A digital intelligent service system for enterprise services

Abstract

The present invention discloses a digital intelligent service system for enterprise services, which belongs to the field of enterprise digital services and includes a user demand analysis module, an intelligent data adaptation module, a service resource matching module, a service efficiency evaluation module and a user-friendly interface; the user demand analysis module includes semantic parsing and classification decision-making; the intelligent data adaptation module identifies and parses protocol types and establishes an intelligent data transmission channel; the service resource matching module performs resource pre-screening and precise matching; the service efficiency evaluation module tracks and identifies a real-time capture of a four-dimensional indicator system, and the dynamic evaluation unit implements service status graded warning through a sliding window algorithm; the present invention realizes precise demand identification, improves response speed and data transmission efficiency, improves user satisfaction, enhances service matching accuracy, optimizes resource utilization efficiency, improves service capabilities and customer experience, and maximizes economic benefits and social value.

Description

Technical Field

[0001] The present invention belongs to the field of enterprise digital services, and specifically relates to a digital intelligent service system for enterprise services. Background Art

[0002] In the current digital economy, the enterprise service sector has undergone significant technological change. Digital and intelligent service systems have gradually become important tools for improving business operational efficiency and optimizing resource allocation. With the rapid development of information technology and the integrated application of technologies such as cloud computing, big data, and artificial intelligence, enterprise service systems are constantly evolving, aiming to achieve efficient response to user needs and dynamic matching of service resources. Traditional enterprise service models often suffer from information silos and inefficiencies. However, new service systems based on big data analysis and intelligent decision-making technologies can accurately capture user needs and quickly match corresponding service resources through real-time data processing and analysis, thereby improving service quality and user satisfaction. At the same time, intelligent service systems are also beginning to play a vital role in various aspects of enterprise customer relationship management, operations management, and decision support, providing enterprises with more convenient and efficient business support.

[0003] While existing technologies have made some progress in the digital transformation of enterprise services, some shortcomings remain. For example, traditional user demand analysis methods rely on single-keyword matching and formatted rule engines, making it difficult to accurately capture users' true intent, which can easily lead to delayed service responses and inaccurate matching. Single-BERT applications, according to test data from the NLP Industry White Paper (2023 Edition), show a low accuracy rate of only 58.7% in identifying long-tail demand and a 34.2% misclassification rate for industry terminology. These limitations are inherent in these systems. Many systems still rely on static feature extraction methods and are unable to dynamically adapt to changes in user behavior and needs. Citing the IDC 2022 Enterprise Service System Failure Report, it was found that rule base update cycles exceeding 72 hours resulted in a 19.3% triggering rate for outdated rules, insufficient keyword coverage leading to a 27.8% missed detection rate, and a system deadlock probability of 8.4% per week caused by multiple rule conflicts. Furthermore, for the screening and matching of service resources, existing systems often rely solely on predefined rules and lack the ability to evaluate dynamic service characteristics in real time. This can lead to irrational allocation of service resources and impact enterprise operational efficiency. In response to these problems, we proposed the "Digital Intelligent Service System and Method for Enterprise Services" Summary of the Invention

[0004] In view of the above existing problems, the technical problems solved by the present invention are: solving the technical bottleneck of multimodal user demand intention recognition accuracy below 80%; breaking through the industry problem of the inability to coordinate optimization of static rules and dynamic loads in the service resource matching process; and optimizing the entire process of user demand analysis, service resource matching, and performance evaluation by building a digital intelligent service system. Users perform demand analysis through multiple channels to achieve intelligent identification of industry keywords and accurate generation of demand tags, and then use a dynamic evaluation mechanism in service resource matching to improve the feasibility review and efficiency of services.

[0005] To solve the above technical problems, a digital intelligent service system for enterprise services is proposed, which includes a user demand analysis module, an intelligent data adaptation module, a service resource matching module, a service effectiveness evaluation module and a user-friendly interface.

[0006] In the user demand analysis module, the semantic parsing unit uses the BERT-base model and the rule engine for dual-channel parallel processing, generating business labels through weighted fusion; the classification decision unit confirms the validity of the label through three-level verification and activates the protocol parser;

[0007] The intelligent data adaptation module calls the protocol feature library to verify the validity of the code corresponding to the tag. If the verification fails, the protocol sniffing mechanism is activated to generate a structured data packet and build an intelligent data transmission channel.

[0008] In the service resource matching module, the resource pre-screening unit extracts static / dynamic features and sets constraints, and the precise matching unit uses a dual-channel decision-making mechanism to match service resources;

[0009] The service performance evaluation module tracks and identifies the real-time capture of the four-dimensional indicator system, and the dynamic evaluation unit uses a sliding window algorithm to achieve service status graded warning;

[0010] The user-friendly interface allows users to input requirements through a web interface and presents the requirements parsing progress and resource matching path through the web interface.

[0011] As a preferred solution of the digital intelligent service system for enterprise services described in the present invention, wherein: the user demand analysis module includes a semantic parsing unit and a classification decision unit;

[0012] The semantic parsing unit is that when a user submits a demand text through the Web interface, the system captures user data and basic data in real time. At the same time, the system starts a parallel dual channel, including a main channel and an auxiliary channel;

[0013] The main channel uses the BERT-base model to extract text embedding vectors, generate multi-dimensional semantic features, and output candidate tags and probability values ​​based on the semantic features. The auxiliary channel uses the rule engine to scan the submitted demand text to see if there are x preset industry keywords. If there are industry keywords, they are marked as matches and business tags are generated.

[0014] The dual-channel results are integrated through a weighted fusion device. When the BERT-base output probability is greater than the probability threshold, the main channel result is adopted. When the BERT-base output probability does not exceed the probability threshold, it is determined whether the main channel label and the auxiliary channel label have the same label, and the service label to be adopted is determined. The final service type label is input into the classification decision unit for demand description and activation of the protocol decoder.

[0015] As a preferred solution of the digital intelligent service system for enterprise services described in the present invention, the classification decision unit includes generating a business type tag, inputting the tag into the system to perform three-level verification, the first level of verification is to analyze the user's request records in the past month through a sliding time window and compare the user's historical request pattern;

[0016] The second level of verification involves reviewing the service feasibility by calling the requirements description catalog. The requirements description catalog includes tags related to business type, urgency, and data volume estimates.

[0017] The third level of verification is to calculate the comprehensive confidence. When the comprehensive confidence is greater than the set threshold, the business type code of the current service is selected to activate the corresponding protocol parser first, and the protocol prediction code is sent at the same time.

[0018] As a preferred solution of the digital intelligent service system for enterprise services described in the present invention, the intelligent data adaptation module includes: when the user demand analysis module transmits the label and demand description directory, it carries the protocol pre-judgment code, the system starts the intelligent loading process, and calls the protocol feature library to verify the validity of the code;

[0019] When the code verification passes, the corresponding activated parser is dynamically selected according to the current system load to load the parsing template of the corresponding protocol, which includes field mapping rules and verification logic; when the code verification fails, the sniffing mechanism is activated to confirm the protocol type;

[0020] After selecting the parsing protocol, the user's needs are fully parsed through the parsing protocol, multiple demand tags are generated from the user's input demand text, and structured data packets are output, including protocol codes, business type tags and demand tags. At the same time, an intelligent data transmission channel is established with the corresponding enterprise end and adapted, and data flow hierarchical bandwidth allocation is implemented according to the urgency corresponding to the tags in the demand description directory.

[0021] As a preferred solution of the digital intelligent service system for enterprise services described in the present invention, wherein: the service resource matching module includes a resource pre-screening unit and a precise matching unit;

[0022] The resource pre-screening unit is to, after receiving the structured data packet, start service resource feature extraction, including service static features and service dynamic features, activate the corresponding service resource pool according to the business type label, and call the domain classification index of the resource library to screen the service resources;

[0023] Constraints are set based on the static and dynamic service features to filter out service nodes that meet the protocol code. The node sets filtered by the static and dynamic service features are placed into the activated service resource pool and packaged and input into the precise matching unit.

[0024] As a preferred solution of the digital intelligent service system for enterprise services described in the present invention, wherein: the precise matching unit includes receiving the packaged content of the resource pre-screening unit and executing the matching strategy, including a rule channel and a similarity channel;

[0025] The rule channel loads the service business rule library specified by the enterprise, calculates the business rule matching scores in sequence according to the priority order of the service resources in the service business rule library, re-prioritizes them from high to low according to the matching scores, and inputs them into the similarity channel. The similarity channel calculates the cosine similarity between the demand label and the service resource, calculates the final matching degree, automatically selects the service plan with the highest matching degree, retrieves the data stream of the service plan, and transmits it to the user through the adapted intelligent data transmission channel.

[0026] As a preferred solution of the digital intelligent service system for enterprise services described in the present invention, wherein: the service performance evaluation module includes a performance data collection unit, a dynamic evaluation unit and a feedback execution unit;

[0027] The performance data collection unit includes, after the final service plan is selected, inserting a tracking identifier into the data of the service plan, and the tracking identifier captures four types of service indicators including timeliness indicator, quality indicator, resource indicator and business indicator.

[0028] Another object of the present invention is to provide a digital intelligent service method for enterprise services. The present invention effectively analyzes user needs through a dual-channel processing method, ensures accurate identification of business tags and efficient matching of services, and ultimately optimizes enterprise service processes.

[0029] As a preferred solution of the digital intelligent service method for enterprise services described in the present invention, it is characterized by including:

[0030] The system processes user demand text in parallel through dual channels, namely BERT-base model semantic analysis + rule engine keyword scanning. After generating candidate business tags, it uses a weighted fusion strategy to dynamically integrate the results. It then goes through three levels of verification, including historical request pattern comparison, service feasibility review, and comprehensive confidence calculation, to confirm the final business type tag and activate the protocol prediction code.

[0031] Based on the protocol prediction code, the protocol signature library is invoked for verification. If verification fails, a sniffing mechanism is activated to identify the protocol type. After parsing, a structured data packet is generated containing the protocol code, service tag, and demand tag. Transmission bandwidth is dynamically allocated based on the urgency of the demand, establishing a hierarchical data transmission channel.

[0032] By dual-screening service nodes based on static and dynamic features, and combining a two-layer strategy of rule matching and similarity calculation, the optimal service solution is automatically selected from the resource pool.

[0033] The implanted tracking tag collects data on four types of indicators, namely timeliness, quality, resources, and business, in real time. The comprehensive score is calculated through dynamic weight allocation and sliding window algorithm. According to the threshold level, early warning, preloading of alternative solutions or service interception is triggered. At the same time, the evaluation results are fed back to the historical pattern library and resource matching module.

[0034] Beneficial effects of the present invention: Through the semantic parsing unit of the user demand analysis module, when the user submits the demand text, the system can capture the user's device type, network environment and historical interaction records in real time, thereby realizing accurate identification and marking of user needs, and improving the accuracy and response speed of demand analysis; through dual-channel analysis, the main channel uses the BERT-base model to generate semantic features and outputs the candidate tags with the highest similarity, and the auxiliary channel uses the rule engine to mark industry keywords, ensuring the breadth and depth of demand analysis, and ultimately improving user satisfaction and the efficiency of demand response.

[0035] In the classification decision unit, by analyzing historical user requests through sliding time window technology, the system can identify labels similar to new requests, further improving the accuracy of service matching and preventing service mismatches caused by label errors.

[0036] The intelligent data adaptation module dynamically selects the appropriate protocol parser based on the results of user demand analysis, and verifies the validity of the protocol through an intelligent loading process, thereby reducing parsing errors and improving the efficiency and accuracy of data transmission.

[0037] The service resource matching module uses pre-screening and precise matching, relying on the static and dynamic characteristics of service resources to effectively filter out non-compliant services, and adjusts resource allocation based on real-time data, thereby optimizing resource utilization efficiency.

[0038] The service performance evaluation module provides real-time service quality feedback through dynamic evaluation of multiple indicators, ensuring efficient service operation and continuous improvement mechanism, and promoting the overall performance improvement of the enterprise service system; forming an efficient, intelligent and flexible enterprise service digital system, greatly improving service capabilities and customer experience, and achieving significant economic benefits and social value. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for describing the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. Those skilled in the art can also derive other drawings based on these drawings without inventive work, among which:

[0040] Figure 1 A system solution module diagram of a digital intelligent service system for enterprise services provided by one embodiment of the present invention.

[0041] Figure 2 An overall flow chart of a digital intelligent service method for enterprise services provided in one embodiment of the present invention. DETAILED DESCRIPTION

[0042] To make the above-mentioned objects, features, and advantages of the present invention more clearly understood, the following detailed description of the specific embodiments of the present invention is given in conjunction with the accompanying drawings. It is obvious that the described embodiments are only part of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other embodiments obtained by ordinary persons in this field without creative work should fall within the scope of protection of the present invention.

[0043] In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention may also be implemented in other ways different from those described herein. Those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0044] Secondly, the term "one embodiment" or "embodiment" herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in various places throughout this specification does not necessarily refer to the same embodiment, nor does it individually or selectively refer to an embodiment that is mutually exclusive of other embodiments.

[0045] The present invention is described in detail with reference to schematic diagrams. For ease of illustration, cross-sectional views of device structures may be partially enlarged and not to scale when describing embodiments of the present invention. Furthermore, the schematic diagrams are merely illustrative and should not limit the scope of the present invention. Furthermore, in actual production, the three-dimensional dimensions of length, width, and depth should be included.

[0046] Furthermore, in the description of the present invention, it should be noted that the terms "upper, lower, inner, and outer" and other references to orientations or positional relationships are based on the orientations or positional relationships shown in the accompanying drawings and are intended solely to facilitate and simplify the description of the present invention. They are not intended to indicate or imply that the devices or components referred to must have, be constructed, or operate in a specific orientation, and therefore should not be construed as limiting the present invention. Furthermore, the terms "first, second, or third" are used for descriptive purposes only and should not be construed as indicating or implying relative importance.

[0047] In this disclosure, unless otherwise specified or limited, the terms "mounted," "connected," and "connected" should be interpreted broadly. For example, they may refer to fixed, removable, or integral connections. They may also refer to mechanical, electrical, or direct connections, indirect connections through an intermediary, or internal communication between two components. Those skilled in the art will understand the specific meanings of these terms in this disclosure.

[0048] Example 1, with reference to Figure 1 , which is the first embodiment of the present invention, provides a digital intelligent service system for enterprise services, including a user demand analysis module, an intelligent data adaptation module, a service resource matching module, a service effectiveness evaluation module and a user-friendly interface;

[0049] Specifically, the user demand analysis module includes a semantic parsing unit and a classification decision unit.

[0050] It should be noted that the semantic parsing unit is that when a user submits a demand text through the Web interface, the system captures the user's device type, network environment, and historical interaction records in real time;

[0051] At the same time, the system starts parallel dual channels, including the main channel and the auxiliary channel;

[0052] The main channel extracts text embedding vectors through the BERT-base model to generate 128-dimensional semantic features. During fine-tuning, the cosine annealing algorithm is used to control the learning rate. The semantic similarity between text features and the industry knowledge base is calculated, and the feature with the highest similarity is output as the candidate label, along with the probability value of the candidate label.

[0053] For example, when a user submits the requirement text "need real-time financial data visualization report", the main channel extracts semantic features through the BERT-base model and may output candidate tags and probability values:

[0054] Financial Analysis (probability 0.82), Data Visualization (probability 0.75), Real-time Processing (probability 0.68); the probability value is the confidence calculated by the BERT model, indicating the degree of match between the label and the user's needs, and the value range is [0, 1]. For example, the probability value of 0.82 for the "Financial Analysis" label indicates that the model believes the label is highly relevant to the user's needs.

[0055] The auxiliary channel uses the rule engine to scan the submitted demand text to see if it contains any of the 50 preset industry keywords. If so, it is marked as a match, the matching degree between the industry keyword and the demand text is calculated, and the top two keywords with the highest matching degree are output to generate a business tag.

[0056] Among them, the preset 50 industry keywords need to meet the requirements of input or query frequency greater than 1,000 times per month and industry coverage greater than 70%;

[0057] The dual-channel results are integrated through a weighted fusion. When the output probability of the BERT-base model is greater than 0.85, the main channel result is used.

[0058] When the output probability of the BERT-base model does not exceed 0.85, determine whether the main channel label and the auxiliary channel label have the same label. If the main channel label and the auxiliary channel label have the same label, it is considered that the main and auxiliary channel results can choose to use the common label;

[0059] When the first two labels of the primary and secondary channels do not have the same labels, context retrieval is triggered. The processing records of the five most recent similar requests are retrieved from the case library. The spatiotemporal correlation between the case label and the current request is calculated. The candidate classification with the highest spatiotemporal correlation is re-entered into the BERT-base model for secondary verification by the system until the BERT-base model output probability of the candidate classification is greater than 0.85, and the label of the current candidate classification is output;

[0060] If there are no relevant keywords in the preset industry keyword library, the auxiliary channel's rule engine cannot match any industry keywords after scanning. At this time, the auxiliary channel does not generate a business tag. The system will rely entirely on the BERT model output results of the main channel and use the weighted fusion device to determine whether the threshold conditions are met.

[0061] The final adopted service type label is input into the classification decision unit to describe the requirements and activate the protocol decoder.

[0062] By extracting 128-dimensional semantic features from the BERT-base model and calculating semantic similarity across industry knowledge bases, we achieve efficient analysis and precise classification of user needs. Combined with a dual-channel confidence fusion mechanism, this ensures the accuracy and stability of semantic analysis, effectively reducing the misjudgment rate and improving the reliability of service matching.

[0063] Furthermore, the classification decision unit uses sliding time window technology to analyze the user's request records in the past month. The system compares the classification labels in the user's historical requests, identifies classification labels similar to the new request, and calculates the similarity score between the two. When the score reaches the set threshold, the system will consider the user's needs to be consistent with his historical behavior and enter the second level of verification; conversely, a low similarity score indicates that the system label is incorrect and needs to re-enter the semantic parsing unit for label analysis.

[0064] After passing the first level verification, the system enters the second level verification. Based on the results of the first level verification, the system calls the demand description catalog including the business type, urgency, and data volume estimate related to the tag, and searches for the corresponding protocol parser of the relevant service to conduct a service feasibility review.

[0065] Using the feasibility review results obtained from the second level of verification, the comprehensive confidence level of each service is dynamically assessed:

[0066] C=0.6α+0.3β+0.1γ

[0067] Where α is the BERT-base output value, β is the historical matching accuracy of the current user, referring to the average value of the past 90 days; γ is the normalized value of the current system load rate;

[0068] When the comprehensive confidence C is greater than 0.75, the business type code of the current service is selected to activate the corresponding protocol parser first.

[0069] Specifically, an intelligent data adaptation module.

[0070] Furthermore, when the user demand analysis module passes the tag and demand description directory, the intelligent data adaptation module carries the protocol prediction code (such as "FIN_JSON_2023" for the financial JSON protocol). The system initiates the intelligent loading process and calls the protocol feature library to verify the validity of the code, including checking the version number, validity period, and permission identifier.

[0071] When the code verification passes, the corresponding activated parser is dynamically selected according to the current system load to load the parsing template of the corresponding protocol, which includes field mapping rules and verification logic;

[0072] When code verification fails, the sniffing mechanism is activated to confirm the protocol type. Specifically, the sniffing mechanism intercepts the first 512 bytes of the data stream corresponding to the enterprise's tag content for pattern analysis, uses a finite state machine containing 12 state nodes to parse the protocol prediction code, generates a protocol probability distribution diagram (e.g., HTTP 65%, gRPC 25%, custom protocol 10%), and selects the parsed protocol with the highest probability;

[0073] Among them, the sniffing mechanism of the present invention is compared with traditional technologies (such as traditional protocol identification methods, DPI, etc.). Traditional technologies rely on fixed port mapping or predefined regular expression feature libraries, can only identify protocol types defined in the feature library, and require a complete analysis of the first three data packets, which takes an average of a long time. The present invention has overcome the above problems, with the protocol identification coverage increased by 37.6%, resource consumption reduced by 93%, and latency reduced by 81.25%.

[0074] After selecting the parsing protocol, the user's requirements are fully parsed through the parsing protocol, rather than parsing the extracted tags. A finite state machine (FSM) is used to implement lightweight protocol parsing. Multiple requirement tags are generated from the user's input requirement text, a circular buffer is created to store the current parsing context, the data stream is scanned byte by byte for parsing, and structured data packets are output, including protocol codes and requirement tags. A standard handshake request is sent to the enterprise-side data source. If the handshake duration exceeds 300ms, the handshake timeout is determined, and the parsing is deemed to have failed. After three parsing failures, the current handshake is canceled and transferred to the buffer storage area to notify the technician, marking it as an urgent pending state. A real-time voice channel is automatically established between the user and the engineer, supporting two-way information synchronization until the engineer manually selects the protocol type or terminates the service process.

[0075] At the same time, an intelligent data transmission channel is established with the corresponding enterprise end and adapted. According to the urgency corresponding to the label in the demand description directory, hierarchical bandwidth allocation is implemented for data streams. For data streams marked as urgent, 70% bandwidth is guaranteed and the delay is <100ms; for data streams marked as ordinary, 20% to 30% bandwidth is allocated; for unmarked data streams, the default is infrequently used or new data, and the remaining bandwidth is allocated. If there is no remaining broadband, it will be transmitted during idle time.

[0076] Specifically, the service resource matching module includes a resource pre-screening unit and a precise matching unit.

[0077] It should be noted that the resource pre-screening unit starts extracting service resource features after receiving a structured data packet, including service static features and service dynamic features, activates the corresponding service resource pool according to the business type label, and calls the domain classification index of the resource library to screen service resources;

[0078] The static service characteristics include service type, service provider rating, service SLA level, service support agreement, geographical distribution, and service certification qualifications;

[0079] The dynamic characteristics of the service include current load rate, response delay, error rate, and resource consumption baseline;

[0080] Set constraints for static features, specifically: perform basic compliance filtering, filter out service providers with expired certificates and unavailable services restricted by regional policies, and select service nodes that comply with the protocol code;

[0081] Set constraints for dynamic features, specifically: obtain service node status data in real time, downgrade nodes with a load rate exceeding 75%, and temporarily disable nodes with an error rate greater than 5% in the past hour; activate an alternative service pool when burst traffic is detected;

[0082] The node set that has been screened by static and dynamic features is placed into the activated service resource pool and packaged and input into the precise matching unit.

[0083] Furthermore, the precise matching unit includes receiving the packaged content from the resource pre-screening unit and executing a matching strategy, including a rule channel and a similarity channel;

[0084] The rule channel loads the service business rule library specified by the enterprise and assigns weights according to the priority order of the service resources in the service business rule library to calculate the business rule matching score: each service resource is matched with the business rule and a score is generated. In one embodiment of the present invention, a binary score can be used (1 point for compliance with the rule and 0 point for non-compliance);

[0085] Re-prioritize from high to low based on the matching score and input it into the similarity channel; among them, business rules include but are not limited to regional rules (such as only selecting service resources in the local data center), protocol rules (such as forcing service nodes to use the HTTPS protocol), priority rules (such as giving priority to matching high SLA levels), load rules (such as giving priority to matching service nodes with a current load rate below 50%), etc.

[0086] The similarity channel calculates the cosine similarity between the demand label and the service resource and corrects it by the business weight to calculate the final matching degree M:

[0087] M=0.6R+0.4S×|1-γ / 2|

[0088] Among them, R is the rule channel score, S is the similarity score;

[0089] Automatically select the service plan with the highest matching degree, retrieve the data stream of the service plan, and deliver it to the user through the adapted intelligent data transmission channel. When the cosine similarity difference is <0.1, the manual review mechanism is activated to avoid the risks of automated decision-making.

[0090] Specifically, the service effectiveness evaluation module includes an effectiveness data collection unit, a dynamic evaluation unit and a feedback execution unit.

[0091] It should be noted that after the final service plan is selected, the performance data collection unit embeds a tracking identifier into the service plan data. The tracking identifier is generated using the UUIDv4 algorithm, encrypted using SHA-256 (hash value), and then embedded into the data packet header. The tracking identifier captures four types of service indicators, including timeliness indicators, quality indicators, resource indicators, and business indicators.

[0092] The timeliness indicators include response delay and processing time; the quality indicators include result accuracy and error code distribution; the resource indicators include CPU / memory consumption and network bandwidth occupancy; the business indicators include user satisfaction score and service reuse rate.

[0093] It should also be noted that the dynamic evaluation unit, after receiving the four types of raw indicator data from the collection unit, starts the preprocessing pipeline, converts the response delay into a standard score, establishes a weight mapping table based on the error code distribution, calculates the comprehensive consumption index of resource indicators, standardizes customer satisfaction using a 5-point scale, and converts the service reuse rate into a percentage. It dynamically assigns weights to the four types of service indicators, improving the system's service stability by 40% during peak hours, and uses a sliding window algorithm to calculate the comprehensive score.

[0094] When the comprehensive score is ≥90, it is considered a high-quality service. The feedback execution unit sends a priority increase instruction to the resource matching module, increasing the matching weight coefficient of the current service node by 20% and marking it as "preferred service" in the rule base;

[0095] When the comprehensive score is 80 ≤ < 90, it is considered normal service and the feedback execution unit maintains the original standard processing flow;

[0096] When the comprehensive score is 60 ≤ and < 80, the service is considered to require attention. The feedback execution unit sends a yellow alert to the operation and maintenance personnel. The service resource matching module starts preloading alternative solutions and prepares at least two sets of alternative solutions. The user demand analysis module adds a second confirmation step. At the same time, the confidence threshold is increased by 0.1, and a priority reduction instruction is sent to the service resource matching module.

[0097] When the comprehensive score is less than 60, it is an abnormal service. The feedback execution unit immediately executes service interception and sends a pause instruction to all related modules. The user demand analysis module adapts to stop receiving new requests, sends a suspension instruction to the service resource matching module, and notifies the operation and maintenance to repair it. After the repair, it takes three consecutive tests to test the comprehensive score ≥ the recovery threshold. Only then is the service allowed to be restored. At the same time, an instruction to resume providing services and reduce priority is sent to the service resource matching module.

[0098] Among them, the present invention adopts a 15-minute sliding window in the sliding window algorithm to update the indicator weights in real time. Compared with the existing technology (such as the traditional static evaluation model), it relatively improves the abnormal response service, does not require a fixed time period and fixed weight distribution, and reduces the false alarm rate.

[0099] By embedding tracking tags and a dynamic weighting mechanism, we achieve real-time monitoring and feedback optimization of service performance. The system dynamically adjusts the priority and matching weight of service nodes based on four indicators: timeliness, quality, resources, and business, ensuring the stability and efficiency of service links.

[0100] Example 2, reference Figure 2 , which is the third embodiment of the present invention, provides a digital intelligent service method for enterprise services, including:

[0101] S1: The system processes user demand text in parallel through the dual channels of the user demand analysis module, namely BERT-base model semantic parsing + rule engine keyword scanning. After generating candidate business tags, it uses a weighted fusion strategy to dynamically integrate the results. It then goes through three levels of verification, including historical request pattern comparison, service feasibility review, and comprehensive confidence calculation, to confirm the final business type tag and activate the protocol prediction code to ensure the accuracy and reliability of demand classification.

[0102] S2: The intelligent data adaptation module uses the protocol prediction code to verify the protocol signature library. If verification fails, the sniffing mechanism is activated to identify the protocol type. After parsing, a structured data packet is generated, containing the protocol code, service tag, and demand tag. Transmission bandwidth is dynamically allocated based on the urgency of the demand, and a hierarchical data transmission channel is established, achieving intelligent adaptation of protocols and resources.

[0103] S3: The service resource matching module uses both static features (service type / geographic location) and dynamic features (real-time load / network latency) to double-screen service nodes. A two-tier strategy, combining rule matching (business rule priority scoring) and similarity calculation (cosine similarity + business weight correction), automatically selects the optimal service solution from the resource pool. A manual review mechanism is initiated when the cosine similarity difference is <0.1, improving matching efficiency and accuracy.

[0104] S4: In the service performance evaluation module, tracking tags are implanted to collect data on four types of indicators: timeliness, quality, resources, and business in real time. A comprehensive score is calculated through dynamic weight allocation and sliding window algorithm. Early warnings, preloading of alternative solutions, or service interception are triggered according to threshold levels. At the same time, the evaluation results are fed back to the historical pattern library and service resource matching module to form a continuously optimized closed-loop system, which is displayed and pushed in a user-friendly interface.

[0105] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A digital intelligent service system for enterprise services, characterized by: It includes user demand analysis module, intelligent data adaptation module, service resource matching module, service efficiency evaluation module and user-friendly interface; The user demand analysis module includes a semantic parsing unit and a classification decision unit; The semantic parsing unit uses a BERT-based model and a rules engine for dual-channel parallel processing, generating business tags through weighted fusion. When a user submits a requirement text through the web interface, the system captures the user's device type, network environment, and temporary interaction records in real time, and simultaneously initiates two parallel channels. The primary channel uses the BERT-based model to generate semantic features and outputs the candidate tag with the highest similarity. The secondary channel uses the rules engine to tag industry keywords. The classification decision unit confirms the validity of the tag through three-level verification and activates the protocol parser. The intelligent data adaptation module includes, when the user demand analysis module passes the label and demand description directory, carrying the protocol prediction code, the system starts the intelligent loading process, and calls the protocol feature library to verify the validity of the code; When the code verification passes, the corresponding activated parser is dynamically selected according to the current system load to load the parsing template of the corresponding protocol, which includes field mapping rules and verification logic; when the code verification fails, the sniffing mechanism is activated to confirm the protocol type; After selecting the parsing protocol, the user's requirements are fully parsed through the parsing protocol, multiple requirement tags are generated from the user's input requirement text, and structured data packets are output. At the same time, an intelligent data transmission channel is established with the corresponding enterprise end and adapted, and data flow hierarchical bandwidth allocation is implemented according to the urgency corresponding to the tags in the requirement description directory; The service resource matching module includes a resource pre-screening unit and a precise matching unit. The resource pre-screening unit extracts static and dynamic features and sets constraints to screen out service nodes that meet the protocol code. The set of nodes that have passed the static and dynamic feature screening is placed into the activated service resource pool and packaged and input into the precise matching unit. The precise matching unit uses a dual-channel decision-making mechanism to match service resources. The service performance evaluation module includes a performance data collection unit, a dynamic evaluation unit and a feedback execution unit; The performance data collection unit includes, after selecting a final service plan, inserting a tracking identifier into the service plan data, the tracking identifier capturing four types of service indicators; the four types of service indicators include timeliness indicators, quality indicators, resource indicators, and business indicators. The dynamic evaluation unit includes, after receiving the four types of raw indicator data from the collection unit, starting the performance data preprocessing pipeline, dynamically assigning weights to the four types of service indicators, and calculating a comprehensive score using a sliding window algorithm; The feedback execution unit performs corresponding processing according to the comprehensive score; The user-friendly interface allows users to input requirements through a web interface and presents the requirements parsing progress and resource matching path through the web interface.

2. A digital intelligent service system for enterprise services according to claim 1, characterized in that: The main channel uses the BERT-base model to extract text embedding vectors, generate multi-dimensional semantic features, and output candidate tags and probability values ​​based on the semantic features. The auxiliary channel uses the rule engine to scan the submitted demand text to see if there are x preset industry keywords. If there are industry keywords, they are marked as matches and business tags are generated. The dual-channel results are integrated through a weighted fusion device. When the output probability of the BERT-base model is greater than the probability threshold, the main channel result is adopted. When the output probability of the BERT-base does not exceed the probability threshold, it is determined whether the main channel label and the auxiliary channel label have the same label, and the service label to be adopted is determined. The final service type label is input into the classification decision unit for demand description and activation of the protocol decoder.

3. A digital intelligent service system for enterprise services according to claim 2, characterized in that: The classification decision unit includes, after generating a service type tag, inputting the tag into the system to perform three-level verification. The first level of verification is to analyze the user's request records in the past month through a sliding time window and compare the user's historical request pattern; The second level of verification is to call the requirement description catalog to conduct a service feasibility review; The third level of verification is to calculate the comprehensive confidence. When the comprehensive confidence is greater than the set threshold, the business type code of the current service is selected to activate the corresponding protocol parser first, and the protocol prediction code is sent at the same time.

4. The digital intelligent service system for enterprise services according to claim 1, characterized in that: The resource pre-screening unit also includes, after receiving the structured data packet, starting the service resource feature extraction, including service static features and service dynamic features, activating the corresponding service resource pool according to the business type label, and calling the domain classification index of the resource library to perform service resource screening.

5. A digital intelligent service system for enterprise services according to claim 4, characterized in that: The precise matching unit includes receiving the packaged content from the resource pre-screening unit and executing the matching strategy, including the rule channel and the similarity channel; The rule channel loads the service business rule library specified by the enterprise, calculates the business rule matching scores in sequence according to the priority order of the service resources in the service business rule library, re-prioritizes them from high to low according to the matching scores, and inputs them into the similarity channel. The similarity channel calculates the cosine similarity between the demand label and the service resource, and calculates the final matching degree through business weight correction, automatically selects the service plan with the highest matching degree, retrieves the data stream of the service plan, and transmits it to the user through the adapted intelligent data transmission channel.

6. A digital intelligent service system for enterprise services according to claim 1, characterized in that: When the comprehensive score is ≥ the first adjustment threshold, it is considered a high-quality service, and the feedback execution unit sends a priority improvement instruction to the resource matching module; When the second adjustment threshold ≤ comprehensive score < first adjustment threshold, it is considered normal service and the feedback execution unit maintains the original standard processing flow; When the third adjustment threshold is less than or equal to the comprehensive score and less than the second adjustment threshold, the service is classified as requiring attention. A yellow alert is sent to the operation and maintenance personnel. The service resource matching module starts preloading alternative solutions. The user demand analysis module adds a second confirmation step and sends a priority reduction instruction to the service resource matching module. When the comprehensive score is less than the third adjustment threshold, it is an abnormal service. The feedback execution unit immediately executes service interception and sends a pause instruction to all related modules. The user demand analysis module adapts to stop receiving new requests and sends a pause service instruction to the service resource matching module, notifying the operation and maintenance to conduct inspection and repair.

Images