Request processing method, processing device, equipment, storage medium and program product

Abstract

The invention provides a request processing method, a processing device, equipment, a storage medium and a program product, and relates to the technical field of computers. The request processing method comprises the following steps: in response to an obtained business request of a rich text format, extracting multiple pieces of reference information from the business request based on multiple format types of the rich text format; performing multi-intention identification on the multiple pieces of reference information based on a multi-intention detection model, and identifying a plurality of associated intentions; constructing an event stream conforming to a multi-intention event stream template set based on the identified associated intentions; splitting the service request into a plurality of request units based on the reference information and the event stream, creating a corresponding intention prompt for each intention, and matching the request units with the intentions one by one; and calling the matched processing module based on the intention prompt, and processing the plurality of request units in sequence according to the execution sequence of the event stream to generate reply information of the service request. Through the technical scheme disclosed by the invention, the processing efficiency of complex user requirements can be improved.

Description

Technical Field

[0001] The present disclosure relates to the field of computer technology, and in particular to a request processing method, a request processing device, an electronic device, a computer-readable storage medium, and a computer program product. Background Art

[0002] The automated business system aims to optimize business processes and improve efficiency and quality through digitalization and automation. With the diversification and comprehensive development of business, it is necessary to handle complex user needs of rich texts with various styles and multimedia elements. These needs may involve different business areas or functional modules. However, at present, the automated business system only has the ability to handle a single task, resulting in low efficiency in handling the above-mentioned complex user needs.

[0003] It should be noted that the information disclosed in the above background technology section is only used to enhance the understanding of the background of the present disclosure, and therefore may include information that does not constitute the prior art known to ordinary technicians in the field. Summary of the invention

[0004] The purpose of the present disclosure is to provide a request processing method, a request processing device, an electronic device, a computer-readable storage medium and a computer program product, which at least to some extent overcome the problem of low efficiency of automated business systems in related technologies in processing complex user needs.

[0005] Other features and advantages of the present disclosure will become apparent from the following detailed description, or may be learned in part by the practice of the present disclosure.

[0006] According to one aspect of the present disclosure, a request processing method is provided, comprising: in response to an acquired business request in a rich text format, extracting multiple reference information from the business request based on multiple format types of the rich text format; performing multi-intent recognition on the multiple reference information based on a multi-intent detection model, and identifying multiple associated intents; constructing an event flow that conforms to a multi-intent event flow template set based on the identified multiple associated intents; decomposing the business request into multiple request units based on the reference information and the event flow, and creating a corresponding intent prompt for each of the intents, wherein the multiple request units are matched one by one with the multiple intents; calling a matching processing module based on the intent prompt, and processing the multiple request units in sequence according to the execution order of the event flow, and generating reply information for the business request.

[0007] In one embodiment of the present disclosure, an event flow that conforms to a multi-intent event flow template set is constructed based on the identified multiple associated intents, including: matching the multiple associated intents with the configured multi-intent event flow template; if the multiple associated intents do not match the multi-intent event flow template, triggering a calibration and recognition operation to calibrate the multiple reference information, and re-executing the multi-intent recognition based on the calibrated multiple reference information until the identified multiple associated intents match the multi-intent event flow template; determining the execution order and transfer parameters of the multiple associated intents based on the multi-intent event flow template; and constructing the multiple associated intents into the event flow based on the execution order and the transfer parameters.

[0008] In one embodiment of the present disclosure, the business request is decomposed into multiple request units based on the reference information and the event stream, including: traversing the multiple intents in the event stream; for each of the intents, searching for relevant request information in the reference information to generate the request unit based on the relevant request information and the intent.

[0009] In one embodiment of the present disclosure, a corresponding intent prompt is created for each of the intents, including: parsing the prompt information required to implement the intent; performing scenario adaptation and content integration on the intent prompt based on the business scenario of the business request and the delivery parameters to obtain the intent prompt.

[0010] In one embodiment of the present disclosure, a matching processing module is called based on the intent prompt, and the multiple request units are processed in sequence according to the execution order of the event flow to generate reply information for the business request, including: calling the processing module based on the intent prompt that matches the request unit, so that the multiple request units correspond one by one to the multiple processing modules based on the execution order; for any adjacent first processing module and second processing module, the first processing result output by the first processing module and the request unit corresponding to the second processing module are input into the second processing module for processing, and a second processing result is output, wherein the processing result output by the last processing module is used as the reply information.

[0011] In one embodiment of the present disclosure, for any adjacent first processing module and second processing module, the first processing result output by the first processing module and the request unit corresponding to the second processing module are input into the second processing module for processing, and the second processing result is output, including: integrating the first processing result and the request unit corresponding to the second processing module based on the transfer parameter to obtain integrated request information; inputting the integrated request information into the second processing module to obtain the second processing result.

[0012] In one embodiment of the present disclosure, the first processing result output by the first processing module and the request unit corresponding to the second processing module are input into the second processing module for processing, and the second processing result is output, which also includes: detecting whether the first processing result matches the input format of the second processing module; if not, format modulating the first processing result based on the event stream delivery mechanism to obtain the first processing result that matches the input format.

[0013] In one embodiment of the present disclosure, before responding to the acquired business request in rich text format, it also includes: detecting whether the original request conforms to the rich text format; if it does not conform to the rich text format, generating modification prompt information, prompting to modify the original request to the business request in the rich text format, or querying a rich text template that matches the original request to convert the original request into the business request in the rich text format based on the rich text template.

[0014] In one embodiment of the present disclosure, multiple reference information is extracted from the business request based on multiple format types of the rich text format, including: parsing the multiple format types in the business request, wherein the multiple format types include text format, link format, table format and multimedia format; based on the parsing result, using an information extraction function to extract type data matching the multiple format types from the business request as the multiple reference information.

[0015] In one embodiment of the present disclosure, before performing multi-intent recognition on the multiple reference information based on a multi-intent detection model and identifying multiple intents, it includes: annotating the collected historical rich text data with multiple associated intents based on different format types to generate intent annotated texts of the different format types; performing format unification processing on the intent annotated texts of the different format types to obtain target format texts, wherein the target format texts include matching intent labels, original format labels, and associated labels with other intents; and fine-tuning a large language model based on the target format text to obtain the multi-intent detection model.

[0016] In one embodiment of the present disclosure, multi-intent recognition is performed on the multiple reference information based on a multi-intent detection model to identify multiple associated intents, including: extracting intent features from the reference information to obtain an intent feature vector; inputting the intent feature vector into the multi-intent detection model to output the identified intent label and the association label between other intents to obtain the multiple associated intents.

[0017] According to another aspect of the present disclosure, a request processing device is provided, including: an extraction module, for extracting multiple reference information from the business request based on multiple format types of the rich text format in response to an acquired business request in a rich text format; an identification module, for performing multi-intent identification on the multiple reference information based on a multi-intent detection model, and identifying multiple associated intents; a construction module, for constructing an event flow that conforms to a multi-intent event flow template set based on the identified multiple associated intents; a disassembly module, for disassembling the business request into multiple request units based on the reference information and the event flow, and creating a corresponding intent prompt for each of the intents, wherein the multiple request units are matched one by one with the multiple intents; a processing module, for calling a matching processing module based on the intent prompt, and processing the multiple request units in sequence according to the execution order of the event flow, and generating reply information for the business request.

[0018] According to another aspect of the present disclosure, an electronic device is provided, including: a processor; and a memory for storing executable instructions of the processor; the processor is configured to execute the above-mentioned request processing method by executing the executable instructions.

[0019] According to another aspect of the present disclosure, 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 above-mentioned request processing method is implemented.

[0020] According to another aspect of the present disclosure, a computer program product is provided, on which a computer program is stored, and when the computer program is executed by a processor, the above-mentioned request processing method is implemented.

[0021] The request processing solution provided by the embodiments of the present disclosure extracts effective reference information from rich texts of various formats, identifies multiple related intents with the help of a multi-intent detection model, builds an event stream that conforms to a template set based on the intent, plans an orderly process for business processing, further decomposes the business request into multiple request units that match the intents one by one, creates accurate intent prompts for each intent, calls the matching processing module based on the intent prompts, processes the request units in sequence according to the event stream order, and generates reply information. It can deeply understand the user's complex needs and realize the processing of tasks with different intentions, thereby enhancing the adaptability and response capabilities to complex business scenarios, which is conducive to improving the processing efficiency of complex user needs.

[0022] It is to be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the present disclosure. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] The accompanying drawings herein are incorporated into the specification and constitute a part of the specification, illustrate embodiments consistent with the present disclosure, and together with the specification are used to explain the principles of the present disclosure. Obviously, the accompanying drawings described below are only some embodiments of the present disclosure, and for ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without creative work.

[0024] Figure 1 A schematic diagram showing a request processing system in an embodiment of the present disclosure;

[0025] Figure 2 A schematic flow chart showing a request processing method in an embodiment of the present disclosure;

[0026] Figure 3 A schematic flow chart showing another request processing method in an embodiment of the present disclosure;

[0027] Figure 4 A schematic flow chart showing another request processing method in an embodiment of the present disclosure;

[0028] Figure 5 A schematic diagram showing another request processing system in an embodiment of the present disclosure;

[0029] Figure 6 A schematic flow chart showing another request processing method in an embodiment of the present disclosure;

[0030] Figure 7 A schematic flow chart showing another request processing method in an embodiment of the present disclosure;

[0031] Figure 8 A schematic diagram of a request processing device in an embodiment of the present disclosure is shown;

[0032] Fig. 9 A structural block diagram of an electronic device in an embodiment of the present disclosure is shown. DETAILED DESCRIPTION

[0033] Example embodiments will now be described more fully with reference to the accompanying drawings. However, example embodiments can be implemented in a variety of forms and should not be construed as limited to the examples set forth herein; rather, these embodiments are provided so that the disclosure will be more comprehensive and complete and to fully convey the concepts of the example embodiments to those skilled in the art. The described features, structures, or characteristics may be combined in any suitable manner in one or more embodiments.

[0034] In addition, the accompanying drawings are only schematic illustrations of the present disclosure and are not necessarily drawn to scale. The same reference numerals in the figures represent the same or similar parts, and their repeated description will be omitted. Some of the block diagrams shown in the accompanying drawings are functional entities and do not necessarily correspond to physically or logically independent entities. These functional entities can be implemented in software form, or implemented in one or more hardware modules or integrated circuits, or implemented in different networks and / or processor devices and / or microcontroller devices.

[0035] The automated business system must not only be able to accurately identify various user needs, but also be able to flexibly call and coordinate different capability modules to achieve step-by-step processing. However, the current multi-intent recognition accuracy and efficiency for user rich text requests are low, making it difficult to call the corresponding capabilities and unable to meet user needs.

[0036] At present, the large language model LLM has achieved good results in various evaluations. With the help of LLM's powerful semantic analysis capabilities, it changes the interaction mode of traditional applications and can quickly realize multi-intent recognition and capability call.

[0037] To facilitate understanding, several terms involved in this application are first explained below.

[0038] Large language model: refers to a natural language processing model based on deep learning technology and trained on large-scale text data. This type of model usually uses a neural network (such as the Transformer structure) to understand, generate and process natural language text. Typical large language models such as GPT-3 and BERT can complete tasks including text generation, language translation, question answering, etc. They have powerful language understanding and generation capabilities by learning language patterns from massive text data.

[0039] Multi-intent recognition: refers to the process of identifying multiple intents or purposes expressed by a user in a sentence or a conversation in natural language processing. This is very common in some dialogue systems or intelligent assistants. For example, a sentence input by a user may contain two intents at the same time: booking a restaurant and asking about the weather. Multi-intent recognition technology can correctly extract multiple independent intents from a sentence so that the system can perform corresponding operations.

[0040] Rich text: refers to text content that contains formatting, usually including text with different fonts, colors, sizes, bold, italics, underlines, hyperlinks, images, tables, etc. Unlike plain text (which only contains characters and no formatting information), rich text can better express and present content. Common rich text formats include HTML, Markdown, etc., which can more flexibly express complex information.

[0041] Content extraction: refers to the technology of automatically extracting useful information from unstructured or semi-structured data (such as text, web pages, documents). This technology is widely used in information retrieval, data mining and other fields, and is often used to extract specific entities, relationships, events, time, place and other information from text. For example, extract the time, place, people involved and other information of an event from a news article. Content extraction helps to convert large amounts of text data into structured information for further analysis and processing.

[0042] Figure 1 1 is a schematic diagram of a business processing system provided by an exemplary embodiment of the present application. The system includes: a plurality of user terminals 120 and a server terminal 140, wherein the user terminal 120 sends a user question and the server terminal 140 responds.

[0043] The user end 120 can be a mobile terminal such as a mobile phone, a game console, a tablet computer, an e-book reader, smart glasses, an MP4 (Moving Picture Experts Group Audio Layer IV) player, a smart home device, an AR (Augmented Reality) device, a VR (Virtual Reality) device, or the user end 120 can also be a personal computer (PC), such as a laptop computer and a desktop computer.

[0044] The user terminal 120 may be installed with an application for providing request processing.

[0045] The user end 120 is connected to the server end 140 via a communication network. Optionally, the communication network is a wired network or a wireless network.

[0046] The server 140 is a server, or is composed of several servers, or is a virtualization platform, or is a cloud computing service center. The server 140 is used to provide background services for applications that provide request processing. Optionally, the server 140 undertakes the main computing work and the user end 120 undertakes the secondary computing work; or, the server 140 undertakes the secondary computing work and the user end 120 undertakes the main computing work; or, the user end 120 and the server 140 adopt a distributed computing architecture for collaborative computing.

[0047] In some optional embodiments, the server 140 is used to store request processing program information.

[0048] Optionally, the logistics user terminals of the application programs installed in different user terminals 120 are the same, or the logistics user terminals of the application programs installed on the two user terminals 120 are logistics user terminals of the same type of application programs on different control system platforms. Based on the different terminal platforms, the specific forms of the logistics user terminals of the application programs may also be different, for example, the logistics user terminals of the application programs may be mobile phone logistics user terminals, PC logistics user terminals, or World Wide Web (Web) logistics user terminals.

[0049] Those skilled in the art will appreciate that the number of the user terminals 120 may be more or less. For example, the terminal may be only one, or the terminal may be dozens or hundreds, or more. The embodiment of the present application does not limit the number and device type of the terminal.

[0050] Optionally, the system may also include a management device ( Figure 1 (not shown), the management device is connected to the server 140 via a communication network. Optionally, the communication network is a wired network or a wireless network.

[0051] Optionally, the above-mentioned wireless network or wired network uses standard communication technology and / or protocol. The network is usually the Internet, but it can also be any network, including but not limited to a local area network (LAN), a metropolitan area network (MAN), a wide area network (WAN), a mobile, wired or wireless network, a dedicated network or any combination of a virtual private network). In some embodiments, the data exchanged through the network is represented by technologies and / or formats including Hyper Text Mark-up Language (HTML), Extensible Markup Language (XML), etc. In addition, conventional encryption technologies such as Secure Socket Layer (SSL), Transport Layer Security (TLS), Virtual Private Network (VPN), Internet Protocol Security (IPsec) can also be used to encrypt all or some links. In other embodiments, customized and / or dedicated data communication technologies can also be used to replace or supplement the above-mentioned data communication technologies.

[0052] like Figure 2 As shown, a request processing method according to an embodiment of the present disclosure includes:

[0053] Step S202 : in response to the acquired service request in the rich text format, extract multiple reference information from the service request based on multiple format types of the rich text format.

[0054] Among them, Rich Text Format (RTF) refers to a text format that can contain information in multiple formats.

[0055] Format types include:

[0056] Text format: that is, ordinary text content, which is the most basic information carrier and is used to convey the main semantic information.

[0057] Link format: It exists in the form of a hyperlink. Users can jump to other web pages, documents or resources by clicking on the link. For example, a rich text introducing a product may contain a link to the product details page.

[0058] Tabular format: Organizing data in a tabular form can clearly display data information with row and column structures.

[0059] Multimedia format: includes multimedia elements such as pictures, audio, and video, which are used to display information more intuitively and vividly.

[0060] Reference information refers to valid information that can be used for intent recognition.

[0061] Step S204: Perform multi-intent recognition on multiple reference information based on the multi-intent detection model to identify multiple associated intents.

[0062] Among them, the multi-intent detection model is a model used to identify multiple intents contained in a text. It can be based on a large language model, machine learning or deep learning technology. By learning a large amount of annotated data, it can analyze the semantic information in the input text and determine the multiple intents contained therein.

[0063] In a business request, there may be a certain logical relationship between multiple intents, and these interrelated intents can together constitute the user's complete needs.

[0064] Step S206, constructing an event flow that conforms to a multi-intent event flow template set based on the identified associated multiple intents.

[0065] Among them, the multi-intent event flow template set refers to a set of pre-configured templates. Each template describes the process and rules of a series of operations that should be performed under a specific combination of multiple intents. The template can be formulated based on business logic and common user intent scenarios.

[0066] The event flow refers to the specific operation execution order determined after matching the corresponding template from the multi-intent event flow template set based on the identified multiple associated intents.

[0067] Step S208, based on the reference information and the event flow, the business request is decomposed into multiple request units, and a corresponding intent prompt is created for each intent, and the multiple request units are matched with the multiple intents one by one.

[0068] Among them, intent prompts refer to prompt information generated according to the specific needs and business scenarios of the intent in order to help users or systems better understand and process each intent. Intent prompts can guide subsequent processing modules to accurately process the intent.

[0069] Step S210, calling a matching processing module based on the intention prompt, and processing multiple request units in sequence according to the execution order of the event flow to generate reply information for the business request.

[0070] Among them, the processing module can be an API (Application Programming Interface). Each module is responsible for processing tasks related to a specific intent and performing corresponding operations according to the intent prompt and the input request unit. For example, the processing module corresponding to the "query product information" intent is responsible for querying detailed information of the product from the database; the processing module corresponding to the "settlement" intent is responsible for processing operations such as order generation, payment, and inventory update. The processing modules coordinate and interact with each other through event streams to jointly complete the processing of business requests and obtain response information for business requests.

[0071] In this embodiment, by extracting effective reference information from rich texts of various formats and using a multi-intent detection model to identify multiple related intents, an event stream that conforms to a template set is constructed based on the intent, an orderly process is planned for business processing, and the business request is further decomposed into multiple request units that match the intents one by one. Accurate intent prompts are created for each intent, and matching processing modules are called based on the intent prompts. The request units are processed in sequence according to the event flow order to generate reply information. This allows for a deep understanding of complex user needs and the processing of tasks with different intents, thereby enhancing the adaptability and response capabilities to complex business scenarios, which is beneficial for improving the processing efficiency of complex user needs.

[0072] Furthermore, please apply the request processing solution to the automated business system, which can adapt to different business scenarios and achieve dynamic processing and rapid response to diverse needs through a customized set of multi-intent templates.

[0073] like Figure 3 As shown, in one embodiment of the present disclosure, an event flow conforming to a multi-intent event flow template set is constructed based on the identified associated multiple intents, including:

[0074] Step S302, matching the associated multiple intents with the configured multi-intent event flow template.

[0075] In some embodiments, each template in the template set is traversed to check whether the current intention combination fully or partially matches a certain template to find an operation process template that conforms to the business logic.

[0076] Step S304, if the associated multiple intents do not match the multi-intent event stream template, trigger the calibration and recognition operation to calibrate the multiple reference information, and re-execute the multi-intent recognition based on the calibrated multiple reference information until the identified associated multiple intents match the multi-intent event stream template.

[0077] In some embodiments, if it is found in the matching operation that the associated multiple intents do not match the existing multi-intent event stream template, the calibration and recognition operation is automatically triggered, and the multiple reference information is calibrated, including but not limited to performing a deeper semantic analysis of the reference information, combining contextual information or using external knowledge sources to clarify ambiguous content. After the calibration is completed, based on the calibrated reference information, the multi-intent recognition process is re-executed, and an attempt is made again to accurately identify the associated multiple intents from this information, and then the newly identified intents are continued to be matched with the multi-intent event stream template, and this cycle is repeated until the identified associated multiple intents are successfully matched with a template in the template set.

[0078] Step S306, determining the execution order and transfer parameters of the associated multiple intents based on the multi-intent event flow template.

[0079] Among them, the transferred parameters can be data or conditions used for information interaction and sharing between operations corresponding to different intents in the process of processing business requests based on the multi-intent event flow template.

[0080] In some embodiments, once a matching multi-intent event flow template is found, the execution order of the associated multiple intents is determined based on the template. The template includes the sequence of each intent in the entire business process, and clarifies the parameters that need to be passed when each intent interacts with each other, so as to ensure that the operations corresponding to each intent can be executed correctly and cooperate with the operations of other intents.

[0081] Step S308, constructing the associated multiple intents into an event stream based on the execution order and the passed parameters.

[0082] In some embodiments, multiple associated intents are constructed into specific event flows according to the determined execution order and transfer parameters. In this process, each intent is arranged in sequence according to the execution order, and the transfer parameters are embedded into the corresponding intent operation links to form a complete and executable business process description, so as to prepare for the subsequent system to process business requests according to the process.

[0083] In this embodiment, by introducing a multi-intent template matching and calibration mechanism, when the identified intent set is not in the predefined template set, multiple identification and calibration are automatically performed until a match is successful, ensuring smooth collaboration between different intents, thereby improving the stability and reliability of the business processing process, reducing the probability of erroneous processing, and improving the accuracy of intent recognition.

[0084] In one embodiment of the present disclosure, a business request is decomposed into multiple request units based on reference information and an event stream, including: traversing multiple intents in an event stream; for each intent, searching for relevant request information in the reference information to generate a request unit based on the relevant request information and the intent.

[0085] In this embodiment, by traversing multiple intents in the event stream and searching for relevant request information in the reference information for each intent to generate a request unit, refined decomposition and structured processing of business requests are achieved. Based on different intents, closely related content can be accurately screened out from complex reference information, and business requests can be divided into multiple request units according to intent, which helps to perform more efficient and targeted processing of each request unit later. At the same time, each request unit corresponds to a specific intent, which enhances the system's understanding and processing capabilities of complex business requests.

[0086] In one embodiment of the present disclosure, a corresponding intent prompt is created for each intent, including: parsing the prompt information required to realize the intent; performing scenario adaptation and content integration on the intent prompt based on the business scenario and delivery parameters of the business request to obtain the intent prompt.

[0087] In this embodiment, by understanding the key operations and information involved in each intent in the business process, the prompt information necessary to realize the intent can be sorted out. These prompt information should be able to guide the user or the system to accurately perform operations related to the intent. Furthermore, since different business scenarios have different requirements for intent prompts, the specific business scenarios in which the business requests are located are analyzed, and scenario information is added to the intent prompts. The transmission parameters may include various information related to the business request. The above information is integrated to obtain intent prompts, which improves the quality and practicality of the intent prompts, helps to guide the processing of business requests more efficiently, improves the accuracy and success rate of business processing, and thus enhances the overall performance of the system and user satisfaction.

[0088] In one embodiment of the present disclosure, a matching processing module is called based on an intention prompt, and multiple request units are processed in sequence according to the execution order of the event flow to generate reply information of the business request, including:

[0089] The processing module is called based on the intention prompt that matches the request unit, so that multiple request units correspond to multiple processing modules one by one based on the execution order; for any adjacent first processing module and second processing module, the first processing result output by the first processing module and the request unit corresponding to the second processing module are input into the second processing module for processing, and the second processing result is output, wherein the processing result output by the last processing module is used as reply information.

[0090] The multiple request units include an adjacent first request unit and a second request unit, and a matching first processing module is called based on an intent prompt corresponding to the first request unit, and a matching second processing module is called based on an intent prompt corresponding to the second request unit.

[0091] Among them, each processing module has its specific function and is specially used to process the request unit of the corresponding intention. For example, the intention prompt corresponding to the first request unit may be "query user basic information", and the first processing module matched with it is the module responsible for querying relevant information from the user information database; if the intention prompt corresponding to the second request unit is "recommend products based on user information", the matching second processing module is the module that recommends products based on user information.

[0092] The first processing module outputs a first processing result. If the first processing module is the first processing module, the first request unit is input to the first processing module. If the first processing module is not the first processing module, the first request unit and the processing result output by the adjacent previous processing module are input to the first processing module.

[0093] In some embodiments, it is determined whether the first processing module is the first processing module. If it is the first processing module, the system directly inputs the first request unit into the first processing module. The first processing module analyzes, calculates or performs other operations on the first request unit according to the processing logic set internally, and finally outputs the first processing result.

[0094] If the first processing module is not the first processing module, the first request unit and the processing result output by the adjacent previous processing module are input to the first processing module together, and the first processing module combines these two parts of input information, processes according to the established logic, and generates a first processing result.

[0095] The first processing result and the second request unit are input into the second processing module to obtain the second processing result, until the last processing module outputs the reply information.

[0096] In this embodiment, except for the first processing module, each processing module takes the output of the previous processing module and the current request unit as input, performs corresponding processing and outputs the result, and repeats this cycle until the last processing module completes the processing of the input information and outputs the final reply information, which is a complete response to the entire business request and contains the processing results that meet the user's needs. By adopting a processing mechanism that calls the capability module step by step, that is, after identifying multiple intentions, each intention is bound to the corresponding processing module, and these modules are called step by step in a certain order, this mechanism ensures that complex requests can be broken down into multiple simple steps, and the processing results of each step can be used as the input of the next step, ensuring the accuracy and consistency of task processing.

[0097] Furthermore, by flexibly calling and coordinating multiple capability modules, different requirements in user requests are processed step by step, thereby effectively responding to complex rich text requests involving multiple business areas or functional modules.

[0098] In one embodiment of the present disclosure, the first processing result output by the first processing module and the request unit corresponding to the second processing module are input into the second processing module for processing to output the second processing result, including: integrating the first processing result and the request unit corresponding to the second processing module based on the transmission parameters to obtain integrated request information; inputting the integrated request information into the second processing module to obtain the second processing result.

[0099] In some embodiments, in an e-commerce scenario, the transmitted parameters may include discount information of the product and promotion activity rules, and the transmitted parameters will affect the way the subsequent processing modules integrate and process the information.

[0100] In this embodiment, the passing parameter is the key information bridge connecting different processing modules. It carries the specific data or conditions required for different stages in the business process, and can flexibly call different API modules according to the identified multiple intentions. Each API module corresponds to a specific processing capability. According to the needs of different intents, the corresponding API is dynamically selected and called, and combined with the results of the previous API output, the information of different stages is integrated by passing parameters, which allows subsequent processing modules to make decisions and operations based on more comprehensive information, ensuring that each subtask can receive the most appropriate processing.

[0101] In one embodiment of the present disclosure, the first processing result output by the first processing module and the request unit corresponding to the second processing module are input into the second processing module for processing, and the second processing result is output, further comprising:

[0102] Detect whether the first processing result matches the input format of the second processing module; if not, format modulate the first processing result based on the event stream transmission mechanism to obtain the first processing result that matches the input format.

[0103] Among them, the event flow transmission mechanism is a set of rules and processes used to manage and coordinate data flow and processing order in the entire system. It is used to define the interaction method, data transmission path and processing order between various processing modules.

[0104] In some implementations, the first processing result is compared with the input format requirements of the second processing module, including but not limited to the operations of checking the data type, verifying the data structure, and checking the format restrictions. When a format mismatch is detected, the specific method of format modulation of the first processing result is determined according to the event stream delivery mechanism and the rules pre-set by the system. This may include operations such as data type conversion, data structure reorganization, data truncation or padding. According to the determined format modulation method, the first processing result is actually modulated. After the format modulation is completed, the modulated first processing result is compared with the input format requirements of the second processing module again, and the above-mentioned format matching detection steps are repeated. If there is still a mismatch, it may be necessary to further adjust the format modulation method, or check whether there are other problems in the event stream delivery mechanism, until the modulated first processing result completely matches the input format of the second processing module.

[0105] In this embodiment, by configuring an efficient event stream delivery mechanism, it is ensured that during multi-intent processing, the output format of each step can be automatically adjusted to meet the input requirements of the next step. This mechanism not only improves the processing efficiency of the system, but also ensures accurate data transmission between modules.

[0106] In one embodiment of the present disclosure, before responding to the acquired service request in rich text format, the method further includes:

[0107] Check whether the original request conforms to the rich text format; if it does not conform to the rich text format, generate a modification prompt message to modify the original request into a business request in the rich text format, or query a rich text template that matches the original request to convert the original request into a business request in the rich text format based on the rich text template.

[0108] In some embodiments, a parsing tool is used to analyze the original request to check whether it complies with predefined rich text format standards, including but not limited to lexical analysis, syntax analysis and other technologies to identify various elements and format information in the request. When it is detected that the original request does not comply with the rich text format, the original request is analyzed in detail to determine the specific non-compliance points. Based on the determined non-compliance points, targeted modification prompt information is generated, and the generated modification prompt information is presented to the user in an appropriate manner, such as popping up a prompt box on the user interface, or recording it in the system log and informing the user to view it.

[0109] In some embodiments, a rich text template library can be pre-established, which contains various types of rich text templates, each template corresponding to a common business request type or format. According to the content and characteristics of the original request, a query is performed in the rich text template library to find a matching template, and the information in the original request is filled into the corresponding fields and positions according to the structure and format requirements of the best matching template. The format of the filled content is adjusted to ensure that it complies with the rich text format standard.

[0110] In this embodiment, by detecting and converting the original request into a rich text format, it is ensured that the input format is a correct rich text format, thereby ensuring that the system can process requests from various sources and formats, and improving the system's compatibility with different types of data.

[0111] In one embodiment of the present disclosure, multiple reference information is extracted from a service request based on multiple format types of a rich text format, including:

[0112] Parse the various format types in the business request, where the various format types include text format, link format, table format and multimedia format; based on the parsing results, use the information extraction function to extract type data matching the various format types from the business request as multiple reference information.

[0113] In this embodiment, by parsing various format types in business requests and using information extraction functions to obtain matching type data as reference information, it is possible to process business requests containing complex formats such as text, links, tables, and multimedia. For data in different formats, matching content can be accurately extracted, greatly improving the accuracy and completeness of information acquisition.

[0114] like Figure 4 As shown, in one embodiment of the present disclosure, before performing multi-intent recognition on multiple reference information based on a multi-intent detection model and identifying multiple intents, the process includes:

[0115] Step S402 , annotating the collected historical rich text data with multiple associated intentions based on different format types, and generating intention annotated texts of different format types.

[0116] In some embodiments, a large amount of historical rich text data is collected, and for each format type of data, the user intent contained therein is analyzed, the intent in the data is annotated, and the rich text data with annotated intent is organized into intent annotated texts of different format types.

[0117] Step S404, formatting the intention annotated texts of different format types is unified to obtain target format texts, which include matching intention tags, original format tags, and association tags with other intentions.

[0118] In some embodiments, intent annotation texts of different format types may differ in structure and representation. In order to facilitate the subsequent processing and utilization of these data, format unification is required. The unified target format includes: matching intent tags, which are used to clearly mark the intent represented by the data; original format tags, which are used to record the original format type of the data (such as text, link, table, etc.); association tags with other intents, which are used to reflect the logical relationship between this intent and other intents.

[0119] The intention annotation text of each format type is converted to conform to a unified target format.

[0120] Step S406: fine-tune the large language model based on the target format text to obtain a multi-intent detection model.

[0121] In some embodiments, the target format text is input into the large language model as fine-tuning data. During the fine-tuning process, the model will learn the patterns and rules contained in the intent labels, original format labels, and intent-related labels in the target format text. By adjusting the parameters of the model, the large language model can better adapt to multi-intent detection tasks and more accurately identify multiple intents associated with different format types in rich text data, as well as the relationships between intents.

[0122] In this embodiment, historical rich text data is annotated with intents based on format types, and the intent information formats behind data of different formats are mined and unified so that the annotated data have a consistent structure, which is convenient for model learning and understanding. The large language model is fine-tuned based on the target format text, effectively utilizing the general capabilities of the large language model and focusing it on the specific task of multi-intent detection, so that the final multi-intent detection model can accurately identify the intents associated with multiple format types in rich text data, as well as the complex relationships between intents, which can help the system understand the complex needs of users more deeply and improve the processing capability and accuracy of rich text business requests, thereby improving user experience and enhancing the adaptability and reliability of the system in complex business scenarios.

[0123] In one embodiment of the present disclosure, multi-intent recognition is performed on multiple reference information based on a multi-intent detection model to identify multiple associated intents, including: extracting intent features from the reference information to obtain an intent feature vector; inputting the intent feature vector into the multi-intent detection model to output an associated label between the identified intent and other intents to obtain multiple associated intents.

[0124] In this embodiment, intent features are extracted from reference information and converted into intent feature vectors, which are then input into a multi-intent detection model to obtain multiple associated intents. Intent feature extraction can accurately extract key intent-related features from complex reference information, and convert diverse information into a vector form that is easy for the model to understand and process, greatly improving the information utilization efficiency and the processing speed of the model. The multi-intent detection model performs analysis based on these intent feature vectors, and can accurately output the identified intent labels and the associated labels with other intents, so that the system can deeply understand the complex intent relationships behind the reference information and effectively tap into the user's potential diverse needs, thereby helping to improve the accuracy and efficiency of business processing and enhance the system's adaptability in complex business scenarios.

[0125] like Figure 5 As shown, according to another embodiment of the present disclosure, a business processing system includes:

[0126] The input verification module 502 is used to ensure that the input format is a correct rich text format.

[0127] The information extraction module 504 is used to extract various types of information from the rich text request, such as plain text, URL links, pictures, tables, etc.

[0128] The multi-intent recognition module 506 is used to use LLM to perform multi-intent recognition on the extracted information to ensure that the recognized intent conforms to the predefined event flow template set.

[0129] The input decomposition and intention matching module 508 is used to decompose the original input into units that match each intention and create a corresponding prompt according to each intention.

[0130] The API calling and data flow transmission module 510 is used to call the corresponding API according to each intent to process the input of the corresponding part, ensure the compatibility of the data formats between the various APIs, so as to achieve smooth transmission of the data flow.

[0131] In this embodiment, by optimizing the identification and processing flow, multiple identification and processing of requests are reduced, the overall response efficiency of the system is improved, and the problem of low identification efficiency in the prior art is solved.

[0132] like Figure 6 As shown, according to another embodiment of the present disclosure, a request processing method includes:

[0133] Get user rich text request and multi-intent recognition Define a multi-intent combined event flow template Ι to check whether the input rich text request r conforms to the rich text format. If not, require re-entry until the format is correct.

[0134] Use the information_extract_func(·) function to extract reference information from various information in the rich text and store them as variables such as text, url, figure, table, and others; text url figure table others.

[0135] Perform multi-intent recognition on reference information, extract and classify multiple intents in rich text requests, and use LLM multi_intent , based on the text, url, figure, table and others information extracted from the input and multi-intent detection Identify multiple intents 1 ,i 2 ,K,i N , and form an event stream I(r) = {i 1 ,i 2 ,K,i N}, verify whether the identified event flow I(r) conforms to the pre-set multi-intention event flow template set Ι. If not, repeat the intent recognition process until I(r)∈Ι.

[0136] Based on the reference information and event stream, the business request is inputted and decomposed to obtain multiple request units and corresponding intent prompts. The intent prompts are used for intent matching and to prepare to call the API of a specific intent. The decompose_func function is used to pass in the extracted rich text information and the identified event stream I(r) to decompose the input into corresponding X(r) and P(r), where X(r) is the set of request units required to call the API for each intent. X(r) = {x 1 ,x 2 ,K,x N}, P(r) contains the prompt required for each intent, that is, P(r) = {p 1 ,p 2 ,K,p N}.

[0137] Call the corresponding processing module API according to different intentions to transfer data streams and ensure that the data format is consistent when transferred between APIs, including:

[0138] Create an API set API = {API 1 ,API 2 ,K,API N}, the nth intent corresponds to the API n =LLM(·|p n ).

[0139] Call the corresponding API for the first intent, that is, API 1 (x 1 ), and obtain the output y 1 .

[0140] Verify 1 Is the format consistent with the API 2 The input format matches the .

[0141] If there is no match, call the API again 1 (x 1 ) until the format matches.

[0142] For each subsequent intent, repeat the above process and call the next API, that is, API n+1 , and verify that its output y n+1 Is the format consistent with the next API n+2 The input format matches the .

[0143] If the format does not match, the current API is called repeatedly until the format matches.

[0144] Repeat the above process until all intents are processed.

[0145] The output of the last API is y N The final response to the user's request is a.

[0146] In this way, the entire algorithm can identify multiple intents from complex rich text requests, call the appropriate APIs in sequence to handle these intents, and finally get a complete answer.

[0147] like Figure 7 As shown, a request processing method according to another embodiment of the present disclosure includes:

[0148] Step S702: Obtain a business request in a rich text format.

[0149] Step S704: extract text, hyperlinks, images and table information in the business request as reference information.

[0150] Step S706, perform multiple intent recognitions on the reference information, and ensure that the event flow obtained by the recognition results is within the given template.

[0151] Step S708: Decompose the service request and the corresponding intention prompt to obtain a request unit.

[0152] Step S710: The large language model calls different API processing request units according to different intentions to obtain corresponding processing results.

[0153] Step S712, perform event stream transfer until the last API outputs reply information.

[0154] It should be noted that the above figures are only schematic illustrations of the processes included in the method according to the exemplary embodiments of the present disclosure, and are not intended to be limiting. It is easy to understand that the processes shown in the above figures do not indicate or limit the time sequence of these processes. In addition, it is also easy to understand that these processes can be performed synchronously or asynchronously, for example, in multiple modules.

[0155] Refer to the following Figure 8 The request processing device 800 according to an embodiment of the present disclosure is described. Figure 8 The request processing device 800 shown is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present disclosure.

[0156] The request processing device 800 is expressed in the form of a hardware module or a software module. The components of the request processing device 800 may include but are not limited to: an extraction module 802, which is used to extract multiple reference information from the business request based on multiple format types of the rich text format in response to the obtained business request in rich text format; an identification module 804, which is used to perform multi-intent identification on multiple reference information based on a multi-intent detection model, and identify multiple associated intents; a construction module 806, which is used to construct an event flow that conforms to the multi-intent event flow template set based on the identified multiple associated intents; a disassembly module 808, which is used to disassemble the business request into multiple request units based on the reference information and the event flow, and create a corresponding intent prompt for each intent, and multiple request units are matched one by one with multiple intents; a processing module 810, which is used to call the matching processing module based on the intent prompt, and process multiple request units in sequence according to the execution order of the event flow to generate reply information for the business request.

[0157] like Fig. 9 As shown, the electronic device 900 is in the form of a general computing device. The components of the electronic device 900 may include but are not limited to: at least one processing unit 910, at least one storage unit 920, and a bus 930 connecting different system components (including the storage unit 920 and the processing unit 910).

[0158] The storage unit stores program codes, which can be executed by the processing unit 910, so that the processing unit 910 performs the steps according to various exemplary embodiments of the present disclosure described in the above “Exemplary Method” section of this specification. For example, the processing unit 910 can perform the following steps: Figure 2 The scheme described in .

[0159] The storage unit 920 may include a readable medium in the form of a volatile storage unit, such as a random access storage unit (RAM) 9201 and / or a cache storage unit 9202 , and may further include a read-only storage unit (ROM) 9203 .

[0160] The storage unit 920 may also include a program / utility 9204 having a set (at least one) of program modules 9205, such program modules 9205 including but not limited to: an operating system, one or more application programs, other program modules, and program data, each of which or some combination may include an implementation of a network environment.

[0161] Bus 930 may represent one or more of several types of bus structures, including a memory unit bus or memory unit controller, a peripheral bus, an accelerated graphics port, a processing unit, or a local bus using any of a variety of bus architectures.

[0162] The electronic device 900 may also communicate with one or more external devices 970 (e.g., keyboards, pointing devices, Bluetooth devices, etc.), one or more devices that enable a user to interact with the electronic device 900, and / or any device that enables the electronic device 900 to communicate with one or more other computing devices (e.g., routers, modems, etc.). Such communication may be performed through an input / output (I / O) interface 950. Furthermore, the electronic device 900 may also communicate with one or more networks (e.g., a local area network (LAN), a wide area network (WAN), and / or a public network, such as the Internet) through a network adapter 960. Fig. 9 As shown, the network adapter 960 communicates with other modules of the electronic device 900 via the bus 930. It should be understood that, although not shown in the figure, other hardware and / or software modules can be used in conjunction with the electronic device 900, including but not limited to: microcode, device drivers, redundant processing units, external disk drive arrays, RAID systems, tape drives, and data backup storage systems.

[0163] Through the description of the above implementation, it is easy for those skilled in the art to understand that the example implementation described here can be implemented by software, or by software combined with necessary hardware. Therefore, the technical solution according to the implementation of the present disclosure can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (which can be a CD-ROM, a USB flash drive, a mobile hard disk, etc.) or on a network, and includes several instructions to enable a computing device (which can be a personal computer, a server, a terminal device, or an electronic device, etc.) to execute the method according to the implementation of the present disclosure.

[0164] In an exemplary embodiment of the present disclosure, a computer-readable storage medium is also provided, on which a program product capable of implementing the above method of the present specification is stored. In some possible implementations, various aspects of the present disclosure may also be implemented in the form of a program product, which includes a program code, and when the program product is run on an electronic device, the program code is used to enable the electronic device to perform the steps according to various exemplary implementations of the present disclosure described in the above "Exemplary Method" section of the present specification.

[0165] According to the program product for implementing the above method according to the embodiment of the present disclosure, it can adopt a portable compact disk read-only memory (CD-ROM) and include program code, and can be run on an electronic device, such as a personal computer. However, the program product of the present disclosure is not limited thereto. In this document, a readable storage medium can be any tangible medium containing or storing a program, which can be used by or in combination with an instruction execution system, an apparatus or a device.

[0166] The program product may use any combination of one or more readable media. The readable medium may be a readable signal medium or a readable storage medium. The readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or device, or any combination of the above. More specific examples (non-exhaustive list) of readable storage media include: an electrical connection with one or more wires, a portable disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above.

[0167] Computer readable signal media may include data signals propagated in baseband or as part of a carrier wave, in which readable program code is carried. Such propagated data signals may take a variety of forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above. Readable signal media may also be any readable medium other than a readable storage medium, which may send, propagate, or transmit a program for use by or in conjunction with an instruction execution system, apparatus, or device.

[0168] The program code embodied on the readable medium may be transmitted using any appropriate medium, including but not limited to wireless, wired, optical cable, RF, etc., or any suitable combination of the foregoing.

[0169] The program code for performing the operations of the present disclosure may be written in any combination of one or more programming languages, including object-oriented programming languages ​​such as Java, C++, etc., and conventional procedural programming languages ​​such as "C" or similar programming languages. The program code may be executed entirely on the user computing device, partially on the IoT terminal, as an independent software package, partially on the user computing device and partially on the remote computing device, or entirely on the remote computing device or server. In the case of a remote computing device, the remote computing device may be connected to the user computing device through any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computing device (e.g., using an Internet service provider to connect through the Internet).

[0170] It should be noted that although several modules or units of the device for action execution are mentioned in the above detailed description, this division is not mandatory. In fact, according to the embodiment of the present disclosure, the features and functions of two or more modules or units described above can be concretized in one module or unit. Conversely, the features and functions of a module or unit described above can be further divided into multiple modules or units to be concretized. In addition, although the various steps of the method in the present disclosure are described in a specific order in the accompanying drawings, this does not require or imply that these steps must be performed in this specific order, or that all the steps shown must be performed to achieve the desired result. In addition or alternatively, some steps can be omitted, multiple steps can be combined into one step for execution, and / or one step can be decomposed into multiple steps for execution, etc.

[0171] Through the description of the above embodiments, it is easy for those skilled in the art to understand that the example embodiments can be implemented by software, or by software combined with necessary hardware. Therefore, the technical solution according to the embodiment of the present disclosure can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (which can be a CD-ROM, a USB flash drive, a mobile hard disk, etc.) or on a network, including several instructions so that a computing device (which can be a personal computer, a server, a mobile terminal, or an electronic device, etc.) executes the method according to the embodiment of the present disclosure. After considering the disclosure of the specification and practicing the disclosure here, it will be easy for those skilled in the art to think of other embodiments of the present disclosure. This application is intended to cover any variation, use or adaptive change of the present disclosure, which follows the general principles of the present disclosure and includes the common knowledge or customary technical means in the technical field that are not disclosed in the present disclosure. The specification and the embodiments are only regarded as exemplary, and the true scope and spirit of the present disclosure are indicated by the attached claims.

Claims

1. A request processing method, characterized in that: include: In response to the acquired business request in rich text format, extracting multiple reference information from the business request based on multiple format types of the rich text format; Performing multi-intent recognition on the multiple reference information based on a multi-intent detection model to identify multiple associated intents; Constructing an event flow conforming to a multi-intent event flow template set based on the identified multiple associated intents; Decomposing the business request into multiple request units based on the reference information and the event stream, and creating a corresponding intent prompt for each of the intents, wherein the multiple request units are matched with the multiple intents one by one; Based on the intention prompt, a matching processing module is called, and the multiple request units are processed in sequence according to the execution order of the event flow to generate reply information for the business request.

2. The request processing method according to claim 1, characterized in that: Constructing an event flow that complies with a multi-intent event flow template set based on the identified multiple associated intents, including: Performing a matching operation on the associated multiple intents and the configured multi-intent event flow template; If the associated multiple intents do not match the multi-intent event stream template, triggering a calibration and recognition operation to calibrate the multiple reference information, and re-performing the multi-intent recognition based on the calibrated multiple reference information until the identified associated multiple intents match the multi-intent event stream template; Determining the execution order and transfer parameters of the associated multiple intents based on the multi-intent event flow template; Based on the execution order and the passed parameters, the associated multiple intents are constructed into the event stream.

3. The request processing method according to claim 2, characterized in that: Decomposing the service request into multiple request units based on the reference information and the event stream includes: Iterating through the plurality of intents in the event stream; For each of the intents, relevant request information is searched in the reference information to generate the request unit based on the relevant request information and the intent.

4. The request processing method according to claim 2, characterized in that: Create a corresponding intent hint for each of the above intents, including: Parsing the prompt information required to achieve the stated intent; Based on the business scenario of the business request and the transmission parameters, the intent prompt is subjected to scenario adaptation and content integration to obtain the intent prompt.

5. The request processing method according to claim 2, characterized in that: Based on the intention prompt, a matching processing module is called, and the multiple request units are processed in sequence according to the execution order of the event flow to generate reply information of the business request, including: Calling the processing module based on the intention prompt matching the request unit, so that the multiple request units correspond to the multiple processing modules one by one based on the execution order; For any adjacent first processing module and second processing module, the first processing result output by the first processing module and the request unit corresponding to the second processing module are input into the second processing module for processing, and the second processing result is output, wherein the processing result output by the last processing module is used as the reply information.

6. The request processing method according to claim 5, characterized in that: For any adjacent first processing module and second processing module, inputting the first processing result output by the first processing module and the request unit corresponding to the second processing module into the second processing module for processing, and outputting the second processing result, comprises: Integrate the first processing result and the request unit corresponding to the second processing module based on the transfer parameter to obtain integrated request information; The integration request information is input into the second processing module to obtain the second processing result.

7. The request processing method according to claim 5, characterized in that: Inputting the first processing result output by the first processing module and the request unit corresponding to the second processing module into the second processing module for processing, and outputting the second processing result, further comprising: Detecting whether the first processing result matches the input format of the second processing module; If there is no match, format modulation is performed on the first processing result based on an event stream delivery mechanism to obtain the first processing result that matches the input format.

8. The request processing method according to claim 1, characterized in that: Before responding to the obtained business request in rich text format, the following is also included: Detecting whether the original request complies with the rich text format; If it does not conform to the rich text format, a modification prompt message is generated to prompt that the original request be modified into a business request in the rich text format, or a rich text template matching the original request is queried to convert the original request into a business request in the rich text format based on the rich text template.

9. The request processing method according to claim 1, characterized in that: Extracting multiple reference information from the service request based on multiple format types of the rich text format includes: Parsing multiple format types in the business request, wherein the multiple format types include text format, link format, table format and multimedia format; Based on the parsing result, an information extraction function is used to extract type data matching the multiple format types from the business request as the multiple reference information.

10. The request processing method according to claim 1, characterized in that: Before performing multi-intent recognition on the multiple reference information based on the multi-intent detection model and identifying the multiple intents, the method includes: Annotating the collected historical rich text data with multiple associated intentions based on the different format types to generate intention annotated texts of the different format types; Performing format unification processing on the intention annotated texts of different format types to obtain target format texts, wherein the target format texts include matching intention tags, original format tags, and association tags with other intentions; The large language model is fine-tuned based on the target format text to obtain the multi-intent detection model.

11. The request processing method according to claim 10, characterized in that: Performing multi-intent recognition on the multiple reference information based on the multi-intent detection model to identify multiple associated intents includes: Extracting intention features from the reference information to obtain an intention feature vector; The intent feature vector is input into the multi-intent detection model to output the associated labels between the identified intent label and other intents, thereby obtaining the associated multiple intents.

12. A request processing device, characterized in that: include: An extraction module, configured to extract, in response to an acquired business request in a rich text format, a plurality of reference information from the business request based on a plurality of format types of the rich text format; An identification module, configured to perform multi-intent identification on the plurality of reference information based on a multi-intent detection model, and identify the associated plurality of intents; A construction module, configured to construct an event flow conforming to a multi-intent event flow template set based on the identified multiple associated intents; A disassembly module, configured to disassemble the business request into a plurality of request units based on the reference information and the event stream, and create a corresponding intent prompt for each of the intents, wherein the plurality of request units are matched one by one with the plurality of intents; A processing module is used to call a matching processing module based on the intention prompt, and process the multiple request units in sequence according to the execution order of the event flow to generate reply information for the business request.

13. An electronic device, characterized in that: include: processor; as well as A memory, configured to store executable instructions of the processor; The processor is configured to execute the request processing method according to any one of claims 1 to 11 by executing the executable instructions.

14. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the request processing method according to any one of claims 1 to 11 is implemented.

15. A computer program product having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the request processing method according to any one of claims 1 to 11 is implemented.

Images