RAG-based multi-round dialogue processing method and system

Abstract

The invention discloses a multi-round dialogue processing method and system based on RAG, and relates to the technical field of dialogue processing. According to the RAG-based multi-round dialogue processing method and system, important information positioning is carried out on the dialogues input in the multi-section dialogue generation process through dialogue content positioning, so that dialogue generation is always carried out around the important information in the dialogue generation process, and the dialogue generation accuracy is ensured; meanwhile, through the setting of a received dialogue input prediction module, when the dialogue content prediction result to be input does not coincide with the dialogue content change prediction information, a model change instruction is sent to a dialogue content generation module, and when the dialogue content prediction result to be input coincides with the dialogue content change prediction information, a model change instruction is sent to the dialogue content generation module. The model change instruction is not sent to the dialogue content generation module, the use effect of the dialogue generation model can be ensured, and the dialogue generation accuracy is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of dialogue processing, and in particular to a RAG-based multi-round dialogue processing method and system. Background Art

[0002] RAG is a natural language processing model that combines retrieval and generation methods. Plugins are modules used to extend the functions of software or applications. Agents are autonomous and active programs or systems that can perceive the environment and provide various services and functions. RAG's approach: store knowledge in a vector database, search for semantically related knowledge from the vector database during user dialogue, build prompts and use LLM to implement dialogue; Plugins' approach: call LLM to decide what API to call and what parameters to pass in to implement the corresponding business logic; Agent's approach: call LLM one or more times, and determine the calling tool or what to do next based on the results until the final result is obtained; However, they are all single-point solutions at the module level, and do not provide complete solutions for multi-tenancy, multi-language and multi-modality; RAG's approach can only reduce some LLM model hallucinations (random answers); Different e-commerce companies and different scenarios have different business processing, and it is necessary to consider deep customization for different business scenarios. Since the enterprise customers of the SAAS platform have various industries and different understanding and processing of business, the application of LLM needs to adapt to different companies and business scenarios.

[0003] Currently, existing technologies lack complete solutions for multi-tenancy, multi-language and multi-modality, and are unable to achieve deep customization for different business scenarios, resulting in cumbersome conversations, low conversation generation accuracy, low processing efficiency, and serious impact on user experience. Summary of the invention

[0004] The purpose of the present invention is to provide a multi-round dialogue processing method and system based on RAG to solve the above-mentioned deficiencies in the prior art.

[0005] In order to achieve the above-mentioned object, the present invention provides the following technical solutions: a multi-round dialogue processing method and system based on RAG, S1, receiving input dialogue content through a content extraction module, and extracting key dialogue data from the input dialogue content and the response dialogue content based on a dialogue information extraction model; S2, receiving the conversation content semantic information and the conversation content emotion level information in S1, and determining the important information location of the conversation content to be input based on the important information content information of the conversation content; S3, sending the important information generated in step S2 to step S6; S4, receiving the conversation content semantic information, conversation content sentiment level information, conversation content important information content information, conversation content to be generated information, and conversation generation work progress information sent by the conversation content extraction module, and judging the change of the conversation content based on the conversation content prediction model, if the conversation content has not changed, executing step S5, if the conversation content has changed, returning to step S4 to re-determine the important information of the input conversation; S5, receiving the important information content information of the conversation content, the information of the conversation content to be generated, and the progress information of the conversation generation work sent by the conversation content extraction module, and predicting the conversation content to be input by the conversation prediction model, and sending the prediction result of the conversation content to be input to step S6; S6, the dialogue output module is also used to receive the dialogue content to be input predicted by the dialogue input prediction module, and judge whether it is necessary to send a model change instruction to the dialogue content generation module through a matching judgment model according to the prediction result of the dialogue content to be input and the dialogue content change prediction information, wherein when the prediction result of the dialogue content to be input and the dialogue content change prediction information do not overlap, the judgment result is yes, that is, step S7 is executed, and when the prediction result of the dialogue content to be input and the dialogue content change prediction information overlap, the judgment result is no, that is, step S8 is executed; S7, sending the modified dialogue generation model to the dialogue content generation module, and executing step S8; S8, based on the input dialogue content, intelligently generates the response dialogue content through the dialogue generation model.

[0006] The multi-round dialogue processing method and system based on RAG include a dialogue content generation module, a dialogue content extraction module, a dialogue content positioning module, a dialogue output module, and a dialogue input prediction module: The dialogue content generation module is used to intelligently generate response dialogue content based on the input dialogue content and through the dialogue generation model, and send the input dialogue content and the response dialogue content to the dialogue content extraction module; The dialogue content extraction module is used to receive the input dialogue content and the response dialogue content input by the dialogue content generation module, and extract key dialogue data from the input dialogue content and the response dialogue content based on the dialogue information extraction model. The dialogue content extraction module is also used to send the extracted key dialogue data, and the key dialogue data includes dialogue content semantic information, dialogue content emotion level information, dialogue content important information content information, dialogue content to be generated information, and dialogue generation work progress information; The dialogue content positioning module is used to receive the dialogue content semantic information and dialogue content emotion level information sent by the dialogue content extraction module, and based on the dialogue content important information content information, determine the important information positioning of the dialogue content to be input, and send the input dialogue important information to the dialogue input prediction module; The dialogue output module is used to receive the dialogue content semantic information, dialogue content emotion level information, dialogue content important information content information, dialogue content to be generated information, and dialogue generation work progress information sent by the dialogue content extraction module, and judge the dialogue content change prediction information based on the dialogue content prediction model. When the dialogue content change prediction information coincides with the input dialogue content, the dialogue generation information is sent to the dialogue content generation module for dialogue content generation; The dialogue input prediction module is used to receive the important information content information of the dialogue content, the dialogue content to be generated, and the dialogue generation work progress information sent by the dialogue content extraction module, and predict the dialogue content to be input through the dialogue prediction model, and send the prediction result of the dialogue content to be input to the dialogue output module.

[0007] Furthermore, the dialogue output module is also used to receive the dialogue content to be input predicted by the dialogue input prediction module, and judge whether it is necessary to send a model change instruction to the dialogue content generation module through a matching judgment model based on the prediction result of the dialogue content to be input and the dialogue content change prediction information, wherein when the prediction result of the dialogue content to be input and the dialogue content change prediction information do not overlap, the judgment result is yes, that is, the model change instruction is sent to the dialogue content generation module, and when the prediction result of the dialogue content to be input and the dialogue content change prediction information overlap, the judgment result is no, that is, the model change instruction is not sent to the dialogue content generation module.

[0008] Furthermore, the dialogue input prediction module includes: A data receiving module, the data receiving module is used to receive the progress information of the dialogue generation work and the important information location of the dialogue content in the dialogue content location module; A data judgment module, wherein the data judgment module is used to judge whether the staff needs to change the conversation content according to the conversation progress information. If the judgment is yes, the conversation content change result is sent to the conversation determination module. If the judgment is no, the conversation content change result is sent to the conversation determination module after the staff completes one conversation. The dialogue determination module is used to determine the dialogue content to be input based on the dialogue content replacement information and dialogue generation work progress information sent by the data judgment module.

[0009] Furthermore, the dialogue output module includes: An emotion level classification module, the emotion level classification module is used to receive dialogue information and classify the dialogue content information into emotion levels, wherein the specific standard of the level classification is that when the dialogue content change prediction information coincides with the input dialogue content, it is first-level generated information, and when the dialogue content change prediction information does not coincide with the input dialogue content, it is second-level generated information; A first-level dialogue generation module, wherein the first-level alarm module is used to receive the first-level generation information sent by the signal level classification module, and the dialogue content generation module generates dialogue according to a predetermined dialogue generation model; The secondary dialogue generation module, the secondary alarm module is used to receive the secondary generation information sent by the signal level classification module, send a model change instruction to the dialogue content generation module, and the dialogue content generation module generates dialogue according to the changed dialogue generation model.

[0010] Furthermore, the dialogue input prediction module includes: A dialogue process prediction module, the work process prediction module is used to receive the dialogue content to be generated information and the dialogue generation work progress information sent by the environment data acquisition module, and determine the completion time of the current dialogue generation progress based on the dialogue process prediction model, so as to obtain the time when the dialogue changes; The motion trajectory simulation module is used to determine the dialogue content to be input through the dialogue logic simulation model based on the time of dialogue content change and the information to be generated of the dialogue content obtained by the dialogue process prediction module.

[0011] Furthermore, the conversation content locating module locates important information in the conversation content, specifically including the following working steps: Using the semantic information of the conversation content and the emotional level of the conversation content, and through the gaze tracking equation, the coordinates of the conversation content landing point are determined as the key points of the conversation content area. The formula is as follows;

[0012] Among them, P i ’ (x, y) is the current conversation content location, H i (x, y) is the position of the current dialogue in the entire dialogue, i∈[1,9], and the coordinate relationship between the current dialogue and the emotion level in the entire dialogue is used to calculate the parameters of the equation, i.e., a, b, c, ..., l; The standard physical model of the decrease in semantic sharpness of peripheral dialogue content is used to determine the size of the dialogue emotion change area, that is, the dialogue emotion change area Pr is 5°, so that the central emotion level size radius corresponding to the entire dialogue can be calculated, and the formula is as follows: , Where R is the radius of the central emotional region of any paragraph, and θ(x, y) is the minimum angular displacement between adjacent emotional key sentences. The formula is as follows: , in , where x t and t Represents the semantic information of the conversation content and the emotional level of the conversation content, X t and Y t They represent the semantic information of the current dialogue content in a paragraph and the emotional level of the dialogue content respectively. ɑ and ω represent the ratio of the horizontal relationship and the vertical relationship between the dialogue paragraph and the full text respectively. Dis is the simulated distance between the dialogue paragraph and the full text, and it is a fixed value.

[0013] Compared with the prior art, the RAG-based multi-round dialogue processing method and system provided by the present invention can locate important information of the dialogue input in the process of generating multiple dialogues by locating the dialogue content, so as to ensure that the dialogue generation is always carried out around important information in the process of dialogue generation, thereby ensuring the accuracy of dialogue generation. At the same time, by receiving the dialogue content to be input predicted by the dialogue input prediction module, and according to the prediction result of the dialogue content to be input and the prediction information of the dialogue content change, a matching judgment model is used to judge whether it is necessary to send a model change instruction to the dialogue content generation module. When the prediction result of the dialogue content to be input and the prediction information of the dialogue content change do not overlap, a model change instruction is sent to the dialogue content generation module. When the prediction result of the dialogue content to be input and the prediction information of the dialogue content change overlap, the model change instruction is not sent to the dialogue content generation module, thereby ensuring the use effect of the dialogue generation model and increasing the accuracy of dialogue generation. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the present invention. For ordinary technicians in this field, other drawings can also be obtained based on these drawings.

[0015] Figure 1 A system structure block diagram provided for an embodiment of the present invention. DETAILED DESCRIPTION

[0016] In order to enable those skilled in the art to better understand the technical solution of the present invention, the present invention will be further described in detail below with reference to the accompanying drawings.

[0017] In the description of the present invention, it should be understood that the terms "center", "longitudinal", "lateral", "length", "width", "thickness", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside", "clockwise", "counterclockwise" and the like indicate orientations or positional relationships based on the orientations or positional relationships shown in the accompanying drawings, and are only for the convenience of describing the present invention and simplifying the description, and do not indicate or imply that the referred device or element must have a specific orientation, be constructed and operated in a specific orientation, and therefore should not be understood as limiting the present invention.

[0018] In addition, the terms "first" and "second" are only used for descriptive purposes and cannot be understood as indicating or implying relative importance or implicitly indicating the number of technical features indicated. Therefore, the features defined as "first" and "second" may explicitly or implicitly include one or more of the features. In the description of the present invention, "multiple" means two or more, unless otherwise clearly and specifically defined. In addition, the terms "installed", "connected" and "connected" should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection, or it can be indirectly connected through an intermediate medium, or it can be the internal connection of two components. For those of ordinary skill in the art, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0019] Example embodiments will be described more fully below with reference to the accompanying drawings, but the example embodiments may be embodied in different forms and should not be construed as limited to the embodiments set forth herein. On the contrary, the purpose of providing these embodiments is to make the present disclosure thorough and complete and to enable those skilled in the art to fully understand the scope of the present disclosure.

[0020] In the absence of conflict, the various embodiments of the present disclosure and the various features therein may be combined with each other.

[0021] As used herein, the term "and / or" includes any and all combinations of one or more of the associated listed items.

[0022] The terms used herein are only used to describe specific embodiments and are not intended to limit the present disclosure. As used herein, the singular forms "a", "an" and "the" are also intended to include the plural forms, unless the context clearly indicates otherwise. It will also be understood that when the terms "comprising" and / or "made of" are used in this specification, the presence of the features, wholes, steps, operations, elements and / or components is specified, but the presence or addition of one or more other features, wholes, steps, operations, elements, components and / or groups thereof is not excluded.

[0023] The embodiments described herein may be described with reference to plan views and / or cross-sectional views by means of idealized schematic diagrams of the present disclosure. Therefore, the example illustrations may be modified according to manufacturing techniques and / or tolerances. Therefore, the embodiments are not limited to the embodiments shown in the accompanying drawings, but include modifications of the configurations formed based on the manufacturing process. Therefore, the regions illustrated in the accompanying drawings have schematic properties, and the shapes of the regions shown in the figures illustrate the specific shapes of the regions of the elements, but are not intended to be limiting.

[0024] Unless otherwise defined, all terms (including technical and scientific terms) used herein have the same meaning as commonly understood by those of ordinary skill in the art. It will also be understood that terms such as those defined in commonly used dictionaries should be interpreted as having a meaning consistent with their meaning in the context of the relevant art and the present disclosure, and will not be interpreted as having an idealized or overly formal meaning unless explicitly defined as such herein.

[0025] See also Figure 1 ,Multi-round dialogue processing method based on RAG,S1,receives input dialogue content through the content extraction module, and extracts key dialogue data from the input dialogue content and the answer dialogue content based on the dialogue information extraction model; S2, receiving the conversation content semantic information and the conversation content emotion level information in S1, and determining the important information location of the conversation content to be input based on the important information content information of the conversation content; S3, sending the important information generated in step S2 to step S6; S4, receiving the conversation content semantic information, conversation content sentiment level information, conversation content important information content information, conversation content to be generated information, and conversation generation work progress information sent by the conversation content extraction module, and judging the change of the conversation content based on the conversation content prediction model, if the conversation content has not changed, executing step S5, if the conversation content has changed, returning to step S4 to re-determine the important information of the input conversation; S5, receiving the important information content information of the conversation content, the information of the conversation content to be generated, and the progress information of the conversation generation work sent by the conversation content extraction module, and predicting the conversation content to be input by the conversation prediction model, and sending the prediction result of the conversation content to be input to step S6; S6, the dialogue output module is also used to receive the dialogue content to be input predicted by the dialogue input prediction module, and judge whether it is necessary to send a model change instruction to the dialogue content generation module through a matching judgment model according to the prediction result of the dialogue content to be input and the dialogue content change prediction information, wherein when the prediction result of the dialogue content to be input and the dialogue content change prediction information do not overlap, the judgment result is yes, that is, step S7 is executed, and when the prediction result of the dialogue content to be input and the dialogue content change prediction information overlap, the judgment result is no, that is, step S8 is executed; S7, sending the modified dialogue generation model to the dialogue content generation module, and executing step S8; S8, based on the input dialogue content, intelligently generates the response dialogue content through the dialogue generation model.

[0026] The multi-round dialogue processing method and system based on RAG include a dialogue content generation module, a dialogue content extraction module, a dialogue content positioning module, a dialogue output module, and a dialogue input prediction module: The dialogue content generation module is used to intelligently generate the response dialogue content based on the input dialogue content and through the dialogue generation model, and send the input dialogue content and the response dialogue content to the dialogue content extraction module; The dialogue content extraction module is used to receive the input dialogue content and the response dialogue content input by the dialogue content generation module, and extract key dialogue data from the input dialogue content and the response dialogue content based on the dialogue information extraction model. The dialogue content extraction module is also used to send the extracted key dialogue data, and the key dialogue data includes dialogue content semantic information, dialogue content emotion level information, dialogue content important information content information, dialogue content to be generated information, and dialogue generation work progress information; The dialogue content positioning module is used to receive the dialogue content semantic information and dialogue content emotion level information sent by the dialogue content extraction module, and based on the dialogue content important information content information, determine the important information positioning of the dialogue content to be input, and send the input dialogue important information to the dialogue input prediction module; The dialogue output module is used to receive the dialogue content semantic information, dialogue content sentiment level information, dialogue content important information content information, dialogue content to be generated information, and dialogue generation work progress information sent by the dialogue content extraction module, and judge the dialogue content change prediction information based on the dialogue content prediction model. When the dialogue content change prediction information coincides with the input dialogue content, the dialogue generation information is sent to the dialogue content generation module for dialogue content generation; The dialogue input prediction module is used to receive the important information content of the dialogue content, the information on the dialogue content to be generated, and the progress information of the dialogue generation work sent by the dialogue content extraction module, and predict the dialogue content to be input through the dialogue prediction model, and send the prediction result of the dialogue content to be input to the dialogue output module.

[0027] The dialogue output module is also used to receive the dialogue content to be input predicted by the dialogue input prediction module, and judge whether it is necessary to send a model change instruction to the dialogue content generation module through a matching judgment model based on the prediction result of the dialogue content to be input and the dialogue content change prediction information. When the prediction result of the dialogue content to be input and the dialogue content change prediction information do not overlap, the judgment result is yes, that is, the model change instruction is sent to the dialogue content generation module; when the prediction result of the dialogue content to be input and the dialogue content change prediction information overlap, the judgment result is no, that is, the model change instruction is not sent to the dialogue content generation module.

[0028] The dialogue input prediction module includes: A data receiving module, the data receiving module is used to receive the progress information of the dialogue generation work and the important information positioning of the dialogue content in the dialogue content positioning module; The data judgment module is used to judge whether the staff needs to change the conversation content according to the conversation progress information. If it is judged to be yes, the result of changing the conversation content is sent to the conversation determination module. If it is judged to be no, the result of changing the conversation content is sent to the conversation determination module after the staff completes a conversation. The dialogue determination module is used to determine the dialogue content to be input based on the dialogue content replacement information and dialogue generation work progress information sent by the data judgment module.

[0029] The dialogue output module includes: The emotion level classification module is used to receive the dialogue information and classify the dialogue content information into emotion levels, wherein the specific standard of the level classification is that when the dialogue content change prediction information coincides with the input dialogue content, it is the first-level generated information, and when the dialogue content change prediction information does not coincide with the input dialogue content, it is the second-level generated information; The first-level dialogue generation module, the first-level alarm module is used to receive the first-level generation information sent by the signal level classification module, and the dialogue content generation module generates dialogue according to a predetermined dialogue generation model; The secondary dialogue generation module and the secondary alarm module are used to receive the secondary generation information issued by the signal level classification module, send the model change instruction to the dialogue content generation module, and the dialogue content generation module generates dialogue according to the changed dialogue generation model.

[0030] The dialogue input prediction module includes: The dialogue process prediction module is used to receive the dialogue content to be generated information and the dialogue generation work progress information sent by the environment data acquisition module, and determine the completion time of the current dialogue generation progress based on the dialogue process prediction model, so as to obtain the time when the dialogue changes; The motion trajectory simulation module is used to determine the dialogue content to be input through the dialogue logic simulation model based on the time of dialogue content change and the information to be generated of the dialogue content obtained by the dialogue process prediction module.

[0031] The conversation content positioning module locates important information in the conversation content, which specifically includes the following steps: Using the semantic information of the conversation content and the emotional level of the conversation content, and through the gaze tracking equation, the coordinates of the conversation content landing point are determined as the key points of the conversation content area. The formula is as follows;

[0032] Among them, P i ’ (x, y) is the current conversation content location, H i (x, y) is the position of the current dialogue in the entire dialogue, i∈[1,9], and the coordinate relationship between the current dialogue and the emotion level in the entire dialogue is used to calculate the parameters of the equation, i.e., a, b, c, ..., l; The standard physical model of the decrease in semantic sharpness of peripheral dialogue content is used to determine the size of the dialogue emotion change area, that is, the dialogue emotion change area Pr is 5°, so that the central emotion level size radius corresponding to the entire dialogue can be calculated, and the formula is as follows: , Where R is the radius of the central emotional region of any paragraph, and θ(x, y) is the minimum angular displacement between adjacent emotional key sentences. The formula is as follows: , in , where x t and t Represents the semantic information of the conversation content and the emotional level of the conversation content, X t and Y t They represent the semantic information of the current dialogue content in a paragraph and the emotional level of the dialogue content respectively. ɑ and ω represent the ratio of the horizontal relationship and the vertical relationship between the dialogue paragraph and the full text respectively. Dis is the simulated distance between the dialogue paragraph and the full text, and it is a fixed value.

[0033] The above description is only by way of illustration of certain exemplary embodiments of the present invention. It is undoubted that those skilled in the art can modify the described embodiments in various ways without departing from the spirit and scope of the present invention. Therefore, the above drawings and descriptions are illustrative in nature and should not be construed as limiting the scope of protection of the claims of the present invention.

Claims

1. A multi-round dialogue processing method based on RAG, characterized in that: The following steps are included: S1, receiving input conversation content through a content extraction module, and extracting key conversation data from the input conversation content and the response conversation content based on a conversation information extraction model; S2, receiving the conversation content semantic information and the conversation content emotion level information in S1, and determining the important information location of the conversation content to be input based on the important information content information of the conversation content; S3, sending the important information generated in step S2 to step S6; S4, receiving the conversation content semantic information, conversation content sentiment level information, conversation content important information content information, conversation content to be generated information, and conversation generation work progress information sent by the conversation content extraction module, and judging the change of the conversation content based on the conversation content prediction model, if the conversation content has not changed, executing step S5, if the conversation content has changed, returning to step S4 to re-determine the important information of the input conversation; S5, receiving the important information content information of the conversation content, the information of the conversation content to be generated, and the progress information of the conversation generation work sent by the conversation content extraction module, and predicting the conversation content to be input by the conversation prediction model, and sending the prediction result of the conversation content to be input to step S6; S6, the dialogue output module is also used to receive the dialogue content to be input predicted by the dialogue input prediction module, and judge whether it is necessary to send a model change instruction to the dialogue content generation module through a matching judgment model according to the prediction result of the dialogue content to be input and the dialogue content change prediction information, wherein when the prediction result of the dialogue content to be input and the dialogue content change prediction information do not overlap, the judgment result is yes, that is, step S7 is executed, and when the prediction result of the dialogue content to be input and the dialogue content change prediction information overlap, the judgment result is no, that is, step S8 is executed; S7, sending the modified dialogue generation model to the dialogue content generation module, and executing step S8; S8, based on the input dialogue content, intelligently generates the response dialogue content through the dialogue generation model.

2. A multi-round dialogue processing system based on RAG, which is applicable to the multi-round dialogue processing method based on RAG according to claim 1, characterized in that: Including dialogue content generation module, dialogue content extraction module, dialogue content positioning module, dialogue output module, dialogue input prediction module: The dialogue content generation module is used to intelligently generate response dialogue content based on the input dialogue content and through the dialogue generation model, and send the input dialogue content and the response dialogue content to the dialogue content extraction module; The dialogue content extraction module is used to receive the input dialogue content and the response dialogue content input by the dialogue content generation module, and extract key dialogue data from the input dialogue content and the response dialogue content based on the dialogue information extraction model. The dialogue content extraction module is also used to send the extracted key dialogue data, and the key dialogue data includes dialogue content semantic information, dialogue content emotion level information, dialogue content important information content information, dialogue content to be generated information, and dialogue generation work progress information; The dialogue content positioning module is used to receive the dialogue content semantic information and dialogue content emotion level information sent by the dialogue content extraction module, and based on the dialogue content important information content information, determine the important information positioning of the dialogue content to be input, and send the input dialogue important information to the dialogue input prediction module; The dialogue output module is used to receive the dialogue content semantic information, dialogue content emotion level information, dialogue content important information content information, dialogue content to be generated information, and dialogue generation work progress information sent by the dialogue content extraction module, and judge the dialogue content change prediction information based on the dialogue content prediction model. When the dialogue content change prediction information coincides with the input dialogue content, the dialogue generation information is sent to the dialogue content generation module for dialogue content generation; The dialogue input prediction module is used to receive the important information content information of the dialogue content, the dialogue content to be generated, and the dialogue generation work progress information sent by the dialogue content extraction module, and predict the dialogue content to be input through the dialogue prediction model, and send the prediction result of the dialogue content to be input to the dialogue output module.

3. The RAG-based multi-round dialogue processing system according to claim 2, characterized in that: The dialogue output module is also used to receive the dialogue content to be input predicted by the dialogue input prediction module, and judge whether it is necessary to send a model change instruction to the dialogue content generation module through a matching judgment model based on the prediction result of the dialogue content to be input and the dialogue content change prediction information, wherein when the prediction result of the dialogue content to be input and the dialogue content change prediction information do not overlap, the judgment result is yes, that is, the model change instruction is sent to the dialogue content generation module, and when the prediction result of the dialogue content to be input and the dialogue content change prediction information overlap, the judgment result is no, that is, the model change instruction is not sent to the dialogue content generation module.

4. The RAG-based multi-round dialogue processing system according to claim 3, characterized in that: The dialogue input prediction module comprises: A data receiving module, the data receiving module is used to receive the progress information of the dialogue generation work and the important information location of the dialogue content in the dialogue content location module; A data judgment module, wherein the data judgment module is used to judge whether the staff needs to change the conversation content according to the conversation progress information. If the judgment is yes, the conversation content change result is sent to the conversation determination module. If the judgment is no, the conversation content change result is sent to the conversation determination module after the staff completes one conversation. The dialogue determination module is used to determine the dialogue content to be input based on the dialogue content replacement information and dialogue generation work progress information sent by the data judgment module.

5. The RAG-based multi-round dialogue processing system according to claim 4, characterized in that: The dialogue output module comprises: An emotion level classification module, the emotion level classification module is used to receive dialogue information and classify the dialogue content information into emotion levels, wherein the specific standard of the level classification is that when the dialogue content change prediction information coincides with the input dialogue content, it is first-level generated information, and when the dialogue content change prediction information does not coincide with the input dialogue content, it is second-level generated information; A first-level dialogue generation module, wherein the first-level alarm module is used to receive the first-level generation information sent by the signal level classification module, and the dialogue content generation module generates dialogue according to a predetermined dialogue generation model; The secondary dialogue generation module, the secondary alarm module is used to receive the secondary generation information sent by the signal level classification module, send a model change instruction to the dialogue content generation module, and the dialogue content generation module generates dialogue according to the changed dialogue generation model.

6. The RAG-based multi-round dialogue processing system according to claim 5, characterized in that: The dialogue input prediction module comprises: A dialogue process prediction module, the work process prediction module is used to receive the dialogue content to be generated information and the dialogue generation work progress information sent by the environment data acquisition module, and determine the completion time of the current dialogue generation progress based on the dialogue process prediction model, so as to obtain the time when the dialogue changes; The motion trajectory simulation module is used to determine the dialogue content to be input through the dialogue logic simulation model based on the time of dialogue content change and the information to be generated of the dialogue content obtained by the dialogue process prediction module.

7. The RAG-based multi-round dialogue processing system according to claim 6, characterized in that: The conversation content positioning module locates important information in the conversation content, specifically including the following working steps: Using the semantic information of the conversation content and the emotional level of the conversation content, and through the gaze tracking equation, the coordinates of the conversation content landing point are determined as the key points of the conversation content area. The formula is as follows: Among them, P i ’ (x, y) is the current conversation content location, H i (x, y) is the position of the current dialogue in the entire dialogue, i∈[1,9], and the coordinate relationship between the current dialogue and the emotion level in the entire dialogue is used to calculate the parameters of the equation, i.e., a, b, c, ..., l; The standard physical model of the decrease in semantic sharpness of peripheral dialogue content is used to determine the size of the dialogue emotion change area, that is, the dialogue emotion change area Pr is 5°, so that the radius of the central emotion level size corresponding to the entire dialogue can be calculated, and the formula is as follows: ， Where R is the radius of the central emotional region of any paragraph, and θ(x, y) is the minimum angular displacement between adjacent emotional key sentences. The formula is as follows: ， in ， where x t and t Represents the semantic information of the conversation content and the emotional level of the conversation content, X t and Y t They represent the semantic information of the current dialogue content in a paragraph and the emotional level of the dialogue content respectively. ɑ and ω represent the ratio of the horizontal relationship and the vertical relationship between the dialogue paragraph and the full text respectively. Dis is the simulated distance between the dialogue paragraph and the full text, and it is a fixed value.

Images