An office service method, apparatus, device and medium

Abstract

This invention relates to the field of intelligent office, and discloses an office service method, apparatus, device, and medium, comprising: receiving an input office task in natural language form; parsing the office task to obtain the task intent; based on the task intent, searching from a preset service capability index system to obtain a target service matching the task intent; and executing the office task through the target service. This solves the problems of inconvenient and inefficient office service interaction in the prior art.

Description

Technical Field

[0001] This invention relates to the field of intelligent office, and more particularly to an office service method, apparatus, equipment and medium. Background Technology

[0002] With the deep penetration and widespread application of artificial intelligence (AI) technology in enterprise office systems, more and more organizations are introducing intelligent office assistants based on natural language interaction such as voice and text. These assistants help employees initiate office tasks in natural language and complete daily operations such as expense approvals, work hour management, and personnel process handling, aiming to optimize the office experience and improve efficiency through intelligent interaction. However, current AI office systems still have two core technical bottlenecks when handling these natural language office tasks, ultimately leading to inconvenient office service interaction and low overall processing efficiency, making it difficult to meet the core needs of enterprises for efficient and intelligent office work.

[0003] First, existing AI-powered office systems struggle to accurately interpret the core intent of tasks, leading to misjudgments and omissions, which in turn result in mismatched or missed services. This forces employees to repeatedly revise their natural language input to clarify task requirements, significantly reducing the convenience of natural language interaction and delaying task processes due to erroneous responses, thus lowering overall office efficiency.

[0004] Secondly, most current AI office systems manage various office services in a decentralized manner, which means that after a natural language office task is initiated, it is necessary to rely on manual judgment or manual retrieval of service entry points. Service matching is time-consuming and has low accuracy, which directly drags down task execution efficiency and further exacerbates the problem of insufficient office convenience.

[0005] In summary, current AI office systems suffer from core problems such as inconvenient office service interaction and low processing efficiency, making it difficult to adapt to the diverse and efficient office needs of enterprises. Summary of the Invention

[0006] This invention provides an office service method, apparatus, computer equipment, and storage medium to solve the problems of inconvenient interaction and low processing efficiency in existing office services.

[0007] Firstly, it provides an office service method, including: Office tasks that receive input in natural language form; Analyze the office tasks to obtain their task intent; Based on the task intent, the system retrieves the target service that matches the task intent from the preset service capability index system. Perform office tasks through targeted services.

[0008] Optionally, the office task can be parsed to obtain the task intent, including: Semantic parsing algorithms are used to segment and recognize semantic entities in office tasks, extracting the core requirement elements in the office tasks. Match the core requirement elements with the preset intent classification template to determine the corresponding intent type; Collect parameter information related to intent type in office tasks as key parameters, store the intent type and key parameters in a structured manner, and generate task intent.

[0009] Optionally, based on the task intent, a search is performed from a pre-defined service capability index system to obtain the target service matching the task intent, including: Using the intent type in the task intent as the search keyword, candidate services with corresponding service functions are selected from the preset service capability index system; The key parameters in the task intent are compared with the input parameter requirements of the candidate services, and the candidate services with a matching degree reaching a preset threshold are selected as the target services.

[0010] Optionally, the key parameters in the task intent are compared with the input parameter requirements of the candidate services, and candidate services with a matching degree reaching a preset threshold are selected as the target service, including: If there is only one target service, then that target service will perform the office tasks according to the task intent; If there are multiple target services, the task intent is broken down into task parameters corresponding to each target service and sent to the corresponding target service. Each target service then performs the task according to the corresponding task parameters.

[0011] Optionally, based on the task intent, a search is performed from a pre-defined service capability index system to obtain the target service matching the task intent, including: If no target service matching the task intent is found, a service matching failure message will be generated, and the content of the task intent and the search log will be recorded. The task intent and retrieval logs are pushed to the preset office service management terminal for updating the service capability index system.

[0012] Optionally, after performing office tasks, the following may also be included: Send the execution results of office tasks; Collect comprehensive feedback information on the execution results and associate the comprehensive feedback information with the corresponding task intent and target service; Based on the comprehensive feedback information, task intent, and target service after association, adjust the matching rules of the service capability index system.

[0013] Optionally, collect comprehensive feedback information regarding the execution results, including: Explicit feedback information regarding the execution results is collected through a pre-defined visual interactive entry point; The algorithm collects implicit operation data related to the execution results based on a preset behavior collection algorithm; The collected explicit feedback information and implicit operational data are processed to obtain comprehensive feedback information.

[0014] Secondly, an office service device is provided, comprising: The receiving module is used to receive office tasks in natural language input. The parsing module is used to parse office tasks and obtain their task intent. The matching module is used to retrieve the target service that matches the task intent from the preset service capability index system based on the task intent; The execution module is used to perform office tasks through the target service.

[0015] Thirdly, a computer device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the above-described office service method.

[0016] Fourthly, a computer-readable storage medium is provided, which stores a computer program that, when executed by a processor, implements the steps of the above-described office service method.

[0017] In the solutions implemented by the aforementioned office service methods, devices, equipment, and storage media, employees can directly initiate office tasks via everyday voice or text, eliminating the need to memorize complex operation paths, fill out standardized forms, or repeatedly correct input to adapt to system requirements. This significantly lowers the barrier to office interaction and improves convenience from the outset. Through in-depth analysis of natural language office tasks, the core requirements of the task are clearly defined, avoiding service mismatches and missed services due to ambiguous intent. Employees are no longer required to repeatedly supplement their explanations, reducing redundant interaction steps. This improves the reliability of natural language interaction and avoids delays due to erroneous responses, thus enhancing office efficiency. Based on the analyzed task intent, suitable services can be quickly located, replacing the traditional method of manual judgment and retrieval of service entry points, significantly shortening service matching time and improving matching accuracy. This effectively solves the problems of inconvenient and inefficient office service interaction in existing technologies. Attached Figure Description

[0018] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the description of the embodiments of the present invention will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0019] Figure 1 This is a flowchart illustrating an embodiment of an office service method according to the present invention; Figure 2 This is a schematic diagram of an office service device according to an embodiment of the present invention. Detailed Implementation

[0020] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0021] It should be understood that, when used in this specification and the appended claims, the term "comprising" indicates the presence of the described features, integrals, steps, operations, elements and / or components, but does not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components and / or collections thereof.

[0022] It should also be understood that the term “and / or” as used in this specification and the appended claims refers to any combination of one or more of the associated listed items and all possible combinations, and includes such combinations.

[0023] As used in this specification and the appended claims, the term "if" may be interpreted, depending on the context, as "when," "once," or "in response to determination." Similarly, the phrase "if determined" or "if matched to [described condition or event]" may be interpreted, depending on the context, as meaning "once determined," "in response to determination," "once matched to [described condition or event]," or "in response to matched to [described condition or event]."

[0024] Furthermore, in the description of this invention and the appended claims, the terms "first," "second," "third," etc., are used only to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0025] References to "one embodiment" or "some embodiments" as described in this specification mean that one or more embodiments of the invention include a specific feature, structure, or characteristic described in connection with that embodiment. Therefore, the phrases "in one embodiment," "in some embodiments," "in other embodiments," "in still other embodiments," etc., appearing in different parts of this specification do not necessarily refer to the same embodiment, but rather mean "one or more, but not all, embodiments," unless otherwise specifically emphasized. The terms "comprising," "including," "having," and variations thereof mean "including but not limited to," unless otherwise specifically emphasized.

[0026] It should be understood that the sequence number of each step in the following embodiments does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0027] To illustrate the technical solution of the present invention, specific embodiments are described below.

[0028] Please see Figure 1 As shown, Figure 1 A flowchart illustrating an office service method provided in an embodiment of the present invention includes the following steps: S11: Office tasks that receive input in natural language form; S12: Analyze the office task to obtain the task intent; S13: Based on the task intent, retrieve the target service that matches the task intent from the preset service capability index system; S14: Perform the office task through the target service.

[0029] For ease of understanding, the following explains some key terms in this embodiment: Natural language-based office tasks refer to instructions or requests issued by users to the office system in everyday human communication languages ​​(such as Chinese, English, etc.). These tasks typically include the office operations the user wants the system to perform and related information. This type of task aims to provide an intuitive and convenient human-computer interaction method, allowing users to communicate with the system without needing to learn specific command syntax.

[0030] Task intent refers to the structured representation of a user's core needs and objectives obtained after parsing an office task in natural language. This intent typically includes the task type, topic, and key information or parameters required to perform the task. The generation of task intent aims to transform unstructured natural language input into standardized data that the system can understand and process, providing a basis for subsequent service matching and execution.

[0031] A service capability index system refers to a pre-established knowledge base or database used to store and manage various office service functions and related information. This system typically includes information such as service name, function description, input parameter requirements, and output result type, and may be organized through tags, categories, or hierarchical structures. The service capability index system aims to provide the system with a comprehensive service search and matching mechanism, ensuring that the appropriate service can be quickly located based on the user's task intent.

[0032] A target service refers to one or more specific office services that can meet the needs of a user's task, retrieved and identified from the service capability index system based on the user's task intent. Identifying target services is a crucial step in automating office tasks, ensuring that user-issued natural language commands are correctly mapped to actual executable system functions.

[0033] This embodiment provides an office service method, which first receives office tasks in natural language input from the user. For example, the user can type "Book a meeting room for 3 PM tomorrow" in a text input box, or verbally say "Check my expense reimbursement progress this month" through a voice input system. The receiving method can be varied, such as obtaining the user's original instructions through keyboard input, voice recognition, handwriting recognition, etc.

[0034] Upon receiving an office task, the system parses it to determine its intent. In one implementation, the system can use a rule-based parsing method, pre-setting a series of keywords and phrases. When a task contains these keywords, the system identifies it as a specific intent. For example, if words like "reservation" or "meeting room" appear in the task, the system initially parses it as an intent to "reserve a meeting room." In another implementation, the system can use a simple statistical model to analyze word frequency in the task text, using frequently occurring words or phrases as clues for intent recognition. For example, if words like "reimbursement" or "progress" appear frequently in the task, the system might parse it as an intent to "inquire about reimbursement."

[0035] Subsequently, based on the task intent, the system searches a pre-defined service capability index to obtain target services that match the task intent. Specifically, the system can use core words or phrases from the parsed task intent as search criteria to find services in the service capability index whose names or descriptions contain these words. For example, if the task intent is parsed as "meeting room booking," the system will search the index for services named "meeting room booking service" or whose descriptions contain "book a meeting room." In another implementation, the system can pre-define a set of tags for each service and compare the information in the task intent with these tags, selecting services with a high tag matching degree as target services. For example, if the task intent contains the tags "reimbursement" and "query," the system will search for services that possess both of these tags.

[0036] Finally, the system executes the office task through the target service. In one implementation, the system directly passes the parsed task intent to the identified target service, which then processes it according to its internal logic. For example, if the target service is a "meeting room booking service," the system sends the task intent (such as "book a meeting room for tomorrow afternoon at 3 PM") directly to the service, which then attempts to complete the booking. In another implementation, the system can convert the information in the task intent into a specific data format according to the target service's interface requirements, and then call the target service's interface to execute the task. For example, if the target service requires structured JSON data as input, the system will convert the task intent into the corresponding JSON format before making the call.

[0037] The above technical solution is illustrated below with a more specific example: Suppose user A wants to check their payslip for this month. User A enters "I want to check my payslip for this month" into the text input box of the office system.

[0038] First, the system receives an office task in natural language form from user A: "I want to check my payslip for this month." This input is captured by the system and used as raw data for subsequent processing.

[0039] Next, the system analyzes the office task to determine its task intent. The system might identify words like "check," "payslip," and "this month" through simple keyword recognition. Based on these keywords, the system initially determines that user A's intent is "check payslip" and attempts to extract the time parameter "this month." At this point, the task intent might be represented as "Intent: Check payslip, Time: This month."

[0040] Then, based on the task intent, the system searches within a pre-defined service capability index to find the target service matching the task intent. The system uses "payroll query" as the search condition and searches the service capability index. Assume there exists a service in the index called "Payroll Query Service," whose function is described as "providing employee payroll query functionality." The system identifies this "Payroll Query Service" as the target service matching the task intent.

[0041] Finally, the system executes the office task through this target service. The system passes the parsed task intent ("Intent: Query payslip, Time: This Month") to the "Payslip Query Service". After receiving the intent, the "Payslip Query Service" queries user A's payslip data for this month according to its internal logic and returns the query results (e.g., displaying user A's salary details for this month) to user A.

[0042] The above embodiments provide a system that receives natural language input from users, performs preliminary analysis to identify task intent, and then searches for and invokes the corresponding target service in a service capability index system based on that intent to execute the task. Compared to traditional office systems that require users to manually select service entry points or input specific commands, the method in this embodiment simplifies the user operation process through automated intent recognition and service matching. For example, in the example of querying a payslip, the user does not need to know the specific "salary query service" entry point or how to invoke it; they only need to express their needs in natural language. This approach effectively solves the problems of inconvenient interaction and low processing efficiency in existing office services. In some embodiments, this application further proposes to parse office tasks to obtain the task intent of the office tasks, including: using a semantic parsing algorithm to segment and recognize semantic entities in the office tasks, extracting the core requirement elements in the office tasks; matching the core requirement elements with a preset intent classification template to determine the corresponding intent type; collecting parameter information related to the intent type in the office tasks as key parameters, and storing the intent type and key parameters in a structured manner to generate the task intent.

[0043] Semantic parsing algorithms are techniques for converting natural language text into a structured representation that machines can understand. Word segmentation breaks down a continuous text sequence into semantically independent word units. For example, "book a meeting room at 3 PM tomorrow" can be broken down into "book," "tomorrow," "afternoon," "3 PM," and "meeting room." Semantic entity recognition identifies entities with specific meanings in text, such as time, location, people, organizations, and products. For example, in "book a meeting room at 3 PM tomorrow," "3 PM tomorrow" can be identified as a time entity, and "meeting room" as a location entity. Core requirement elements refer to words or phrases that express the user's main intent and key information in an office task. For example, in "book a meeting room for me," "book" is an action, and "meeting room" is an object. Rule-based methods can be used, with predefined dictionaries and grammatical rules, for word segmentation and entity recognition; alternatively, machine learning methods, such as Conditional Random Fields (CRF), Recurrent Neural Networks (RNN), and Transformers, can be used for training and recognition.

[0044] Intent classification templates are predefined, structured patterns representing different categories of office tasks, such as "book a meeting room," "send an email," and "create a document." The matching process aims to compare the core requirement elements extracted from the office task with these templates to identify the overall category of the user's request. Intent types are labels determined after classification, representing the user's core intent, such as "meeting room booking intent" and "email sending intent." Keyword matching or rule matching can be used to compare the extracted core requirement elements with keywords or patterns in the templates; alternatively, text classification models (such as Support Vector Machines (SVM) or deep learning classifiers) can be used for training, taking the core requirement elements as input and outputting the corresponding intent type.

[0045] Key parameters refer to the specific information necessary to complete a task under a specific intent type. For example, for the intent of "booking a meeting room," key parameters may include "time," "location," "participants," and "meeting topic." Collecting this parameter information involves extracting specific values ​​or descriptions related to the identified intent type from the original office task text. Structured storage organizes and saves the intent type and key parameters in a machine-readable and processable format (such as JSON, XML, or database records) to form a complete task intent. Structured storage can use key-value pairs, such as {"intent type":"book meeting room","time":"tomorrow afternoon 3 pm","location":"meeting room A"}.

[0046] The following is a concrete example. When the system receives an office task in natural language format, "Please help me book a meeting tomorrow afternoon at 3 PM in conference room A on the third floor. The topic is a project kickoff meeting, and the participants are Zhang San and Li Si," the semantic parsing algorithm first segments the task, identifying words such as "book," "tomorrow afternoon at 3 PM," "conference room A on the third floor," "project kickoff meeting," "Zhang San," and "Li Si." Next, through semantic entity recognition, "tomorrow afternoon at 3 PM" is identified as a time entity, "conference room A on the third floor" as a location entity, and "Zhang San" and "Li Si" as personnel entities. These identified entities, along with the key verb "book," constitute the core requirement elements. Subsequently, the system matches these core requirement elements with a preset intent classification template. For example, it identifies elements such as "book" and "conference room" as highly consistent with the intent type "conference room booking," thus determining the intent type of this office task as "conference room booking." Based on this, the system will further collect parameter information related to the "meeting room reservation" intent type. For example, "tomorrow afternoon at 3 pm" will be used as the "time" parameter, "meeting room A on the third floor" as the "location" parameter, "project kickoff meeting" as the "topic" parameter, and "Zhang San and Li Si" as the "participants" parameter. Finally, the intent type "meeting room reservation" and these key parameters will be stored in a structured manner to generate a complete task intent, which can be represented as {"intent type":"meeting room reservation","time":"tomorrow afternoon at 3 pm","location":"meeting room A on the third floor","topic":"project kickoff meeting","participants":["Zhang San","Li Si"]}`.

[0047] Through the aforementioned technical solution, this application overcomes the complexity of natural language understanding and achieves accurate parsing of office tasks. By employing word segmentation, semantic entity recognition, and intent classification, the system can accurately extract core requirement elements from unstructured natural language and transform them into structured intent types and key parameters. This not only significantly improves the accuracy and efficiency of task intent recognition and avoids service matching errors caused by fuzzy understanding, but also provides clear and standardized input for subsequent automated service execution, thereby enhancing the intelligence level and user experience of the entire office service process.

[0048] In some embodiments, this application further proposes to retrieve target services that match the task intent from a preset service capability index system based on the task intent. This step includes: using the intent type in the task intent as the search keyword, filtering out candidate services with corresponding service functions from the preset service capability index system; comparing the key parameters in the task intent with the input parameter requirements of the candidate services, and selecting candidate services with a matching degree reaching a preset threshold as target services.

[0049] Specifically, the intent type in a task intent is an abstract summary of the core purpose of a user's request after semantic parsing, such as "meeting room reservation" or "document search." Using this as a search keyword enables preliminary screening of the service capability index system. This screening can be based on database index queries, using the intent type as a query condition to quickly locate all service records that claim to support that intent type; or, in a knowledge graph-based service capability index system, the intent type can be used as a node or attribute, and related service nodes can be found through graph traversal algorithms. The service capability index system is a structured information repository containing detailed descriptions of various available office services, such as service functions, supported intent types, required input parameters and their data types and constraints, and expected output results. Comparing the key parameters in the task intent with the input parameter requirements of candidate services refers to a detailed comparison of the specific data items in the task intent (such as date, time, location, topic, etc.) with the parameters necessary for the candidate service to perform its function. This comparison can include parameter name matching, data type compatibility checks, value range verification, and even the use of semantic similarity algorithms to handle variations in parameter names. A matching degree reaching a preset threshold refers to the degree of conformity between key parameters and input parameter requirements, assessed quantitatively. When this quantitative value reaches or exceeds a preset standard, the candidate service is considered qualified. The matching degree can be a percentage, calculated by considering the number of matching parameters, weighted by importance, etc.; or it can be a Boolean value indicating whether all necessary parameters are fully met. Selecting candidate services with a matching degree reaching the preset threshold as target services means ultimately determining one or more services best suited for performing the current office task from the candidate services that have undergone parameter comparison, providing specific service interfaces for subsequent office task execution.

[0050] Upon receiving the parsed and structured task intent, the proposed solution first uses the intent type within the task intent as the primary search keyword. This allows for initial screening from a pre-defined service capability index system, quickly identifying all candidate services with functionalities matching the intent type. This intent type-based matching effectively narrows down the scope of subsequent refined comparisons, significantly improving retrieval efficiency. Subsequently, for these initially screened candidate services, the solution further compares the key parameters contained in the task intent with the input parameter requirements declared by each candidate service. This step aims to ensure that the selected services not only match functionally but also can handle specific data and contextual information. By calculating the matching degree between the key parameters and the input parameter requirements and comparing it with a pre-defined threshold, the solution can accurately identify services that not only meet functional requirements but also data compatibility requirements, ultimately selecting them as the target services.

[0051] As a specific implementation, suppose a user inputs an office task: "Please reserve meeting room 301 for my departmental meeting next Tuesday from 10:00 AM to 11:00 AM." After the aforementioned parsing steps, the system generates a task intent, where the intent type is "Reserve Meeting Room," and the key parameters include "Date: Next Tuesday," "Time: 10:00 AM to 11:00 AM," "Meeting Room: 301," and "Purpose: Departmental Meeting." At this point, the system uses "Reserve Meeting Room" as a search keyword to filter all candidate services related to meeting room reservations from the service capability index system, such as "Meeting Room Management System API" and "Smart Office Assistant Service." Subsequently, the system compares the key parameters in the task intent (such as "Next Tuesday," "10:00 AM to 11:00 AM," "301," and "Departmental Meeting") with the input parameter requirements of these candidate services. For example, "Meeting Room Management System API" might require "meetingDate," "startTime," "endTime," "roomNumber," and "purpose," while "Smart Office Assistant Service" might require "date," "timeRange," "location," and "eventDescription." The system calculates the matching degree between each candidate service and the task intent. For example, if the parameter names and data types of the "Meeting Room Management System API" are highly consistent with the key parameters in the task intent, its matching degree may reach 95%; while the "Intelligent Office Assistant Service" requires some parameter mapping or conversion, and its matching degree is 80%. If the preset matching degree threshold is 90%, then the "Meeting Room Management System API" will be selected as the target service for performing the office task.

[0052] Through the above technical solution, this application can efficiently and accurately retrieve target services that highly match the structured task intent from a pre-set service capability index system. First, coarse-grained filtering using intent type significantly improves retrieval efficiency; then, fine-grained comparison of key parameters with input parameter requirements ensures service functionality and data compatibility, thereby improving matching accuracy. This effectively solves the problem of quickly locating the most suitable target service, laying the foundation for the smooth execution of subsequent office tasks.

[0053] In some embodiments, this application further proposes to compare the key parameters in the task intent with the input parameter requirements of the candidate services, and select the candidate services with a matching degree reaching a preset threshold as the target services, including: if there is only one target service, the target service performs the office task according to the task intent; if there are multiple target services, the task intent is decomposed into task parameters corresponding to each target service, and sent to the corresponding target services respectively, and each target service performs the office task according to the corresponding task parameters.

[0054] The scenario of "if there is only one target service" refers to the situation where, after searching and filtering from a pre-defined service capability index system based on the task intent, the system identifies only one target service that can fully meet or best match the needs of the current office task. This usually means that the intent of the office task is singular and clear, or that only one service in the service capability index system possesses the required functionality. "Then the target service will execute the office task according to the task intent" means that, in the case of only one target service, that target service will directly receive and process the complete task intent. Execution can be achieved by calling the target service's API interface or by sending structured instructions and data to the target service. The target service will complete the specified office task based on its own functional logic, utilizing the intent type and key parameters in the task intent.

[0055] The scenario of "if there are multiple target services" refers to a situation where, during the retrieval and filtering process, the system discovers multiple target services that can meet or partially meet the needs of the current office task, and these services handle different aspects of the task intent or provide different execution paths. This typically occurs when the task intent is complex, contains multiple sub-intents, or has multiple implementation methods. "Then the task intent is broken down into task parameters corresponding to each target service" means that when there are multiple target services, the original, complete task intent needs to be refined and decomposed. The decomposition process can involve analyzing the intent type and key parameters in the task intent to identify which parts can be handled by which target service. For example, a complex task intent contains two sub-intents: "create document" and "send email." The original task intent needs to be broken down into parameters for the "create document" service and parameters for the "send email" service. The basis for decomposition can be preset rules, service input parameter requirements, or machine learning models. "Distributed separately to the corresponding target service" means that the decomposed task parameters are no longer a single unit, but rather customized instructions and data for each target service. These customized task parameters are sent independently to their respective target services. The distribution method can be through message queues, remote procedure calls (RPC), or direct API calls. "Each target service executes its assigned tasks based on the corresponding task parameters" means that each target service that receives specific task parameters will independently execute its assigned sub-tasks according to the received parameters.

[0056] This application's solution, after retrieving and filtering target services from a pre-defined service capability index system based on task intent, intelligently adjusts the task execution strategy according to the number of search results. When the matching process identifies only one target service, the system directly transmits the complete task intent to that single target service. This target service will independently and completely execute the task based on its functions and the received intent information, thus ensuring the direct and efficient processing of a single, clearly defined task. However, when the matching process identifies multiple target services, the system does not simply select one service but refines the original, overall task intent. Specifically, the system intelligently breaks down the task intent into multiple smaller, more specific task parameters, each corresponding to a specific target service. This breakdown is based on a deep understanding of the original task intent and accurate identification of the capabilities of each target service. Subsequently, these decomposed task parameters are sent separately and independently to their corresponding target services. Each target service only needs to focus on and execute the sub-task indicated by the specific task parameter it receives. This mechanism enables complex and multifaceted office tasks to be effectively decomposed, processed in parallel, or specialized by the most suitable service. It avoids the inefficiency or errors that can result from a single service handling multiple intentions, significantly improving the accuracy, efficiency, and overall processing capacity of the system. In this way, the proposed solution can adaptively handle office tasks of varying complexity, ensuring both precision and efficiency in task execution.

[0057] As a specific implementation, suppose a user inputs an office task in natural language: "Please help me book a meeting room for tomorrow afternoon from 3 PM to 4 PM and notify all attendees." First, the system receives and parses the task to obtain the task intent, which includes the intent type "Meeting Management" and key parameters: "Time: Tomorrow afternoon from 3 PM to 4 PM," "Location: Meeting Room," "Action: Booking," "Target: Attendees," and "Action: Notification." Next, the system searches its service capability index based on this task intent. If the search results show only one target service, "Intelligent Meeting Assistant," capable of handling all sub-tasks of "Meeting Management" (including booking meeting rooms and sending notifications), the system will directly send the complete task intent to the "Intelligent Meeting Assistant" service. This service will then automatically complete the meeting room booking and attendee notification based on the task intent. If the search results show two target services, one named "Meeting Room Booking System" specifically for booking meeting rooms, and the other named "Enterprise Communication Platform" specifically for sending notifications, the system will then break down the original task intent. For the "Meeting Room Reservation System," task parameters including "Time: Tomorrow afternoon, 3 PM to 4 PM," "Location: Meeting Room," and "Action: Reservation" will be generated. For the "Enterprise Communication Platform," task parameters including "Target: Attendees" and "Action: Notification" will be generated. These two separate task parameters will then be sent to the "Meeting Room Reservation System" and the "Enterprise Communication Platform," respectively. The "Meeting Room Reservation System" will independently complete the meeting room reservation, while the "Enterprise Communication Platform" will independently notify the attendees.

[0058] Through the above technical solution, this application can effectively address the various outcomes that may occur during the matching of office task services, significantly improving the flexibility and accuracy of task execution. When only one target service is matched, the system can directly and efficiently utilize that service to complete the task, avoiding unnecessary complexity. When multiple target services are matched, by intelligently decomposing complex task intentions into task parameters corresponding to each target service and issuing them for execution separately, it can not only fully utilize the professional capabilities of different services to achieve parallel or distributed processing of office tasks, but also avoid execution deviations or efficiency bottlenecks that may result from a single service handling multiple intentions. This enables the system to respond more accurately to users' complex needs, improving the success rate of automated office task processing and user experience.

[0059] In some embodiments, this application further proposes that when a target service matching the task intent is obtained by searching from a preset service capability index system based on the task intent, if no target service matching the task intent is found, a service matching failure prompt message is generated, and the content of the task intent and the search log are recorded; the task intent and the search log are pushed to a preset office service management terminal for updating the service capability index system.

[0060] If no target service matching the task intent is found, a service matching failure message is generated, and the content of the task intent and the retrieval log are recorded. This feature describes the system's response when service matching fails. It ensures that users or system administrators can be promptly informed of task execution obstacles and provides a data foundation for subsequent troubleshooting and system optimization. The system can automatically generate a message containing an error code, task ID, and a brief reason for failure, and send it to relevant parties via the user interface, email, or instant messaging tools. Alternatively, the system can generate a structured log entry containing detailed information about the task intent (such as intent type and key parameters), intermediate results during the retrieval process, the reason for failure, and a timestamp, and store it in a database or log file. Pushing the task intent and retrieval log to a pre-defined office service management terminal to update the service capability index system describes how service matching failure information can be used to improve the system. By feeding back failure cases to the management terminal, the system can learn and optimize itself, thereby improving the future service matching success rate. The office service management terminal can be a dedicated backend management system that receives these push messages and allows administrators to manually analyze them, add new service capabilities, or adjust the matching rules for existing services. Alternatively, the office service management terminal can integrate a machine learning module to automatically analyze the task intent and retrieval logs, identify missing or inaccurate parts in the service capability index system, and automatically generate update suggestions or directly adjust the rules.

[0061] As a specific implementation, suppose a user inputs an office task in natural language: "Reserve meeting room number 3 for me at 10:00 AM tomorrow and notify all members of the marketing department to attend." After receiving and parsing the task, the system identifies the task intent as "reserving a meeting room" and "notifying attendees," with key parameters including "10:00 AM tomorrow," "meeting room number 3," and "all members of the marketing department." However, when searching the service capability index system based on this task intent, the system finds that there is no single service in the current service capability index system that can simultaneously satisfy both the intents of "reserving a meeting room" and "notifying all members of the marketing department," or the information for "meeting room number 3" is not accurately included in the existing services, resulting in the failure to find a completely matching target service. At this time, the system will immediately generate a service matching failure message, such as "Sorry, no completely matching service could be found to perform your meeting room reservation and notification task. Please try splitting the task or providing more detailed information." At the same time, the system will record the complete content of the task intent (including intent type and key parameters) and a detailed log of this search for the target service (such as the services attempted to match, the reason for the matching failure, etc.). Subsequently, these recorded task intentions and retrieval logs are automatically pushed to a pre-defined office service management terminal. This management terminal can be a backend system monitored by operations and maintenance personnel, or a machine learning module that can detect functional deficiencies (such as missing combined services) or incomplete data (such as outdated meeting room information) in the service capability index system, update or add new service capabilities, or adjust the matching rules of existing services.

[0062] In some embodiments, this application further proposes that after performing office tasks, the process includes: sending the execution results of the office tasks; collecting comprehensive feedback information on the execution results and associating the comprehensive feedback information with the corresponding task intent and target service; and adjusting the matching rules of the service capability index system based on the associated comprehensive feedback information, task intent and target service.

[0063] Sending the results of office tasks can be achieved by displaying task completion prompts on the user interface, generating and sending reports containing task details, or triggering subsequent automated processes. The purpose is to ensure that users are informed of task progress in a timely manner.

[0064] Collecting comprehensive feedback information on execution results aims to obtain user or system evaluation data on the task execution effectiveness. This comprehensive feedback information can take various forms. For example, it can directly collect explicit user evaluations of the execution results through pre-defined visual interactive entry points, such as satisfaction ratings and text comments; or it can be based on pre-defined behavior collection algorithms to obtain implicit operational data by analyzing user actions after the task execution results are presented, such as whether secondary modifications are made, whether similar tasks are repeatedly submitted, and the duration of user interaction. These explicit feedback and implicit operational data, after processing, together constitute a comprehensive evaluation of the task execution results.

[0065] Linking comprehensive feedback information to the corresponding task intent and target service refers to establishing a logical connection between the collected feedback information and the original task intent that triggered the execution of this task, as well as the target service that actually performs the task. This connection can be achieved by recording the corresponding task intent identifier and service identifier simultaneously when storing feedback information in the database, ensuring that each piece of feedback can be traced back to the specific request and processing procedure.

[0066] Adjusting the matching rules of the service capability index system based on the associated comprehensive feedback information, task intent, and target service refers to optimizing the matching logic between task intents and target services within the service capability index system using this feedback data with contextual information. For example, if a task intent receives frequent negative feedback after being executed through a target service, the system can lower the matching priority between that target service and the task intent, or adjust the parsing rules of the task intent to make it more inclined to match other better-performing services. Conversely, if positive feedback is consistently received, its matching weight can be strengthened. This adjustment may involve updating the parameters of the matching algorithm, modifying the weights of the intent classification model, or directly updating the mapping relationships in the service capability index system.

[0067] The following example illustrates this. Suppose a user inputs "Reserve meeting room B for tomorrow afternoon at 2 PM." The system first parses the task intent as "Reserve meeting room," with the key parameters "Time: Tomorrow afternoon at 2 PM, Location: Meeting Room B." Based on this intent, the system retrieves the "Meeting Room Reservation Service" from its service capability index as the target service and executes it. The execution result is "Meeting Room B successfully reserved." Subsequently, the system sends this result to the user. After receiving the successful reservation notification, the user may click the "Satisfied" button on the interface, or may not make any modifications or cancellations to the meeting room reservation for a period of time. These explicit or implicit feedback messages are collected and associated with the task intent of "Reserve Meeting Room" and the "Meeting Room Reservation Service." If the system consistently observes high user satisfaction after the "Reserve Meeting Room" intent is executed through the "Meeting Room Reservation Service," the system may increase the matching weight between this intent and the service. Conversely, if users frequently cancel or modify their bookings immediately after booking, the system will analyze the reasons and may adjust the matching rules. For example, when matching "book a meeting room" intents, it may prioritize services that provide real-time meeting room occupancy information, or add a confirmation step before matching to avoid unnecessary bookings and cancellations.

[0068] Through the above technical solution, the office service method of this application can transform from static rule matching to dynamic adaptive learning. The system no longer relies solely on preset rules to execute tasks, but can learn from actual user interactions and task execution results, continuously optimizing its service matching strategy. This significantly improves the accuracy of office task execution and user satisfaction, enabling the office service system to become more intelligent and efficient with use, effectively solving the problem that traditional office service methods struggle to continuously optimize and improve service quality when facing complex and ever-changing user needs.

[0069] In some embodiments, this application further proposes collecting comprehensive feedback information on the execution results, including: collecting explicit feedback information on the execution results through a preset visual interactive entry point; collecting implicit operation data on the execution results based on a preset behavior collection algorithm; and processing the collected explicit feedback information and implicit operation data to obtain comprehensive feedback information.

[0070] Explicit feedback refers to user evaluations or opinions on service delivery results expressed through proactive actions. This feedback is collected through a pre-defined visual interactive entry point, which can be an automatically pop-up evaluation window upon service completion. This window includes a star rating, satisfaction options (e.g., "Very Satisfied," "Satisfied," "Neutral," "Dissatisfied," "Very Dissatisfied"), and a text comment box for users to directly input their feedback. Alternatively, this entry point can also be an interactive element within the service history interface allowing users to retrospectively evaluate completed tasks, such as providing "like" or "dislike" buttons, or allowing users to categorize specific service results with tags.

[0071] Implicit operational data refers to behavioral data automatically recorded by the system during user interaction with service results, which is not actively expressed by the user. This data can indirectly reflect user satisfaction with the service execution results or usage habits. Based on a preset behavior collection algorithm, it can monitor user dwell time on the service result display page, mouse click heatmaps, page scrolling behavior, and copying, pasting, or downloading operations on the result content. In addition, the algorithm can also analyze user behavior patterns after receiving the service result, such as whether they immediately close the page, frequently switch to other applications, repeatedly perform similar tasks, or edit or modify the service result.

[0072] Comprehensive feedback information is a more complete and objective feedback result obtained by fusing and analyzing the collected explicit feedback information and implicit operational data. The process may include cleaning and standardizing the explicit feedback data, and then weighting and fusing it with implicit operational data. For example, a weighted average can be taken from user-given star ratings and implicit indicators such as user dwell time on the results page and number of downloads to obtain a comprehensive satisfaction score. Furthermore, natural language processing techniques can be used to perform sentiment analysis on explicit text comments and cross-validate them with implicit operational data (such as negative operational behaviors) to more accurately identify users' true emotional tendencies and needs.

[0073] This application's solution overcomes the limitations of a single feedback format by combining explicit feedback information with implicit operational data, thereby obtaining more comprehensive and accurate integrated feedback information. Specifically, after an office task is completed and the execution result is sent, the system first directly obtains the user's subjective evaluation of the execution result through a preset visual interaction entry point, i.e., explicit feedback information. Simultaneously, based on a preset behavior collection algorithm, the system silently monitors and records the user's interaction behavior with the service result in the background, generating implicit operational data. This implicit operational data provides objective evidence of user behavior, such as the user's level of attention to the result, acceptance level, or subsequent handling methods. Subsequently, the system performs in-depth processing on these two types of feedback data from different sources and of different natures. This processing is not a simple data aggregation, but rather an effective fusion and calibration of the subjectivity of explicit feedback and the objectivity of implicit operations through intelligent algorithms. For example, when the explicit feedback is a neutral evaluation, if the implicit operational data shows that the user frequently modifies or abandons the use of the result, the integrated feedback will tend to have a lower satisfaction level. Conversely, if the user does not provide explicit feedback, but the implicit operational data shows that they have efficiently utilized the service result, the integrated feedback will tend to have a higher satisfaction level. In this way, the system can generate more accurate and reliable comprehensive feedback information, which can more realistically reflect the user's satisfaction with the service execution results and the performance of the service itself.

[0074] The following example illustrates this. Suppose a user requests a task in natural language: "Generate a report on the company's annual financial analysis." After parsing, matching, and executing the task, the system provides the user with a financial analysis report. To collect comprehensive feedback on this result, the system first displays a brief rating window on the report's display interface as a pre-set interactive visual entry point. This window prompts the user to rate the report (e.g., 1-5 stars) and provides a text box for specific comments. The user selects 3 stars and enters in the text box, "The report content is comprehensive, but the data visualization needs improvement." Simultaneously, the system's pre-set behavior tracking algorithm begins recording the user's interaction with the report. The algorithm detects that the user spent approximately 2 minutes on the report page, scrolling multiple times but not clicking the download button, copying, or sharing. Furthermore, after closing the report, the user immediately initiated another task: "Please create charts from the key data in the financial report." The system then processes this explicit feedback (3-star rating and text comment) along with implicit operational data (time spent on the report, failure to download / share, initiation of a new task). The processing module's analysis suggests that although the user gave a moderate rating, their actions of not downloading, not sharing, and immediately initiating a chart creation task strongly indicate that the user is indeed dissatisfied with the report's data visualization effects and requires further processing. Therefore, the system's final overall feedback will focus more on the issue of "poor data visualization effects" and treat it as the key feedback point.

[0075] Through the above technical solution, this application can collect feedback information on the results of office task execution more comprehensively and accurately. By combining explicit feedback information actively expressed by users with implicit operational data automatically monitored by the system, it effectively compensates for the potential bias or inaccuracy of a single form of feedback. Explicit feedback directly reflects the user's subjective intentions, while implicit operational data provides objective behavioral evidence. The two corroborate and complement each other, making the final comprehensive feedback information more authentic and reliable. Based on this high-quality comprehensive feedback information, when the system adjusts the matching rules of the service capability index system according to the above comprehensive feedback information, task intent, and target service, it can achieve more accurate and effective rule optimization. This not only improves the adaptability and intelligence level of the service capability index system but also significantly improves the user experience, enabling subsequent office task execution to better meet the user's actual needs, thereby improving the efficiency and user satisfaction of the entire office service method.

[0076] In one embodiment, this application also provides an office service device. For example... Figure 2 As shown, it includes a receiving module 21, a parsing module 22, a matching module 23, and an execution module 24. Detailed descriptions of each functional module are as follows: The receiving module 21 is used to receive office tasks in natural language input. Parsing module 22 is used to parse office tasks and obtain the task intent of the office tasks; Matching module 23 is used to retrieve the target service that matches the task intent from a preset service capability index system based on the task intent; Execution module 24 is used to perform office tasks through the target service.

[0077] This invention also provides an electronic device, including a memory, a processor, and a program stored in the memory and executable on the processor. When the processor executes the program, it implements the aforementioned office service method; to avoid repetition, this will not be described again here. Alternatively, the electronic device can implement the functions of each module in this embodiment of the office service device; this will also not be described again here.

[0078] This invention also provides a readable storage medium storing a program. When the program is executed by a processor, it implements the aforementioned office service method. To avoid repetition, this will not be described again here. Alternatively, when the program is executed by a processor, it implements the functions of each module in this embodiment of the office service device, which will also not be described again here.

[0079] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included within the protection scope of the present invention.

[0080] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), Rambus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0081] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional units and modules is used as an example. In practical applications, the above functions can be assigned to different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above.

Claims

1. An office service method, characterized in that, include: Office tasks that receive input in natural language form; The office task is analyzed to obtain the task intent of the office task; Based on the task intent, a search is performed from the preset service capability index system to obtain the target service that matches the task intent; The office tasks are performed through the target service.

2. The office service method according to claim 1, characterized in that, The process of parsing the office task to obtain its task intent includes: A semantic parsing algorithm is used to perform word segmentation and semantic entity recognition on the office task to extract the core requirement elements in the office task. The core requirement elements are matched with preset intent classification templates to determine the corresponding intent types; Collect parameter information related to the intent type in the office task as key parameters, store the intent type and key parameters in a structured manner, and generate the task intent.

3. The office service method according to claim 2, characterized in that, The step of retrieving a target service matching the task intent from a preset service capability index system includes: Using the intent type in the task intent as the search keyword, candidate services with corresponding service functions are selected from the preset service capability index system; The key parameters in the task intent are compared with the input parameter requirements of the candidate services, and the candidate services with a matching degree reaching a preset threshold are selected as the target services.

4. The office service method according to claim 3, characterized in that, The step of comparing the key parameters in the task intent with the input parameter requirements of the candidate services and selecting candidate services with a matching degree reaching a preset threshold as the target service includes: If there is only one target service, then that target service will perform the office task according to the task intent; If there are multiple target services, the task intent is broken down into task parameters corresponding to each target service and sent to the corresponding target service respectively. Each target service executes the office task according to the corresponding task parameters.

5. The office service method according to claim 1, characterized in that, The step of retrieving a target service matching the task intent from a preset service capability index system includes: If no target service matching the task intent is found, a service matching failure message is generated, and the content of the task intent and the search log are recorded. The task intent and retrieval logs are pushed to a preset office service management terminal for updating the service capability index system.

6. The office service method according to claim 1, characterized in that, After performing the office task, the process also includes: Send the execution results of the aforementioned office task; Collect comprehensive feedback information on the execution results, and associate the comprehensive feedback information with the corresponding task intent and target service; Based on the comprehensive feedback information, task intent, and target service after association, adjust the matching rules of the service capability index system.

7. The office service method according to claim 6, characterized in that, The collection of comprehensive feedback information regarding the execution result includes: Explicit feedback information regarding the execution results is collected through a preset visual interactive entry point; Implicit operation data related to the execution result is collected based on a preset behavior acquisition algorithm; The collected explicit feedback information and implicit operational data are processed to obtain comprehensive feedback information.

8. An office service device, characterized in that, include: The receiving module is used to receive office tasks in natural language input. The parsing module is used to parse the office task to obtain the task intent of the office task; The matching module is used to retrieve a target service that matches the task intent from a preset service capability index system based on the task intent. The execution module is used to execute the office tasks through the target service.

9. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the office service method as described in any one of claims 1 to 7.

10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by a processor, it implements the steps of the office service method as described in any one of claims 1 to 7.

Images