Processing method, device and system for intelligent interaction based on intelligent AI model

Abstract

The invention relates to the technical field of intelligent dialogue interaction, in particular to a processing method, device and system for intelligent interaction based on an intelligent AI model. According to the method, the current interaction processing path and the current dialogue interaction state matched with the current intention are analyzed, accurate adjustment of multi-round dialogue paths and response logic is achieved, accurate analysis is conducted on the current intention and the current text data according to the current interaction processing path, and a text analysis result corresponding to the current interaction processing path is obtained. And finally, deep semantic analysis is performed on a text analysis result and a current dialogue interaction state based on context interaction management and control data obtained by interaction between the same object and the intelligent AI model, and leap-type upgrade from traditional rule driving to cognitive driving is realized, so that efficient, intelligent and extensible new-generation AI intelligent dialogue interaction is realized, and the interaction efficiency is improved. And the accuracy and efficiency of intelligent dialogue interaction are improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of intelligent dialogue interaction, and in particular to a processing method, device and system for intelligent interaction based on an intelligent AI model. BACKGROUND

[0002] With the rapid development of intelligent technology and the rapid progress of society, intelligent customer service systems have been widely applied in various fields.

[0003] At present, in the intelligent customer service system, customer service interaction is often carried out in the following way. Specifically, the keywords of the text input by the user are extracted, and the extracted keywords are compared with the predefined keywords to determine the user's intention, and then the customer service interaction is carried out. Alternatively, a machine learning model (such as support vector machine, random forest, naive Bayes) is used to analyze the text input by the user to determine the user's intention, and then the customer service interaction is carried out.

[0004] In practical applications, it is found that based on the machine learning model, the accuracy is improved to some extent compared with the keyword comparison method for intention analysis. However, there are still limitations in understanding complex language structures, sentiment analysis and deep semantic analysis, especially when dealing with diversified customer inquiries. The current machine learning still stays in the shallow semantic understanding, resulting in low efficiency and accuracy of customer service interaction. Therefore, it is urgent to propose a new intelligent dialogue interaction method to improve the efficiency and accuracy of intelligent dialogue interaction, thereby improving the user interaction experience. SUMMARY

[0005] The present application provides a processing method, device and system for intelligent interaction based on an intelligent AI model, which can improve the efficiency and accuracy of intelligent dialogue interaction, thereby improving the user interaction experience.

[0006] To solve the above technical problems, the first aspect of the embodiment of the present application discloses a processing method for intelligent interaction based on an intelligent AI model, which comprises: analyzing the current intention of the target user according to the current text data input by the target user; determining the current interaction processing path matching the current intention and the current dialogue interaction state of the target user according to the current intention of the target user and the current text data input by the target user; analyzing the current intention of the target user and the current text data input by the target user according to the current interaction processing path corresponding to the current intention, to obtain the text analysis result corresponding to the current interaction processing path; According to the text analysis result corresponding to the current interaction processing path, the current dialogue interaction state of the target user, and the pre-constructed interaction management data corresponding to the target user, a current dialogue result matched with the target user is generated. The interaction management data corresponding to the target user includes all context interaction data obtained by the target user in this interaction with the same object and the intelligent AI model.

[0007] As an optional implementation, in the first aspect of the present application, the current intention of the target user is analyzed according to the current text data input by the target user, including: According to the current text data input by the target user, the scene parameters of the current business scenario are determined, and the scene parameters of the current business scenario include the scene type. According to the scene parameters of the current business scenario, a plurality of current intention labels matched with the scene parameters of the current business scenario are obtained. According to the current text data input by the target user, context interaction data matched with the current text data is determined from the pre-constructed interaction management data corresponding to the target user. According to the current text data input by the target user, a current intention prompt label matched with the current text data is obtained, and according to the current text data input by the target user, a current intention output requirement matched with the target user is determined. According to the current intention prompt label corresponding to the target user, the current text data input by the target user, all the current intention labels, the current intention output requirement corresponding to the target user, and the context interaction data matched with the current text data, a current intention prompt parameter of the current text data input by the target user is generated. The current intention prompt parameter corresponding to the target user is analyzed based on the pre-trained intention analysis model to obtain the current intention of the target user.

[0008] As an optional implementation, in the first aspect of the present application, the current dialogue result matched with the target user is generated according to the text analysis result corresponding to the current interaction processing path, the current dialogue interaction state of the target user, and the pre-constructed interaction management data corresponding to the target user, including: determine an interaction credibility of the text analysis result corresponding to each of the current interaction processing paths according to the current text data input by the target user, the pre-constructed interaction management data corresponding to the target user, and the current dialog interaction state of the target user, and select a current interaction processing path with the highest interaction credibility from all the current interaction processing paths as a target current interaction processing path of the target user according to the interaction credibility of each of the current interaction processing paths, and determine the text analysis result of the target current interaction processing path as a current dialog result matched with the target user, wherein the number of the current interaction processing paths is greater than or equal to 1; and / or determine a current parameter corresponding to the target user according to the current text data input by the target user, and determine retrieval data matched with the target user according to the current parameter corresponding to the target user; and generate a current dialog result matched with the target user according to the current text data input by the target user, the retrieval data corresponding to the target user, the text analysis result corresponding to the current interaction processing path, the current dialog interaction state of the target user, and the pre-constructed interaction management data corresponding to the target user.

[0009] The second aspect of the embodiment of the present application discloses a processing device for intelligent interaction based on an intelligent AI model, which comprises: an analysis module configured to analyze a current intention of a target user according to current text data input by the target user; a determination module configured to determine a current interaction processing path matched with the current intention and a current dialog interaction state of the target user according to the current intention of the target user and the current text data input by the target user; The analysis module is further configured to analyze the current intention of the target user and the current text data input by the target user according to the current interaction processing path corresponding to the current intention, and obtain a text analysis result corresponding to the current interaction processing path. a generation module configured to generate a current dialog result matched with the target user according to the text analysis result corresponding to the current interaction processing path, the current dialog interaction state of the target user, and pre-constructed interaction management data corresponding to the target user. The interaction management data corresponding to the target user comprises all context interaction data obtained by the target user in this interaction with the intelligent AI model for the same object.

[0010] As an optional implementation, in the second aspect of the present application, the specific manner in which the analysis module analyzes the current intention of the target user according to the current text data input by the target user comprises: According to the current text data input by the target user, scene parameters of a current service scenario are determined, the scene parameters of the current service scenario including a scene type; According to the scene parameters of the current service scenario, a plurality of current intent labels matching the scene parameters of the current service scenario are obtained; According to the current text data input by the target user, context interaction data matching the current text data is determined from the pre-constructed interaction control data corresponding to the target user; According to the current text data input by the target user, a current intent prompt label matching the current text data is obtained, and according to the current text data input by the target user, a current intent output requirement matching the target user is determined; According to the current intent prompt label corresponding to the target user, the current text data input by the target user, all the current intent labels, the current intent output requirement corresponding to the target user, and the context interaction data matching the current text data, a current intent prompt parameter of the current text data input by the target user is generated; Based on the pre-trained intent analysis model, the current intent prompt parameter corresponding to the target user is analyzed to obtain the current intent of the target user.

[0011] As an optional implementation, in the second aspect of the present application, the generation module generates the specific manner of the current conversation result matching the target user according to the text analysis result corresponding to the current interaction processing path, the current conversation interaction state of the target user, and the pre-constructed interaction control data corresponding to the target user, including: According to the current text data input by the target user, the pre-constructed interaction control data corresponding to the target user, and the current conversation interaction state of the target user, the interaction credibility of the text analysis result corresponding to each current interaction processing path is determined, and according to the interaction credibility of the text analysis result corresponding to each current interaction processing path, the current interaction processing path with the highest interaction credibility is selected from all the current interaction processing paths as the target current interaction processing path of the target user, and the text analysis result of the target current interaction processing path is determined as the current conversation result matching the target user, the number of the current interaction processing paths being greater than or equal to 1; and / or According to the current text data input by the target user, a current parameter corresponding to the target user is determined, and search data matched with the target user is determined according to the current parameter corresponding to the target user; according to the current text data input by the target user, the search data corresponding to the target user, the text analysis result corresponding to the current interaction processing path, the current conversation interaction state of the target user, and the pre-constructed interaction control data corresponding to the target user, a current conversation result matched with the target user is generated.

[0012] The third aspect of the present application discloses an intelligent interaction system, the system comprises: a memory storing executable program codes; a processor coupled with the memory; The processor invokes the executable program codes stored in the memory to execute part or all steps of any one of the processing methods for intelligent interaction based on intelligent AI model disclosed in the first aspect of the present application.

[0013] The fourth aspect of the present application discloses a computer storage medium, the computer storage medium stores computer instructions, when the computer instructions are invoked, part or all steps of any one of the processing methods for intelligent interaction based on intelligent AI model disclosed in the first aspect of the present application are executed.

[0014] Compared with the prior art, the embodiments of the present application have the following beneficial effects: In the embodiments of the present application, the current text data input by the user is analyzed for current intention, and based on the current intention of the user and the current text data input by the user, the current interaction processing path and the current conversation interaction state matched with the current intention are accurately analyzed, so that the multi-round conversation path and the response logic are accurately adjusted, and according to the current interaction processing path corresponding to the current intention, the current intention of the user and the current text data input by the user are accurately analyzed to obtain the text analysis result corresponding to the current interaction processing path, and finally the text analysis result and the current conversation interaction state of the user are analyzed for deep semantic analysis based on the pre-constructed context interaction control data obtained by the user interacting with the intelligent AI model for the same object, that is, from the traditional "rule-driven" to the "cognitive-driven" leapfrog upgrade, realizing the efficient, intelligent and expandable new generation AI intelligent conversation interaction, improving the accuracy and efficiency of the intelligent conversation interaction, and enhancing the personalized service capability, thereby improving the user interaction experience. BRIEF DESCRIPTION OF DRAWINGS

[0015] In order to make the technical solutions in the embodiments of the present application clearer, the accompanying drawings needed in the embodiments description will be briefly introduced as follows. Obviously, the accompanying drawings in the following description are only some embodiments of the present application, and all other embodiments obtained by those of ordinary skill in the art without any creative work on the basis of these drawings are within the protection scope of the present application.

[0016] Figure 1 is a flow diagram of a processing method for intelligent interaction based on an intelligent AI model according to an embodiment of the present application; Figure 2 is a flow diagram of another processing method for intelligent interaction based on an intelligent AI model according to an embodiment of the present application Figure 3 is a structural diagram of a processing device for intelligent interaction based on an intelligent AI model according to an embodiment of the present application; Figure 4 is a structural diagram of another processing device for intelligent interaction based on an intelligent AI model according to an embodiment of the present application; Figure 5 is a structural diagram of an intelligent interaction system according to an embodiment of the present application. DETAILED DESCRIPTION

[0017] In order to make the technical solutions in the embodiments of the present application clearer, the accompanying drawings needed in the embodiments description will be briefly introduced as follows. Obviously, the accompanying drawings in the following description are only some embodiments of the present application, and all other embodiments obtained by those of ordinary skill in the art without any creative work on the basis of these drawings are within the protection scope of the present application.

[0018] The terms "first", "second", etc. in the specification and claims of the present application and the above-mentioned accompanying drawings are used to distinguish different objects, and are not used to describe a specific order. In addition, the terms "include" and "have" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, device, product or equipment including a series of steps or units is not limited to the listed steps or units, but can optionally include steps or units not listed or can optionally include other steps or units inherent to the process, method, product or equipment.

[0019] Reference herein to "an embodiment" means that a particular feature, structure, or characteristic described in connection with the embodiment can be included in at least one embodiment of the application. The appearances of the phrase in various places in the specification are not necessarily all referring to the same embodiment, nor are they necessarily mutually exclusive of one another. As will be apparent to those of ordinary skill in the art, embodiments described herein can be combinable with other embodiments.

[0020] The application discloses a processing method, device and system for intelligent interaction based on an intelligent AI model. The current intention of a user input current text data is analyzed, and based on the current intention of the user and the current text data input by the user, a current interaction processing path and a current dialogue interaction state matching the current intention are accurately analyzed, thereby realizing accurate adjustment of multi-round dialogue path and response logic, and accurate analysis of the current intention of the user and the current text data input by the user according to the current interaction processing path corresponding to the current intention, obtaining a text analysis result corresponding to the current interaction processing path, and finally performing deep semantic analysis on the text analysis result and the current dialogue interaction state of the user based on context interaction control data obtained by the user interacting with the intelligent AI model for the same object, that is, a leapfrog upgrade from traditional "rule-driven" to "cognitive-driven", thereby realizing a new generation of AI intelligent dialogue interaction that is efficient, intelligent and scalable, improving the accuracy and efficiency of intelligent dialogue interaction, and enhancing personalized service capabilities, thereby improving user interaction experience. The following will be described in detail.

[0021] Embodiment one Please refer to Figure 1 , Figure 1 is a flowchart of a processing method for intelligent interaction based on an intelligent AI model disclosed by the embodiment of the application. The method can be applied to any scenario requiring intelligent dialogue interaction, such as customer service scenarios in the fields of finance, e-commerce, government affairs, etc. The scenario is provided with one of the corresponding intelligent interaction system (local system or cloud system), intelligent dialogue interaction device and intelligent dialogue interaction server (local server or cloud server), and is used to realize intelligent dialogue interaction according to the method. The intelligent dialogue interaction device includes but is not limited to a mobile phone, a computer, a smart bracelet and other devices capable of intelligent dialogue interaction. As shown in Figure 1 The method can include the following operations: 101. Analyzing the current intention of a target user according to the current text data input by the target user; determining the current interaction processing path matching the current intention and the current dialogue interaction state of the target user according to the current intention of the target user and the current text data input by the target user.

[0022] In the embodiment of the present application, optionally, the current text data input by the target user can be understood as at least one of the following: literal data directly input by the target user, literal data converted from image data and / or voice data input by the user. Further optionally, different current intents correspond to different current interactive processing paths. Specifically, if the current intent is customer service question and answer, such as lease process, delivery time limit, etc., the corresponding retrieval interactive processing path is used; if the current intent is tool name or function name, such as querying order list, querying logistics information, the corresponding tool interactive processing path is used; if the current intent is the name of a work flow, such as canceling an order, urging an audit work flow, the corresponding work flow processing path is used.

[0023] 102. According to the current interactive processing path corresponding to the current intent, the current intent of the target user and the current text data input by the target user are analyzed to obtain a text analysis result corresponding to the current interactive processing path.

[0024] 103. According to the text analysis result corresponding to the current interactive processing path, the current conversation interactive state of the target user and the pre-constructed interactive control data corresponding to the target user, a current conversation result matched with the target user is generated.

[0025] Optionally, the interaction control data corresponding to the target user in the embodiment of the present application includes all context interaction data obtained by the target user interacting with the intelligent AI model for the same object this time. Further optionally, for the same object this time, the context interaction control data is constructed for the entire dialogue interaction process of the target user from the first text data input by the target user for the object, i.e., through the session management unit, until the end of the current intelligent dialogue interaction. The context interaction control data for the current intelligent dialogue interaction is described in detail as follows: according to the dialogue identifier of the target user (such as an order identifier), the historical dialogue interaction data of the target user (including the first text data input by the user and the dialogue result corresponding to each input text data), the current workflow analysis result, the intent analysis result, the text analysis result, the current intent (such as logistics delay consultation), the current dialogue interaction state (such as the need to explain the delay reason and provide a solution), and the current dialogue result are used to construct the context interaction control data. Further, before constructing the context interaction control data, if the formats of the current workflow analysis result, the intent analysis result, and the text analysis result are not uniform, they can be converted into a uniform format, and then the context interaction control data is constructed. It should be noted that if the current dialogue has not ended, steps 101-103 are continued to be executed until the end of the current intelligent interaction for the same object, so as to realize the deep integration of the semantic understanding, multi-modal data processing, and dynamic knowledge reasoning capabilities of the intelligent AI model, construct an intelligent dialogue engine for complex customer service scenarios, and realize a leap from traditional "rule-driven" to "cognitive-driven" upgrade. The same object can be understood as all objects required to be queried in the current intelligent AI model interaction. The intelligent AI model includes all models mentioned in the present application.

[0026] It can be seen that the implementation Figure 1 The described method accurately analyzes the current interaction processing path and the current dialogue interaction state matching the current intent based on the current intent of the user and the current text data input by the user, realizes accurate adjustment of multi-round dialogue path and response logic, accurately analyzes the current intent of the user and the current text data input by the user according to the current interaction processing path corresponding to the current intent, obtains the text analysis result corresponding to the current interaction processing path, and finally performs deep semantic analysis on the text analysis result and the current dialogue interaction state of the user based on the context interaction control data obtained by the user interacting with the intelligent AI model for the same object, i.e., from traditional "rule-driven" to "cognitive-driven" leap, realizes a new generation of AI intelligent dialogue interaction that is efficient, intelligent, and extensible, improves the accuracy and efficiency of intelligent dialogue interaction, and enhances personalized service capabilities, thereby improving the user interaction experience.

[0027] In the embodiments of the present application, optionally, the current intention of the target user is analyzed according to the current text data input by the target user, including: According to the current text data input by the target user, the scene parameters of the current business scene are determined, and the scene parameters of the current business scene include a scene type; According to the scene parameters of the current business scene, a plurality of current intention labels matching the scene parameters of the current business scene are obtained; According to the current text data input by the target user and all current intention labels, the current intention prompt parameters of the current text data input by the target user are generated; The current intention of the target user is obtained by performing intention analysis on the current intention prompt parameters corresponding to the target user based on the pre-trained intention analysis model; In which, according to the current text data input by the target user and all current intention labels, the current intention prompt parameters of the current text data input by the target user are generated, including: According to the current text data input by the target user, the context interaction data matching the current text data is determined from the pre-constructed interaction control data corresponding to the target user; According to the current text data input by the target user, the current intention prompt label matching the current text data is obtained, and the current intention output requirement matching the target user is determined according to the current text data input by the target user; According to the current intention prompt label corresponding to the target user, the current text data input by the target user, all current intention labels, the current intention output requirement corresponding to the target user and the context interaction data matching the current text data, the current intention prompt parameters of the current text data input by the target user are generated.

[0028] In the embodiments of the present application, optionally, the scene type includes but is not limited to one or more of the consultation type scene, the transaction type scene, the after-sales type scene and the complaint type scene. Wherein, the current intention label is determined by the scene parameters of the current business scene. For example, for the consultation type scene and the after-sales type scene, the corresponding current intention label includes but is not limited to one or more of the order status query, the application for return and exchange, the invoice related question, the payment exception processing and the commodity consultation.

[0029] In the embodiments of the present application, optionally, the current intention prompt label can be understood as a pre-determined structured prompt template matching the current text data. Further optionally, different current text data corresponds to different current intention output requirements, such as requiring to directly output the intention name without explanation of the intention name. Further optionally, the intention analysis model can be any model capable of realizing intention analysis, such as LLM model.

[0030] In the embodiment of the present application, further optionally, the context interaction data matched by the current text data can be understood as all context interaction data obtained by the same object interacting with the intelligent AI model, or can be understood as the intent keyword corresponding to the target user in the current text data and the context sentence corresponding to the intent keyword.

[0031] It can be seen that the embodiment of the present application can also analyze the current intent prompt parameters of the intent label of the current business scenario matched with the user and the current text data input by the user, which is helpful for the intent analysis model to analyze and understand the input text in depth, so as to accurately understand the actual intent of the user and improve the accuracy and efficiency of intent recognition. In addition, by fusing the multi-modal input of the user, the context interaction control data, and combining the self-attention mechanism of the intent recognition model to capture the context relationship in the text, the understanding ability under fuzzy expression, complex sentence and cross-domain problem is significantly improved, which can accurately identify the actual intent of the user from the dialogue or current text data. Even in the face of complex context or polysemous words, the real actual intent of the user can be efficiently and accurately analyzed, thereby further improving the accuracy and reliability of intelligent dialogue interaction.

[0032] In the embodiment of the present application, optionally, according to the text analysis result corresponding to the current interaction processing path, the current dialogue interaction state of the target user and the pre-constructed interaction control data corresponding to the target user, a current dialogue result matched with the target user is generated, including: According to the current text data input by the target user, the pre-constructed interaction control data corresponding to the target user and the current dialogue interaction state of the target user, the interaction credibility of the text analysis result corresponding to each current interaction processing path is determined, and according to the interaction credibility corresponding to each current interaction processing path, the current interaction processing path with the highest interaction credibility is selected from all current interaction processing paths as the target current interaction processing path of the target user, and the text analysis result of the target current interaction processing path is determined as the current dialogue result matched with the target user, and the number of current interaction processing paths is greater than or equal to 1; and / or According to the current text data input by the target user, the current parameter corresponding to the target user is determined, and according to the current parameter corresponding to the target user, the retrieval data matched with the target user is determined; according to the current text data input by the target user, the retrieval data corresponding to the target user, the text analysis result corresponding to the current interaction processing path, the current dialogue interaction state of the target user and the pre-constructed interaction control data corresponding to the target user, a current dialogue result matched with the target user is generated; wherein the current parameter corresponding to the target user includes the attribute parameter of the target user and / or the current state parameter of the target object required to be understood by the target user.

[0033] In the embodiment of the present application, optionally, the higher the interaction credibility is, the higher the corresponding current dialogue result credibility is. For example, it is assumed that the user asks: "The logistics shows 'in transit', but the original plan is to deliver today, is it a lost piece?". Among them, the text analysis result obtained by searching the interaction processing path is "recent heavy rain caused the distribution center to be stranded, and it may be delayed for 1-2 days"; the text analysis result obtained by the tool interaction processing path is "the package is stranded due to the distribution center, and is expected to be delivered on June 20, and there is no record of lost pieces"; and the text analysis result obtained by the workflow processing path is "we are verifying the logistics situation, and will process it for you as soon as possible". Among them, based on the current intention of the user to confirm the logistics status, the interaction credibility of the tool interaction processing path, the search interaction processing path and the workflow processing path is obviously decreasing in turn, and the interaction credibility is in turn the first credibility greater than the second credibility greater than the third credibility. At this time, "recent heavy rain caused the distribution center to be stranded, and it may be delayed for 1-2 days" is taken as the current dialogue result of the user, and is output to the user.

[0034] In the embodiment of the present application, optionally, the attribute parameter of the target user includes a user level. Further, it also includes at least one of a user gender, a user age and a user occupation. The current state parameter corresponding to the target object includes a current location, a time of arriving at the current location, a time of sending to a next logistics location and a predicted arrival time. For example, it is assumed that the user asks: "Can't I return the goods I bought?", it is analyzed that the user is a normal user (non-VIP) without VIP permission; and the goods purchased by the user are "customized goods", which are non-returnable goods, and the order does not exceed 7 days, within the returnable time range, and combined with the foregoing return application request of the user, further analysis is performed on these data, and the current dialogue result is obtained as "Hello, the goods you purchased are customized goods, and the platform does not support return service according to the regulations, thank you for your understanding."

[0035] It should be noted that the above two current dialogue result determination methods can be cooperative. For example, still assuming that the user asks: "Can I return the goods I bought?", at this time, through the second method, i.e., the retrieval interaction processing path, it is analyzed that the goods purchased by the user are "customized goods" and belong to non-returnable goods, and through the tool interaction processing path, it is analyzed that the order is more than 7 days and cannot be returned. It is preliminarily judged that it cannot be returned. Then through the first method, and through the tool interaction processing path, it is analyzed that the goods are of the customized type, the signing date is June 10, and the interaction credibility is 0.85, through the retrieval interaction processing path, it is analyzed that the customized goods / over 7 days cannot be returned, and the interaction credibility is 0.65, the workflow processing path analyzes that "contact customer service for confirmation", and the interaction credibility is 0.3, and the tool interaction processing path and the retrieval interaction processing path both analyze that it cannot be returned. At this time, the analysis results obtained by the first method and the second method are comprehensively analyzed with the interaction control data, such as customized goods and application for return, to obtain the current dialogue result: the goods purchased by you belong to customized goods, and have exceeded the 7-day no-reason return period (the signing date is June 10), according to the platform policy, this kind of situation does not support the return service, thank you.

[0036] It can be seen that the embodiment of the present application can also analyze the current dialogue result through the interaction credibility method, improve the response speed, analysis controllability and output stability of the intelligent dialogue interaction result analysis; through the retrieval analysis method, the current dialogue result is analyzed, while ensuring the controllability and output stability of the intelligent dialogue interaction result, the multi-source data processing capability, intelligence and flexibility of the intelligent dialogue interaction are improved; and through the above two methods, different sources of information such as the cooperative retrieval mechanism of dynamic knowledge graph and business rule library are fused, combined with the powerful vectorization semantic matching capability of the model, not only the efficient call of structured knowledge is supported, but also massive unstructured documents (such as platform rule files, product instructions, historical orders, etc.) can be real-time semantic retrieval, natural and smooth content that meets the context is generated, more intelligent, context coherent and personalized response output is realized, the accuracy, reliability and flexibility of intelligent dialogue interaction can be further improved, and more complex interaction scenarios can be coped with, which is conducive to further enhancing the personalized service capability, thereby improving the user interaction experience.

[0037] In the embodiment of the present application, according to the current interaction processing path corresponding to the current intention, the current intention of the target user and the current text data input by the target user are analyzed to obtain a text analysis result corresponding to the current interaction processing path, including: When the current interaction processing path corresponding to the current intention includes a retrieval interaction processing path, a vector conversion operation is performed on the current intention of the target user to obtain an intention feature vector of the current intention; Based on the predetermined index relationship, similarity analysis is performed on the intention feature vector of the current intention and the plurality of index segments determined in advance, to obtain the similarity between the intention feature vector of the current intention and each index segment; According to the similarity corresponding to the plurality of index segments, a target index segment with the maximum similarity is screened out from all the index segments, and context text data of the target user is generated according to the target index segment and the current text data input by the target user; Text analysis is performed on the context text data of the target user based on the first text analysis model trained in advance, to obtain a text analysis result corresponding to the target user; The text analysis result corresponding to the current interaction processing path includes the text analysis result corresponding to the target user.

[0038] In the embodiments of the present application, optionally, the search interaction processing path is used to represent that information retrieval is performed on the current intention of the target user and the current text data input by the target user based on an external knowledge source (such as a document, a database or the Internet), and the retrieved information is combined with the first text analysis model to obtain the text analysis result corresponding to the target user.

[0039] In the embodiments of the present application, optionally, before searching, any question that a user may ask and a corresponding answer are indexed and associated to obtain a corresponding index segment, such as using a BM25 and / or vector embedding method for indexing and association. Further optionally, for the same question, there may be multiple answers, which all need to be indexed and associated. Still further optionally, for any index segment, a preset character is used for segmentation identification, such as using “###” or “**” for segmentation identification. Further optionally, for any index segment, the number of characters of the corresponding answer is less than or equal to a preset number of characters, such as 3000.

[0040] In the embodiments of the present application, optionally, the first text analysis model is any model that realizes text analysis, such as a BART model.

[0041] As can be seen, the embodiments of the present application can also first perform vector conversion on the text data input by the user at present, so as to compare and analyze the content with the index segment, improve the accuracy and efficiency of index analysis, and perform context analysis on the most matched index segment analyzed and the current text data input by the user, and finally analyze the analyzed context data based on a text analysis model to generate coherent and relevant text, further improving the result analysis accuracy of the current text data of the user, thereby facilitating further improvement of the accuracy of intelligent dialogue interaction.

[0042] In the embodiment of the present application, optionally, according to the current interaction processing path corresponding to the current intention, the current intention of the target user and the current text data input by the target user are analyzed to obtain a text analysis result corresponding to the current interaction processing path, including: When the current interaction processing path corresponding to the current intention includes a tool interaction processing path, the current text data input by the target user and the current intention of the target user are analyzed based on a second pre-trained text analysis model to obtain an intention parameter of the target user; According to the current intention of the target user, a target interaction tool matching the current intention is determined, and the intention parameter of the target user is analyzed based on the target interaction tool matching the current intention to obtain an intention analysis parameter corresponding to the intention parameter; The intention analysis parameter corresponding to the intention parameter is analyzed based on the second text analysis model to obtain an intention analysis result corresponding to the intention parameter; The text analysis result corresponding to the current interaction processing path includes the intention analysis result corresponding to the intention parameter.

[0043] In the embodiment of the present application, the tool interaction processing path is used to represent a manner of dynamically calling external services or internal function modules according to the current text data of the target user and the context interaction data to determine the intention analysis result. Among them, the corresponding API is set for the corresponding intention to ensure that the necessary credentials and permissions for accessing these APIs are obtained, the functions, input parameters, output formats and possible error codes of the API are understood, and then the current text data analysis is performed through the corresponding API to obtain the corresponding intention analysis result.

[0044] In the embodiment of the present application, optionally, the second text analysis model is any model capable of text analysis, such as an LLM model, which analyzes the current text data and the current intention to obtain a corresponding intention analysis parameter. For example, if user Zhang San inputs "view logistics information of order number 2025061116480001", the LLM will identify the intention parameters: query logistics information, order number 20250615ABCD1234, and select an interaction tool matching the query logistics information to analyze the intention parameters and the current text data input by the user to obtain an intention analysis parameter. Then, the LLM analyzes the intention analysis parameter to obtain an intention analysis result: order number: 20250615ABCD1234; customer name: Zhang San; contact number: 000000000; delivery address: ********; product details: wireless Bluetooth earphones x2; logistics status: in transit.

[0045] As can be seen, the embodiment of the present application can also call an interaction tool according to the corresponding intention to analyze the current text data input by the user and the current intention of the user. Figure 1And analysis is carried out, so as to accurately and efficiently analyze the text data that can be understood by the user, thereby facilitating further improvement of the accuracy and reliability of intelligent dialogue interaction.

[0046] In the embodiment of the application, optionally, according to the current interaction processing path corresponding to the current intention, the current intention of the target user and the current text data input by the target user are analyzed to obtain a text analysis result corresponding to the current interaction processing path, including: When the current interaction processing path corresponding to the current intention includes a workflow processing path, according to the current intention of the target user, a target node matched with the current intention is determined; According to the target node matched with the current intention, the current intention of the target user is subjected to workflow analysis to obtain a workflow analysis result of the current intention; The current intention of the target user is updated to the workflow analysis result of the current intention, and the operation of determining the target node matched with the current intention according to the current intention of the target user and the operation of subjecting the current intention of the target user to workflow analysis according to the target node matched with the current intention to obtain the workflow analysis result of the current intention are re-executed until a target workflow analysis result of the current intention is obtained; The text analysis result corresponding to the current interaction processing path includes the target workflow analysis result of the current intention.

[0047] In the embodiment of the application, optionally, the workflow processing path is used to represent a mode of automatically processing a service request of the target user. The mode includes a plurality of workflows, and each workflow is composed of a plurality of work links, and the work links are connected through nodes for data flow analysis. For any one workflow, there can be at least one next work link for a previous work link. Each work link has a corresponding decision condition as an analysis basis for subsequent data flow direction. Further, the corresponding workflow can be displayed to the target user.

[0048] It can be seen that the embodiment of the application can also analyze the corresponding text data analysis in a workflow manner, so that the analysis of the text data can be orderly cleaned and analyzed according to the preset workflow, the entire life cycle from user input text data to final solution is automatically managed, the execution accuracy and efficiency of the workflow analysis result are improved, and not only the processing efficiency of the workflow is improved, but also the consistency and traceability of the processing quality are ensured.

[0049] In an optional embodiment, the method can further include the following steps: When the current data input by the target user includes voice data, a data enhancement operation is performed on the voice data input by the target user to obtain data-enhanced voice data input by the target user, and an endpoint detection is performed on the voice data input by the target user to obtain target voice data with blank content removed; based on a preset format, format conversion is performed on the target voice data corresponding to the target user to obtain format-converted target voice data, and an acoustic feature of the target voice data is recognized based on a pre-trained voice recognition model; the recognized acoustic feature is analyzed to obtain current text data corresponding to the target voice data, and a correction operation is performed on the current text data based on a pre-trained text correction model to obtain corrected current text data as the current text data input by the target user; wherein the correction operation includes at least one of a wrong word correction operation, a symbol correction operation, a semantic correction operation, and a format correction operation. When the current data input by the target user includes image data, a data enhancement operation is performed on the image data input by the target user to obtain data-enhanced image data corresponding to the target user, and the image data is recognized based on a pre-trained image classification model to obtain the type of the image data; based on a pre-trained target detection model, a region detection is performed on the type of the image data and the image data corresponding to the target user to obtain a key region of the image data, and a recognition operation is performed on the key region of the image data based on a pre-trained text recognition model to obtain target text content of the image data; according to the collected text question data input by the target user and the target text content of the image data, a corresponding dialogue result is generated for the target user as the current text data input by the target user.

[0050] In this optional embodiment, optionally, the text question data input by the target user can include the current text data input by the user, or the text data input after the recognition of the key region.

[0051] In this optional embodiment, for the voice data, optionally, the voice data input by the target user can be voice data in any language, such as Chinese, English, Cantonese, Sichuanese, Northeastern Mandarin, and / or, professional terminology data from any field, such as e-commerce. Further optionally, the data augmentation operation can include any operation capable of achieving voice noise reduction, such as using Wiener filtering to remove background noise (such as current noise, environmental noise, etc.). The endpoint detection method can include any method capable of voice endpoint detection, such as voice activity detection methods, such as short-time energy detection, Gaussian mixture model, etc.; and blank content can be understood as content without speech, such as silent parts, pauses with a duration greater than or equal to a preset duration. The preset format can be any pre-determined format, such as a mono WAV file. Further optionally, the voice recognition model can include any pre-trained model capable of voice recognition, such as the ISS model. It can recognize acoustic features in the target voice data, such as Mel spectrum, without manually generating traditional features such as MFCC to obtain the current text data corresponding to the target voice data. The target speech data can be stored as text data or a timestamped text sequence. Optionally, the text correction model can include any model capable of text correction, such as the BERT model or a model trained based on the BERT model. Optionally, symbol correction can be understood as adding or modifying symbols in the current text data that do not conform to semantics / context, such as changing "When will my order arrive?" to "When will my order arrive?". Semantic correction can be understood as correcting any current text data that does not conform to semantics / context, such as a user's actual speech: "I bought customized headphones (background includes bus announcements), can I return them after 7 days?", the speech recognition model recognizes: "I bought customized headphones, can I return them after 7 days, Mom?", and after text correction, it becomes: "I bought customized headphones, can I return them after 7 days?". Format correction can be understood as conforming to a specified format, and different text information has different format requirements, such as verifying whether the order number conforms to company rules; if not, format correction is performed.

[0052] In the optional embodiment, for the image data, the optional data enhancement operation includes, but is not limited to, at least one of denoising, graying / binarization, size adjustment, contrast enhancement, cropping / rotation. Further optionally, the image classification model can be any model capable of image classification. The type of image data includes, but is not limited to, one or more of the types of ID card, invoice, commodity, etc. Further optionally, the target detection module can be any model that realizes target detection. The key area of the image data includes, but is not limited to, the amount box and / or the two-dimensional code. Further optionally, the text recognition model can be any model capable of text recognition, such as an OCR recognition model.

[0053] It can be seen that the optional embodiment realizes voice purification by enhancing the voice data input by the user to remove noise in the voice, performing endpoint detection to remove blank content, reducing the data volume while ensuring the correctness of the voice data, and converting the format to the required format, thereby facilitating the improvement of voice recognition efficiency and accuracy; and based on the voice recognition model, the text data recognition of the purified voice data is realized, which accurately converts the audio signal into text data; and through the text correction model, the current text data obtained by converting the voice data is corrected in terms of symbols, semantics, format, and misspelled words, which can improve the readability of the text, thereby facilitating the further improvement of the accuracy and reliability of intent analysis and intelligent dialogue interaction. By performing enhancement operations on the image data input by the user, the image data is cleaner and more standardized, which improves the quality of the image data, thereby facilitating the further improvement of the recognition accuracy of the image; and by sequentially performing image classification, target detection, and OCR recognition on the preprocessed image data, the preliminary analysis of the user's intent is finally performed on the recognized text data and the text question data input by the user as the current text data of the user, which realizes preliminary intelligent dialogue interaction, improves the determination accuracy and reliability of the current text data, and facilitates the further improvement of the accuracy and reliability of intent analysis and intelligent dialogue interaction.

[0054] In another optional embodiment, the method can further include the following steps: segmenting the current text data input by the target user into text segments to obtain target text expressing the emotion of the target user; analyzing the target text of the target user to obtain a first emotional feature of the target user; obtaining pre-constructed interaction control data, and analyzing a second emotional feature of the target user according to the interaction control data and the target text; determining a current emotion analysis result of the target user according to the first emotional feature of the target user and the second emotional feature of the target user; According to the text analysis result corresponding to the current interaction processing path, the current dialogue interaction state of the target user, and the pre-constructed interaction control data corresponding to the target user, a current dialogue result matched with the target user is generated, including: According to the text analysis result corresponding to the current interaction processing path, the current dialogue interaction state of the target user, the pre-constructed interaction control data corresponding to the target user, and the current emotion analysis result of the target user, a current dialogue result matched with the target user is generated.

[0055] It can be seen that the optional embodiment performs current emotion analysis on the current text data input by the user and the context interaction control data, improves the analysis accuracy of the current emotion of the user, and combines the analyzed current emotion into the text analysis result, the current dialogue interaction state, and the interaction control data to analyze the current dialogue result together, further improving the analysis accuracy of the current dialogue result, thereby facilitating further improvement of the smoothness and accuracy of intelligent dialogue interaction.

[0056] In yet another optional embodiment, the method can further include the following steps: Analyzing the target text of the target user to obtain the current vocabulary situation for expressing emotion, such as the number of vocabularies and the type of vocabularies; Filtering the context interaction data of the last preset number (such as 2 times) from the interaction control data, and analyzing the context vocabulary situation for expressing emotion, such as the number of context vocabularies and the type of context vocabularies, according to the context interaction data; According to the current vocabulary situation of the target user and the context vocabulary situation of the target user, setting corresponding emotion weights for the first emotion feature of the target user and the second emotion feature of the target user, wherein the stronger the negative emotion feature, the higher the corresponding emotion weight; According to the first emotion feature of the target user and the second emotion feature of the target user, the current emotion analysis result of the target user is determined, including: Performing emotion parameter quantization on the first emotion feature of the target user to obtain a first emotion quantization value of the target user, and generating a first emotion value of the target user according to the first emotion quantization value of the target user and the corresponding emotion weight; Performing emotion parameter quantization on the second emotion feature of the target user to obtain a second emotion quantization value of the target user, and generating a second emotion value of the target user according to the second emotion quantization value of the target user and the corresponding emotion weight; According to the first emotion value of the target user and the second emotion value of the target user, a target emotion value of the target user is generated as the current emotion analysis result of the target user.

[0057] In the optional embodiment, the words for expressing emotions include, but are not limited to, slow, unprofessional, poor, complaint, no longer needed, return, and the like negative words. Among them, for slow, unprofessional, and poor, further divided into very slow / unprofessional / poor, very slow / unprofessional / poor.

[0058] It can be seen that the optional embodiment improves the analysis accuracy and reliability of the current emotion of the user by respectively performing emotion weight analysis on the current text data and the context interaction management data input by the user to combine the corresponding emotion features, thereby further improving the analysis accuracy of the current dialogue result.

[0059] Embodiment two Please refer to Figure 2 , Figure 2 is another flowchart of the processing method for intelligent interaction based on an intelligent AI model disclosed in the embodiments of the present application. The method can be applied to any scenario requiring intelligent dialogue interaction, such as customer service scenarios in the fields of finance, e-commerce, government affairs, etc. The scenario is provided with one of the corresponding intelligent interaction system (local system or cloud system), intelligent dialogue interaction device, and intelligent dialogue interaction server (local server or cloud server), and is used to realize intelligent dialogue interaction according to the method. The intelligent dialogue interaction device includes, but is not limited to, a mobile phone, a computer, a smart bracelet, and the like device capable of intelligent dialogue interaction. As shown in Figure 2 the method can include the following operations: 201. Determine the scene parameters of the current business scenario according to the current text data input by the target user, wherein the scene parameters of the current business scenario include the scene type.

[0060] 202. Obtain a plurality of current intent labels matching the scene parameters of the current business scenario according to the scene parameters of the current business scenario.

[0061] 203. Generate the current intent prompt parameters of the current text data input by the target user according to the current text data input by the target user and all current intent labels. 204. Perform intent analysis on the current intent prompt parameters corresponding to the target user based on the pre-trained intent analysis model to obtain the current intent of the target user.

[0062] In the embodiments of the present application, optionally, the scene type includes one or more of the following: a consultation scenario, a transaction scenario, an after-sales scenario, and a complaint type scenario. The current intent label is determined by the scene parameters of the current business scenario. For example, for the consultation scenario and the after-sales scenario, the corresponding current intent labels include, but are not limited to, querying the order status, applying for return or exchange, invoice-related questions, payment exception handling, and product consultation.

[0063] Optionally, in the embodiment of the present application, the current intention prompt label can be understood as a preset determined structured prompt template matched with the current text data. Different current text data corresponds to different current intention output requirements, such as requiring to directly output the intention name without explanation of the intention name. Further optionally, the intention analysis model can be any model capable of realizing intention analysis.

[0064] In the embodiment of the present application, further optionally, the context interaction data matched with the current text data here can be understood as all context interaction data obtained by the same object interacting with the intelligent AI model this time, or can be understood as the intention keyword corresponding to the target user in the current text data and the context words and sentences corresponding to the intention keyword.

[0065] It should be noted that the description of other related content can be referred to the description of related content in Embodiment One, which will not be repeated here.

[0066] As can be seen, the embodiment of the present application can analyze the current intention prompt parameters by the intention label matched with the user and the current text data input by the user in the current business scenario matched with the user, which is helpful for the intention analysis model to deeply analyze and understand the input text, so as to accurately understand the actual intention of the user and improve the accuracy and efficiency of intention recognition; and by fusing the multi-modal input of the user, the context interaction control data, and combining the self-attention mechanism of the intention recognition model to capture the context relationship in the text, the actual intention of the user can be accurately recognized from the dialogue or the current text data, even in the face of complex context or polysemous words, the real actual intention of the user can be efficiently and accurately analyzed, thereby being conducive to further improving the accuracy and reliability of intelligent dialogue interaction.

[0067] Embodiment Three Please refer to Figure 3 , Figure 3 is a structural schematic diagram of a processing device for intelligent interaction based on an intelligent AI model disclosed by the embodiment of the present application. The device can be applied to any scene requiring intelligent dialogue interaction, such as customer service scenes in the fields of finance, e-commerce, government affairs, etc. The device can include one of an intelligent interaction system (local system or cloud system), an intelligent dialogue interaction device, and an intelligent dialogue interaction server (local server or cloud server). The intelligent dialogue interaction device includes but is not limited to a mobile phone, a computer, a smart bracelet, etc. capable of intelligent dialogue interaction. As shown in Figure 3 The device includes: An analysis module 301, configured to analyze the current intention of the target user according to the current text data input by the target user. The determining module 302 is configured to determine a current interaction processing path matched with the current intention of the target user and a current dialog interaction state of the target user according to the current intention of the target user and the current text data input by the target user. The analyzing module 301 is further configured to analyze the current intention of the target user and the current text data input by the target user according to the current interaction processing path corresponding to the current intention, to obtain a text analysis result corresponding to the current interaction processing path. The generating module 303 is configured to generate a current dialog result matched with the target user according to the text analysis result corresponding to the current interaction processing path, the current dialog interaction state of the target user and the interaction control data corresponding to the target user which is constructed in advance. The interaction control data corresponding to the target user includes all context interaction data obtained by the target user in this time for interacting with the intelligent AI model for the same object.

[0068] It can be seen that the implementation Figure 3 The described device can accurately analyze the current interaction processing path matched with the current intention and the current dialog interaction state of the target user according to the current intention analysis on the current text data input by the user and based on the current intention of the user and the current text data input by the user, accurately adjust the multi-round dialog path and response logic, accurately analyze the current intention of the user and the current text data input by the user according to the current interaction processing path corresponding to the current intention, obtain the text analysis result corresponding to the current interaction processing path, and finally perform deep semantic analysis on the text analysis result and the current dialog interaction state of the user based on the context interaction control data obtained by the user for interacting with the intelligent AI model for the same object which is constructed in advance, that is, from the traditional 'rule-driven' to 'cognitive-driven' leapfrog upgrade, realize the efficient, intelligent and expandable new generation AI intelligent dialog interaction, improve the accuracy and efficiency of the intelligent dialog interaction, and enhance the personalized service capability, thereby improving the user interaction experience.

[0069] In the embodiment of the application, optionally, the analyzing module 301 analyzes the specific manner of the current intention of the target user according to the current text data input by the target user, which includes: According to the current text data input by the target user, the scene parameters of the current business scene are determined, and the scene parameters of the current business scene include a scene type. According to the scene parameters of the current business scene, a plurality of current intention labels matched with the scene parameters of the current business scene are obtained. According to the current text data input by the target user and all current intention labels, the current intention prompt parameters of the current text data input by the target user are generated. performing intent analysis on the current intent prompt parameter corresponding to the target user according to the pre-trained intent analysis model, to obtain the current intent of the target user; The specific manner in which the analysis module generates the current intent prompt parameter of the current text data input by the target user according to the current text data input by the target user and all current intent labels includes: According to the current text data input by the target user, the context interaction data matched with the current text data is determined from the pre-constructed interaction control data corresponding to the target user; According to the current text data input by the target user, the current intent prompt label matched with the current text data is obtained, and the current intent output requirement matched with the target user is determined according to the current text data input by the target user; According to the current intent prompt label corresponding to the target user, the current text data input by the target user, all current intent labels, the current intent output requirement matched with the target user, and the context interaction data matched with the current text data, the current intent prompt parameter of the current text data input by the target user is generated.

[0070] It can be seen that the embodiment of the present application can also analyze the current intent prompt parameter by analyzing the intent label of the current business scenario matched with the user and the current text data input by the user, which helps the intent analysis model to analyze and understand the input text in depth, so as to accurately understand the actual intent of the user and improve the accuracy and efficiency of intent recognition. In addition, by fusing the multi-modal input of the user, the context interaction control data, and combining the self-attention mechanism of the intent recognition model to capture the context relationship in the text, the actual intent of the user can be accurately recognized from the dialogue or the current text data, even in the face of complex context or polysemous words, the real actual intent of the user can be efficiently and accurately analyzed, thereby further improving the accuracy and reliability of intelligent dialogue interaction.

[0071] In the embodiment of the present application, the specific manner in which the generation module 303 generates the current dialogue result matched with the target user according to the text analysis result corresponding to the current interaction processing path, the current dialogue interaction state of the target user, and the pre-constructed interaction control data corresponding to the target user includes: According to the current text data input by the target user, the pre-constructed interaction management data corresponding to the target user, and the current dialogue interaction state of the target user, the interaction credibility of the text analysis result corresponding to each current interaction processing path is determined, and according to the interaction credibility corresponding to each current interaction processing path, the current interaction processing path with the highest interaction credibility is selected from all current interaction processing paths as the target current interaction processing path of the target user, and the text analysis result of the target current interaction processing path is determined as the current dialogue result matched with the target user, and the number of current interaction processing paths is greater than or equal to 1; and / or According to the current text data input by the target user, the current parameter corresponding to the target user is determined, and according to the current parameter corresponding to the target user, the retrieval data matched with the target user is determined; according to the current text data input by the target user, the retrieval data corresponding to the target user, the text analysis result corresponding to the current interaction processing path, the current dialogue interaction state of the target user, and the pre-constructed interaction management data corresponding to the target user, the current dialogue result matched with the target user is generated; wherein the current parameter corresponding to the target user includes the attribute parameter of the target user and / or the current state parameter of the target object required to be understood by the target user.

[0072] It can be seen that the embodiment of the present application can also analyze the current dialogue result by the interaction credibility mode, improve the response speed, analysis controllability and output stability of the intelligent dialogue interaction result analysis; and by the retrieval analysis mode, the current dialogue result is analyzed, while ensuring the controllability and output stability of the intelligent dialogue interaction result, the multi-source data processing capability, intelligence and flexibility of the intelligent dialogue interaction are improved; and by the above two modes, different sources of information such as the cooperative retrieval mechanism of the dynamic knowledge graph and the business rule library are fused, combined with the powerful vectorization semantic matching capability of the model, not only the efficient call of structured knowledge is supported, but also real-time semantic retrieval of massive unstructured documents (such as platform rule files, product instructions, historical orders, etc.) is performed, natural and smooth content conforming to the context is generated, more intelligent, context coherent and personalized response output is realized, the accuracy, reliability and flexibility of the intelligent dialogue interaction are further improved, and more complex interaction scenarios can be coped with, which is beneficial to further enhancing the personalized service capability, thereby improving the user interaction experience.

[0073] In the embodiment of the present application, optionally, the analysis module 301 analyzes the current intention of the target user and the current text data input by the target user according to the current interaction processing path corresponding to the current intention, to obtain the specific manner of the text analysis result corresponding to the current interaction processing path, including: When the current interaction processing path corresponding to the current intention includes the retrieval interaction processing path, a vector conversion operation is performed on the current intention of the target user to obtain an intention feature vector of the current intention; Based on the predetermined index relationship, similarity analysis is performed on the intention feature vector of the current intention and the plurality of index segments determined in advance to obtain the similarity between the intention feature vector of the current intention and each index segment; According to the similarity corresponding to the plurality of index segments, a target index segment with the maximum similarity is selected from all index segments, and context text data of the target user is generated according to the target index segment and the current text data input by the target user; Text analysis is performed on the context text data of the target user based on the first text analysis model trained in advance to obtain a text analysis result corresponding to the target user; The text analysis result corresponding to the current interaction processing path includes the text analysis result corresponding to the target user.

[0074] It can be seen that the embodiment of the present application can also convert the text data input by the user into a vector to compare and analyze the content with the index segment, improve the accuracy and efficiency of index analysis, analyze the most matched index segment and the current text data input by the user, and finally analyze the analyzed context data based on the text analysis model to generate coherent and relevant text, thereby further improving the result analysis accuracy of the current text data of the user, thereby facilitating further improvement of the accuracy of intelligent dialogue interaction.

[0075] In the embodiment of the present application, optionally, the analysis module 301 analyzes the current intention of the target user and the current text data input by the target user according to the current interaction processing path corresponding to the current intention to obtain a specific manner of the text analysis result corresponding to the current interaction processing path, including: When the current interaction processing path corresponding to the current intention includes the tool interaction processing path, the current text data input by the target user and the current intention of the target user are analyzed based on the second text analysis model trained in advance to obtain an intention parameter of the target user; According to the current intention of the target user, a target interaction tool matched with the current intention is determined, and the intention parameter of the target user is analyzed based on the target interaction tool matched with the current intention to obtain an intention analysis parameter corresponding to the intention parameter; The intention analysis parameter corresponding to the intention parameter is analyzed based on the second text analysis model to obtain an intention analysis result corresponding to the intention parameter; The text analysis result corresponding to the current interaction processing path includes the intention analysis result corresponding to the intention parameter.

[0076] It can be seen that the embodiment of the present application can also call the interaction tool according to the corresponding intention to analyze the current text data input by the user and the current intention of the user Figure 1 and analyze, so as to accurately and efficiently analyze the text data that can be understood by the user, thereby facilitating further improvement of the accuracy and reliability of intelligent dialogue interaction.

[0077] In the embodiment of the present application, optionally, the analysis module 301 analyzes the current intention of the target user and the current text data input by the target user according to the current interaction processing path corresponding to the current intention, and obtains a specific manner of the text analysis result corresponding to the current interaction processing path, including: When the current interaction processing path corresponding to the current intention includes a workflow processing path, according to the current intention of the target user, a target node matched with the current intention is determined; According to the target node matched with the current intention, the current intention of the target user is analyzed by workflow, and a workflow analysis result of the current intention is obtained; The current intention of the target user is updated to the workflow analysis result of the current intention, and the operation of determining the target node matched with the current intention according to the current intention of the target user and analyzing the current intention of the target user by workflow according to the target node matched with the current intention to obtain the workflow analysis result of the current intention is re-executed until the target workflow analysis result of the current intention is obtained; The text analysis result corresponding to the current interaction processing path includes the target workflow analysis result of the current intention.

[0078] It can be seen that the embodiment of the present application can also analyze the corresponding text data analysis by workflow, so that the analysis of the text data can be orderly cleaned and analyzed according to the preset workflow, realizes the automatic management of the entire life cycle from the user input text data to the final solution, improves the execution accuracy and efficiency of the workflow analysis result, and not only improves the processing efficiency of the workflow, but also ensures the consistency and traceability of the processing quality.

[0079] In an optional embodiment, Figure 4 Another structure schematic diagram of the processing device for intelligent interaction based on the intelligent AI model disclosed by the embodiment of the present application is shown in the figure, and the device can also include: Figure 4 ​The data processing module 304 is configured to, when the current data input by the target user includes voice data, perform a data enhancement operation on the voice data input by the target user to obtain voice data input by the target user after data enhancement, perform endpoint detection on the voice data input by the target user to obtain target voice data with blank content removed, perform format conversion on the target voice data corresponding to the target user based on a preset format to obtain target voice data after format conversion, identify acoustic features of the target voice data based on a pre-trained voice recognition model, analyze the identified acoustic features to obtain current text data input by the target voice data, and perform a correction operation on the current text data based on a pre-trained text correction model to obtain corrected current text data as current text data input by the target user; wherein the correction operation includes at least one of a wrong word correction operation, a symbol correction operation, a semantic correction operation, and a format correction operation. The data processing module 304 is configured to, when the current data input by the target user includes image data, perform a data enhancement operation on the image data input by the target user to obtain image data corresponding to the target user after data enhancement, identify the image data based on a pre-trained image classification model to obtain a type of the image data, perform region detection on the type of the image data and the image data corresponding to the target user based on a pre-trained target detection model to obtain a key region of the image data, perform an identification operation on the key region of the image data based on a pre-trained text recognition model to obtain target text content of the image data, and generate a corresponding dialogue result for the target user based on the collected text question data input by the target user and the target text content of the image data as current text data input by the target user.

[0080] It can be seen that the optional embodiment realizes voice purification by enhancing the voice data input by the user to remove noise in the voice, performing endpoint detection to remove blank content, reducing the data volume while ensuring the correctness of the voice data, and converting the format to the required format, thereby facilitating the improvement of voice recognition efficiency and accuracy; and based on the voice recognition model, text data recognition is performed on the purified voice data, thereby realizing accurate conversion of the audio signal into text data; and through the text correction model, the current text data obtained by converting the voice data is corrected in terms of symbols, semantics, format, and misspelled words, thereby improving the readability of the text, thereby facilitating the further improvement of the accuracy and reliability of the intent analysis and intelligent dialogue interaction. By performing enhancement operations on the image data input by the user, the image data is made cleaner and more standardized, thereby improving the quality of the image data, thereby facilitating the further improvement of the image recognition accuracy; and by sequentially performing image classification, target detection, and OCR recognition on the preprocessed image data, and finally performing preliminary analysis of the user intent on the recognized text data and the text question data input by the user as the current text data of the user, preliminary intelligent dialogue interaction is realized, the determination accuracy and reliability of the current text data are improved, thereby facilitating the further improvement of the accuracy and reliability of the intent analysis and intelligent dialogue interaction.

[0081] Embodiment Four Please refer to Figure 5 , Figure 5 is a structural schematic diagram of an intelligent interaction system disclosed by an embodiment of the present application. The intelligent interaction system can be applied to any scene requiring intelligent dialogue interaction, such as customer service scenes in the fields of finance, e-commerce, government affairs, etc. As shown in Figure 5 , the device can include: a memory 401 storing executable program codes; a processor 402 coupled with the memory 401; Further, it can further include an input interface 403 and an output interface 404 coupled with the processor 402; The processor 402 invokes the executable program codes stored in the memory 401, and is configured to execute the steps in the processing method for intelligent interaction based on the intelligent AI model described in Embodiment One or Embodiment Two.

[0082] Embodiment Five The embodiments of the present application disclose a computer storage medium storing computer instructions, which, when invoked, are configured to execute the steps in the processing method for intelligent interaction based on the intelligent AI model described in Embodiment One or Embodiment Two.

[0083] Embodiment Six The embodiment of the present application discloses a computer program product, which comprises a non-transitory computer-readable storage medium storing a computer program, and the computer program is operable to cause a computer to execute steps in the processing method for intelligent interaction based on an intelligent AI model described in embodiment one or embodiment two.

[0084] The device embodiments described above are only schematic, wherein the modules illustrated as separate components can or can not be physically separated, and the components illustrated as modules can or can not be physical modules, i.e., can be located in one place or distributed on multiple network modules. Part or all of the modules can be selected to achieve the purpose of the embodiment scheme according to actual needs. Those skilled in the art can understand and implement it without creative labor.

[0085] Through the specific description of the above embodiments, those skilled in the art can clearly understand that each embodiment can be realized by means of software and the necessary general hardware platform, and of course, it can also be realized by hardware. Based on such understanding, the above technical solutions can be embodied in the form of software products, and the computer software product can be stored in a computer readable storage medium, including a read-only memory (Read-Only Memory, ROM), a random access memory (Random Access Memory, RAM), a programmable read-only memory (Programmable Read-only Memory, PROM), an erasable programmable read-only memory (Erasable Programmable Read Only Memory, EPROM), a one-time programmable read-only memory (One-time Programmable Read-Only Memory, OTPROM), an electrically erasable programmable read-only memory (Electrically-Erasable Programmable Read-Only Memory, EEPROM), a compact disc read-only memory (Compact Disc Read-Only Memory, CD-ROM) or other optical disk storage, a magnetic disk storage, a magnetic tape storage, or any other computer readable medium that can be used to carry or store data.

[0086] Finally, it should be noted that: the processing method, device and system for intelligent interaction based on intelligent AI model disclosed in the embodiments of the present application are only the preferred embodiments of the present application, and are used to illustrate the technical solutions of the present application, but not to limit them; although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand; the technical solutions recorded in the foregoing embodiments can still be modified, or some technical features can be replaced; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A processing method for intelligent interaction based on an intelligent AI model, characterized in that, The method includes: Analyze the target user's current intent based on the current text data input by the target user; Based on the target user's current intent and the target user's current text input, determine the current interaction processing path that matches the current intent and the target user's current dialogue interaction state; Based on the current interaction processing path corresponding to the current intent, analyze the current intent of the target user and the current text data input by the target user to obtain the text analysis result corresponding to the current interaction processing path; Based on the text analysis results corresponding to the current interaction processing path, the current dialogue interaction state of the target user, and the pre-constructed interaction control data corresponding to the target user, a current dialogue result matching the target user is generated. The interaction control data corresponding to the target user includes all contextual interaction data obtained from the target user's interaction with the intelligent AI model on the same object this time.

2. The processing method for intelligent interaction based on an intelligent AI model according to claim 1, characterized in that, The step of analyzing the target user's current intent based on the current text data input by the target user includes: Based on the current text data input by the target user, determine the scenario parameters of the current business scenario, wherein the scenario parameters of the current business scenario include the scenario type; Based on the scenario parameters of the current business scenario, obtain multiple current intent tags that match the scenario parameters of the current business scenario; Based on the current text data input by the target user, determine the contextual interaction data that matches the current text data from the pre-constructed interaction control data corresponding to the target user; Based on the current text data input by the target user, obtain the current intent prompt label that matches the current text data, and determine the current intent output requirement that matches the target user based on the current text data input by the target user; Based on the current intent prompt annotation corresponding to the target user, the current text data input by the target user, all the current intent tags, the current intent output requirements corresponding to the target user, and the context interaction data matching the current text data, the current intent prompt parameters of the current text data input by the target user are generated. Based on the pre-trained intent analysis model, the current intent prompt parameters corresponding to the target user are analyzed to obtain the current intent of the target user.

3. The processing method for intelligent interaction based on an intelligent AI model according to claim 1 or 2, characterized in that, The step of generating a current dialogue result matching the target user based on the text analysis result corresponding to the current interaction processing path, the current dialogue interaction state of the target user, and the pre-constructed interaction control data corresponding to the target user includes: Based on the current text data input by the target user, the pre-constructed interaction control data corresponding to the target user, and the target user's current dialogue interaction state, the interaction credibility of the text analysis results corresponding to each current interaction processing path is determined. Then, based on the interaction credibility of each current interaction processing path, the current interaction processing path with the highest interaction credibility is selected from all current interaction processing paths as the target current interaction processing path for the target user. Finally, the text analysis results of the target current interaction processing path are determined as the current dialogue result matching the target user; and / or Based on the current text data input by the target user, determine the current parameters corresponding to the target user, and based on the current parameters corresponding to the target user, determine the search data matching the target user; based on the current text data input by the target user, the search data corresponding to the target user, the text analysis results corresponding to the current interaction processing path, the current dialogue interaction state of the target user, and the pre-constructed interaction control data corresponding to the target user, generate the current dialogue result matching the target user.

4. The processing method for intelligent interaction based on an intelligent AI model according to claim 1 or 2, characterized in that, The step of analyzing the target user's current intent and the target user's current text data based on the current interaction processing path corresponding to the current intent, and obtaining the text analysis result corresponding to the current interaction processing path, includes: When the current interaction processing path corresponding to the current intent includes a retrieval interaction processing path, a vector transformation operation is performed on the current intent of the target user to obtain the intent feature vector of the current intent; Based on the pre-determined index relationship, a similarity analysis is performed on the intent feature vector of the current intent and multiple pre-determined index segments to obtain the similarity between the intent feature vector of the current intent and each index segment; Based on the similarity of multiple index segments, the target index segment with the highest similarity is selected from all the index segments, and the context text data of the target user is generated based on the target index segment and the current text data input by the target user. Based on the pre-trained first text analysis model, the contextual text data of the target user is analyzed to obtain the text analysis results corresponding to the target user; The text analysis results corresponding to the current interaction processing path include the text analysis results corresponding to the target user.

5. The processing method for intelligent interaction based on an intelligent AI model according to claim 1 or 2, characterized in that, The step of analyzing the target user's current intent and the target user's current text data based on the current interaction processing path corresponding to the current intent, and obtaining the text analysis result corresponding to the current interaction processing path, includes: When the current interaction processing path corresponding to the current intent includes the tool interaction processing path, the current text data input by the target user and the current intent of the target user are analyzed based on the pre-trained second text analysis model to obtain the intent parameters of the target user; Based on the current intent of the target user, a target interaction tool matching the current intent is determined, and the intent parameters of the target user are analyzed based on the target interaction tool matching the current intent to obtain the intent analysis parameters corresponding to the intent parameters; Based on the second text analysis model, the intent analysis parameters corresponding to the intent parameters are analyzed to obtain the intent analysis results corresponding to the intent parameters; The text analysis results corresponding to the current interaction processing path include the intent analysis results corresponding to the intent parameters.

6. The processing method for intelligent interaction based on an intelligent AI model according to claim 1 or 2, characterized in that, The step of analyzing the target user's current intent and the target user's current text data based on the current interaction processing path corresponding to the current intent, and obtaining the text analysis result corresponding to the current interaction processing path, includes: When the current interaction processing path corresponding to the current intent includes a workflow processing path, a target node matching the current intent is determined based on the target user's current intent. Based on the target node that matches the current intent, perform workflow analysis on the target user's current intent to obtain the workflow analysis result of the current intent; The current intent of the target user is updated to the workflow analysis result of the current intent, and the operations of determining the target node matching the current intent based on the current intent, and performing workflow analysis on the current intent of the target user based on the target node matching the current intent to obtain the workflow analysis result of the current intent are re-executed until the target workflow analysis result of the current intent is obtained. The text analysis results corresponding to the current interaction processing path include the target workflow analysis results of the current intent.

7. The processing method for intelligent interaction based on an intelligent AI model according to claim 1 or 2, characterized in that, The method further includes: When the current data input by the target user includes voice data, a data augmentation operation is performed on the voice data input by the target user to obtain data-augmented voice data. Endpoint detection is then performed on the voice data to obtain target voice data with blank content removed. Based on a preset format, the target voice data is converted to obtain format-converted target voice data. Based on a pre-trained speech recognition model, the acoustic features of the target voice data are identified. The identified acoustic features are analyzed to obtain the current text data corresponding to the target voice data. Based on a pre-trained text correction model, a correction operation is performed on the current text data to obtain corrected current text data, which is used as the current text data input by the target user. When the current data input by the target user includes image data, a data augmentation operation is performed on the image data input by the target user to obtain augmented image data. The image data is then identified based on a pre-trained image classification model to determine its type. Based on a pre-trained target detection model, region detection is performed on the type of the image data and the image data corresponding to the target user to obtain the key regions of the image data. Based on a pre-trained text recognition model, a recognition operation is performed on the key regions of the image data to obtain the target text content of the image data. Based on the collected text question data input by the target user and the target text content of the image data, a corresponding dialogue result is generated for the target user as the current text data input by the target user.

8. A processing device for intelligent interaction based on an intelligent AI model, characterized in that, The device includes: The analysis module is used to analyze the current intent of the target user based on the current text data input by the target user; The determination module is used to determine the current interaction processing path matching the current intent and the current dialogue interaction state of the target user based on the target user's current intent and the current text data input by the target user. The analysis module is further configured to analyze the target user's current intent and the target user's current text data based on the current interaction processing path corresponding to the current intent, and obtain the text analysis result corresponding to the current interaction processing path; The generation module is used to generate a current dialogue result that matches the target user based on the text analysis results corresponding to the current interaction processing path, the current dialogue interaction state of the target user, and the pre-built interaction control data corresponding to the target user. The interaction control data corresponding to the target user includes all contextual interaction data obtained from the target user's interaction with the intelligent AI model on the same object this time.

9. An intelligent interactive system, characterized in that, The system includes: Memory containing executable program code; A processor coupled to the memory; The processor calls the executable program code stored in the memory to execute the processing method for intelligent interaction based on the intelligent AI model as described in any one of claims 1-7.

10. A computer storage medium, characterized in that, The computer storage medium stores computer instructions, which, when invoked, are used to execute the processing method for intelligent interaction based on an intelligent AI model as described in any one of claims 1-7.

Images