Intelligent service method and computer equipment applied to smart community

Through adaptive decomposition algorithm and unsupervised learning algorithm, a service topic association network diagram is built, and a knowledge base and mapping library is generated, which solves the information asymmetry and service quality problems in community services, and realizes the accuracy of service requests and resources. Matching and efficient matching improve the overall quality and efficiency of community services.

CN118552205BActive Publication Date: 2024-11-22SHENZHEN RONGXINCHENG TECH CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202410947177.6
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-07-16
Publication Date
2024-11-22
Estimated Expiration
2044-07-16

AI Technical Summary

Technical Problem

The existing community service model has problems such as resource dispersion, information asymmetry, and uneven service quality, which is difficult to meet the growing service requests of residents. The service provider lacks effective demand collection and analysis methods, and the service requesting party lacks convenient services. Query and matching channels.

Method used

Adaptive decomposition algorithm is used to process original data, calculate the service impact entropy of service information components, extract key information components, perform data cleaning and formatting processing, use unsupervised learning algorithm to build service topic association network diagrams, generate service Q&A knowledge base and Topic - keyword mapping library, calculate semantic matching degree, if high, enter the service orchestration process, and if low, automatic question and answer.

Benefits of technology

It improves the data processing efficiency and accuracy of community services, realizes accurate matching of service requests and resources, enhances semantic understanding and intelligent matching capabilities, and improves service quality and utilization.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118552205B_ABST
    Figure CN118552205B_ABST
Patent Text Reader

Abstract

The present application relates to an intelligent service method and computer equipment applied to a smart community. The method comprises: using an adaptive decomposition algorithm to process raw data to obtain a service information component set; calculating the service impact entropy of each service information component in the service information component set to extract key information components from the service information component set; performing data cleaning and formatting on the key information components to obtain a structured target data set; using an unsupervised learning algorithm to process the target data set to construct a service theme association network diagram; constructing a standardized service question-answer pair knowledge base based on the service theme association network diagram, and generating a service theme-keyword mapping library; calculating the semantic matching degree between community service resource data and resident service request data; if the semantic matching degree is higher than a preset matching degree threshold, entering the service orchestration process. Improves the efficiency and accuracy of raw data processing of community services. Enhances the semantic understanding and intelligent matching capabilities of community services.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present application relates to the field of intelligent service technology, and in particular to an intelligent service method and computer equipment applied to a smart community. Background Art

[0002] With the rapid development of social economy and the continuous improvement of people's living standards, the demand of community residents for life services is growing, showing the characteristics of personalization, diversification and refinement. However, current community services still face many problems such as scattered resources, information asymmetry, uneven service quality, etc., which makes it difficult to meet the growing service requests of residents.

[0003] The traditional community service model mainly relies on offline manual services, which have problems such as slow service response, difficult to ensure service quality, and low utilization of service resources. Service providers lack effective means of demand collection and analysis, and it is difficult to accurately grasp residents' service requests. Service requesters lack convenient service query and request channels, and it is difficult to quickly match appropriate service resources. Summary of the invention

[0004] Based on this, it is necessary to provide an intelligent service method and computer equipment for smart communities that is efficient, high-quality, high-utilization and can accurately grasp service requests in response to the above technical problems.

[0005] An intelligent service method applied to a smart community comprises the following steps:

[0006] Collect the original data in the smart community and use the adaptive decomposition algorithm to process the original data to obtain the service information component set; the original data includes residents' service request data and community service resource data;

[0007] Calculating the service impact entropy of each service information component in the service information component set, and extracting key information components from the service information component set according to a preset impact entropy range and the service impact entropy to form a key information component set;

[0008] Perform data cleaning and formatting on the key information component set to obtain a structured target data set;

[0009] An unsupervised learning algorithm is used to process the target data set to obtain a service theme-keyword set, and the association strength between service themes is calculated based on the service theme-keyword set to construct a service theme association network diagram;

[0010] Based on the service theme association network diagram, extract the title and body text under the service theme, build a standardized service question-answer knowledge base, and generate a service theme-keyword mapping library;

[0011] Based on the service question-answer knowledge base and the service topic-keyword mapping library, the semantic matching degree between community service resource data and resident service request data is calculated;

[0012] If the semantic matching degree is higher than the preset matching degree threshold, the service orchestration process will be entered; if the semantic matching degree is lower than the preset matching degree threshold, the automatic question-answering algorithm program will be carried out.

[0013] In one embodiment, if the semantic matching degree is higher than the preset matching degree threshold, the service orchestration process is entered, including the steps of:

[0014] In the service orchestration process, a service orchestration plan is generated, and the service logic consistency and compliance check are performed on the service orchestration plan to verify and optimize the service orchestration plan.

[0015] In one embodiment, in the service orchestration process, after the steps of generating a service orchestration plan and performing service logic consistency verification and compliance check on the service orchestration plan to verify and optimize the service orchestration plan, the steps include:

[0016] Execute the verified and optimized service orchestration plan, and generate service execution logs in real time through data stream adaptive parsing method;

[0017] Based on the service execution log, the execution process of the service orchestration plan is dynamically monitored, and abnormal situations are identified and recorded in real time.

[0018] In one embodiment, after the step of dynamically monitoring the execution process of the service orchestration solution according to the service execution log and identifying and recording abnormal situations in real time, the steps include:

[0019] Analyze abnormal situations using abnormality diagnosis algorithms to determine the causes of abnormalities;

[0020] Retrieve the exception handling strategy library based on the cause of the exception, obtain the corresponding exception handling plan and execute it.

[0021] In one embodiment, if the semantic matching degree is lower than the preset matching degree threshold, after the step of performing the automatic question-answering algorithm procedure, the following steps are included:

[0022] If the semantic match is lower than a preset match threshold after a preset round of interaction based on the automatic question-answering algorithm program, a prompt to request a redescription or to seek other help will be issued.

[0023] In one embodiment, the steps of collecting raw data in the smart community and processing the raw data using an adaptive decomposition algorithm to obtain a service information component set include the following steps:

[0024] After obtaining the raw data, the raw data is initially cleaned and filtered;

[0025] An adaptive decomposition algorithm is used to process the raw data after cleaning and filtering to convert the unstructured and semi-structured data in the raw data into a structured service information component set.

[0026] In one embodiment, the step of calculating the service impact entropy of each service information component in the service information component set, and extracting key information components from the service information component set according to a preset impact entropy range and the service impact entropy to form a key information component set includes the steps of:

[0027] By statistically analyzing the service history data, the probability distribution of the service information components is obtained;

[0028] According to the probability distribution of values, the service impact entropy of each service information component in the service information component set is calculated;

[0029] If the service impact entropy is within a preset impact entropy range, the corresponding service information component is extracted as a key information component to form a key information component set.

[0030] In one embodiment, the step of processing the target data set using an unsupervised learning algorithm, obtaining a service theme-keyword set, and calculating the association strength between service themes based on the service theme-keyword set, and constructing a service theme association network diagram includes the following steps:

[0031] Process the target data set based on the latent Dirichlet allocation model to obtain a service topic-keyword set;

[0032] Based on the service topic-keyword set, the co-occurrence frequency and / or mutual information between the service topics are calculated to obtain the correlation strength between the service topics;

[0033] According to the service theme-keyword set and the association strength, a theme association matrix is ​​constructed to obtain the service theme association network diagram.

[0034] In one embodiment, based on the service theme association network diagram, extracting the title and body text under the service theme, building a standardized service question-answer pair knowledge base, and generating a service theme-keyword mapping library, the steps include:

[0035] Traversing the service topics in the service topic association network graph, extracting the titles and body texts of the service topics, organizing the titles and body texts into question-answer pairs, and organizing the question-answer pairs into a service question-answer pair knowledge base;

[0036] The service topics in the service topic association network diagram are traversed, and K keywords associated with the service topics are extracted from large to small relevance to form a service topic-keyword mapping library.

[0037] A computer device comprises a memory and a processor, wherein the memory stores a computer program and the processor implements the steps of the above method when executing the computer program.

[0038] One of the above technical solutions has the following advantages and beneficial effects:

[0039] The present application is applied to the intelligent service method of the smart community, through the following steps: collecting the original data in the smart community, and using the adaptive decomposition algorithm to process the original data to obtain a service information component set; calculating the service impact entropy of each service information component in the service information component set, and extracting the key information component from the service information component set according to the preset impact entropy range and service impact entropy; performing data cleaning and formatting on the key information components to obtain a structured target data set; using an unsupervised learning algorithm to process the target data set, obtaining a service theme-keyword set, and calculating the association strength between service themes based on the service theme-keyword set, and constructing a service theme association network diagram; based on the service theme association network diagram, extracting the title and body text under the service theme, constructing a standardized service question-answer pair knowledge base, and generating a service theme-keyword mapping library; based on the service question-answer pair knowledge base and the service theme-keyword mapping library, calculating the semantic matching degree between community service resource data and resident service request data; if the semantic matching degree is higher than the preset matching degree threshold, entering the service orchestration process; if the semantic matching degree is lower than the preset matching degree threshold, performing an automatic question-answer algorithm program. The present application method introduces an adaptive decomposition algorithm and service impact entropy, intelligently disassembles the resident service request data, and extracts key information component elements. Data cleaning technology is used to automatically identify and correct data quality issues, provide high-quality data support for intelligent services, and improve the efficiency and accuracy of raw data processing for community services. Unsupervised learning algorithms are used to automatically discover the service theme-keyword set implicit in the raw data and reveal the service theme association structure. Multi-dimensional semantic similarity indicators such as theme similarity and keyword similarity are integrated to achieve accurate matching of service requests and candidate resources, and enhance the semantic understanding and intelligent matching capabilities of community services. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] Figure 1 This is an application environment diagram of the intelligent service method applied to a smart community in an embodiment of the present application.

[0041] Figure 2 Schematic diagram of the process of the intelligent service method applied to the smart community in the embodiment of the present application.

[0042] Figure 3This is a flowchart of the steps of obtaining a service information component set in an embodiment of the present application.

[0043] Figure 4 Schematic diagram of the process of obtaining key information component sets in an embodiment of the present application.

[0044] Figure 5 A flowchart of the steps for obtaining a service theme association network diagram in an embodiment of the present application.

[0045] Figure 6 This is a flowchart of the steps of obtaining a service theme-keyword mapping library in an embodiment of the present application.

[0046] Figure 7 Schematic diagram of the abnormal monitoring process in the embodiment of the present application.

[0047] Figure 8 Schematic diagram of the process of handling exceptions in the embodiment of the present application.

[0048] Fig. 9 Schematic diagram of the internal structure of a computer device in an embodiment of the present application. DETAILED DESCRIPTION

[0049] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0050] The intelligent service method for smart community provided in this application can be applied to Figure 1 In the application environment shown. The terminal 102 communicates with the server 104 through the network. The terminal 102 obtains raw data, the server 104 collects raw data from the terminal 102, and the server 104 runs its own method to process the raw data. The terminal 102 can be, but is not limited to, various personal computers, laptops, smart phones, tablet computers, and portable wearable devices, and the server 104 can be implemented as an independent server or a server cluster consisting of multiple servers.

[0051] In one embodiment, Figure 2 As shown, a smart service method for smart communities is provided. Figure 1 The server in the example is used to illustrate the following steps:

[0052] Step S210, collecting raw data in the smart community, and using an adaptive decomposition algorithm to process the raw data to obtain a service information component set.

[0053] The original data comes from a variety of data such as resident service request data, community service resource data, and resident feedback and evaluation data. Among them, resident service request data includes various service requests submitted by residents through smart community service platforms, mobile applications, call centers and other channels, such as home repairs, community activities, property complaints, etc.; community service resource data includes service resources available in the community, such as information and service scope of repairmen, volunteers, public facilities, etc.; resident feedback and evaluation data includes residents' satisfaction scores and text comments on the services provided. These original data usually exist in unstructured or semi-structured forms such as text, voice, and images. The data volume is large, the format is diverse, and the quality is uneven. Effective preprocessing and conversion are required to provide a high-quality data foundation for subsequent analysis and application. In one example, the original data includes resident service request data and community service resource data.

[0054] In one example, if Figure 3 As shown, the steps of collecting raw data in the smart community and processing the raw data using an adaptive decomposition algorithm to obtain a service information component set include the following steps:

[0055] Step S310, after obtaining the original data, preliminary cleaning and filtering are performed on the original data.

[0056] Step S320: Adopt an adaptive decomposition algorithm to process the cleaned and filtered raw data, so as to convert the unstructured and semi-structured data in the raw data into a structured service information component set.

[0057] Specifically, after obtaining the raw data, the data needs to be cleaned and filtered to remove obvious noise data, such as duplicate records, meaningless characters, and incorrect formats. Then, the unstructured and semi-structured raw data are converted into a structured service information component set using an adaptive decomposition algorithm. Among them, the adaptive decomposition algorithm is an algorithm that can automatically adjust the decomposition granularity and strategy according to the intrinsic characteristics and structure of the data. For example, for a resident service request "My faucet is broken and needs repair", the adaptive decomposition algorithm can split it into multiple keywords such as "faucet", "broken", and "repair", each of which represents a service information component of the request. The core of the adaptive decomposition algorithm is to identify key entities, key attributes, key operations, and other information elements in the request through multiple levels of analysis such as lexical analysis, grammatical analysis, and semantic analysis. At the same time, the adaptive decomposition algorithm can also dynamically adjust the decomposition granularity and method according to factors such as the language style and context of the resident service request. For example, for the colloquial expression "My water pipe burst, please come and fix it", the algorithm can convert it into a more standardized expression "The water pipe burst and needs to be repaired" through semantic understanding. For another example, for some highly professional requests, such as "The compressor of the air conditioner outdoor unit is faulty and needs to be replaced", the resident service request can be split into more fine-grained information components based on the domain knowledge base, such as "air conditioner", "outdoor unit", "compressor", "fault", "replacement", etc. This adaptive decomposition process can be expressed by the following formula:

[0058] ServiceRequest=Decompose(RawData, KnowledgeBase, ContextInfo)

[0059] Among them, ServiceRequest represents the service information component set obtained after decomposition, RawData represents the original data, KnowledgeBase represents the domain knowledge base, ContextInfo represents the context information of the request, and Decompose represents the adaptive decomposition function.

[0060] Through adaptive decomposition, unstructured and semi-structured raw data can be transformed into a set of structured service information components, each of which represents a key information component element of the request.

[0061] Step S220, calculating the service impact entropy of each service information component in the service information component set, and extracting key information components from the service information component set according to a preset impact entropy range and the service impact entropy to form a key information component set.

[0062] After completing the preprocessing of the original data, the service information component set is obtained. The service information component set contains a large number of service information components, such as service subject, service object, service time, service location, etc. However, not all service information components have the same impact on service quality and efficiency. In order to further refine key information components and optimize service processes, it is necessary to identify those key information components that have an impact on the service from the service information component set. To this end, the service impact entropy is used as a metric to quantify the degree of impact of each service information component on the service. The concept of service impact entropy is derived from the concept of entropy in information theory, which represents the uncertainty or importance of a service information component in the service process. Specifically, the service impact entropy H(X) of a service information component X can be calculated using the following formula:

[0063] H(X)=-∑P(x)logP(x)

[0064] Among them, x represents a possible value of the service information component X, and P(x) represents the probability of the value x appearing. It can be seen that the size of the service impact entropy H(X) depends on the value distribution of the service information component X. If the values ​​of the service information component X are concentrated on a few values, then H(X) is small, indicating that the impact of the service information component X on the service is relatively certain and important; conversely, if the values ​​of the service information component X are relatively scattered, then H(X) is large, indicating that the impact of the service information component X on the service is relatively uncertain and minor. For example, in home maintenance services, the values ​​of the service information component "service type" are concentrated on a few types such as "water and electricity maintenance" and "home appliance maintenance", so its service impact entropy is small and it is a key information component; while the values ​​of the information component "service time" are relatively scattered and can be any time of the day, so its service impact entropy is large and it is a minor information.

[0065] In one example, if Figure 4 As shown, the step of calculating the service impact entropy of each service information component in the service information component set, and extracting key information components from the service information component set according to a preset impact entropy range and the service impact entropy to form a key information component set includes the steps of:

[0066] Step S410, obtaining the value probability distribution of the service information component by statistically analyzing the service history data.

[0067] Step S420: Calculate the service impact entropy of each service information component in the service information component set according to the probability distribution of values.

[0068] Step S430: If the service impact entropy is within a preset impact entropy range, the corresponding service information component is extracted as a key information component to form a key information component set.

[0069] In actual calculations, by statistically analyzing the service history data, we can obtain the value probability distribution of each service information component, and then calculate the service impact entropy of each service information component. For example, for the service information component "service type", we can count the frequency of occurrence of various service types in the service history data, and obtain its probability distribution as follows:

[0070] Water and electricity maintenance: 0.4 Home appliance maintenance: 0.3 Door and window maintenance: 0.2 Other maintenance: 0.1

[0071] According to the above formula, the service impact entropy of "service type" can be calculated as:

[0072] H(service type)=-(0.4log0.4+0.3log0.3+0.2log0.2+0.1log0.1)≈1.846

[0073] Similarly, the service impact entropy of other service information components can be calculated. After obtaining the service impact entropy of all information components, it is necessary to set a preset impact entropy range [Hmin, Hmax], and determine the service information components whose service impact entropy falls within the preset impact entropy range as key information components. The selection of the preset impact entropy range needs to comprehensively consider factors such as the characteristics of the service, the quality of the data, and the efficiency of the calculation, and can usually be determined through expert experience or data experiments. For example, the median of the service impact entropy can be selected as the midpoint of the preset impact entropy range, and then the upper and lower limits of the threshold can be adjusted according to actual needs. Assuming that the preset impact entropy range is set to [1.5, 2.0], the service impact entropy of the above-mentioned "service type" information component of 1.846 falls within this range and is therefore determined to be a key information component.

[0074] After determining the key information components, these key information components can be extracted from the service information component set to form a key information component set. This key information component set not only greatly reduces the dimension and complexity of the information, but also retains the core information elements that affect the service quality and efficiency, providing high-quality information input for subsequent service matching, service orchestration and other tasks. For example, in home maintenance services, key information components such as "service type", "faulty equipment", and "fault description" may be extracted from a large amount of information components to form the following key information component set:

[0075] {"Service Type":"Water and Electricity Repair","Faulty Equipment":"Faucet","Fault Description":"Water Leakage"}

[0076] This key information component set represents the core requirements of the service request in a structured form.

[0077] Step S230, performing data cleaning and formatting processing on the key information components to obtain a structured target data set.

[0078] After the key information components are extracted, a key information component set containing the core elements of the service request is obtained. However, these key information components often have some quality problems, such as information redundancy, inconsistent formats, and value errors, which will affect the accuracy of subsequent data analysis and service matching. In order to further improve the data quality, the key information component set needs to be cleaned and formatted. Data cleaning refers to identifying and correcting errors, inconsistencies, duplications, and other problems in the key information component set to make the data more accurate, complete, and consistent. Specifically, a series of data cleaning technologies can be used, such as data deduplication, outlier detection, and missing value filling. For example, for the "service type" information in the key information component set, data deduplication technology can be used to identify and merge service types with the same semantics but different expressions, such as "water and electricity repair" and "water and electricity maintenance". For another example, for the "faulty equipment" information in the key information component set, outlier detection technology can be used to identify and correct some obviously wrong equipment names, such as "faucet" being mistakenly written as "water faucet". Outlier detection can be based on statistical methods, such as z-score, box plot, etc., or on machine learning methods, such as isolation forest, local anomaly factor, etc.

[0079] Another important task of data cleaning is missing value processing. In actual service requests, the information provided by users is often incomplete, and some key information is missing. For example, users may only provide "service type" and "faulty equipment" but not "fault description". For these missing values, a variety of filling strategies can be used, such as mean filling, median filling, mode filling, regression filling, etc. The specific filling strategy needs to be selected based on factors such as the type of missing values, the distribution of data, and the characteristics of the business. For example, for numerical "fault duration" information, the mean or median can be used to fill the missing value; for categorical "fault cause" information, the mode or regression model can be used to fill the missing value. Filling missing values ​​can not only improve the integrity of the data, but also provide more feature information for subsequent data analysis and mining.

[0080] In addition to data cleaning, the key information component set also needs to be formatted to conform to certain data formats and standards for subsequent storage, query and analysis. Formatting mainly includes two aspects: data type conversion and data structure normalization. Data type conversion refers to converting the value of key information into an appropriate data type, such as converting a string type date into a date type, converting a string type value into a value type, etc. This makes it easy to compare, sort, calculate and other operations on the data. Data type conversion can be achieved using regular expressions, parsing functions and other methods. For example, for the key information "fault occurrence time", the corresponding regular expression can be used to extract and convert the date value and convert it into a standard date and time type.

[0081] Data structure normalization refers to converting key information component sets into structured data that conforms to predefined patterns, such as JSON, XML, relational tables, etc. This makes it easy to store, transmit, and query data. Data structure normalization needs to be designed according to specific application scenarios and technology stacks. For example, in RESTful APIs based on HTTP protocol, JSON format is usually used to represent structured data. The key information component set of a home maintenance service can be represented in a specific JSON format, with the key information of the service request represented in the form of a key-value pair, where the key represents the type of information and the value represents the content of the information. This structured data format not only facilitates data exchange and storage, but also allows for efficient data parsing and manipulation using a rich JSON processing library.

[0082] Through data cleaning and formatting, a structured target data set is finally obtained.

[0083] Step S240, using an unsupervised learning algorithm to process the target data set, obtaining a service theme-keyword set, and calculating the association strength between service themes based on the service theme-keyword set to construct a service theme association network diagram.

[0084] In a specific example, if Figure 5 As shown, the steps of using an unsupervised learning algorithm to process a target data set, obtaining a service theme-keyword set, and calculating the association strength between service themes based on the service theme-keyword set, and constructing a service theme association network diagram include the following steps:

[0085] Step S510: Process the target data set based on the latent Dirichlet allocation model to obtain a service topic-keyword set.

[0086] Step S520: based on the service topic-keyword set, calculate the co-occurrence frequency and / or mutual information between the service topics to obtain the association strength between the service topics.

[0087] Step S530, constructing a topic association matrix according to the service topic-keyword set and the association strength to obtain a service topic association network diagram.

[0088] In this example, the unsupervised learning algorithm is the Latent Dirichlet Allocation model. The Latent Dirichlet Allocation model (LDA) is used to implement service topic mining. It assumes that each service request is generated by a mixture of multiple topics, and each topic is composed of multiple keywords. By estimating the conditional probability distribution of service request-topic and topic-keyword, the Latent Dirichlet Allocation model (LDA) can automatically infer the service topic and keyword distribution (service topic-keyword set) in the data set.

[0089] In one example, the Latent Dirichlet Allocation (LDA) model can be described by the following generative process:

[0090] 1. For each service topic k∈{1,…,K}, generate a service topic-keyword distribution φk∼Dir(α) from the Dirichlet distribution α

[0091] 2. For each service request d∈{1,…,D}: a. Generate a service request-topic distribution θd∼Dir(η) from the Dirichlet distribution η b. For each word position n∈{1,…,Nd} in the service request d: i. Generate a topic index zdn∼Multinomial(θd) from the multinomial distribution θd ii. Generate a word wdn∼Multinomial(φzdn) from the multinomial distribution φzdn

[0092] Where K is the number of topics, D is the number of service requests, and Nd is the number of words in service request d. φk is the keyword distribution of the kth topic, θd is the topic distribution of the dth service request, zdn is the topic index of the nth word in the dth service request, and wdn is the nth word in the dth service request. α and η are prior parameters that control the prior shape of the topic-keyword distribution and service request-topic distribution.

[0093] By learning the parameters φk and θd of the latent Dirichlet allocation model LDA model, we can obtain the service topic-keyword distribution and service request-topic distribution in the dataset, and then realize the automatic discovery of service topics and keywords. For example, in the field of home maintenance services, the following topic-keyword distribution may be found:

[0094] Topic 1: (faucet: 0.3, leak: 0.2, valve: 0.1, ...)

[0095] Topic 2: (lamp: 0.4, switch: 0.2, short: 0.1, ...)

[0096] Topic 3: (Air conditioning: 0.3, Refrigeration: 0.2, Compressor: 0.1, ...)

[0097] This shows that there are multiple service topics in the data set, such as "water and electricity maintenance", "lighting maintenance", and "air conditioning maintenance", and each service topic contains some typical fault keywords. At the same time, the topic distribution of each service request can also be obtained, such as:

[0098] Request 1: (topic 1: 0.6, topic 2: 0.3, topic 3: 0.1)

[0099] Request 2: (topic 2: 0.8, topic 1: 0.1, topic 3: 0.1)

[0100] Request 3: (topic 3: 0.7, topic 1: 0.2, topic 2: 0.1)

[0101] This shows that request 1 mainly belongs to the "water and electricity maintenance" topic, request 2 mainly belongs to the "lighting maintenance" topic, and request 3 mainly belongs to the "air conditioning maintenance" topic. Through this soft clustering method, the subject attribution of service requests can be more flexibly and accurately characterized.

[0102] After obtaining the service topics and keyword distribution, we can further calculate the correlation strength between different topics and construct a service topic correlation network diagram. Specifically, we can measure the correlation strength between topics by using indicators such as co-occurrence frequency and mutual information between topics. For example, we can define the co-occurrence frequency Cooc(i, j) of topic i and topic j as:

[0103] Cooc(i,j)=∑dp(i|d)p(j|d)

[0104] Among them, p(i|d) represents the posterior probability that service request d belongs to topic i, and p(j|d) represents the posterior probability that service request d belongs to topic j. The co-occurrence frequency measures the frequency with which two topics appear simultaneously in the same service request, reflecting the correlation between them. Similarly, the mutual information MI(i, j) between topic i and topic j can also be defined as:

[0105] MI(i,j)=∑dp(i,j|d)log(p(i,j|d) / (p(i|d)p(j|d)))

[0106] Where p(i, j|d) represents the posterior probability that service request d belongs to both topic i and topic j. Mutual information measures the difference between the co-occurrence probability of two topics and their independent occurrence probability, reflecting their mutual dependence.

[0107] By calculating the co-occurrence frequency or mutual information between different topics, a topic association matrix can be constructed, and then a service topic association network diagram can be generated. In the service topic association network diagram, the node represents the service topic, the edge represents the association strength between the topics, and the edge weight represents the magnitude of the association strength. Through the visual service topic association network diagram, the association structure and dependency relationship between service topics can be intuitively revealed, and some potential service patterns and rules can be discovered. For example, it may be found that the association between the "water and electricity maintenance" topic and the "lamp maintenance" topic is strong because they both involve circuit problems; the association between the "air conditioning maintenance" topic and the "refrigerator maintenance" topic is strong because they both involve refrigeration problems. These findings can help better understand the knowledge structure of the service field and optimize the classification and organization of services.

[0108] Step S250, based on the service theme association network diagram, extract the title and body text under the service theme, build a standardized service question and answer knowledge base, and generate a service theme-keyword mapping library.

[0109] After completing the service topic mining, a service topic association network diagram is obtained, which reveals the semantic association and dependency relationship between different service topics in the service field. This service topic association network diagram provides a high-level service knowledge structure, but it still lacks specific service content and question and answer information. In order to further enrich and refine service knowledge so that it can be directly applied to the processing and response of service requests, it is necessary to build a structured and standardized service question and answer knowledge base based on the service topic association network diagram.

[0110] In one example, if Figure 6 As shown, based on the service theme association network diagram, the title and body text under the service theme are extracted, a standardized service question-answer pair knowledge base is constructed, and a service theme-keyword mapping library is generated, including the steps of:

[0111] Step S610, traverse the service topics in the service topic association network diagram, extract the titles and body texts of the service topics, organize the titles and body texts into question-answer pairs, and organize the question-answer pairs into a service question-answer pair knowledge base.

[0112] Step S620, traverse the service themes in the service theme association network diagram, extract K keywords associated with the service themes in descending order of relevance, and form a service theme-keyword mapping library.

[0113] Specifically, we can traverse each service topic node in the service topic association network diagram, extract the titles and body texts of all service requests under the service topic, and organize them into a standardized question-answer pair format. For example, for the topic "water and electricity maintenance", we can extract the following service request titles and body texts:

[0114] Title 1: What to do if the faucet leaks? Content 1: Faucet leakage is a common household maintenance problem. Generally, there are several ways to deal with it: 1) Replace the faucet valve core; 2) Replace the faucet seal; 3) Tighten the faucet joint; 4) Clean the faucet outlet. If the above methods cannot solve the leakage problem, it is recommended to contact a professional maintenance personnel for repair.

[0115] Title 2: How to tell if a water pipe is leaking? Content 2: There are several ways to tell if a water pipe is leaking: 1) Observe the wall or floor for water stains or mildew; 2) Observe the water meter for abnormal rotation; 3) Use a stethoscope or a receiver to listen to the water pipe for abnormal sounds; 4) Use a leak detector to detect if the water pipe is leaking. If a water pipe is found to be leaking, close the water valve immediately and contact a professional repairman for repair.

[0116] These titles and texts can be organized into structured service question-answer pairs according to certain formats and specifications, forming the following knowledge base entries:

[0117] <qa> <question> What should I do if the faucet is leaking?< / question> <answer> Faucet leakage is a common household maintenance problem. Generally, there are several ways to deal with it: 1) Replace the faucet valve core; 2) Replace the faucet seal; 3) Tighten the faucet joint; 4) Clean the faucet outlet. If the above methods cannot solve the leakage problem, it is recommended to contact a professional maintenance personnel for repair.< / answer> < / qa> .

[0118] <qa> <question> How to determine if a water pipe is leaking?< / question> <answer> There are several ways to determine whether a water pipe is leaking: 1) Observe whether there are water stains or mildew on the wall or floor; 2) Observe whether the water meter is rotating abnormally; 3) Use a stethoscope or a receiver to listen to whether there are abnormal sounds in the water pipe; 4) Use a leak detector to detect whether the water pipe has leakage current. If a water pipe is found to be leaking, the water valve should be closed immediately and a professional maintenance person should be contacted for repair.< / answer> < / qa> .

[0119] In this way, the knowledge scattered in different service requests can be extracted and organized into a structured and modular service question-answer pair knowledge base. The service question-answer pair knowledge base not only contains rich service questions and answer information, but also has high knowledge density and reusability, and can be directly applied to the matching and response of service requests.

[0120] While building the service question-answer knowledge base, you can also use the service topic association network diagram to generate a service topic-keyword mapping library. Specifically, for each topic node in the network diagram, you can extract its associated Top-K keywords (K keywords with decreasing relevance) to form a service topic-keyword mapping library table. For example, for the topic "water and electricity maintenance", you can extract the following keyword mapping:

[0121] Water and electricity maintenance: {faucet: 0.3, water leakage: 0.2, valve: 0.1, water pipe: 0.1, water meter: 0.05, ...}

[0122] This mapping table reflects the most important and relevant keywords and their weight distribution under each service topic. This mapping table can be used to quickly and accurately classify service requests and extract keywords. For example, for a new service request "the faucet valve is broken, causing water leakage", its keywords "faucet", "valve", and "water leakage" can be extracted first, and then the topic "water and electricity maintenance" can be quickly located by querying the service topic-keyword mapping library table. This request matching method based on topics and keywords can not only improve the efficiency of matching, but also improve the accuracy and coverage of matching.

[0123] In one example, the service theme-keyword mapping table can be represented as a matrix M, where M(i, j) represents the correlation weight between the i-th service theme and the j-th keyword. Given a service request q, the service theme matching based on service theme and keyword can be achieved through the following calculation steps:

[0124] 1. Extract service topics and keywords from service request q to obtain service request vector v_q;

[0125] 2. Calculate the similarity between the service request vector v_q and the keyword vector v_t of each service topic, and obtain the service request-service topic similarity vector s_q: s_q(i)=sim(v_q, v_t(i))=v_q·M(i,:) / (|v_q||M(i,:)|) where v_t(i) represents the keyword vector of the i-th topic, M(i,:) represents the i-th row of the topic-keyword matrix, · represents the vector dot product, and || represents the vector modulus.

[0126] 3. Select the Top-K service topics with the highest similarity as candidate matching service topics for service request q.

[0127] Through this service request-service topic matching method, service requests can be quickly and accurately classified into the most relevant service topics, and then accurate service responses can be achieved using the service question and answer knowledge base.

[0128] Step S260, based on the service question-answer pair knowledge base and the service topic-keyword mapping library, calculate the semantic matching degree between the community service resource data and the resident service request data.

[0129] Based on the semantic matching method, the similarity between the resident service request data and the community service resource data (i.e., semantic matching) is calculated, and the service resource with the highest similarity is selected as the candidate matching result. The semantic matching here not only considers the literal similarity between the service community service resource data and the resident service request data, but also considers their similarity at the semantic level, such as the subject category to which they belong, keyword distribution, etc. Through semantic matching, the limitations of keyword matching can be overcome to achieve more intelligent and flexible service matching.

[0130] In an example, given a service request q and a service resource r, service resource matching based on semantic matching can be achieved through the following calculation steps:

[0131] 1. Perform topic classification and keyword extraction on service request q and service resource r respectively to obtain their topic vectors v_q and v_r, and keyword vectors w_q and w_r. The topic vector represents the distribution of requests or resources in the topic space, and the keyword vector represents the distribution of requests or resources in the keyword space. This step can be achieved using the aforementioned service topic-keyword mapping library.

[0132] 2. Calculate the similarity s_topic(q, r) between service request q and service resource r in the topic space. You can use cosine similarity and other measurement methods: s_topic(q, r) = cos(v_q, v_r) = v_q·v_r / (|v_q||v_r|) where · represents the vector dot product and | | represents the vector modulus. The similarity in the topic space reflects the matching degree of the service request and service resource in terms of semantic topics.

[0133] 3. Calculate the similarity s_keyword(q, r) between service request q and service resource r in the keyword space. You can use Jaccard similarity and other measurement methods: s_keyword(q, r) = |w_q∩w_r| / |w_q∪w_r|, where ∩ represents set intersection, ∪ represents set union, and | | represents set cardinality. The similarity in the keyword space reflects the degree of match between the request and the resource at the keyword level.

[0134] 4. Calculate the comprehensive semantic similarity s(q, r) between the service request q and the service resource r. Fusion methods such as weighted average can be used: s(q, r) = λ*s_topic(q, r) + (1-λ)*s_keyword(q, r) where λ is the weight coefficient of topic similarity, which can be adjusted according to service scenarios and requirements. The comprehensive semantic similarity reflects the overall matching degree (i.e. semantic matching degree) of the service request and service resource at the service topic and keyword levels.

[0135] 5. For a given service request q, calculate its comprehensive semantic similarity with all candidate service resources, and select the Top-K service resources with the highest similarity as the matching results.

[0136] The following is a specific example to illustrate the process of service resource matching. Assume that a service request q is received: "The faucet valve is broken, causing water leakage", and the most relevant service resources need to be matched from the service resource library. First, the request q is subject-classified and keyword-extracted to obtain its subject vector v_q and keyword vector w_q:

[0137] v_q=(0.8, 0.1, 0.1, ...) / / indicates that the request mainly belongs to the topic of "water and electricity maintenance"

[0138] w_q={faucet, valve, leak}

[0139] Then, candidate service resources are selected from the service resource library and their semantic similarity with the request q is calculated. For example, for the service resource r1 "replace the faucet valve", its topic vector v_r1 and keyword vector w_r1 are obtained:

[0140] v_r1=(0.9, 0.05, 0.05, ...) / / indicates that the resource mainly belongs to the topic of "water and electricity maintenance"

[0141] w_r1={faucet, valve, replacement}

[0142] The topic similarity, keyword similarity, and comprehensive semantic similarity of request q and resource r1 can be calculated:

[0143] s_topic(q,r1)=0.8*0.9+0.1*0.05+0.1*0.05=0.73

[0144] s_keyword(q,r1)=2 / 4=0.5

[0145] s(q,r1)=0.6*0.73+0.4*0.5=0.638

[0146] Similarly, the semantic similarity between request q and other candidate resources can be calculated, and the resource with the highest similarity can be selected as the matching result. For example, suppose the following matching results are obtained:

[0147] r1: Replace the faucet valve, s(q, r1)=0.638

[0148] r2: Repair water pipe leakage, s(q, r2)=0.583

[0149] r3: clean the faucet outlet, s(q, r3)=0.472...

[0150] r1 and r2 can be used as the best matching resources for request q and recommended to the service requester. The key here is to use the subject and keyword information of the service resources to perform multi-dimensional and multi-level semantic matching with the request, rather than just simple keyword matching, thus achieving more accurate and intelligent service resource matching.

[0151] Step S270, if the semantic matching degree is higher than the preset matching degree threshold, the service arrangement process is entered; if the semantic matching degree is lower than the preset matching degree threshold, the automatic question-answering algorithm program is performed.

[0152] After completing the service resource matching, the semantic matching degree between the community service resource data and the resident service request data is obtained. This semantic matching degree reflects the matching degree between the service request and the service resource, and is the key basis for determining the next service process. Specifically, a preset matching degree threshold can be set to divide the service requests into two categories: requests with a semantic matching degree higher than the threshold indicate that the request has obtained a satisfactory resource match and can directly enter the subsequent service orchestration process; requests with a semantic matching degree lower than the threshold indicate that the request is not matched enough with the existing service resources, and further clarification or optimization of the request description is needed to improve the matching accuracy. For the second type of requests, further interaction with the requesting user is required, and reply suggestions are given through the automatic question-and-answer algorithm to guide the user to optimize the request description, thereby improving the matching degree between the request and the resource.

[0153] For the first type of request, in one example, if the semantic matching degree is higher than a preset matching degree threshold, the steps of entering the service orchestration process include the following steps: in the service orchestration process, generating a service orchestration plan, and performing service logic consistency verification and compliance checking on the service orchestration plan to verify and optimize the service orchestration plan.

[0154] Specifically, after obtaining a highly matched service request through service request interaction, the request content must first be converted into a set of structured service requests to clarify the service objectives, constraints, priorities, and other information. For example, for the request "the drain pipe of the kitchen sink is blocked, causing the sink to be unable to drain", the following service requests can be extracted:

[0155] Service goal: Unblock the kitchen sink drain pipe

[0156] ·Service object: kitchen sink and its drainage pipes

[0157] Service time: as soon as possible

[0158] Service location: User's home

[0159] Service requirements: On-site service by professional dredging service personnel

[0160] With structured service requests, we can start generating service orchestration solutions. First, we need to retrieve relevant candidate services from the service resource library based on the service request, such as "pipeline dredging", "sewer cleaning", "pipeline detection", etc. Then, we need to analyze the relationships between these candidate services, such as dependency, combination, and mutual exclusion, and build a service relationship graph. In the service relationship graph, each node represents a candidate service, and each edge represents the relationship between services. The type and weight of the edge reflect the nature and strength of the relationship. For example, for the "pipeline dredging" service, it may depend on the "pipeline detection" service, because only when the location and cause of the blockage are determined can the dredging be carried out in a targeted manner; it may be combined with the "sewer cleaning" service to form a complete dredging and maintenance service package; it may be mutually exclusive with the "pipeline modification" service, because modification and dredging are usually two different stages of work.

[0161] On the basis of constructing the service relationship graph, you can use graph search, combinatorial optimization and other algorithms to search for the optimal service combination solution on the service relationship graph. The "optimal" here is a multi-objective concept that needs to comprehensively consider multiple dimensions such as service function, quality, cost, time, etc. The service combination problem can be modeled as a multi-objective optimization problem, and the objective function can be the weighted sum of service utilities:

[0162] max U = ∑ wi * Ui

[0163] Among them, U represents the comprehensive utility of the service combination scheme, Ui represents the utility value of the i-th service, and wi represents the weight of the i-th service, which reflects the contribution of different services to the overall utility. The utility value can be calculated based on the attributes of the service and the user's preferences. For example, for the "pipeline dredging" service, its utility value may be related to factors such as the success rate, speed, and cost of the dredging. The service weight can be set according to the user's preferences and needs. For example, if the user is more concerned about the timeliness of the service, then the weight of "quick response" may be higher.

[0164] When searching for service combination solutions, you also need to consider various constraints, such as service availability, mutual exclusivity, and dependencies. These constraints can be expressed as logical rules or mathematical inequalities, which limit the feasible space of service combination solutions. For example, for the "pipeline dredging" service, there may be the following constraints:

[0165] Availability constraints: The "Pipeline Unclogging" service must be available within the time period specified by the user

[0166] Mutually exclusive constraint: "Pipeline dredging" service and "Pipeline transformation" service cannot appear in the same solution at the same time

[0167] Dependency constraint: If the solution includes the "Pipeline Unclogging" service, the "Pipeline Detection" service must be executed first.

[0168] By solving the multi-objective optimization problem under constraints, we can get a set of optimal service combination solutions, that is, under the premise of satisfying the constraints, there is no other solution that can be better than this solution in all objectives at the same time. Then, according to the user's preferences, we can select a solution that best meets the user's needs from the optimal solution set as the final service orchestration solution.

[0169] After generating a service orchestration plan, the plan needs to be verified and optimized. On the one hand, the logical consistency of the plan needs to be verified to ensure that the dependencies and execution order between services comply with logical rules and that there are no circular dependencies, deadlocks, and other problems. This can be achieved by building a service flow chart and using graph theory algorithms to perform path analysis and reachability analysis. On the other hand, the compliance of the plan needs to be verified to ensure that the combination and execution of services meet the normative requirements of relevant laws, regulations, industry standards, service agreements, and other normative requirements. This can be achieved by formalizing the normative requirements into a set of rules or constraints and using a rule engine or constraint solver for automated verification.

[0170] Through logical consistency verification and compliance checking, we can determine whether there are any problems with the service orchestration solution, and make corresponding corrections and optimizations. For example, if a logical circular dependency is found in the solution, the cycle can be eliminated by adjusting the execution order of the services or introducing intermediate services; if a non-compliant service combination is found in the solution, the compliance requirements can be met by replacing or deleting the relevant services. Through iterative optimization, the quality and reliability of the service orchestration solution can be continuously improved until an optimal solution that meets various constraints and achieves the expected results is obtained.

[0171] Finally, the optimized service orchestration plan is transformed into a specific service process specification, which clearly defines the input, output, execution conditions, exception handling and other details of each service to form a standardized service orchestration template.

[0172] In one example, if Figure 7 As shown, in the service orchestration process, after generating a service orchestration plan and performing service logic consistency verification and compliance check on the service orchestration plan to verify and optimize the service orchestration plan, the steps include:

[0173] Step S710, executing the verified and optimized service orchestration solution, and generating a service execution log in real time through a data stream adaptive parsing method.

[0174] Step S720: dynamically monitor the execution process of the service orchestration solution according to the service execution log, and identify and record abnormal situations in real time.

[0175] For service execution tasks, it is necessary to implement calls and data exchanges between service components according to the process specifications and interface protocols defined in the service orchestration scheme. This usually requires a service orchestration engine that can interpret and execute service orchestration specifications and provide a unified service scheduling and management platform. For example, when executing the "pipeline dredging" service, the service orchestration engine first needs to call the "pipeline detection" service to obtain the location and cause information of the pipeline blockage, and then select the appropriate dredging tools and methods based on this information, and dispatch dredging service personnel to provide door-to-door service. The entire execution process needs to strictly follow the process logic and constraints defined in the service orchestration scheme to ensure the correctness and consistency of service execution.

[0176] While the service is being executed, the execution process needs to be monitored in real time to promptly detect and handle possible abnormal situations. This requires the introduction of data stream adaptive parsing technology to achieve real-time collection, parsing, and analysis of service execution data. Data stream adaptive parsing can dynamically adjust data collection and parsing strategies according to different service scenarios and data types, and select appropriate data collection interfaces, parsing algorithms, and verification rules, thereby achieving efficient, accurate, and comprehensive data collection and parsing.

[0177] Through data stream adaptive parsing, various raw data generated during service execution can be converted into structured and semantic service execution logs. Service execution logs record various links and key indicators of service execution in a standardized format, such as service task ID, execution time, input parameters, output results, execution status, etc. At the same time, they can also include some contextual information, such as service request content, user feedback, environmental parameters, etc., to facilitate subsequent analysis and optimization.

[0178] With structured service execution logs, we can conduct multi-dimensional and fine-grained analysis and mining to achieve comprehensive monitoring and management of the service execution process. On the one hand, we can use technologies such as rule engines and threshold detection to scan service execution logs in real time, automatically identify some predefined abnormal patterns, and generate corresponding alarm information based on the type and severity of the abnormality, triggering corresponding processing procedures such as retry, rollback, and manual intervention. On the other hand, we can also use data statistics, correlation analysis and other technologies to aggregate and analyze service execution logs, mine some key indicators and trends in the service execution process, evaluate the overall quality and effect of service execution, and discover weak links and optimization space in the service process.

[0179] Through service execution and monitoring, we can ensure the reliable execution of service orchestration solutions in the actual environment and improve the efficiency and quality of service delivery. At the same time, we can also continuously optimize and improve service orchestration solutions through analysis and feedback of execution data, forming a closed-loop service optimization mechanism.

[0180] In one example, if Figure 8 As shown, according to the service execution log, the execution process of the service orchestration solution is dynamically monitored, and the steps of identifying and recording abnormal situations in real time include the following steps:

[0181] Step S810, using an abnormality diagnosis algorithm to analyze the abnormal situation and determine the cause of the abnormality.

[0182] Step S820, search the exception handling strategy library according to the exception cause, obtain the corresponding exception handling plan and execute it.

[0183] During the service execution and monitoring process, various abnormal situations that occur during the service execution process are captured and recorded in real time through technologies such as data stream adaptive parsing and anomaly detection. These abnormal situations may arise from various reasons, such as errors or incomplete service requests, failures or unavailability of service components, interference or changes in the external environment, etc. They will cause interruptions or failures in service execution and affect the reliability and continuity of the service. In order to minimize the impact of abnormalities on service quality, it is necessary to introduce a service abnormality analysis and disposal mechanism in the service orchestration process to quickly and accurately diagnose and handle monitored service abnormalities to ensure that the service can run continuously and stably.

[0184] The goal of service anomaly analysis is to infer the root cause of the anomaly based on the anomaly symptoms and related context. This usually requires the use of a pre-defined anomaly cause knowledge base, which summarizes and generalizes the correspondence between common anomaly symptoms and anomaly causes based on historical anomaly cases and expert experience. When an anomaly occurs, the anomaly cause knowledge base can be quickly retrieved to find the anomaly cause that best matches the current anomaly symptom.

[0185] After clarifying the cause of the abnormality, it is necessary to further collect and analyze other data related to the cause, such as service execution logs, user feedback, environmental monitoring data, etc., to verify and refine the initial abnormal cause. This may require service personnel to conduct comprehensive analysis and reasoning on various data based on their professional experience and industry best practices to obtain the final abnormal cause diagnosis results.

[0186] For example, suppose that during the monitoring of the "pipeline dredging" service, it is found that the dredging effect is not ideal, and the water flow rate in the pipeline is still very slow, which is far from the normal level. By querying the abnormal cause knowledge base and analyzing the relevant data, it is preliminarily determined that the causes of the abnormality may be: 1) Improper dredging method selection, failure to effectively clear the pipeline blockage; 2) There are multiple blockages in the pipeline, and some blockages have not been discovered and handled; 3) The pipeline has structural problems such as damage or aging, resulting in water flow obstruction.

[0187] Based on the above possible reasons, the service personnel can further check whether the selection of dredging tools is reasonable, and conduct another comprehensive inspection of the pipeline to check whether there are any missed blockages. If the dredging method and the possibility of missed blockages are ruled out, it is likely that there is a problem with the pipeline itself, and it is necessary to consider replacing or repairing the pipeline.

[0188] After determining the cause of the abnormality, you can enter the abnormality handling stage. It is necessary to match the predefined abnormality handling strategy library according to the diagnosed abnormality cause, and automatically generate the corresponding handling plan. For example, if the abnormality cause is "improper selection of dredging method", the corresponding handling strategy may be: 1) Replace other types of dredging tools, such as high-pressure water jets or electric dredging machines; 2) Contact more experienced dredging experts and ask them to provide technical guidance and solution support; 3) Retrain service personnel in dredging skills to improve the professionalism and accuracy of dredging method selection. If the abnormality cause is "structural problems in the pipeline", then the handling strategy may become: 1) Contact a professional pipeline engineering team to repair or replace the pipeline; 2) If the pipeline problem is serious, it may be necessary to stop the water supply and notify the user; 3) Develop a pipeline modification plan to avoid similar problems from happening again.

[0189] After generating abnormal disposal plans, the system can automatically execute these plans and coordinate the arrangement of relevant service resources and personnel. For example, it can notify dredging experts, apply for new dredging tools, contact pipeline engineering teams, etc. At the same time, the system will continue to monitor the implementation process and effect of the disposal plan, and dynamically optimize the disposal plan according to the actual situation until the problem is completely solved and the normal operation of the service is restored.

[0190] This application method automatically extracts multi-dimensional constraints of service requirements, models service orchestration as a constraint optimization problem, and solves the optimal service process. It introduces intelligent operation and maintenance technologies such as data stream adaptive parsing, rule engine, and abnormal diagnosis algorithm to achieve full-process automation and intelligent management of services, thereby improving the automation and intelligence level of community service processes.

[0191] For the second type of request, the automatic question-answering algorithm is an important branch of artificial intelligence and natural language processing. Its goal is to automatically generate relevant answers or suggestions based on the user's questions or requests. In the scenario of service request interaction, the automatic question-answering algorithm can be used to analyze the content and semantics of the user's request, find out the unclear, incomplete or non-standard parts, and give corresponding optimization suggestions. For example, for the request "My water pipe is leaking", the automatic question-answering algorithm can generate the following reply suggestions:

[0192] "Hello, in order to provide you with better service, could you please provide more information, such as the specific location of the leak, the severity of the leak, when the leak was discovered, etc.? This information can help more accurately match service resources and improve service quality."

[0193] Through such interactive questions and answers, users can be guided to provide more detailed and accurate request descriptions, thereby improving the matching degree between requests and resources.

[0194] In one example, the process of service request interaction can be described as a Markov decision process (MDP). In the Markov decision process (MDP), the state space S represents different description states of the service request, the action space A represents different reply suggestions generated by the automatic question-answering algorithm, the transition function P(s'|s, a) represents the probability of transitioning to state s' after taking action a in state s, and the reward function R(s, a) represents the benefit of taking action a in state s. The goal is to find an optimal question-answering strategy π(a|s) to maximize the cumulative reward. The Markov decision process (MDP) can be solved by methods such as reinforcement learning to obtain the optimal question-answering strategy π(a|s).

[0195] In one example, if the semantic matching degree is lower than the preset matching degree threshold, after the step of performing the automatic question-answering algorithm program, the following steps are included: if the semantic matching degree is lower than the preset matching degree threshold after the preset rounds of interaction based on the automatic question-answering algorithm program, a prompt to re-describe the request or a prompt to seek other help is issued. In practical applications, a preset round K (such as K=3) can be set, indicating that the service request goes through a maximum of K rounds of question-answering interaction. In each round of interaction, the semantic similarity s(q, r) between the request and the resource is calculated and compared with the threshold λ: if s(q, r)≥λ, it means that the request has been satisfactorily matched, the interaction can be terminated and the service orchestration process can be entered; if s(q, r)<λ, it means that the request needs to be further optimized and needs to continue to interact. If after K rounds of interaction, the matching degree between the request and the resource is still lower than the threshold, it means that the request may not be able to get satisfactory service, and the user needs to be prompted to re-describe the request or seek other help. This can avoid excessive invalid interactions and improve service efficiency and user experience.

[0196] For example, suppose we receive a service request q: "The water pipe at home is broken and leaking badly", set the similarity threshold λ=0.6, and the maximum number of interactions K=3. In the first round of interaction, find the service resource r that best matches the request, calculate their semantic similarity s(q, r)=0.55<0.6, and give a reply suggestion:

[0197] "Hello, in order to provide you with better service, in which room and where is the water leak? Can you describe the broken water pipe?"

[0198] The user updates the request description q' according to the suggestion: "The connection between the water pipe and the wall in the master bedroom bathroom is broken and water is spraying out." The similarity between the request q' and the resource r is recalculated s(q', r) = 0.68>0.6, reaching the matching threshold, so the interaction is terminated and the service orchestration process is entered.

[0199] For example, suppose another service request q is received: "The water pipe is blocked". After three rounds of interaction, the request is updated to q': "The kitchen sink's drain pipe is blocked, causing the sink to be unable to drain", but the matching degree with the resource is still low, s(q', r) = 0.58 < 0.6. At this time, the user is prompted:

[0200] "Sorry, your request may require more details, or you may want to try another type of service, such as professional unblocking services. If you have any further questions, please continue to contact us."

[0201] This can help users understand the difficulty of processing the current request, reasonably adjust their expectations, and avoid excessive unnecessary interactions.

[0202] The present application is applied to the intelligent service method of the smart community, through the following steps: collecting the original data in the smart community, and using the adaptive decomposition algorithm to process the original data to obtain a service information component set; calculating the service impact entropy of each service information component in the service information component set, and extracting the key information component from the service information component set according to the preset impact entropy range and service impact entropy; performing data cleaning and formatting on the key information components to obtain a structured target data set; using an unsupervised learning algorithm to process the target data set, obtaining a service theme-keyword set, and calculating the association strength between service themes based on the service theme-keyword set, and constructing a service theme association network diagram; based on the service theme association network diagram, extracting the title and body text under the service theme, constructing a standardized service question-answer pair knowledge base, and generating a service theme-keyword mapping library; based on the service question-answer pair knowledge base and the service theme-keyword mapping library, calculating the semantic matching degree between community service resource data and resident service request data; if the semantic matching degree is higher than the preset matching degree threshold, entering the service orchestration process; if the semantic matching degree is lower than the preset matching degree threshold, performing an automatic question-answer algorithm program. The present application method introduces an adaptive decomposition algorithm and service impact entropy, intelligently disassembles the resident service request data, and extracts key information component elements. Data cleaning technology is used to automatically identify and correct data quality issues, provide high-quality data support for intelligent services, and improve the efficiency and accuracy of raw data processing for community services. Unsupervised learning algorithms are used to automatically discover the service theme-keyword set implicit in the raw data and reveal the service theme association structure. Multi-dimensional semantic similarity indicators such as theme similarity and keyword similarity are integrated to achieve accurate matching of service requests and candidate resources, and enhance the semantic understanding and intelligent matching capabilities of community services.

[0203] It should be understood that although Figure 2-Figure 8 The steps in the flowchart are shown in sequence as indicated by the arrows, but these steps are not necessarily executed in the order indicated by the arrows. Unless otherwise specified in this document, there is no strict order restriction for the execution of these steps, and these steps can be executed in other orders. Moreover, Figure 2-Figure 8 At least part of the steps may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily executed at the same time, but can be executed at different times. The execution order of these sub-steps or stages is not necessarily sequential, but can be executed in turn or alternately with other steps or at least part of the sub-steps or stages of other steps.

[0204] In one embodiment, a computer device is provided. The computer device may be a server, and its internal structure diagram may be as follows: Fig. 9As shown. The computer device includes a processor, a memory, a network interface and a database connected through a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store data such as original data, service information component sets, key information component sets, target data sets, etc. The network interface of the computer device is used to communicate with an external terminal through a network connection. The computer program is executed by the processor to implement an intelligent service method applied to a smart community.

[0205] Those skilled in the art will understand that Fig. 9 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0206] In one embodiment, a computer device is provided, including a memory and a processor, wherein a computer program is stored in the memory, and when the processor executes the computer program, the following steps are implemented:

[0207] Collect the original data in the smart community and use the adaptive decomposition algorithm to process the original data to obtain the service information component set; the original data includes residents' service request data and community service resource data;

[0208] Calculating the service impact entropy of each service information component in the service information component set, and extracting key information components from the service information component set according to a preset impact entropy range and the service impact entropy to form a key information component set;

[0209] Perform data cleaning and formatting on the key information component set to obtain a structured target data set;

[0210] An unsupervised learning algorithm is used to process the target data set to obtain a service theme-keyword set, and the association strength between service themes is calculated based on the service theme-keyword set to construct a service theme association network diagram;

[0211] Based on the service theme association network diagram, extract the title and body text under the service theme, build a standardized service question-answer knowledge base, and generate a service theme-keyword mapping library;

[0212] Based on the service question-answer knowledge base and the service topic-keyword mapping library, the semantic matching degree between community service resource data and resident service request data is calculated;

[0213] If the semantic matching degree is higher than the preset matching degree threshold, the service orchestration process will be entered; if the semantic matching degree is lower than the preset matching degree threshold, the automatic question-answering algorithm program will be carried out.

[0214] In one embodiment, a computer readable storage medium is provided, on which a computer program is stored, and when the computer program is executed by a processor, the following steps are implemented:

[0215] Collect the original data in the smart community and use the adaptive decomposition algorithm to process the original data to obtain the service information component set; the original data includes residents' service request data and community service resource data;

[0216] Calculating the service impact entropy of each service information component in the service information component set, and extracting key information components from the service information component set according to a preset impact entropy range and the service impact entropy to form a key information component set;

[0217] Perform data cleaning and formatting on the key information component set to obtain a structured target data set;

[0218] An unsupervised learning algorithm is used to process the target data set to obtain a service theme-keyword set, and the association strength between service themes is calculated based on the service theme-keyword set to construct a service theme association network diagram;

[0219] Based on the service theme association network diagram, extract the title and body text under the service theme, build a standardized service question-answer knowledge base, and generate a service theme-keyword mapping library;

[0220] Based on the service question-answer knowledge base and the service topic-keyword mapping library, the semantic matching degree between community service resource data and resident service request data is calculated;

[0221] If the semantic matching degree is higher than the preset matching degree threshold, the service orchestration process will be entered; if the semantic matching degree is lower than the preset matching degree threshold, the automatic question-answering algorithm program will be carried out.

[0222] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided in this application may include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. As an illustration and not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).

[0223] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0224] The above-described embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be construed as limiting the scope of the patent application. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the patent application shall be subject to the attached claims.

Claims

1. An intelligent service method applied to a smart community, characterized in that: The following steps are involved: Collecting raw data in the smart community and processing the raw data using an adaptive decomposition algorithm to obtain a service information component set; the raw data includes resident service request data and community service resource data; The resident service request data refers to various service requests submitted by residents through the smart community service platform, mobile applications or call centers; The community service resource data includes information and service scope of maintenance workers, volunteers and public facilities available in the community; Calculating the service impact entropy of each service information component in the service information component set, and extracting key information components from the service information component set according to a preset impact entropy range and the service impact entropy to form a key information component set; the service impact entropy is used to quantify the degree of influence of the service information component on the service; The service information component includes service subject, service object, service time and service location; Performing data cleaning and formatting processing on the key information component set to obtain a structured target data set; An unsupervised learning algorithm is used to process the target data set to obtain a service theme-keyword set, and the association strength between the service themes is calculated based on the service theme-keyword set to construct a service theme association network diagram; the service theme association network diagram is used to reveal the association structure and dependency relationship between the service themes; Based on the service theme association network diagram, extract the title and body text under the service theme, build a standardized service question-answer pair knowledge base, and generate a service theme-keyword mapping library; Calculating the semantic matching degree between the community service resource data and the resident service request data based on the service question-answer pair knowledge base and the service theme-keyword mapping library; The semantic matching degree is the similarity between the resident service request data and the community service resource data; If the semantic matching degree is higher than the preset matching degree threshold, the service orchestration process is entered; If the semantic matching degree is lower than a preset matching degree threshold, an automatic question-answering algorithm program is performed; the automatic question-answering algorithm program is used to give reply suggestions through the automatic question-answering algorithm to guide the user to optimize the request description; The step of processing the target data set with an unsupervised learning algorithm, obtaining a service theme-keyword set, and calculating the association strength between the service themes based on the service theme-keyword set, and constructing a service theme association network diagram includes the following steps: Processing the target data set based on a latent Dirichlet allocation model to obtain a service topic-keyword set; Based on the service theme-keyword set, calculating the co-occurrence frequency and / or mutual information between the service themes to obtain the association strength between the service themes; Constructing a theme association matrix according to the service theme-keyword set and the association strength to obtain the service theme association network diagram; Based on the service theme association network diagram, extracting the title and body text under the service theme, building a standardized service question-answer pair knowledge base, and generating a service theme-keyword mapping library, the steps include: Traversing the service topics in the service topic association network graph, extracting the title and the body text of the service topic, organizing the title and the body text into question-answer pairs, and organizing the question-answer pairs into the service question-answer pair knowledge base; The service themes in the service theme association network diagram are traversed, and K keywords associated with the service themes are extracted in descending order of relevance to form the service theme-keyword mapping library.

2. The intelligent service method applied to a smart community according to claim 1, characterized in that: If the semantic matching degree is higher than the preset matching degree threshold, the service arrangement process is entered, including the following steps: In the service orchestration process, a service orchestration plan is generated, and the service logic consistency verification and compliance check are performed on the service orchestration plan to verify and optimize the service orchestration plan.

3. The intelligent service method applied to a smart community according to claim 2, characterized in that: In the service orchestration process, after the steps of generating a service orchestration plan and performing service logic consistency verification and compliance check on the service orchestration plan to verify and optimize the service orchestration plan, the steps include: Execute the verified and optimized service orchestration scheme, and generate service execution logs in real time through a data stream adaptive parsing method; According to the service execution log, the execution process of the service orchestration solution is dynamically monitored, and abnormal situations are identified and recorded in real time.

4. The intelligent service method applied to a smart community according to claim 3, characterized in that: After the step of dynamically monitoring the execution process of the service orchestration solution according to the service execution log and identifying and recording abnormal situations in real time, the following steps are included: Analyzing the abnormal situation using an abnormality diagnosis algorithm to determine the cause of the abnormality; According to the cause of the exception, the exception handling strategy library is retrieved, and the corresponding exception handling plan is obtained and executed.

5. The intelligent service method applied to a smart community according to claim 1, characterized in that: If the semantic matching degree is lower than the preset matching degree threshold, after the step of performing the automatic question-answering algorithm program, the following steps are included: If the semantic matching degree is lower than the preset matching degree threshold after the preset rounds of interaction based on the automatic question-answering algorithm program, a prompt for a redescription request or a prompt for seeking other help is issued.

6. The intelligent service method applied to a smart community according to any one of claims 1 to 4, characterized in that: The steps of collecting raw data in the smart community and processing the raw data using an adaptive decomposition algorithm to obtain a service information component set include the following steps: After obtaining the raw data, performing preliminary cleaning and filtering on the raw data; An adaptive decomposition algorithm is used to process the cleaned and filtered raw data to convert unstructured and semi-structured data in the raw data into a structured set of service information components.

7. The intelligent service method applied to a smart community according to any one of claims 1 to 4, characterized in that: The step of calculating the service impact entropy of each service information component in the service information component set, and extracting key information components from the service information component set according to a preset impact entropy range and the service impact entropy to form a key information component set comprises the steps of: Obtaining a probability distribution of values ​​of the service information component by statistically analyzing service history data; Calculating the service impact entropy of each service information component in the service information component set according to the value probability distribution; If the service impact entropy is within the preset impact entropy range, the corresponding service information component is extracted as the key information component to form the key information component set.

8. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 7 are implemented.