to a preset text structure, and generate a comprehensive optimization prompt word including demand background, current content, existing problems and optimization direction.

[0157] Construct an optimization prompt word template, which contains a fixed text structure, such as the demand background part, the current content part, the existing problem part and the optimization direction part.

[0158] Fill the marketing knowledge demand description into the demand background part so that the big model can understand the original demand for knowledge construction; fill the candidate knowledge content into the current content part to display the basic content that needs to be optimized; fill the optimization instructions into the existing problems part and the optimization direction part respectively, clearly pointing out the problems with the content and the specific directions that need to be optimized.

[0159] Through the above filling, a comprehensive optimization prompt word is generated, which provides comprehensive optimization context information for the large model.

[0160] Step S143: configuring the optimization parameters of the large model instruction fine-tuning interface, wherein the optimization parameters include learning rate, number of training rounds, context window size, and output content length limit.

[0161] Configure optimization parameters for the large model instruction fine-tuning interface. The learning rate controls the magnitude of model parameter updates, affecting the speed and effectiveness of model optimization. The number of training rounds refers to the number of times the model trains on data during the optimization process. An appropriate number of training rounds ensures that the model fully learns the optimization instructions. The context window size determines the scope of context the model can consider when processing text. A larger window size helps the model understand the logic of longer texts. The output content length limit ensures that the optimized knowledge content is within a reasonable length range.

[0162] According to the characteristics of the candidate knowledge content and optimization requirements, set appropriate optimization parameter values ​​to achieve the best optimization effect.

[0163] Step S144: inputting the comprehensive optimization prompt words and the optimization parameters into the instruction fine-tuning interface of the large model, triggering the large model to execute the knowledge content optimization operation, and obtaining preliminary optimized knowledge content.

[0164] The comprehensive optimization prompt words and optimization parameters are input into the instruction fine-tuning interface of the large model. The instruction fine-tuning interface parses and processes these input information and passes it to the optimization module inside the large model.

[0165] The large model performs knowledge content optimization based on the information in the comprehensive optimization prompts and the settings of optimization parameters. During the optimization process, the model will modify, supplement, and adjust the candidate knowledge content according to the optimization direction based on existing problems to generate preliminary optimized knowledge content.

[0166] Step S145: performing a secondary quality assessment on the preliminary optimized knowledge content to generate a secondary knowledge quality assessment result of the preliminary optimized knowledge content.

[0167] A secondary quality assessment is performed on the preliminary optimized knowledge content using the same knowledge quality assessment criteria and assessment method as step S130.

[0168] The evaluation process includes checking aspects such as content completeness, expression accuracy and application adaptability, and generating secondary knowledge quality evaluation results. The secondary knowledge quality evaluation results are used to determine whether the preliminary optimized knowledge content has met the expected quality requirements.

[0169] Step S146: Compare the secondary knowledge quality assessment result with the knowledge quality assessment standard. If the secondary knowledge quality assessment result meets the knowledge quality assessment standard, determine the preliminary optimized knowledge content as the optimized knowledge content; if the secondary knowledge quality assessment result does not meet the knowledge quality assessment standard, use the secondary knowledge quality assessment result as the new knowledge quality assessment result, and repeat the optimization processing steps until the optimized knowledge content that meets the knowledge quality assessment standard is obtained.

[0170] Compare the secondary knowledge quality assessment results with the knowledge quality assessment standards to see whether the comprehensive score of each dimension reaches the threshold specified in the standards.

[0171] If the criteria are met, the preliminary optimized knowledge content meets the requirements and is determined to be optimized knowledge content. If the criteria are not met, the secondary knowledge quality assessment results are used as the new knowledge quality assessment results, and the process returns to step S141, where the instruction conversion process is repeated to generate new optimization instructions. Subsequently, according to the process from steps S142 to S145, the comprehensive optimization prompt words are reconstructed, the optimization parameters are configured, the instruction fine-tuning interface is input, the optimization operation is executed, and the quality assessment is repeated again.

[0172] This cycle repeats, with each iteration targeting the issues identified in the previous assessment. This continues until the results of a secondary knowledge quality assessment meet the knowledge quality assessment criteria. The initial optimized knowledge content at this point becomes the final optimized knowledge content. For example, during the optimization of "battery life and charging facility related knowledge," if the initial optimization reveals that the knowledge point "The impact of extreme weather on battery life and charging facility efficiency" is still missing, this result is used as the new assessment result, and the optimization steps are repeated, supplementing the relevant content and reassessing until all issues are resolved.

[0173] Step S150: performing knowledge structuring and organization processing on the optimized knowledge content to generate an automobile marketing knowledge unit including a knowledge subject identifier, content hierarchical relationship, and associated index information, and adding the automobile marketing knowledge unit to an automobile marketing knowledge base.

[0174] The optimized knowledge content is organized and processed into a structured form. Through a series of processing steps, the knowledge content, originally presented in natural language form, is transformed into knowledge units with fixed structures and relationships. This structured processing method facilitates the storage, retrieval, and application of knowledge. Finally, the generated automotive marketing knowledge units are added to the automotive marketing knowledge base, enriching the knowledge base's content.

[0175] Step S151: performing subject division processing on the optimized knowledge content, decomposing the optimized knowledge content into multiple knowledge sub-modules according to the content subject, and each knowledge sub-module is developed around a core subject.

[0176] To organize the optimization knowledge content into topics, first read through the entire optimization knowledge content and identify the core topics involved. For example, for "Battery life and charging infrastructure knowledge," core topics can be categorized as "Battery life basic principles," "Factors affecting battery life," "Charging facility types and characteristics," and "Battery life and charging infrastructure matching strategies."

[0177] Based on these core themes, the optimization knowledge content is decomposed into multiple knowledge sub-modules. Each knowledge sub-module focuses on a core theme and contains all knowledge content related to the theme, ensuring that the content of each sub-module has strong relevance and independence.

[0178] Step S152: Assign a unique knowledge subject identifier to each knowledge submodule. The knowledge subject identifier is encoded using an alphanumeric combination, including a field classification code, a subject category code, and a unique serial number.

[0179] Each knowledge submodule is assigned a unique knowledge subject identifier, using an alphanumeric encoding method. The domain classification code identifies the broad domain to which the knowledge belongs, such as "XNYQC" for "new energy vehicle marketing." The subject category code distinguishes different subject categories, such as "DX" for "battery life" and "CD" for "charging facility." The unique serial number is a unique numerical number assigned to each submodule to ensure unique identification.

[0180] For example, the knowledge subject identifier of the "Basic Principles of Battery Life" sub-module can be "XNYQC-DX-001", and the identifier of the "Charging Facility Types and Characteristics" sub-module can be "XNYQC-CD-001", etc.

[0181] Step S153: Analyze the logical relationship between each knowledge sub-module, determine the parent-child relationship, parallel relationship, reference relationship and supplement relationship between the knowledge sub-modules, and construct a knowledge content hierarchical structure tree.

[0182] Analyze the logical relationship between each knowledge sub-module, and determine the relationship type between them by comparing and analyzing the core content and themes of each sub-module.

[0183] A parent-child relationship means that the content of one submodule contains the content of another submodule. For example, the "Factors Affecting Battery Life" submodule and the "Impact of Temperature on Battery Life" submodule have a parent-child relationship, and the latter is part of the former.

[0184] A parallel relationship means that multiple sub-modules are at the same level, and are related to each other in content but do not have an inclusion relationship. For example, the sub-modules "Characteristics of fast charging facilities" and "Characteristics of slow charging facilities" are in a parallel relationship.

[0185] A reference relationship means that the content of one submodule references the content of another submodule. For example, the "matching strategy of battery life and charging facilities" submodule may reference the content of the "charging facility types and characteristics" submodule.

[0186] A supplementary relationship means that the content of one sub-module supplements the content of another sub-module, such as the "Impact of battery maintenance on battery life" sub-module supplements the "Factors affecting battery life" sub-module.

[0187] Based on these relationships, a knowledge content hierarchy structure tree is constructed to clearly display the hierarchy and relationships between each knowledge sub-module.

[0188] For example, step S1531: extracting the content summary of each knowledge sub-module to generate a module summary that can reflect the core content of the knowledge sub-module.

[0189] Content summary is extracted for each knowledge submodule, and a text summary generation algorithm is used to analyze and refine the content of each submodule.

[0190] First, perform word segmentation and keyword extraction on the text to identify the core vocabulary and key information in the submodule. Then, based on this information, construct a module summary that summarizes the core content of the submodule. The module summary should be concise and clear, accurately reflecting the main content and theme of the submodule.

[0191] For example, the module summary of the sub-module "The impact of temperature on battery life" can be extracted as "The specific impact of different temperature conditions on the battery life of new energy vehicles and related principles are explained."

[0192] Step S1532: Calculate the semantic similarity between the module summaries of any two knowledge sub-modules and generate a knowledge sub-module similarity matrix.

[0193] To calculate the semantic similarity between the module summaries of any two knowledge submodules, each module summary is first converted into a vector form, and the dimension of the vector is determined based on the lexical semantic features in the summary.

[0194] Semantic similarity is then determined by calculating the degree of correlation between the two vectors. A higher correlation indicates closer semantics between the two module summaries. The semantic similarity calculation results between all two submodules are organized into a matrix, the knowledge submodule similarity matrix, where each element represents the semantic similarity between the two submodules.

[0195] Step S1533: Based on the knowledge sub-module similarity matrix, a hierarchical clustering algorithm is used to perform cluster analysis on the knowledge sub-modules, and the knowledge sub-modules with semantic similarity greater than a set similarity threshold are classified into one category to form a preliminary knowledge module group.

[0196] Based on the knowledge submodule similarity matrix, a hierarchical clustering algorithm is used for cluster analysis. The hierarchical clustering algorithm starts with each knowledge submodule as a separate cluster, and then gradually merges clusters with similarity greater than the set similarity threshold in descending order to form larger clusters.

[0197] Through this method, knowledge submodules with high semantic similarity are grouped together to form preliminary knowledge module groups. For example, submodules related to battery life are clustered into one group, and submodules related to charging facilities are clustered into another group.

[0198] Step S1534: Analyze the semantic relationship between the knowledge sub-modules within each knowledge module group, identify the knowledge sub-module pairs with inclusion relationship, determine the included knowledge sub-module as the child node, determine the knowledge sub-module that includes other knowledge sub-modules as the parent node, and establish a parent-child relationship.

[0199] The semantic relationship between each knowledge sub-module within each knowledge module group is analyzed, and by comparing the module summary and core content of the sub-module, the knowledge sub-module pairs with inclusion relationship are identified.

[0200] If the content of a submodule is completely covered by the content of another submodule, the included submodule is the child node, and the submodule that includes it is the parent node, thus establishing a parent-child relationship. For example, in the group of factors affecting battery life, the "Effect of Low Temperature on Battery Life" submodule is included by the "Effect of Temperature on Battery Life" submodule. The former is the child node, and the latter is the parent node.

[0201] Step S1535: Identify knowledge sub-modules in the same level that do not have a containment relationship but are semantically related, and determine them as a parallel relationship.

[0202] In the same knowledge module group or the same level, identify those knowledge sub-modules that do not have an inclusion relationship but are semantically related to each other and determine them as a parallel relationship.

[0203] These sub-modules complement or relate to each other in terms of content, but each has independent themes and content. For example, the sub-modules "The impact of battery capacity on battery life" and "The impact of battery aging on battery life" belong to the same level of the "Factors affecting battery life" group. There is no inclusion relationship, but they are semantically related, so they are in a parallel relationship.

[0204] Step S1536: Search for instances in which the content of a knowledge sub-module explicitly references the content of other knowledge sub-modules, and establish a reference relationship.

[0205] Check the content of each knowledge sub-module one by one to find any explicit mention or citation of the content of other knowledge sub-modules.

[0206] When a submodule cites concepts, data, or conclusions from another submodule during its explanation, a reference relationship is established between the two submodules, and the specific content and location of the reference is recorded. For example, if the "Battery Life and Charging Facility Matching Strategy" submodule cites the "Charging Speed ​​of Fast Charging Facilities," a reference relationship is established between the two submodules.

[0207] Step S1537: Identify the knowledge sub-module that provides supplementary explanation or expanded elaboration on the content of any knowledge sub-module and establish a supplementary relationship.

[0208] Identify those knowledge sub-modules that supplement or expand the content of other knowledge sub-modules, and establish a complementary relationship between them when the content of one sub-module can provide additional information, details or extended explanation for the content of another sub-module.

[0209] For example, the “Battery Life Test Standards” sub-module provides supplementary explanations of the test basis for the life data mentioned in the “Battery Life Basic Principles” sub-module, thus establishing a complementary relationship between the two.

[0210] Step S1538: Based on the knowledge module group, according to the determined parent-child relationship, parallel relationship, reference relationship and supplementary relationship, a knowledge content hierarchical structure tree with a multi-tree structure is constructed, in which the parent node is located at the upper layer, the child node is located at the lower layer, the parallel nodes are located at the same layer, and the reference relationship and supplementary relationship are represented by additional edges.

[0211] Based on the knowledge module group, each group is regarded as a main branch of the knowledge content hierarchy tree. According to the determined parent-child relationship, the parent node is placed in the upper layer and the child node is placed in the lower layer to form a hierarchical structure of the tree.

[0212] Submodules in a parallel relationship are placed at the same level, maintaining their parallel positions. For reference and supplementary relationships, additional edges are added between the corresponding child nodes to clearly demonstrate the non-hierarchical associations between submodules.

[0213] Finally, a knowledge content hierarchy structure tree with a multi-tree structure is constructed, which comprehensively and intuitively reflects the various logical relationships between the knowledge sub-modules.

[0214] Step S154: Extract key concepts, core terms and important data in each knowledge sub-module, establish a concept association network within the knowledge sub-module and a cross-reference index between knowledge sub-modules, and generate association index information, which includes the associated knowledge subject identifier, association type and association strength.

[0215] Extract key concepts from each knowledge sub-module, such as "battery energy density" and "charging efficiency"; core terms, such as the various professional terms mentioned above; important data, such as the charging time range under different charging methods.

[0216] Establish a concept association network within the knowledge sub-module and analyze the relationship between key concepts, core terms and important data within the sub-module. For example, there is a positive correlation between "battery energy density" and "battery range".

[0217] At the same time, a cross-reference index is established between knowledge submodules, recording the associations between different submodules based on previously determined reference and supplement relationships. The generated association index information includes the associated knowledge topic identifier (i.e., the knowledge topic identifier of the associated submodule); the association type (e.g., reference, supplement, etc.); and the association strength (e.g., strong, medium, weak, etc.), which is categorized based on the closeness and importance of the association.

[0218] Step S155: Integrate and encapsulate the knowledge subject identifier, knowledge submodule content, knowledge content hierarchical relationship and associated index information to generate a structured automobile marketing knowledge unit, which is stored in an extensible markup language format.

[0219] The knowledge subject identifier, knowledge submodule content, knowledge content hierarchical relationship, and associated index information are integrated and encapsulated. According to the syntax rules of the Extensible Markup Language, corresponding tags are defined for each part, such as <subject identifier>, <submodule content>, <hierarchical relationship>, <associated index>, etc.

[0220] By filling each part of the content under the corresponding tag, a clearly structured and well-formatted XML document is formed, which is a structured automotive marketing knowledge unit. This format is easy for computers to recognize and process, and is conducive to the storage and management of knowledge.

[0221] Step S156: calling the management interface of the automobile marketing knowledge base, adding the automobile marketing knowledge unit to the corresponding category directory of the knowledge base, and updating the global index and knowledge relationship map of the automobile marketing knowledge base.

[0222] Call the management interface of the automotive marketing knowledge base, which provides functions such as adding knowledge units and updating indexes. Based on the knowledge subject identifier and content theme of the automotive marketing knowledge unit, it is added to the corresponding category directory in the knowledge base, such as the "battery life knowledge" directory or the "charging facility knowledge" directory.

[0223] Once the addition is complete, the global index of the automotive marketing knowledge base is updated. This global index records the storage location and key information of all knowledge units, facilitating quick retrieval. Simultaneously, the knowledge base's knowledge relationship graph is updated based on the hierarchical relationships and associated index information within the knowledge units. This ensures that the graph accurately reflects the relationships between the newly added knowledge units and existing ones, ensuring the integrity and relevance of the knowledge base.

[0224] Step S210: Model training step. The method also includes a step of training a large model for knowledge content generation and optimization.

[0225] To ensure that the large model is more suitable for generating and optimizing knowledge content in the automotive marketing field, it needs to be trained specifically. The training process will combine professional data from the automotive marketing field, adjust model parameters, and improve the model's knowledge generation and optimization capabilities in this field.

[0226] Step S211: collecting professional corpus data in the field of automobile marketing, wherein the professional corpus data includes automobile marketing knowledge documents, industry reports, marketing cases, and user consultation records.

[0227] Collect professional corpus data in the field of automotive marketing. These data come from a wide range of sources, including marketing knowledge documents released by automobile manufacturers, automobile marketing industry reports issued by industry research institutions, marketing success cases of various automobile brands, and historical user consultation records.

[0228] During the collection process, the legality and compliance of the data must be ensured. For data such as consultation records involving user privacy, data desensitization technology is used to remove the user's personal identity information, such as name, contact information, ID number, etc., to protect user privacy.

[0229] Step S212: pre-processing the collected professional corpus data, including data cleaning, deduplication, format unification and word segmentation.

[0230] The collected professional corpus data is preprocessed. Data cleaning is to remove noise and irrelevant information in the data, such as pop-up advertising content and malformed characters in the document; deduplication is to delete duplicate data to avoid repeated training; format unification is to convert documents of different formats into a unified text format to facilitate subsequent processing; word segmentation is to divide continuous text into independent words in preparation for model training.

[0231] Step S213: constructing a training dataset and a validation dataset, and dividing the preprocessed professional corpus data into the training dataset and the validation dataset according to a set ratio.

[0232] The preprocessed professional corpus data is divided into a training dataset and a validation dataset according to a set ratio, such as an 8:2 ratio. The training dataset is used to learn the model parameters, and the validation dataset is used to evaluate the model performance during training and adjust the training strategy in a timely manner.

[0233] Step S214: configuring the training parameters of the large model, wherein the training parameters include the training batch size, the number of training iterations, the learning rate and the regularization coefficient.

[0234] Configure the training parameters of the large model. The training batch size refers to the number of data samples input into the model each time; the number of training iterations refers to the number of times the model completes training on the training dataset; the learning rate is used to control the step size of the model parameter update; the regularization coefficient is used to prevent the model from overfitting and improve the model's generalization ability.

[0235] Set appropriate training parameter values ​​according to the scale and characteristics of professional corpus data.

[0236] Step S215: Input the training data set into the large model, train the large model according to the configured training parameters, use the validation data set to evaluate the performance of the model during the training process, and adjust the training parameters according to the evaluation results.

[0237] The training dataset is fed into the large model, which is then trained according to the configured training parameters. After each training iteration, the model's performance is evaluated using the validation dataset, with metrics such as the relevance, accuracy, and fluency of the generated text.

[0238] Based on the evaluation results, if the model performance does not meet expectations, such as the generated text has low relevance to the topic, the training parameters are adjusted, such as reducing the learning rate, increasing the number of training iterations, etc., until the model performance on the validation dataset reaches a satisfactory level.

[0239] Step S216: After the training is completed, the large model is tested, and the test data set is used to evaluate the generation effect and optimization ability of the model. If the test results meet the preset requirements, it is determined to be a trained large model; if not, the training parameters are readjusted or the training data is supplemented, and training is performed again.

[0240] After training is completed, the large model is tested using an independent test dataset. The test dataset does not participate in the training and verification process of the model, which can more objectively evaluate the model performance.

[0241] Evaluate the model's performance on the test dataset, such as the quality of the generated knowledge content and its compatibility with requirements, as well as its optimization capabilities, such as its ability to correct problematic knowledge content. If the test results meet the preset requirements, the model is considered a well-trained large model. If not, analyze the reasons, readjust the training parameters, or add more professional corpus data, and repeat the training steps until the model passes the test.

[0242] Figure 2 The following diagram illustrates exemplary hardware and software components of a system 100 for constructing a marketing knowledge base based on a large model, which can implement the concepts of the present application, according to some embodiments of the present application. For example, the processor 120 can be used in the system 100 for constructing a marketing knowledge base based on a large model and for executing the functions of the present application.

[0243] The large-model-based marketing knowledge base construction system 100 can be a general-purpose server or a special-purpose server, both of which can be used to implement the large-model-based marketing knowledge base construction method of this application. Although only one server is shown in this application, for convenience, the functions described in this application can be implemented in a distributed manner on multiple similar platforms to balance the processing load.

[0244] For example, the large-model-based marketing knowledge base construction system 100 may include a network port 110 connected to a network, one or more processors 120 for executing program instructions, a communication bus 130, and various forms of storage media 140, such as a disk, ROM, or RAM, or any combination thereof. Exemplarily, the large-model-based marketing knowledge base construction system 100 may also include program instructions stored in ROM, RAM, or other types of non-transitory storage media, or any combination thereof. The method of the present application may be implemented based on these program instructions. The large-model-based marketing knowledge base construction system 100 also includes an I / O interface 150 between the computer and other input and output devices.

[0245] For ease of explanation, only one processor is described in the large model-based marketing knowledge base construction system 100. However, it should be noted that the large model-based marketing knowledge base construction system 100 in this application may also include multiple processors, so the steps performed by one processor described in this application may also be performed jointly or individually by multiple processors. For example, if the processor of the large model-based marketing knowledge base construction system 100 executes step A and step B, it should be understood that step A and step B may also be executed jointly by two different processors or individually in one processor. For example, the first processor executes step A, the second processor executes step B, or the first processor and the second processor execute steps A and B together.

[0246] In addition, an embodiment of the present invention further provides a readable storage medium, in which computer-executable instructions are preset. When a processor executes the computer-executable instructions, the above-mentioned method for constructing a marketing knowledge base based on a large model is implemented.

[0247] It should be noted that in order to simplify the description of the present invention and thus help understand one or more embodiments of the invention, in the foregoing description of the embodiments of the present invention, multiple features are sometimes combined into one embodiment, figure or description thereof.

Claims

1. A method for constructing a marketing knowledge base based on a large model, characterized in that: The method comprises: Determine the knowledge construction needs in the automotive marketing field and generate a marketing knowledge needs description, which includes the marketing subject scope, knowledge content type, and knowledge application scenario limitations; Input the marketing knowledge demand description into the pre-trained big model, call the text generation interface of the big model to perform preliminary knowledge content generation processing, and obtain candidate knowledge content corresponding to the marketing knowledge demand description; Acquiring a preset knowledge quality assessment standard, performing a knowledge quality assessment process on the candidate knowledge content according to the knowledge quality assessment standard, and generating a knowledge quality assessment result; Inputting the marketing knowledge requirement description, the candidate knowledge content, and the knowledge quality assessment result into the macro model, calling the instruction fine-tuning interface of the macro model to perform knowledge content optimization processing, and obtaining optimized knowledge content that meets the knowledge quality assessment standard; Performing knowledge structuring and organization processing on the optimized knowledge content to generate an automobile marketing knowledge unit including a knowledge subject identifier, content hierarchical relationship, and associated index information, and adding the automobile marketing knowledge unit to an automobile marketing knowledge base; The obtaining of a preset knowledge quality assessment standard, performing a knowledge quality assessment process on the candidate knowledge content according to the knowledge quality assessment standard, and generating a knowledge quality assessment result includes: Retrieving preset knowledge quality assessment standards from the standard database of the knowledge management system, parsing the knowledge quality assessment standards, and determining specific assessment indicators and indicator weights included in the content completeness assessment dimension, the expression accuracy assessment dimension, and the application adaptability assessment dimension; In terms of content completeness assessment, the candidate knowledge content is tested for knowledge point coverage to identify whether the candidate knowledge content contains all the core knowledge points required in the marketing knowledge demand description, and the number of missing knowledge points and the adequacy of the explanation of each knowledge point are counted; In terms of expression accuracy assessment, the candidate knowledge content is checked for terminology standardization, logical coherence, and data accuracy, and incorrect terminology, logically contradictory sentences and paragraphs, and inaccurate data expressions are marked; In terms of application adaptability, the candidate knowledge content is analyzed for matching with the knowledge application scenarios defined in the marketing knowledge requirement description to assess the knowledge content's support for specific business scenarios, its applicability to the target user group, and its operability in actual marketing activities. Based on the specific evaluation indicator scores and indicator weights of each evaluation dimension, the weighted summation method is used to calculate the comprehensive score of content completeness, the comprehensive score of expression accuracy, and the comprehensive score of application adaptability; The comprehensive scores of each dimension, scores of specific evaluation indicators, lists of missing knowledge points, error marking information and scenario matching analysis results are summarized and organized to generate knowledge quality assessment results that include scoring results, problem descriptions and improvement suggestions.

2. The method for constructing a marketing knowledge base based on a large model according to claim 1, characterized in that: Determining the knowledge construction requirements in the automotive marketing field and generating a marketing knowledge requirements description includes: Obtaining a preliminary knowledge requirement document provided by the automotive marketing business department, performing text parsing on the preliminary knowledge requirement document, and extracting content type indication information therein, the content type indication information including requirement keywords, topic phrases, and scenario description paragraphs; Perform word frequency statistics and semantic clustering on the extracted demand keywords, identify core demand themes and secondary demand themes, and generate a demand theme hierarchical structure; Performing scenario feature extraction processing on the extracted scenario description paragraphs to determine scenario feature parameters of knowledge application, wherein the scenario feature parameters include specific business scenarios, target user groups, and expected application methods; Integrate the demand subject hierarchical structure, the content type indication information and the scenario characteristic parameters to generate a preliminary marketing knowledge demand description; Feedback the preliminary marketing knowledge demand description to the automobile marketing business department for demand confirmation, receive demand modification opinions returned by the business department, iteratively adjust the preliminary marketing knowledge demand description based on the demand modification opinions, and generate a final marketing knowledge demand description.

3. The method for constructing a marketing knowledge base based on a large model according to claim 1, characterized in that: The step of inputting the marketing knowledge requirement description into a pre-trained large model and calling the text generation interface of the large model to perform preliminary knowledge content generation processing to obtain candidate knowledge content corresponding to the marketing knowledge requirement description includes: Performing demand feature vectorization processing on the marketing knowledge demand description, converting the demand description in text form into a demand feature vector that conforms to the input format of the large model; Calling the domain knowledge awakening module of the large model, associating and matching the demand feature vector with the pre-trained knowledge graph in the automotive marketing field, and activating the domain knowledge parameters in the large model related to the marketing knowledge demand description; Configuring a generation parameter set for the text generation interface, the generation parameter set including a topic relevance weight, a content depth coefficient, a professional term density threshold, and an output length range; Input the demand feature vector and the generation parameter set into the text generation interface, trigger the large model to perform the knowledge content generation operation, and obtain the initial knowledge text; Performing redundant information filtering on the initial knowledge text, deleting repeated expressions, extended contents irrelevant to the requirements, and conflicting and contradictory contents, to obtain a purified intermediate knowledge text; The intermediate knowledge text is checked for content integrity. If there are obvious missing paragraphs, the missing paragraph identifiers and the original requirement feature vectors are re-entered into the text generation interface for supplementary generation. The supplementary generated content is spliced ​​and fused with the intermediate knowledge text to obtain candidate knowledge content.

4. The method for constructing a marketing knowledge base based on a large model according to claim 1, characterized in that: Regarding the content completeness assessment dimension, the candidate knowledge content is tested for knowledge point coverage to identify whether the candidate knowledge content contains all the core knowledge points required in the marketing knowledge demand description, and the number of missing knowledge points and the adequacy of the explanation of each knowledge point are counted, including: Extract core knowledge points from the marketing knowledge demand description and, based on the automotive marketing domain knowledge system, determine a set of core knowledge points corresponding to the marketing knowledge demand description. Each core knowledge point includes a topic name, necessary elaboration aspects, and required level of detail. Constructing a knowledge point recognition model, wherein the knowledge point recognition model uses a bidirectional long short-term memory network combined with a conditional random field algorithm to scan the candidate knowledge content section by section, and identify the knowledge points contained in the candidate knowledge content and the actual elaboration of each knowledge point; Compare and match the identified included knowledge points with the core knowledge point set, determine the core knowledge points that are not included, and count the number of missing knowledge points; For each included core knowledge point, match its actual elaboration aspects with the necessary elaboration aspects and calculate the elaboration aspect coverage; Calculate the adequacy score of each knowledge point based on the coverage of the elaboration aspects and the level of detail of each elaboration aspect according to the preset scoring rules, wherein the adequacy score of the elaboration aspect is positively correlated with the coverage of the elaboration aspects and the level of detail; The number of missing knowledge points and the score of the adequacy of the explanation of each knowledge point are recorded in the content completeness assessment sub-result as the basis for calculating the comprehensive content completeness score.

5. The method for constructing a marketing knowledge base based on a large model according to claim 1, characterized in that: Regarding the dimension of expression accuracy assessment, the candidate knowledge content is checked for terminology standardization, logical coherence, and data accuracy, marking incorrect professional terms, logically contradictory sentences and paragraphs, and inaccurate data expression content fragments, including: Retrieving a set of standard terms from a professional automotive marketing terminology database, wherein the set of standard terms includes term names, standard definitions, correct usage scenarios, and examples of common incorrect usage; Perform term extraction on the candidate knowledge content, identify all professional terms appearing in the text, match and compare the extracted professional terms with the standard term set, check the correctness of term spelling, consistency of definition and adaptability to usage scenarios, and mark incorrect terms that do not meet the standards; Perform sentence-level logical relationship analysis on the candidate knowledge content, using dependency syntax analysis to identify logical connections between adjacent sentences, including causal relationships, progressive relationships, transitional relationships, and parallel relationships, check whether there are improper use of logical connectives or logical contradictions, and mark sentences and paragraphs with logical contradictions; Extracting data representations from the candidate knowledge content and comparing the extracted data representations with pre-set authoritative automotive marketing data sources to check whether the data values ​​are accurate, whether the units are consistent, and whether the descriptions are objective, and marking content segments with inaccurate data representations. The data representations include verifiable information such as values, percentages, time, location, and event descriptions; The markers for incorrect terminology, logically contradictory paragraphs, and data error fragments are classified and counted, and the terminology error rate, logical contradiction incidence rate, and data error ratio are calculated as the basis for calculating the comprehensive score of expression accuracy.

6. The method for constructing a marketing knowledge base based on a large model according to claim 1, characterized in that: Regarding the application adaptability evaluation dimension, the candidate knowledge content is analyzed for matching with the knowledge application scenarios defined in the marketing knowledge requirement description to evaluate the knowledge content's support for specific business scenarios, its applicability to the target user group, and its operability in actual marketing activities, including: Analyze the marketing knowledge demand description and extract scenario characteristic parameters of the knowledge application scenario, wherein the scenario characteristic parameters include business scenario type, target user group characteristics, marketing activity purpose, knowledge application method and expected effect indicators; Extracting application scenario-related information from the candidate knowledge content to identify the applicable scenario descriptions, recommended usage objects, suggested application methods, and expected goals implicit in the knowledge content; Calculate the similarity between the applicable scenario description of the candidate knowledge content and the business scenario type in the scenario feature parameters to evaluate the support degree of the knowledge content for the specific business scenario; Matching and analyzing the recommended users of the candidate knowledge content with the target user group characteristics in the scenario characteristic parameters to evaluate the applicability of the knowledge content to the target user group; Evaluate the operability of the knowledge content in actual marketing activities based on the degree of fit between the recommended application method and expected goal of the candidate knowledge content and the marketing activity objectives, knowledge application method and expected effect indicators in the scenario characteristic parameters; The scenario support score, user applicability score, and operability score are weighted and calculated according to preset weights to obtain a comprehensive application adaptability score, which reflects the overall matching level between the candidate knowledge content and the application scenario.

7. The method for constructing a marketing knowledge base based on a large model according to claim 1, characterized in that: The step of inputting the marketing knowledge requirement description, the candidate knowledge content, and the knowledge quality assessment result into the macro model, calling the instruction fine-tuning interface of the macro model to perform knowledge content optimization processing, and obtaining optimized knowledge content that meets the knowledge quality assessment criteria includes: Performing instruction conversion processing on the knowledge quality assessment results to convert the problem descriptions and improvement suggestions therein into optimization instructions that conform to the large model instruction format. The optimization instructions include supplementary instructions for content completeness, correction instructions for expression accuracy, and adjustment instructions for application adaptability. Constructing an optimization prompt word template, filling the marketing knowledge demand description, the candidate knowledge content, and the optimization instructions into the optimization prompt word template according to a preset text structure, and generating a comprehensive optimization prompt word that includes demand background, current content, existing problems, and optimization direction; Configuring optimization parameters of the large model instruction fine-tuning interface, wherein the optimization parameters include learning rate, number of training rounds, context window size, and output content length limit; Inputting the comprehensive optimization prompt words and the optimization parameters into the instruction fine-tuning interface of the large model to trigger the large model to perform knowledge content optimization operations to obtain preliminary optimized knowledge content; Performing a secondary quality assessment on the preliminary optimized knowledge content to generate a secondary knowledge quality assessment result of the preliminary optimized knowledge content; The secondary knowledge quality assessment result is compared with the knowledge quality assessment standard. If the secondary knowledge quality assessment result meets the knowledge quality assessment standard, the preliminary optimized knowledge content is determined as the optimized knowledge content; if the secondary knowledge quality assessment result does not meet the knowledge quality assessment standard, the secondary knowledge quality assessment result is used as the new knowledge quality assessment result, and the optimization processing steps are repeated until the optimized knowledge content that meets the knowledge quality assessment standard is obtained.

8. The method for constructing a marketing knowledge base based on a large model according to claim 1, characterized in that: The performing of knowledge structuring and organization processing on the optimized knowledge content to generate an automobile marketing knowledge unit including a knowledge subject identifier, a content hierarchical relationship, and associated index information, and adding the automobile marketing knowledge unit to an automobile marketing knowledge base includes: Performing thematic division processing on the optimized knowledge content, decomposing the optimized knowledge content into multiple knowledge sub-modules according to the content theme, each knowledge sub-module revolving around a core theme; Assign a unique knowledge subject identifier to each knowledge submodule. The knowledge subject identifier is encoded using an alphanumeric combination, including a field classification code, a subject category code, and a unique serial number. Analyze the logical relationship between each knowledge sub-module, determine the parent-child relationship, parallel relationship, reference relationship and supplementary relationship between knowledge sub-modules, and construct a knowledge content hierarchical structure tree; Extract key concepts, core terms, and important data from each knowledge submodule, establish a concept association network within the knowledge submodule and a cross-reference index between knowledge submodules, and generate association index information, which includes the associated knowledge subject identifier, association type, and association strength; Integrate and encapsulate knowledge subject identifiers, knowledge submodule content, knowledge content hierarchical relationships, and associated index information to generate structured automotive marketing knowledge units, which are stored in an extensible markup language format; The management interface of the automobile marketing knowledge base is called to add the automobile marketing knowledge unit to the corresponding category directory of the knowledge base, and update the global index and knowledge relationship map of the automobile marketing knowledge base.

9. A marketing knowledge base construction system based on a large model, characterized by: It includes a processor and a memory, the memory is connected to the processor, the memory is used to store programs, instructions or codes, and the processor is used to execute the programs, instructions or codes in the memory to implement the marketing knowledge base construction method based on a large model as described in any one of claims 1 to 8.

Images