Big model-based customer intelligent recommendation method and device, and computer equipment

Abstract

This invention discloses a customer intelligent recommendation method, apparatus, and computer equipment based on a large-scale language model. The method includes: acquiring users' natural language business opportunity needs and converting them into text information in a unified format; using a large-scale language model to perform deep analysis of the text information, extracting user need types and related core business entities; judging the information completeness of the preliminary structured information based on the large-scale language model's generation capabilities, and supplementing missing or ambiguous information through multi-round interactive dialogue to obtain structured data; converting the structured data into a domain-specific DSL query, verifying it, and sending it to a downstream system to filter and match customers, and generating matching results; displaying the results in a user-friendly manner through a customizable wbit module, and providing visual annotations. Implementing this invention reduces manual intervention and improves processing efficiency and accuracy.

Description

Technical Field

[0002] This invention relates to computers, and more specifically to a method, apparatus, and computer equipment for intelligent customer recommendation based on large models. Background Technology

[0004] In modern business operations, customer referral systems have become an important tool for improving business efficiency and expanding market share. Traditionally, these systems rely primarily on user interaction via page clicks, requiring users to manually select structured filtering criteria such as industry, region, and customer size. This interaction method not only demands that users have a thorough understanding of the system's field settings, increasing the learning curve, but also leads to a prolonged demand delivery cycle due to its reliance on manual input or selection.

[0005] While some systems have attempted to incorporate natural language processing (NLP) techniques to simplify user workflows, these improvements are mostly limited to simple keyword matching and fail to deeply understand the complex business opportunities users present. Such systems struggle with unstructured data, unable to accurately parse users' natural language expressions, resulting in low accuracy in demand conversion. Furthermore, when users provide vague or incomplete information, existing systems lack the ability to proactively guide users to supplement key information, further limiting recommendation efficiency and accuracy.

[0006] Therefore, it is necessary to design a new method that can automate the entire process from unstructured natural language to structured business data, possessing powerful natural language understanding and processing capabilities. This method can not only accurately capture users' complex needs but also proactively prompt users to supplement relevant details when information is insufficient, thereby reducing human intervention and improving processing efficiency and accuracy. Summary of the Invention

[0008] The purpose of this invention is to overcome the shortcomings of the prior art and provide a customer intelligent recommendation method, apparatus and computer equipment based on a large model.

[0009] To achieve the above objectives, the present invention adopts the following technical solution: a customer intelligent recommendation method based on a large model, comprising:

[0010] Obtain users' natural language business opportunity needs and convert them into text information in a unified format;

[0011] The text information is analyzed in depth using a large language model to extract user demand types and related core business entities, forming preliminary structured information.

[0012] The completeness of the preliminary structured information is judged based on the ability to generate large-scale language models, and missing or ambiguous information is supplemented through multi-round interactive dialogue to obtain structured data.

[0013] The structured data is transformed into a domain-specific DSL query, verified, and then sent to the downstream system to filter and match customers, and generate matching results.

[0014] The matching results are displayed in a user-friendly manner and visually annotated using a customizable wbit module.

[0015] The further technical solution is as follows: after displaying the matching results in a user-friendly manner through a customizable wbit module and performing visual annotation, it also includes:

[0016] Collect user feedback and interaction data as training samples to continuously fine-tune a large language model.

[0017] The further technical solution is as follows: obtaining the user's natural language business opportunity requirements and converting them into text information in a unified format includes:

[0018] The system obtains natural language business opportunity requirements input by users through a user interface; the user interface includes at least one of a web interface, a mobile APP interface, a mini-program interface, and a voice interaction interface.

[0019] The further technical solution is as follows: The text information is deeply analyzed using a large-scale language model to extract user demand types and related core business entities, forming preliminary structured information, including:

[0020] The text information is preprocessed to obtain the preprocessing result;

[0021] The preprocessed results are used to determine the specific business opportunity demand type of the user using a large language model, and key business entities are extracted from the preprocessed results using named entity recognition technology to obtain the analysis results.

[0022] Based on the analysis results, preliminary structured information is constructed, which includes user needs and intentions, core business entities, and specific business scenario parameters.

[0023] The further technical solution is: to perform preprocessing operations on the text information to obtain preprocessing results;

[0024] The text information is processed by word segmentation, part-of-speech tagging, dependency parsing, and semantic role tagging to obtain preprocessing results.

[0025] Its further technical solution is as follows: The method of judging the information completeness of the preliminary structured information based on the generation capability of a large language model, and supplementing missing or ambiguous information through multi-round interactive dialogue to obtain structured data, including:

[0026] The integrity of the preliminary structured information is assessed using the generative capabilities of a large language model to obtain the assessment results;

[0027] When the evaluation results are ambiguous or lack key information, guiding questions are automatically generated, and multiple rounds of interactive dialogue are initiated through the user interface to collect more information. In each round of dialogue, the user's response is analyzed in conjunction with the context to supplement and clarify the data in the preliminary structured information until all necessary information is complete and clear, so as to obtain the structured data.

[0028] The further technical solution is as follows: The process of converting the structured data into a domain-specific DSL query, verifying it, sending it to the downstream system to filter and match customers, and generating matching results includes:

[0029] The structured data is converted into a domain-specific language to obtain the converted DSL;

[0030] The converted DSL is validated, including checks on syntax correctness, semantic consistency, and business logic legality, to obtain the validation results.

[0031] When the verification result is successful, the converted DSL is sent to the downstream API interface and passed to the downstream customer matching system, so that the customer matching system can filter potential customers that meet the conditions from the database and generate detailed filtering results containing basic customer information, matching score and core matching dimensions to obtain matching results.

[0032] The further technical solution is as follows: The matching results are displayed in a user-friendly manner through a customizable wbit module, and visual annotations are performed, including:

[0033] Retrieve user-defined content, including display parameters for layout, matching weight, and annotation styles;

[0034] Based on user-defined content, the matching results are transformed into an intuitive display format including sorting, visual annotation, and structured pop-ups.

[0035] This invention also provides a customer intelligent recommendation system based on a large model, comprising:

[0036] The acquisition and conversion unit is used to acquire the user's natural language business opportunity requirements and convert them into text information in a unified format;

[0037] The extraction unit is used to perform deep analysis of the text information using a large language model, extract user demand types and related core business entities, and form preliminary structured information.

[0038] The improvement unit is used to determine the information completeness of the preliminary structured information based on the generation capability of a large language model, and to supplement missing or ambiguous information through multi-round interactive dialogue to obtain structured data.

[0039] The matching unit is used to transform the structured data into a domain-specific DSL query, verify it, send it to the downstream system to filter and match customers, and generate matching results.

[0040] The display unit is used to present the matching results in a user-friendly manner through a customizable wbit module and to perform visual annotations.

[0041] The present invention also provides a computer device, the computer device including a memory and a processor, the memory storing a computer program, and the processor executing the computer program to implement the above-described method.

[0042] The advantages of this invention compared to existing technologies are as follows: This invention transforms acquired user natural language business opportunity needs into text information in a unified format, and uses a large-scale language model for deep analysis to extract user need types and core business entities to form preliminary structured information. Subsequently, based on the model's generation capabilities, the completeness of the information is assessed, and missing or ambiguous information is supplemented through multiple rounds of interaction to ensure the accuracy and completeness of the data. Next, the complete structured data is converted into a domain-specific DSL query, which, after verification, is sent to downstream systems to filter and match customers and generate results. Finally, a customizable wbit module is used to humanize the display of the matching results, realizing fully automated processing from unstructured natural language to structured business data. This method not only possesses powerful natural language understanding and processing capabilities, accurately capturing complex user needs, but also proactively prompts users to supplement relevant details when information is insufficient, thereby reducing manual intervention and significantly improving processing efficiency and accuracy.

[0043] The present invention will be further described below with reference to the accompanying drawings and specific embodiments. Attached Figure Description

[0045] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0046] Figure 1 A flowchart illustrating the customer intelligent recommendation method based on a large model provided in an embodiment of the present invention;

[0047] Figure 2 A schematic block diagram of a customer intelligent recommendation device based on a large model provided in an embodiment of the present invention;

[0048] Figure 3 A schematic block diagram of a computer device provided for an embodiment of the present invention. Detailed Implementation

[0050] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0051] It should be understood that, when used in this specification and the appended claims, the terms "comprising" and "including" indicate the presence of the described features, integrals, steps, operations, elements and / or components, but do not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components and / or collections thereof.

[0052] It should also be understood that the terminology used in this specification is for the purpose of describing particular embodiments only and is not intended to limit the invention. As used in this specification and the appended claims, the singular forms “a,” “an,” and “the” are intended to include the plural forms unless the context clearly indicates otherwise.

[0053] It should also be further understood that the term "and / or" as used in this specification and the appended claims refers to any combination of one or more of the associated listed items and all possible combinations, and includes such combinations.

[0054] Please see Figure 1 , Figure 1This is a flowchart illustrating the customer intelligent recommendation method based on a large model provided in this embodiment of the invention. This method, applied to a server, automates the entire process from unstructured natural language business opportunity requirements to structured business data. First, it acquires user natural language input through various user interfaces and converts it into text information in a unified format. Then, it uses a large language model to perform deep analysis of the text, extracting user demand types and related core business entities to form preliminary structured information. After assessing the completeness of the information, if missing or ambiguous information is found, it is supplemented through multiple rounds of interactive dialogue to ensure accurate structured data. Subsequently, this data is transformed into domain-specific DSL queries and sent to downstream systems to filter and match customers and generate matching results. Finally, the results are displayed intuitively and visually annotated using a customizable wbit module. The entire process possesses powerful natural language understanding and processing capabilities, not only capturing complex user needs but also proactively prompting for supplementary details, thereby reducing manual intervention and improving processing efficiency and accuracy. Furthermore, it continuously fine-tunes the model by collecting user feedback to further enhance service accuracy and user experience.

[0055] Figure 1 This is a flowchart illustrating the customer intelligent recommendation method based on a large model provided in an embodiment of the present invention. Figure 1 As shown, the method includes the following steps S110 to S160.

[0056] S110: Obtain the user's natural language business opportunity requirements and convert them into text information in a unified format.

[0057] In this embodiment, natural language business opportunity demand refers to the description of business opportunities input by users through various interfaces (such as web interface, mobile APP interface, mini-program interface, voice interaction interface, etc.). These descriptions may include, but are not limited to, the desire to find potential customers in a specific industry, seeking partners for business expansion, or understanding the opportunities of a certain product or service in a specific geographic market. This demand is usually expressed through the user's natural language, which can be in the form of text input or voice input.

[0058] Text information refers to converting the natural language business opportunity requests input by users through the aforementioned interface into a standardized text representation. This process is crucial to ensuring that subsequent processing steps can correctly parse and understand user needs.

[0059] Specifically, the system obtains natural language business opportunity requirements input by users through a user interface; the user interface includes at least one of a web interface, a mobile APP interface, a mini-program interface, and a voice interaction interface.

[0060] In this embodiment, users can submit their business opportunity requirements in various ways. For example, they can directly enter a text description on a webpage or mobile application; use the voice input function of a mobile application and then convert the speech into text using speech-to-text technology; or submit their requirements using a mini-program and a specially designed voice interaction interface.

[0061] Regardless of how a user submits their request, the system needs to convert it into a standardized text format. This helps ensure consistency and reliability in subsequent semantic parsing and structured processing. For example, if a user inputs their request via voice, the audio signal must first be converted into text using speech recognition technology; then, further processing may be required to remove unnecessary punctuation, correct spelling errors, etc., thereby generating a standard text format that is easy for large language models to process.

[0062] In summary, step S110 emphasizes how to effectively collect users' original business opportunity needs and transform them into text information in a unified format that is easy for computers to process, laying the foundation for subsequent deep semantic analysis and structured prototype extraction (step S120). This process not only involves technical implementation but also reflects a focus on user experience, aiming to enable users to express their business intentions in the most natural and convenient way.

[0063] S120. Utilize a large-scale language model to perform deep analysis on the text information, extract user demand types and related core business entities, and form preliminary structured information.

[0064] In this embodiment, the preliminary structured information refers to a data structure that includes the user's intent type regarding business opportunity needs, related core business entities, and specific business scenario parameters. This data structure aims to transform the user's natural language description into a machine-understandable form, enabling subsequent steps to perform more accurate processing and customer matching based on this information.

[0065] In one embodiment, step S120 described above may include steps S121 to S123.

[0066] S121. Perform preprocessing operations on the text information to obtain preprocessing results.

[0067] In this embodiment, the preprocessing result refers to the intermediate data generated after performing a series of basic processing steps on the original text information. These processing steps include, but are not limited to, word segmentation (splitting sentences into words or phrases), part-of-speech tagging (identifying the grammatical role of each word), dependency parsing (determining the dependencies between words), and semantic role tagging (identifying the function of each component in the sentence). Through these processing steps, the grammatical structure and semantic relationships of the text can be revealed, laying the foundation for subsequent deep analysis.

[0068] Specifically, the text information is processed by word segmentation, part-of-speech tagging, dependency parsing, and semantic role tagging to obtain preprocessing results.

[0069] S122. A large language model is used to determine the user's specific business opportunity demand type from the preprocessed results, and named entity recognition technology is used to extract key business entities from the preprocessed results to obtain analysis results.

[0070] In this embodiment, the analysis results are information derived from further processing of the preprocessing results. Specifically, this step first uses a large language model for intent recognition to clarify the type of business opportunity demand raised by the user (e.g., customer expansion, cooperation negotiation, etc.). Next, named entity recognition technology is applied to extract entity information closely related to the business from the text (such as industry category, product name, customer size, geographical scope, etc.). The final analysis results include the judgment of the user's demand type and the key business entities extracted from it.

[0071] S123. Based on the analysis results, construct preliminary structured information containing user needs and intentions, core business entities, and specific business scenario parameters.

[0072] In this step, based on the analysis results obtained in the first two steps, the system constructs a detailed preliminary structured information framework. This framework not only clarifies the user's intended needs (i.e., the purpose the user hopes to achieve through this query), but also lists all identified core business entities and their attributes, and combines them with specific business scenario parameters (such as budget constraints, time periods, etc.) to form a complete dataset. This preliminary structured information provides the necessary input for the next step of information completion and clarification, ensuring the coherence and accuracy of the entire process.

[0073] In this embodiment, step 2 aims to use a large language model to perform in-depth analysis of the received natural language business opportunity requirements, thereby extracting a preliminary structured information containing user demand intent type, core business entities, and business scenario parameters.

[0074] First, a series of preprocessing operations are performed on the input natural language business opportunity requirements to clarify the grammatical structure and semantic relationships of the text. These preprocessing operations include, but are not limited to:

[0075] Word segmentation: dividing a sentence into words or phrases.

[0076] Part-of-speech tagging: Marks the grammatical role (such as noun, verb, etc.) of each word in a sentence.

[0077] Dependency parsing: Identifies dependencies between words to help understand sentence structure.

[0078] Semantic role labeling: Determine the function of each component in a sentence (such as who is the performer of the action and who is the receiver) in order to gain a deeper understanding of the meaning of the sentence.

[0079] Through the above preprocessing operations, a basic understanding framework can be built to support further semantic analysis.

[0080] Based on the preprocessing results, a large language model is then used for intent recognition. The goal of this step is to determine the type of business opportunity the user needs. For example, a user's request might be related to customer expansion, business negotiations, or market research. Through deep semantic analysis of the text, the model can accurately identify the user's actual intent, which is the first step towards achieving precise recommendations.

[0081] Named Entity Recognition (NER) is used to extract core business-related entities from text. These entities include, but are not limited to:

[0082] Industry type: The specific industry sector that the user is interested in.

[0083] Product Name: The specific name of the product or service.

[0084] Customer size: The size or number range of the target customers.

[0085] Geographic scope: The geographical area where the user wishes to conduct business.

[0086] Partnership Budget: The amount of money users plan to invest in the partnership.

[0087] Timeframe: The time limit for project implementation or the expected completion date.

[0088] NER technology can accurately extract this key information from a large amount of natural language descriptions, serving as important input for subsequent steps.

[0089] Finally, based on the results of the previous two steps—intent recognition and named entity recognition—the system constructs a preliminary structured information prototype. This prototype not only includes the user's specific needs and intents (such as the goal they hope to achieve through this query), but also integrates all identified core business entities and their attributes, combined with specific business scenario parameters (such as budget constraints, time periods, etc.). This structured representation makes the data more orderly, easier to understand and process, laying the foundation for the next step of information completion and clarification.

[0090] In summary, step S120, through a series of meticulous operations, transforms the user's natural language description into clear, ordered, and structured information, laying a solid foundation for further customer recommendation processes. This process demonstrates the powerful capabilities of modern artificial intelligence technology in understanding and processing complex natural language tasks.

[0091] S130. Based on the large-scale language model generation capability, determine the information completeness of the preliminary structured information, and supplement the missing or ambiguous information through multi-round interactive dialogue to obtain structured data.

[0092] In this embodiment, structured data refers to the collection of formatted information that is transformed from initially identified information into a clear and accurate set of information containing all necessary key details through supplementation and clarification.

[0093] In this embodiment, step S130 aims to evaluate the preliminary structured information by utilizing the generation capabilities of a large language model, and to supplement missing or ambiguous information through multi-turn interactive dialogue with the user when necessary, thereby forming complete structured data.

[0094] In one embodiment, step S130 described above may include steps S131 to S132.

[0095] S131. The integrity of the preliminary structured information is evaluated using the generation capability of a large language model to obtain the evaluation result.

[0096] In this embodiment, the evaluation result refers to the comprehensive judgment given by the large language model after checking the completeness and clarity of the preliminary structured information, regarding which parts of the information are complete and which need further supplementation or clarification.

[0097] First, the large language model checks the completeness of the initial structured information based on preset core information fields (which vary depending on the business scenario). Core information fields may include key elements such as industry type, product name, and customer size.

[0098] In addition to checking the completeness of information, the model also evaluates the clarity and accuracy of existing information. For example, some descriptions may be too broad or ambiguous and require further clarification.

[0099] Based on the above two assessments, the model will provide a comprehensive evaluation result, indicating which parts are complete and which parts require further information collection.

[0100] S132. When the evaluation results are ambiguous or key information is missing, guiding questions are automatically generated, and multiple rounds of interactive dialogue are initiated through the user interface to collect more information. In each round of dialogue, the user's response is analyzed in conjunction with the context to supplement and clarify the data in the preliminary structured information until all necessary information is complete and clear, so as to obtain the structured data.

[0101] In this embodiment, if the evaluation results show that there is ambiguity or missing key information, the system will automatically generate a series of guiding questions. These questions are designed to guide the user to provide more specific and clear information in order to fill in the gaps or clarify the ambiguities.

[0102] The system employs a context-aware mechanism for multi-turn interactive dialogues. This means that in each round of dialogue, the large language model not only considers the current question and answer but also refers to all previous dialogue content to ensure that information that has already been explicitly asked is not repeatedly asked.

[0103] As the conversation progresses, the system continuously updates and optimizes the initial structured information. Each time new information is obtained from the user, the overall completeness is reassessed, and a decision is made as to whether further questions should be asked or if the process can proceed to the next step.

[0104] This process will continue until all necessary information has been collected and expressed clearly and accurately enough. At this point, the initial structured information transforms into complete structured data, containing all the key details of the user's business opportunity needs, providing a solid foundation for subsequent steps.

[0105] Specifically, the system leverages the generation capabilities of a large language model to assess the completeness and accuracy of the preliminary structured information obtained in step S120. This process determines whether the structured information covers all necessary information points based on the core information field set preset by the business scenario configuration module. If ambiguous information or missing key information is found, the system automatically generates guiding questions and initiates a dialogue with the user through the user interface to solicit more information. This interaction process is multi-round and context-aware, ensuring that already clear information is not repeatedly asked. Each user response is analyzed within the existing dialogue context to supplement and improve the initial structured information until it reaches a complete and clear standard. Once the information is confirmed to be complete and accurate, the system proceeds to the next step.

[0106] In summary, step S130, by intelligently utilizing the capabilities of a large language model, not only efficiently identifies deficiencies in the initial structured information but also effectively resolves these issues through intelligent interactive methods, ultimately ensuring that the output data is both comprehensive and accurate. This method significantly improves the efficiency and accuracy of processing natural language business opportunity requests.

[0107] S140. The structured data is converted into a domain-specific DSL query, verified, and then sent to the downstream system to filter and match customers, and a matching result is generated.

[0108] In this embodiment, the matching result refers to the list of potential customers who meet the criteria selected from the customer database based on the converted DSL query, including the customer's detailed information, matching score, and core matching dimensions.

[0109] In one embodiment, step S140 described above may include steps S141 to S143.

[0110] S141. The structured data is converted into a domain-specific language to obtain the converted DSL.

[0111] In this embodiment, the transformed DSL refers to a query language expression that is converted from structured business data according to the syntax rules and instruction templates of a specific domain, and can be recognized and executed by the downstream customer matching system.

[0112] Specifically, based on the business opportunity requirements input by the user, a series of processing steps culminate in a list of qualified potential customers selected from the database by the downstream customer matching system. This list includes not only detailed basic information about the customers but also a matching score for each customer and analysis of core matching dimensions, helping users quickly identify the customers best suited to their needs.

[0113] S142. Verify the converted DSL, including checking syntax correctness, semantic consistency, and business logic legality, to obtain the verification result.

[0114] In this embodiment, the verification result refers to the conclusion drawn after checking the syntax correctness, semantic consistency, and business logic legality of the converted DSL, in order to confirm whether it is suitable for further processing and practical application.

[0115] S143. When the verification result is successful, the converted DSL is sent to the downstream API interface and passed to the downstream customer matching system, so that the customer matching system can filter potential customers that meet the conditions from the database and generate detailed filtering results containing basic customer information, matching score and core matching dimensions to obtain matching results.

[0116] This step emphasizes that only after successful verification will the next step—sending the DSL query to the downstream system to perform specific customer matching tasks—begin. This process ensures that only valid and compliant queries are further processed, thereby improving the accuracy and efficiency of the entire recommendation process. The final matching results provide users with an intuitive and useful way to evaluate and select the most suitable business partners.

[0117] In this embodiment, the final structured business data, after completion and clarification in step S130, is first converted into a Domain-Specific Language (DSL). This language is specifically designed for customer matching business and includes relevant grammar rules and instruction templates, enabling it to accurately express query logic based on structured business data. Next, a preset DSL parser performs a series of checks on the converted DSL, including grammatical correctness, semantic consistency, and business logic validity. If the DSL fails these checks, it needs to return to step S130, where a large language model optimizes the parsing results based on specific error messages and retryes the DSL conversion. Once the DSL passes all checks, it is sent to the downstream customer matching system API interface. This system filters target customers from the database based on the query conditions in the DSL and generates a list of filtered results containing basic customer information, matching scores, and core matching dimensions. Thus, the entire process ensures the accuracy and effectiveness from the original input to the final customer filtering results.

[0118] S150. The matching results are displayed in a user-friendly manner through a customizable wbit module, and visual annotations are provided.

[0119] The wbit module's user-friendly display features present matched customer information filtered by the downstream customer matching system to users in an intuitive and user-friendly format. The wbit module integrates display format configuration and result rendering units, aiming to provide highly customizable display options to meet the specific needs of different users.

[0120] In one embodiment, step S150 described above may include steps S151 to S152.

[0121] S151. Obtain user-defined content, including display parameters for layout, matching weight, and annotation style.

[0122] First, in step S151, the system collects the user's specific preference settings for the display effect. This includes, but is not limited to:

[0123] Display layout: Users can choose different view modes to view the matching results, such as list view, card view, or chart view.

[0124] Matching weight settings: Allows users to adjust the relative importance of various matching dimensions (such as geographical location, historical transaction volume, customer feedback rating, etc.) according to their business priorities, thereby affecting the final ranking.

[0125] Visual annotation styles: Users can specify how to highlight key matching dimensions or points of particular interest using color, icons, or other visual elements.

[0126] These settings offer great flexibility, allowing the final presentation to precisely meet the user's expectations and needs.

[0127] S152. Based on the user-defined content, the matching results are transformed into an intuitive display format including sorting, visual annotation, and structured pop-ups.

[0128] Next, in step S152, based on the user-defined parameters obtained from step S151, the result rendering unit begins its work. It transforms the processed matching customer data into the following intuitive display formats:

[0129] Sorting and Display: All matched customers are reordered according to the matching weight set by the user, ensuring that the most important potential customers always appear in the most prominent position.

[0130] Visual annotation: The core matching dimensions are marked using user-selected styles, enabling users to quickly identify each customer's unique strengths or the best fit with their business needs.

[0131] Structured pop-ups: Provide users with a convenient way to quickly access detailed information about any customer. A small window containing basic customer information, interaction history, and other relevant data will pop up when clicked.

[0132] This process not only improves the readability and comprehensibility of information, but also greatly enhances the user experience, enabling users to more efficiently evaluate and select suitable matching clients.

[0133] S160: Collect user feedback and interaction data as training samples to continuously fine-tune the large language model.

[0134] In this embodiment, the first step involves systematically collecting user feedback on the recommendation results. This feedback may include, but is not limited to:

[0135] Whether users accept the recommended customers: The degree to which users accept the customers recommended by the system directly reflects the accuracy and relevance of the recommendation algorithm.

[0136] Adjusting requirements: If users' business needs change (such as expanding the target market or changing product preferences), they may request corresponding adjustments to the recommendation logic.

[0137] In addition to explicit user feedback, it is also necessary to record various data generated throughout the interaction process, which mainly includes:

[0138] Natural Language Requirement Text: The initial description of the requirements submitted by the user in natural language.

[0139] Multi-turn dialogue records: Content from multiple interactions between the user and the system, which helps to refine the user's specific needs.

[0140] Analysis results: The understanding results obtained by the system after performing semantic analysis on the natural language input of the user.

[0141] DSL Conversion Result: The result of processing user requirements into domain-specific language (DSL) expressions to facilitate subsequent calculations and queries.

[0142] Once this data is collected, it can serve as a valuable training resource for fine-tuning large-scale language models. Specifically:

[0143] Continuously optimize semantic understanding accuracy: By learning from data samples containing both correct and incorrect interpretations, the model can better grasp the meaning of words in different contexts and improve its overall semantic understanding ability.

[0144] Intent recognition accuracy: Based on the comparison between the user's true intent and the feedback they provide, the model can learn to more accurately identify the user's actual needs.

[0145] Information completion efficiency: When the information provided by the user is incomplete, the model needs to be able to infer the missing parts from the existing data. Through continuous iterative training, this "gap-filling" skill will gradually be strengthened.

[0146] Ultimately, this series of optimizations aims to improve the system's intelligence and service quality, thereby providing users with a more accurate and personalized service experience. Over time, with technological advancements, the model will continue to evolve, adapting to new challenges and meeting growing user expectations. Furthermore, this approach encourages a positive feedback loop, where every interaction becomes an opportunity to improve future performance.

[0147] This embodiment uses a large language model as the core of intelligent processing, constructing an end-to-end intelligent recommendation architecture of "natural language reception - deep semantic parsing - proactive completion and clarification - DSL conversion - user-friendly display," achieving high-precision automated conversion from unstructured natural language to structured business data. Based on the generative capabilities of the large language model, this embodiment designs a context-aware multi-turn interactive completion mechanism that can automatically identify ambiguous or missing information and proactively guide users to complete it. This method solves the problem of existing technologies being unable to proactively complete and clarify information. This embodiment introduces a customized DSL to achieve accurate mapping of structured data to downstream system instructions, and uses syntax validation to ensure the accuracy of filtering. Simultaneously, the wbit module provides an advanced display format, supports user-defined configurations, and improves the efficiency of result usage.

[0148] In summary, the method of this embodiment not only achieves multiple innovations at the technical level, but also demonstrates significant advantages in practical applications, providing users with a more convenient, efficient, and accurate service experience.

[0149] In this embodiment, a large-scale language model is used as the core, allowing users to directly input their business opportunity needs through natural language, without the need for manual input or clicking to select structured filtering conditions. This method breaks through the click-based interaction limitations of traditional supplier pages, significantly reducing the operational difficulty and learning cost for users, and improving operational convenience and user experience.

[0150] Leveraging the powerful capabilities of large-scale language models, this embodiment achieves deep semantic understanding. Combined with intent recognition and named entity recognition technologies, it can accurately extract the initial form of structured information. Compared to traditional simple keyword matching methods, this approach offers deeper semantic understanding and more precise demand transformation, laying a solid foundation for improving customer matching accuracy.

[0151] Leveraging the generative capabilities of large models, this embodiment possesses the ability to proactively initiate multiple rounds of interactive completion and clarification. It can automatically identify and correct ambiguous and missing key information, ensuring the integrity and accuracy of structured information. This avoids matching errors caused by incomplete information, further improving the accuracy of recommendations.

[0152] Through DSL (Domain-Specific Language) conversion and syntax verification mechanisms, this embodiment achieves fully automated conversion from unstructured natural language to structured business data and then to downstream system-recognizable instructions. The entire process requires no manual intervention, significantly improving the automation level and processing efficiency of customer recommendations while reducing enterprise labor costs.

[0153] Leveraging the personalized display format provided by the wbit module, this embodiment allows users to customize display parameters, intuitively presenting customer match ranking and core matching dimensions, enabling users to quickly filter out high-quality customers. This efficiency in viewing and using results greatly enhances the user experience.

[0154] This embodiment adds a model optimization step, which continuously fine-tunes the model based on user feedback and interaction data, enabling it to adapt to different user expression habits and diverse business scenarios, thereby improving the versatility and adaptability of the method.

[0155] The aforementioned customer intelligent recommendation method based on a large model transforms acquired user natural language business opportunity needs into text information in a unified format. A large language model is then used for deep analysis to extract user need types and core business entities, forming preliminary structured information. Subsequently, the model's generation capabilities are used to assess information completeness, and multiple rounds of interaction are used to supplement missing or ambiguous information, ensuring data accuracy and integrity. Next, the complete structured data is converted into a domain-specific DSL query, validated, and then sent to downstream systems to filter and match customers and generate results. Finally, a customizable wbit module is used to humanize the display of matching results, achieving fully automated processing from unstructured natural language to structured business data. This method not only possesses powerful natural language understanding and processing capabilities, accurately capturing complex user needs, but also proactively prompts users to supplement relevant details when information is insufficient, thereby reducing manual intervention and significantly improving processing efficiency and accuracy.

[0156] Figure 2 This is a schematic block diagram of a customer intelligent recommendation device 300 based on a large model provided in an embodiment of the present invention. Figure 2 As shown, corresponding to the above-described customer intelligent recommendation method based on a large model, the present invention also provides a customer intelligent recommendation device 300 based on a large model. This customer intelligent recommendation device 300 includes a unit for executing the above-described customer intelligent recommendation method based on a large model, and the device can be configured in a server. Specifically, please refer to... Figure 2 The customer intelligent recommendation device 300 based on a large model includes an acquisition and conversion unit 301, an extraction unit 302, a refinement unit 303, a matching unit 304, a display unit 305, and a fine-tuning unit 306.

[0157] The system comprises the following components: A conversion unit 301 acquires users' natural language business opportunity needs and converts them into text information in a unified format; an extraction unit 302 uses a large-scale language model to perform deep analysis on the text information, extracting user need types and related core business entities to form preliminary structured information; a refinement unit 303 judges the completeness of the preliminary structured information based on the large-scale language model's generation capabilities and supplements missing or ambiguous information through multi-round interactive dialogue to obtain structured data; a matching unit 304 converts the structured data into a domain-specific DSL query, verifies it, and sends it to the downstream system to filter and match customers, generating matching results; a display unit 305 displays the matching results in a user-friendly manner using a customizable wbit module and provides visual annotations; and a fine-tuning unit 306 collects user feedback and interaction data as training samples to continuously fine-tune the large-scale language model.

[0158] In one embodiment, the acquisition and conversion unit 301 is used to acquire natural language business opportunity requirements input by the user through a user interface; the user interface includes at least one of a web interface, a mobile APP interface, a mini-program interface, and a voice interaction interface.

[0159] In one embodiment, the extraction unit 302 includes:

[0160] The preprocessing subunit is used to preprocess the text information to obtain preprocessing results; the analysis subunit is used to use a large language model to determine the user's specific business opportunity demand type from the preprocessing results, and to use named entity recognition technology to extract key business entities from the preprocessing results to obtain analysis results; the construction subunit is used to construct preliminary structured information containing user demand intent, core business entities and specific business scenario parameters based on the analysis results.

[0161] In one embodiment, the preprocessing subunit is used to perform word segmentation, part-of-speech tagging, dependency parsing, and semantic role tagging on the text information to obtain preprocessing results.

[0162] In one embodiment, the improvement unit 303 includes:

[0163] The evaluation subunit is used to evaluate the completeness of the preliminary structured information using the generative capabilities of a large language model to obtain an evaluation result. The information improvement subunit is used to automatically generate guiding questions when the evaluation result is ambiguous or key information is missing, and to initiate multi-round interactive dialogues through the user interface to collect more information. In each round of dialogue, the subunit combines context analysis of the user's response to supplement and clarify the data in the preliminary structured information until all necessary information is complete and clear, so as to obtain the structured data transformation.

[0164] In one embodiment, the matching unit 304 includes:

[0165] The transformation subunit is used to transform the structured data into a domain-specific language to obtain a transformed DSL; the verification subunit is used to verify the transformed DSL, including checking syntactic correctness, semantic consistency, and business logic legality, to obtain a verification result; the matching subunit is used to send the transformed DSL to the downstream API interface and pass it to the downstream customer matching system when the verification result is successful, so that the customer matching system can filter potential customers that meet the conditions from the database and generate detailed filtering results containing basic customer information, matching score, and core matching dimensions to obtain a matching result.

[0166] In one embodiment, the display unit 305 includes: a content acquisition unit, used to acquire user-defined content, including display parameters such as layout, matching degree weight, and annotation style; and a sorting display unit 305, used to convert the matching results into an intuitive display format of sorting display, visual annotation, and structured pop-up based on the user-defined content.

[0167] It should be noted that those skilled in the art can clearly understand that the specific implementation process of the above-mentioned customer intelligent recommendation device 300 based on a large model and each unit can be referred to the corresponding description in the foregoing method embodiments. For the sake of convenience and brevity, it will not be repeated here.

[0168] The aforementioned customer intelligent recommendation device 300 based on a large model can be implemented as a computer program, which can, for example... Figure 3 It runs on the computer device shown.

[0169] Please see Figure 3 , Figure 3 This is a schematic block diagram of a computer device provided in an embodiment of this application. The computer device 500 can be a server, wherein the server can be a standalone server or a server cluster composed of multiple servers.

[0170] See Figure 3 The computer device 500 includes a processor 502, a memory, and a network interface 505 connected via a system bus 501. The memory may include a non-volatile storage medium 503 and internal memory 504.

[0171] The non-volatile storage medium 503 may store an operating system 5031 and a computer program 5032. The computer program 5032 includes program instructions that, when executed, cause the processor 502 to perform a customer intelligent recommendation method based on a large model.

[0172] The processor 502 provides computing and control capabilities to support the operation of the entire computer device 500.

[0173] The internal memory 504 provides an environment for the execution of the computer program 5032 in the non-volatile storage medium 503. When the computer program 5032 is executed by the processor 502, the processor 502 can execute a customer intelligent recommendation method based on a large model.

[0174] This network interface 505 is used for network communication with other devices. Those skilled in the art will understand that... Figure 3 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device 500 to which the present application is applied. The specific computer device 500 may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0175] The processor 502 is used to run a computer program 5032 stored in the memory to implement all the steps of the customer intelligent recommendation method based on the large model.

[0176] It should be understood that in the embodiments of this application, the processor 502 may be a central processing unit (CPU), or it may be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or any conventional processor.

[0177] It will be understood by those skilled in the art that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program includes program instructions and can be stored in a storage medium, which is a computer-readable storage medium. The program instructions are executed by at least one processor in the computer system to implement the process steps of the embodiments of the above methods.

[0178] Therefore, the present invention also provides a storage medium. This storage medium can be a computer-readable storage medium. The storage medium stores a computer program, wherein when executed by a processor, the computer program causes the processor to perform all the steps of the large-model-based intelligent customer recommendation method.

[0179] The storage medium can be any computer-readable storage medium capable of storing program code, such as a USB flash drive, portable hard drive, read-only memory (ROM), magnetic disk, or optical disk.

[0180] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention.

[0181] In the several embodiments provided by this invention, it should be understood that the disclosed apparatus and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative. For example, the division of each unit is merely a logical functional division, and there may be other division methods in actual implementation. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed.

[0182] The steps in the method of this invention can be adjusted, merged, or reduced in order according to actual needs. The units in the device of this invention can be merged, divided, or reduced according to actual needs. Furthermore, the functional units in the various embodiments of this invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0183] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, a terminal, or a network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention.

[0184] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any person skilled in the art can easily conceive of various equivalent modifications or substitutions within the technical scope disclosed in the present invention, and these modifications or substitutions should all be covered within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A customer intelligent recommendation method based on a large model, characterized in that, include: Obtain users' natural language business opportunity needs and convert them into text information in a unified format; The text information is analyzed in depth using a large language model to extract user demand types and related core business entities, forming preliminary structured information. The completeness of the preliminary structured information is judged based on the ability to generate large-scale language models, and missing or ambiguous information is supplemented through multi-round interactive dialogue to obtain structured data. The structured data is transformed into a domain-specific DSL query, verified, and then sent to the downstream system to filter and match customers, and generate matching results. The matching results are displayed in a user-friendly manner and visually annotated using a customizable wbit module.

2. The customer intelligent recommendation method based on a large model according to claim 1, characterized in that, After displaying the matching results in a user-friendly manner using the customizable wbit module and performing visual annotations, the method further includes: Collect user feedback and interaction data as training samples to continuously fine-tune a large language model.

3. The customer intelligent recommendation method based on a large model according to claim 1, characterized in that, The process of acquiring users' natural language business opportunity needs and converting them into text information in a unified format includes: The system obtains natural language business opportunity requirements input by users through a user interface; the user interface includes at least one of a web interface, a mobile APP interface, a mini-program interface, and a voice interaction interface.

4. The customer intelligent recommendation method based on a large model according to claim 1, characterized in that, The process involves using a large language model to perform deep analysis of the text information, extracting user demand types and related core business entities, and forming preliminary structured information, including: The text information is preprocessed to obtain the preprocessing result; The preprocessed results are used to determine the specific business opportunity demand type of the user using a large language model, and key business entities are extracted from the preprocessed results using named entity recognition technology to obtain the analysis results. Based on the analysis results, preliminary structured information is constructed, which includes user needs and intentions, core business entities, and specific business scenario parameters.

5. The customer intelligent recommendation method based on a large model according to claim 4, characterized in that, The text information is preprocessed to obtain a preprocessing result; The text information is processed by word segmentation, part-of-speech tagging, dependency parsing, and semantic role tagging to obtain preprocessing results.

6. The customer intelligent recommendation method based on a large model according to claim 1, characterized in that, The process involves determining the completeness of the preliminary structured information based on the generation capabilities of a large language model, and supplementing missing or ambiguous information through multi-round interactive dialogue to obtain structured data, including: The integrity of the preliminary structured information is assessed using the generative capabilities of a large language model to obtain the assessment results; When the evaluation results are ambiguous or lack key information, guiding questions are automatically generated, and multiple rounds of interactive dialogue are initiated through the user interface to collect more information. In each round of dialogue, the user's response is analyzed in conjunction with the context to supplement and clarify the data in the preliminary structured information until all necessary information is complete and clear, so as to obtain the structured data.

7. The customer intelligent recommendation method based on a large model according to claim 1, characterized in that, The process of converting the structured data into a domain-specific DSL query, verifying it, sending it to the downstream system for filtering and matching customers, and generating matching results includes: The structured data is converted into a domain-specific language to obtain the converted DSL; The converted DSL is validated, including checks on syntax correctness, semantic consistency, and business logic legality, to obtain the validation results. When the verification result is successful, the converted DSL is sent to the downstream API interface and passed to the downstream customer matching system, so that the customer matching system can filter potential customers that meet the conditions from the database and generate detailed filtering results containing basic customer information, matching score and core matching dimensions to obtain matching results.

8. The customer intelligent recommendation method based on a large model according to claim 1, characterized in that, The method of displaying the matching results in a user-friendly manner through a customizable wbit module and providing visual annotations includes: Retrieve user-defined content, including display parameters for layout, matching weight, and annotation styles; Based on user-defined content, the matching results are transformed into an intuitive display format including sorting, visual annotation, and structured pop-ups.

9. A customer intelligent recommendation system based on a large model, characterized in that: include: The acquisition and conversion unit is used to acquire the user's natural language business opportunity requirements and convert them into text information in a unified format; The extraction unit is used to perform deep analysis of the text information using a large language model, extract user demand types and related core business entities, and form preliminary structured information. The improvement unit is used to determine the information completeness of the preliminary structured information based on the generation capability of a large language model, and to supplement missing or ambiguous information through multi-round interactive dialogue to obtain structured data. The matching unit is used to transform the structured data into a domain-specific DSL query, verify it, send it to the downstream system to filter and match customers, and generate matching results. The display unit is used to present the matching results in a user-friendly manner through a customizable wbit module and to perform visual annotations.

10. A computer device, characterized in that, The computer device includes a memory and a processor, the memory storing a computer program, and the processor executing the computer program to implement the method as described in any one of claims 1 to 8.

Images