Dissident communication method and device based on role driving characteristics, equipment and medium

Abstract

The application relates to the technical field of voice semantics, and discloses a role-driven feature-based objection communication method, device, equipment and medium, which comprises the following steps: obtaining objection information, generating structured objection data according to text, object image data and interactive configuration data; generating a communication script and communication audio, extracting emotion features and rhythm features, and generating an audio feature sequence; determining role basic features and generating a role-driven feature sequence, and further generating a communication animation; generating objection communication content according to display content and fusing the objection communication content with the communication animation; and outputting the objection communication content, processing an interactive request, generating interactive processing data and evidence data. The application can be applied to business scenarios such as financial technology and medical health, and can continuously connect the display, communication content, interactive processing and evidence process by associating and processing multi-source information and combining audio feature-driven role animation, so that the consistency and continuity of multi-modal objection communication content are improved.

Description

Technical Field

[0001] This invention relates to the field of speech and semantic technology, and in particular to a method, apparatus, device, and medium for objection communication based on role-driven features. Background Technology

[0002] Existing technologies for communication regarding objections typically employ various information formats such as text, voice, animation, and interactive pages to deliver content. However, these formats lack a unified data association and temporal correspondence mechanism. When objections require explanation based on specific facts and supporting evidence, issues arise such as inconsistencies between the supporting evidence and the objection itself, a lack of synchronization between the emotion and rhythm of the voice and the character's facial expressions and actions, and discontinuities between the displayed evidence and the communication process. This makes it difficult to maintain consistency and continuity in the multimodal content of objection communication. Therefore, existing technologies urgently need to address the lack of unified association and collaborative processing among multimodal content in objection communication.

[0003] In the fintech business, scenarios such as insurance claim objections, denial explanations, policy liability clarifications, and appeal communication typically involve claim information, insurance contract terms, customer information, and subsequent interactions. Existing technologies often generate separate terms explanations, communication voice messages, virtual character animations, and interactive pages. These various types of content lack a continuous connection based on the same claim objection, easily leading to issues such as inaccurate correspondence between insurance terms and specific claim facts, asynchrony between voice expression and character actions, and asynchrony between term display and animated communication. This inconsistency is particularly pronounced when the communication content is lengthy or when continued customer feedback is required.

[0004] In the healthcare sector, scenarios such as explaining medical disputes, communicating medical service controversies, interpreting test results, and notifying users of supplementary materials also require content expression that combines specific facts, medical evidence, and user information. Existing technologies typically handle the presentation of medical evidence, voice explanations, virtual character expressions, and user interaction separately. The lack of continuous data correspondence between these different processing stages easily leads to discrepancies between medical evidence and specific facts, voice emotion and character performance, and the presented content and the interaction process. This makes it difficult to create a consistent and continuous multimodal dispute communication process. Summary of the Invention

[0005] The main objective of this invention is to provide a method, apparatus, device, and storage medium for objection communication based on role-driven characteristics, aiming to solve the technical problem that existing technologies have difficulty in uniformly associating and collaboratively processing multi-source information and multimodal communication content during objection communication, resulting in insufficient consistency and continuity of content during the communication process.

[0006] To achieve the above objectives, the present invention provides a method for objection communication based on role-driven characteristics, comprising: Obtain objection information, supporting text, object profile data, and interaction configuration data for the target matter; verify the correspondence between the objection information and the supporting text; and generate structured objection data. A communication script is generated based on the structured objection data, a communication audio is generated based on the communication script, and emotional and rhythmic features are extracted from the communication audio to generate an audio feature sequence. Based on the object profile data, the basic characteristics of the character are determined, and based on the emotional characteristics in the audio feature sequence and the preset emotional reference feature set, a character-driven feature sequence is generated. Based on the character's basic features, the character's driving feature sequence, the audio feature sequence, and the communication audio, multiple animation clips are generated, and the multiple animation clips are connected to generate a communication animation; Based on the structured objection data and the basis text, the basis display content is generated, and the basis display content is integrated with the communication animation to generate objection communication content; The objection communication content is output based on the interaction configuration data, the interaction request for the objection communication content is received and the interaction is processed, the interaction processing data is generated, and the evidence storage data is generated based on the objection communication content and the interaction processing data.

[0007] Furthermore, to achieve the above objectives, the present invention provides a role-driven objection communication device, comprising: The objection information verification module is used to obtain objection information, basis text, object profile data and interaction configuration data of the target matter, verify the correspondence between the objection information and the basis text, and generate structured objection data. The audio feature generation module is used to generate a communication script based on the structured objection data, generate communication audio based on the communication script, and extract emotional features and rhythm features from the communication audio to generate an audio feature sequence. The character-driven generation module is used to determine the basic characteristics of the character based on the object profile data, and to generate a character-driven feature sequence based on the emotional features in the audio feature sequence and a preset set of emotional reference features. The communication animation generation module is used to generate multiple animation clips based on the character's basic features, the character's driving feature sequence, the audio feature sequence, and the communication audio, and to connect the multiple animation clips to generate a communication animation. The display fusion module is used to generate display content based on the structured objection data and the objection text, and to fuse the display content with the communication animation to generate objection communication content. The interactive evidence storage module is used to output the objection communication content based on the interactive configuration data, receive interactive requests for the objection communication content and perform interactive processing, generate interactive processing data, and generate evidence storage data based on the objection communication content and the interactive processing data.

[0008] Furthermore, to achieve the above objectives, the present invention also provides a computer device, the computer device including a memory, a processor, and a role-driven objection communication program stored in the memory and executable on the processor, wherein when the role-driven objection communication program is executed by the processor, it implements the steps of the role-driven objection communication method as described above.

[0009] Furthermore, to achieve the above objectives, the present invention also provides a computer-readable storage medium storing a role-driven objection communication program, wherein the role-driven objection communication program, when executed by a processor, implements the steps of the role-driven objection communication method described above.

[0010] Beneficial Effects: This invention relates to the field of speech and semantic technology, and discloses a method, apparatus, device, and medium for objection communication based on role-driven features. The method includes: acquiring objection information; generating structured objection data based on text, object profile data, and interaction configuration data; generating a communication script and communication audio; extracting emotional and rhythmic features to generate an audio feature sequence; determining basic role features and generating a role-driven feature sequence, further generating a communication animation; generating content based on the presentation and integrating it with the communication animation to form objection communication content; outputting the objection communication content and processing interaction requests, generating interaction processing data and evidence storage data. This invention can be applied to business scenarios such as fintech and healthcare. By associating multi-source information and combining audio feature-driven role animation, it continuously connects the presentation of information, communication content, interaction processing, and evidence storage processes, thereby improving the consistency and continuity of multimodal objection communication content. Attached Figure Description

[0011] The present invention will be further described below with reference to the accompanying drawings and embodiments. In the accompanying drawings: Figure 1 This is a schematic diagram of an application environment for a role-driven objection communication method according to an embodiment of the present invention; Figure 2 This is a flowchart illustrating an embodiment of the objection communication method based on role-driven characteristics according to the present invention. Figure 3 for Figure 2 A detailed flowchart of step S10; Figure 4 for Figure 2 A detailed flowchart of step S20; Figure 5 for Figure 2 A detailed flowchart of step S30; Figure 6 for Figure 2 A detailed flowchart of step S40; Figure 7 for Figure 2 A detailed flowchart of step S50; Figure 8 for Figure 2 A detailed flowchart of step S60; Figure 9 This is a schematic diagram of the functional modules of a preferred embodiment of the objection communication device based on role-driven features of the present invention. Figure 10 This is a schematic diagram of the structure of a computer device according to an embodiment of the present invention; Figure 11 This is another structural schematic diagram of a computer device according to one embodiment of the present invention. Detailed Implementation

[0012] It should be understood that the specific embodiments described herein are for illustrative purposes only and are not intended to limit the scope of the invention.

[0013] The objection communication method based on role-driven characteristics provided in this invention can be applied to, for example... Figure 1 In this application environment, the client communicates with the server via a network. The server can obtain objection information from the client, generate structured objection data based on text, object profile data, and interaction configuration data; generate communication scripts and communication audio, extract emotional and rhythmic features, and generate audio feature sequences; determine basic character features and generate character-driven feature sequences, further generating communication animations; generate content based on the displayed information and integrate it with the communication animations to form objection communication content; output objection communication content and process interaction requests, generating interaction processing data and evidence storage data. This invention can be applied to business scenarios such as fintech and healthcare. By associating multi-source information and combining audio feature-driven character animations, it continuously connects the display of information, communication content, interaction processing, and evidence storage processes, thereby improving the consistency and continuity of multimodal objection communication content. The client can be, but is not limited to, various personal computers, laptops, smartphones, tablets, and portable wearable devices. The server can be implemented using a standalone server or a server cluster consisting of multiple servers. The invention will be described in detail below through specific embodiments.

[0014] Please see Figure 2 , Figure 2This is a flowchart illustrating an embodiment of the role-driven feature-based objection communication method provided by the present invention. It should be noted that although the logical order is shown in the flowchart, in some cases, the steps shown or described may be performed in a different order than that shown here.

[0015] like Figure 2 As shown, the objection communication method based on role-driven characteristics proposed in this invention includes the following steps: S10: Obtain objection information, basis text, object profile data, and interaction configuration data for the target matter; verify the correspondence between the objection information and the basis text; and generate structured objection data. In this embodiment, data ownership relationships are established around the same target matter. Objection information, basis text, object profile data, and interaction configuration data are retrieved respectively through matter identifier and object identifier. Objection information can come from text input, business records, or speech-to-text results. Based on the text, retrieval is performed according to matter type, version tag, and validity period. Object profile data records object attributes and interaction preferences, and interaction configuration data records available channels, terminal parameters, and interaction methods.

[0016] When verifying the correspondence between objection information and the supporting text, the issue type, factual fields, time information, and content keywords in the objection information are standardized. The supporting text is divided according to title, paragraph, or semantic scope. The correspondence is then determined by combining the consistency of issue type, time validity, version consistency, and content matching degree. Correspondence verification can employ field value comparison, keyword matching, synonym mapping, and text position retrieval, and the matching results, corresponding text positions, and abnormal statuses are recorded.

[0017] Structured objection data organizes the validated data according to a unified field structure, retaining the identification information of the target matter, the objection content, the corresponding basis, the validation status, and the association identifier. For cases where one objection corresponds to multiple supporting text fragments, a one-to-many relationship can be maintained; for cases where multiple objections correspond to the same supporting text fragment, the same supporting identifier can be reused, ensuring consistent attribution for data from different sources.

[0018] When the objection information and the supporting text have fixed fields, a field mapping table is used to unify the field names. The corresponding record is retrieved according to the item identifier, and then the item type, version mark, time range and key fields are compared item by item. This method is suitable for environments with stable data structures and clear text versions.

[0019] When the objection information is free text and the supporting text is long text, text segmentation, keyword extraction, and synonym mapping are used to establish a corresponding relationship. The objection information is split according to sentence and paragraph boundaries, and the supporting text is divided according to title level and paragraph boundaries. Candidate text fragments are filtered by issue keywords, factual terms, and time information, and then the verification is completed by combining version tags.

[0020] In the fintech business, after receiving objections from customers regarding the claim results, the insurance claims platform reads the case records, insurance contract terms, customer profile data, and channel configuration data based on the claim case number. It then performs corresponding verifications based on the type of insurance, insurance period, terms version, and objection content, organizing the case information, objection content, corresponding terms, and verification status into structured objection data.

[0021] In the field of healthcare, after receiving objections from users regarding health assessment results, internet health management platforms read service records, health service basis text, user profile data, and interaction configuration data based on service order identifiers. They then perform corresponding verifications based on service type, service time, content version, and objection content, organizing the service items, objection content, corresponding basis, and verification status into structured objection data.

[0022] This embodiment unifies the association of multi-source data through item identifiers and object identifiers, and verifies the correspondence between objection information and supporting text by combining item type, time, version and content information. This can reduce the mixing of cross-item data and mismatch of supporting text. By storing the verification results and corresponding relationships through a unified field structure, the consistency and accuracy of structured objection data can be improved.

[0023] S20, generate a communication script based on the structured objection data, generate communication audio based on the communication script, and extract emotional features and rhythm features from the communication audio to generate an audio feature sequence; In this embodiment, structured objection data is read according to field relationships, including objection facts, supporting content, processing requirements, and related status information. The field content is deduplicated, sentence-completed, and semantically merged, then converted into continuous text according to the content organization relationships required for communication. The communication script carries the complete language content for object-oriented output. It can be generated by combining preset content templates with structured fields, or by sentence reorganization, referential completion, and tone adjustment. When field content is missing, only existing valid content is organized to avoid supplementing without basis.

[0024] The communication script is converted into communication audio via text-to-speech. During the text-to-speech process, the pronunciation content, pause positions, and audio duration are determined based on text units, punctuation positions, and sentence lengths, and speech data with a continuous timeline is output. The communication audio can be saved as waveform data, encoded audio streams, or time-stamped audio segments.

[0025] This study extracts emotional features reflecting the vocal expression state and rhythmic features reflecting temporal organization from communication audio. Emotional features can be formed by fundamental frequency variations, energy variations, spectral distribution, and sound intensity variations, while rhythmic features can be formed by speech rate, pronunciation duration, pause duration, and stress placement. These features are aligned according to their temporal positions in the audio and organized chronologically into an audio feature sequence, ensuring a continuous correspondence between vocal states at different time points.

[0026] Given that the structured objection data fields are stable and the content types are relatively fixed, a content template and field mapping method is used to generate communication scripts. Different fields are written into corresponding text positions, and continuous text is formed by sentence splicing and tone parameter adjustment. The communication audio is generated using a text-to-speech module. Emotional and rhythmic features are extracted according to fixed time windows, and then organized into an audio feature sequence based on the window time markers.

[0027] When structured objection data exhibits significant variations in length or complex sentence organization, semantic segmentation and text recombination are employed to generate communication scripts. Similar information is merged based on content relevance, and sentence length is controlled according to text boundaries. After the communication audio is generated, analysis intervals are divided according to phoneme boundaries, word boundaries, or pause positions. Emotional and rhythmic features are extracted within each analysis interval, and then audio feature sequences are formed based on corresponding time positions.

[0028] In the fintech business, insurance claims platforms generate claims dispute communication scripts based on the claims facts, insurance terms, and processing requirements in structured dispute data, and convert them into audio. The system extracts tone variations, vocal energy, speech rate, and pauses from the audio, forming an audio feature sequence in chronological order to accurately record the vocal expression states corresponding to different communication content.

[0029] In the healthcare sector, internet-based health management platforms generate communication scripts based on service items, health information descriptions, and feedback content from structured objection data, and then convert them into audio. The system extracts information such as fundamental frequency, sound intensity, pronunciation duration, and pause duration from the audio to form a continuous audio feature sequence, ensuring that the content of the health service objection is consistent with the corresponding audio expression in time.

[0030] This embodiment converts structured objection data into a continuous communication script and generates communication audio. Then, it extracts emotional and rhythmic features with time correspondence from the communication audio. This ensures that the communication content is consistent with the state of voice expression and that the emotional and rhythmic changes in the audio form a continuous feature representation, thereby improving the consistency and usability of audio information in multimodal communication processing.

[0031] S30, determine the basic characteristics of the role based on the object profile data, and generate a role-driven feature sequence based on the emotional features in the audio feature sequence and a preset emotional reference feature set; In this embodiment, object profile data is provided by data records associated with the target object, used to determine the basic characteristics of the role that do not change instantly with the current communication audio. Object profile data may include content reception preferences, interaction habits, age range, language preferences, and auxiliary presentation needs. After encoding, normalizing, and mapping fields for different types of data, it is converted into role attribute parameters, thereby determining the role's appearance category, expression intensity range, movement range, and overall presentation style, forming the role's basic characteristics. These basic characteristics are stored in the form of parameter vectors, attribute sets, or feature encodings, ensuring that the role's expression corresponding to the same object remains stable.

[0032] Emotional features in the audio feature sequence are arranged according to the temporal position of the communication audio, reflecting the emotion category, intensity, and changing state at different time intervals. A preset emotion reference feature set stores multiple reference emotional states and their corresponding feature representations. This set can be established by feature encoding of pre-collected emotion data or by configuring reference vectors based on preset emotion categories. The emotional features in the audio feature sequence are matched with the preset emotion reference feature set. Through similarity comparison, interval mapping, or category mapping, the corresponding emotion-driving parameters are determined, and then organized according to the original temporal position to form a role-driven feature sequence. The role-driven feature sequence preserves the correspondence between emotional states and temporal positions, allowing role expression parameters to continuously change with emotional changes in the communication audio.

[0033] When object profile data is primarily composed of discrete attributes, an attribute coding table is used to convert language preferences, interaction habits, and presentation preferences into discrete codes. Based on these code combinations, corresponding parameters are retrieved from a character parameter library to form the character's basic characteristics. Emotional features are represented using category labels and intensity values. An emotion reference feature set is pre-defined through an emotion category index, and the corresponding driving parameters are adjusted based on the intensity values.

[0034] When the object profile data is mainly composed of continuous vectors, feature normalization and vector mapping are used to transform data of different dimensions into a unified feature space, and the character attribute parameters are determined based on vector distance. Emotional features are represented by continuous vectors, and the matching results are determined from a preset emotional reference feature set through vector similarity. The corresponding driving parameters are organized according to audio time stamps to form a character driving feature sequence.

[0035] In the fintech business, insurance claims platforms determine the basic characteristics of roles based on language preferences, interaction habits, and presentation preferences in customer profiles. They also match the emotional characteristics in the audio of claims objection communication with a preset set of emotional reference characteristics, forming a role-driven feature sequence according to the audio time position, so that the role's expression state corresponding to different claims communication content is consistent with the audio emotion.

[0036] In the field of healthcare, internet health management platforms determine the basic characteristics of roles based on the content receiving preferences, language preferences, and auxiliary presentation needs in user profiles, and generate role-driven feature sequences based on the emotional changes in audio communication of health service objections, so that the role expression parameters in different time intervals keep in continuous correspondence with the corresponding audio emotions.

[0037] This embodiment determines stable basic characteristics of a role through object profile data and matches the emotional features in the communication audio with a preset set of emotional reference features. This allows the role attributes to correspond to the differences between the object and the role's driving state to change with the changes in the audio's emotional state, thereby improving the consistency between the role's expression and the object's needs and the communication audio.

[0038] S40, based on the character's basic features, the character's driving feature sequence, the audio feature sequence, and the communication audio, generate multiple animation clips, and connect the multiple animation clips to generate a communication animation; In this embodiment, the character's basic features are used to maintain the consistency of the character's static attributes across various animation segments, and may include character appearance parameters, facial structure parameters, skeletal structure parameters, and basic posture parameters. The character-driven feature sequence records changes in the character's state according to time position, and may include facial expression amplitude, head posture, limb posture, and local motion parameters. The audio feature sequence provides emotional and rhythmic changes corresponding to the time position of the communication audio, which provides the complete playback duration and sound content. When generating animation segments, the character-driven feature sequence and audio feature sequence are mapped to the timeline of the communication audio. The corresponding character state parameters are called according to the time interval, and frame-by-frame rendering is performed in conjunction with the character's basic features, ensuring that the same character maintains a consistent appearance across different time intervals while simultaneously presenting a dynamic state that changes with the audio.

[0039] Multiple animated segments can be divided according to the duration, sentence boundaries, pause positions, or range of characteristic changes in the communication audio. Each animated segment retains its corresponding time interval, start state, and end state, and maintains a correspondence with the corresponding time range in the communication audio. When connecting segments, the boundary states of adjacent animated segments are processed to ensure continuity in character postures, facial expressions, and screen time. Then, the animated segments are combined according to the chronological order of the communication audio to form the communication animation.

[0040] When the communication audio duration is short and the state changes are relatively gradual, a fixed time window is used to divide the animation segments. The character-driven feature sequence and the audio feature sequence are sampled according to the same time window, and a continuous frame sequence is generated within each time window. Adjacent animation segments are spliced ​​together according to timestamps, and the character state at the boundary is smoothly transitioned to reduce abrupt pose changes.

[0041] When the communication audio is long or the state changes significantly, the segment range is determined by combining sentence boundaries, pause positions, and feature change positions, so that each animation segment covers relatively complete audio content. The basic characteristics of the characters remain unchanged, and the dynamic parameters within the segment are updated according to the character-driven feature sequence and audio feature sequence within the corresponding time range. The transition duration of the segment boundary is adjusted according to the difference in state before and after, and then the combination is completed according to the time sequence of the communication audio.

[0042] In the fintech business, insurance claims platforms divide lengthy audio communication messages about claims disputes into multiple time intervals. Based on the basic characteristics of the customer's role, the platform maintains consistency in the appearance of the role and generates animation clips by combining the role-driven feature sequences and audio feature sequences within each time interval. These clips are then connected in chronological order to form a continuous animation of claims dispute communication.

[0043] In the field of healthcare, internet health management platforms divide audio communication regarding health service objections into multiple time intervals based on sentence boundaries and changes in voice state. Within each time interval, corresponding character animations are generated, and the boundary states of adjacent animation segments are processed for continuity, enabling longer health service communication content to be presented in the form of continuous animation.

[0044] This embodiment maintains consistency in character attributes across multiple animation segments by leveraging basic character features. It also utilizes character-driven feature sequences, audio feature sequences, and communication audio to jointly control the dynamic state across different time intervals, thereby reducing the time discrepancy between character performance and audio content. Furthermore, by continuously connecting multiple animation segments, it reduces action jumps and screen breaks during long-term communication, thus improving the continuity of the communication animation.

[0045] S50, Based on the structured objection data and the basis text, generate basis display content, and integrate the basis display content with the communication animation to generate objection communication content; In this embodiment, the structured objection data stores data fields associated with the current objection and their corresponding relationships, and provides text content that can be displayed based on the text. Based on the association information in the structured objection data, the text range corresponding to the objection content is determined from the text, and the text content is formatted, paragraphed, and configured with display parameters to generate the displayed content. The displayed content may include text, tag information, and layout parameters, and retains the association relationship with the corresponding objection content.

[0046] When integrating evidence-based content with communication animations, the display area and time period of the evidence-based content are configured according to the screen size, duration, and content position of the communication animation. Then, image overlay, area juxtaposition, or layered composition methods are used to create a combined image. The integrated objection communication content simultaneously includes the communication animation and evidence-based content related to the current objection, ensuring that the two types of content maintain correspondence within the same presented data.

[0047] Given a relatively fixed text structure, text indexing and field mapping are used to determine the text range that matches the structured objection data. The corresponding text is then converted into display content according to a preset format and superimposed or arranged side by side according to the screen area of ​​the communication animation.

[0048] When the text is long or the content is scattered, text segmentation and content matching are used to determine the relevant text segments. The display area, font size and page switching method are adjusted according to the segment length. Then, the display time of each text segment is configured according to the duration of the communication animation to complete the image composition.

[0049] In the fintech business, the insurance claims platform uses the information related to claims in the structured objection data to determine the text content related to the current objection from the insurance contract terms, generates the basis display content, and integrates it with the claims objection communication animation to form objection communication content that includes both animation explanation and corresponding terms.

[0050] In the field of healthcare, internet-based health management platforms use service item association information from structured objection data to determine the content related to the current objection from the health service basis text, generate basis display content, and integrate it with health service objection communication animations, so that service descriptions and corresponding basis are presented in the same communication content.

[0051] This embodiment determines the relevant content in the supporting text by using structured objection data and generates supporting content for display, which can reduce the correspondence deviation between the objection content and the displayed supporting evidence. By integrating the supporting content with the communication animation in terms of visuals and time, the consistency between the explanation content and the supporting content during the objection communication process can be improved.

[0052] S60, based on the interaction configuration data, output the objection communication content, receive the interaction request for the objection communication content and perform interaction processing, generate interaction processing data, and generate evidence storage data based on the objection communication content and the interaction processing data.

[0053] In this embodiment, the interaction configuration data is used to control the output method of the objection communication content, and may include available channels, terminal capabilities, page size, audio playback status, subtitle support status, and interaction interface information. Based on the interaction configuration data, output parameters matching the current interaction environment are selected, and the objection communication content is converted into corresponding page data, audio / video data, or combined data, and sent to the corresponding terminal, ensuring that the content format is consistent with the current output environment.

[0054] Interaction requests are initiated by an object in response to previously outputted objection communication content. These requests can include requests for content viewing, information feedback, supplementary submissions, application processing, or status inquiries. Upon receiving an interaction request, the system reads the content identifier, request content, and related information from the request. Based on the request content, it determines the corresponding processing procedure and records the request reception time, processing status, and processing result, forming interaction processing data. This interaction processing data maintains a connection with the objection communication content, ensuring that one or more interactions can be categorized under a specific objection communication process.

[0055] The evidence data is formed by linking and organizing the content of objection communication and interaction processing data. It can save content version, output records, interaction content, processing status, processing results, and time information. During the generation process, the linked data undergoes format uniformity and integrity checks, and is organized according to the data identifier of the same matter, so that the communication content and subsequent interaction records can be saved in the same data object.

[0056] When the type of interactive terminal is relatively fixed, pre-configured output parameters are used to match the terminal capabilities. The content for communication regarding objections is output based on page size, audio playback capabilities, and interactive interface configuration. Interactive requests are received via page controls or interface messages. The corresponding processing flow is invoked according to the request type, and the request content and return result are written to the same interaction record.

[0057] When there are multiple interaction channels, a channel capability table is used to match the output format of the objection communication content, generating suitable content formats for different channels. Interaction requests from different channels are merged using a unified item identifier, and the processing results are written into the interaction processing data in chronological order, which is then combined with the corresponding objection communication content to generate evidence data.

[0058] In the fintech business, the insurance claims platform outputs the content of claims objection communication according to the customer terminal and service channel configuration, receives feedback, supplementary applications or status query requests submitted by customers, records the corresponding processing content and processing results, and associates the communication content and interaction records according to the claims matters to form evidence data.

[0059] In the field of healthcare, internet-based health management platforms output communication content regarding health service objections based on the capabilities of user terminals, receive feedback or review requests from users regarding service descriptions, record the content of requests and processing results, and associate and save relevant communication content and interaction processing records according to health service items.

[0060] This embodiment controls the output format of objection communication content through interactive configuration data, which can reduce the problem of content format mismatch with the interactive environment; by continuously associating interaction requests, processing results and objection communication content and organizing them into evidence data, the consistency and traceability of content records and interaction records during the objection communication process can be improved.

[0061] In one embodiment, such as Figure 3 As shown, step S10 above includes: S101, Obtain objection information for the target matter, extract the matter identifier and object identifier from the objection information, obtain the basis text based on the matter identifier, and obtain object profile data and interaction configuration data based on the object identifier; S102, extract the objection reason information, the factual information of the matter, the information of the submitted materials and the information of subsequent processing requirements from the objection information, and extract multiple basis units and material constraint content from the basis text. Each basis unit includes a corresponding basis identifier, basis content and definition content. S103, Based on the objection reason information, candidate basis units are selected from the multiple basis units, and the factual information of the matter is semantically matched with the definition content in each candidate basis unit to determine the target basis unit; S104, Based on the subsequent processing requirement information, candidate material requirements are selected from the material constraint content, the candidate material requirements are matched with the submitted material information to determine the material requirements to be supplemented, and the objection information and the basis text are verified based on the target basis unit and the material requirements to be supplemented to generate corresponding verification results; S105, based on the corresponding verification results, the matter identifier, the objection reason information, the matter fact information, the submitted material information, the subsequent processing requirement information, the target basis unit, and the supplementary material requirements are organized according to the field correspondence to generate structured objection data.

[0062] In this embodiment, the target item is the data ownership object, and the objection information carries an item identifier and an object identifier. The item identifier is used to retrieve the supporting text that matches the current item, and the search criteria can include item category, text version, and validity status. The object identifier is used to query object profile data and interaction configuration data, enabling data from different sources to be associated with the same item and the same object. When format differences exist, character standardization, encoding conversion, and field mapping can be performed before the search is completed.

[0063] Objection information is extracted according to content type, including the reason for the objection, factual information, submitted materials, and subsequent processing requirements. The basis text is divided into multiple basis units based on heading level, paragraph boundaries, numbering structure, or semantic scope, and material constraints are extracted. Each basis unit stores corresponding basis identifiers, basis content, and definitions, maintaining a correlation between basis location, specific text, and applicable meaning. Material constraints can record material name, material type, submission conditions, and completeness requirements.

[0064] Candidate basis units are selected from multiple basis units based on the information regarding the reasons for objection. Then, the factual information of the matter is semantically matched with the definition content in each candidate basis unit. During the matching process, term normalization and semantic segmentation can be performed on the content of both parties, and the degree of consistency in objects, behaviors, conditions, time, and states can be compared. The target basis unit is determined based on the degree of keyword overlap, synonym mapping results, and semantic similarity. The target basis unit continues to retain the correspondence between the basis identifier, basis content, and definition content.

[0065] Based on subsequent processing requirements, candidate material requirements are selected from the material constraints. These candidate requirements are then matched with submitted material information in terms of name, type, status, and completeness. Material requirements that are not met or only partially met are identified as supplementary material requirements. The target basis unit is used to verify the correspondence between the objection reasons and facts and the basis text. Supplementary material requirements are used to verify the consistency between the submitted material status and the material constraints, and to generate corresponding verification results for record matching status, basis location, material missing status, and exception fields.

[0066] Based on the corresponding verification results, the objection identifier, objection reason information, objection fact information, submitted material information, subsequent processing requirement information, target basis unit, and supplementary material requirements are organized into structured objection data according to preset field relationships. For cases where one objection fact information corresponds to multiple basis units, a one-to-many relationship record can be saved; for cases where multiple objection fact information corresponds to the same target basis unit, the same basis identifier can be reused, thus preserving the correspondence between objection content, objection facts, material status, processing requirements, and basis content.

[0067] This embodiment associates data from different sources by using item identifiers and object identifiers, and breaks down the objection information into content and organizes the supporting text into units. Then, it combines the objection reason information, the factual information of the matter, the information on subsequent processing requirements, and the information on submitted materials to determine the target supporting unit and the requirements for supplementary materials. This can reduce mismatches between objection information and supporting text and deviations in the judgment of material status. By organizing the structured objection data according to the corresponding verification results and using a unified field relationship, the accuracy of the correspondence between objection content, factual information of the matter, supporting content, and material requirements can be improved.

[0068] In one embodiment, such as Figure 4 As shown, step S20 above includes: S201, extract the objection reason information, the factual information of the matter, the target basis unit, the supplementary material requirements and the subsequent processing requirements from the structured objection data; S202, generate emotional expression content based on the objection reason information and the factual information of the matter, generate basis explanation content based on the target basis unit and the factual information of the matter, and generate subsequent guidance content based on the supplementary material requirements and the subsequent processing requirements. Organize the emotional expression content, the basis explanation content and the subsequent guidance content in the order of content to generate a candidate communication script. S203, extract the reference content and material guidance content from the candidate communication script, match the reference content with the target reference unit, and match the material guidance content with the material requirements to be supplemented, modify the candidate communication script based on the matching results, and generate a communication script; S204, Generate communication audio based on the communication script, divide the communication script into multiple script content units, and determine the audio time interval corresponding to each script content unit in the communication audio. S205, the communication audio is divided into multiple audio segments according to the multiple audio time intervals, the emotional features and rhythmic features of each audio segment are extracted, and the corresponding emotional features and rhythmic features are organized according to the time sequence of each audio segment to generate an audio feature sequence.

[0069] In this embodiment, objection reason information, factual information, target basis unit, supplementary material requirements, and subsequent processing needs information are read from structured objection data, while preserving the relationships between each field. The objection reason information reflects the content targeted by the objection; the factual information provides factual data related to the objection judgment; the target basis unit stores the basis identifier, basis content, and definition content; the supplementary material requirements record material conditions that have not yet been met; and the subsequent processing needs information records processing items that need to be completed. During the reading process, null value checks, field format standardization, and duplicate content merging can be performed to prevent the same content from being reused during script generation.

[0070] When organizing emotional expression content based on information about the reasons for objections and the facts of the matter, the content requiring a response is identified from the reasons for objections, and the scope of expression is determined in conjunction with the facts of the matter, ensuring that the generated content responds to the objections without altering the factual data. When generating explanatory content based on target basis units and factual information, semantic connections are established between the facts of the matter and the definitions and basis content in the target basis units, converting content applicable to the current facts into continuous text. When generating follow-up guidance content based on supplementary material requirements and subsequent processing needs, unmet material requirements and matters requiring further processing are extracted and organized into directly communicable explanatory content. The emotional expression content, explanatory content, and follow-up guidance content are arranged and connected according to a preset content order to form a candidate communication script.

[0071] Structured objection data can be input into a Large Language Model (LLM) specifically designed for insurance scenarios to generate communication text. This LLM can be fine-tuned based on insurance contract terms, claims processing guidelines, compliance communication texts, and historical approved texts, enabling the model to identify semantic relationships between objection reasons, factual information, target basis units, required supplementary materials, and follow-up processing needs. The model input can be configured with separate input fields for objection reasons, factual information, basis content, material requirements, and follow-up processing needs, along with content generation constraints. This ensures the output content is organized according to an empathetic, explanatory, and guiding content structure. The empathetic component combines objection reasons and factual information to generate emotional expression content, limiting the generation of additional commitments or information exceeding the facts. The explanatory component combines target basis units and factual information to generate explanatory content, retaining basis identifiers, basis content, and corresponding content related to the facts. The guiding component combines required supplementary materials and follow-up processing needs to generate subsequent guiding content, clarifying the materials needed and matters requiring further processing. Each section can also set compliance constraints to limit vague commitments, broaden interpretations, or add processing content that is not present in the structured objection data, thereby forming candidate communication scripts.

[0072] Identify the supporting evidence cited in the candidate communication scripts, as well as the guiding content for supplementary materials. When matching the cited evidence with the target supporting unit, compare the evidence identifier, key semantics, applicable conditions, and scope of facts. When matching the guiding content for supplementary materials with the required materials, compare the material name, material type, submission conditions, and completeness requirements. If there are inconsistencies in the cited objects, missing content, or the guiding scope exceeds the corresponding data, adjust the relevant text according to the matching results to form a communication script, ensuring that the supporting evidence and material content in the script are consistent with the structured objection data.

[0073] The communication script is processed through text normalization, pronunciation conversion, and speech synthesis to form the communication audio. Text normalization can handle numbers, symbols, and abbreviations, while speech synthesis generates continuous audio data based on the text content. The communication script is divided into multiple script content units according to semantic boundaries, sentence boundaries, or paragraph boundaries, and the start and end times of each script content unit in the communication audio are determined using the time alignment results between the text and audio, forming corresponding audio time intervals.

[0074] The communication audio is divided into time intervals, ensuring that each audio segment is temporally associated with its corresponding script content unit. Emotional characteristics can be formed by changes in fundamental frequency, energy distribution, spectral variation, and sound intensity, while rhythmic characteristics can be formed by speech rate, pronunciation duration, pause duration, and stress placement. After time stamping and dimensional unification, the emotional and rhythmic characteristics of each audio segment are organized according to the chronological order of the audio segments, forming an audio feature sequence that allows for the continuous recording of sound states across different time intervals.

[0075] For example, the formulas for representing the emotional and rhythmic features of an audio segment are:

[0076] in, This represents the audio feature representation corresponding to the i-th audio segment. i represents the time sequence number of the audio segment. The first emotional feature component of the i-th audio segment can represent the degree of empathy. The second emotional feature component of the i-th audio segment can represent the level of professionalism. The rhythmic feature component of the i-th audio segment can be used to represent speech rate. This indicates that the current audio features are represented by a three-dimensional real number vector.

[0077] This embodiment extracts the objection content, facts, target basis, and follow-up processing information from the structured objection data, and matches and corrects the basis references and material guidance content in the candidate communication script. This reduces the content deviation between the communication script and the structured objection data. By establishing the correspondence between the script content units and the communication audio time intervals, and organizing the emotional and rhythmic characteristics of each audio segment in chronological order, the temporal consistency between the communication content and the audio state can be improved.

[0078] In one embodiment, such as Figure 5 As shown, step S30 above includes: S301, extract object attribute data and historical interaction preference data from the object profile data, determine target character features from the preset character feature library based on the object attribute data and the historical interaction preference data, and generate basic character features based on the target character features; S302, for each emotional feature in the audio feature sequence, select multiple target emotional reference features from the preset emotional reference feature set, and generate the emotional fusion weight corresponding to each target emotional reference feature; S303, for each emotion feature, based on the emotion fusion weights corresponding to the emotion features, multiple target emotion reference features corresponding to the emotion features are weighted and fused to generate fused emotion features, and the fused emotion features are mapped to the same feature dimension as the character's basic features to generate character emotion features corresponding to the emotion features; S304, Generate emotional transition features based on the character emotional features corresponding to adjacent emotional features in the audio feature sequence; S305, organize the character's emotional features according to the time order of each emotional feature in the audio feature sequence, and insert corresponding emotional transition features between adjacent character emotional features to generate a character-driven feature sequence.

[0079] In this embodiment, object attribute data and historical interaction preference data are extracted from object profile data. Object attribute data can be represented by discrete attribute values, continuous numerical values, or attribute vectors to describe stable information related to character presentation. Historical interaction preference data can be statistically analyzed based on content selection, presentation mode selection, and interaction feedback from past interaction records to form preference tags or preference weights. A preset character feature library stores multiple candidate character features and their corresponding attribute and preference conditions. Target character features are determined through attribute matching, weight scoring, or vector similarity comparison. Then, the target character features are encoded and their dimensions are unified to generate basic character features, enabling the basic attributes of the character to be represented in a unified data format.

[0080] Each emotional feature in the audio feature sequence reflects the emotional state at the corresponding time position. A preset emotional reference feature set stores the reference feature representations corresponding to different emotional states. For each emotional feature, emotional category, emotional intensity, or feature similarity can be compared to select multiple target emotional reference features from the preset emotional reference feature set, and corresponding emotional fusion weights are generated based on the degree of matching. The emotional fusion weights can be normalized so that multiple target emotional reference features participate in subsequent fusion according to their different contributions.

[0081] For each emotional feature, multiple target emotional reference features are weighted and fused according to emotional fusion weights to obtain a fused emotional feature. When the fused emotional feature is inconsistent with the character's basic feature dimension, it can be transformed to the same feature dimension through a mapping matrix, feature projection, or piecewise mapping to generate a character emotional feature corresponding to the current emotional feature. This ensures that character information and emotional information from different sources have a consistent data dimension and preserves the correspondence between each character's emotional feature and the audio time position.

[0082] Adjacent emotional features correspond to the emotional features of characters at adjacent time points. Emotional transition features are generated based on the feature differences between the preceding and following character emotional features. Interpolation, piecewise gradation, or difference mapping can be used to determine intermediate states, creating a continuous change between adjacent emotional states. The emotional transition features retain the corresponding preceding and following emotional states and their time points, representing the process of change between the emotional features of two characters.

[0083] The emotional features of each character are arranged in chronological order within the audio feature sequence, with corresponding emotional transition features inserted between adjacent character emotional features. The process preserves the correspondence between character emotional features, emotional transition features, and temporal positions, forming a character-driven feature sequence that allows the character's state to be continuously represented according to the emotional changes in the audio.

[0084] For example, the formula for the emotion reference feature vector in the preset emotion reference feature set is:

[0085] in, This represents the k-th emotion reference feature in the preset emotion reference feature set. k represents the number of the emotion reference feature. K represents the number of emotion reference features in the preset emotion reference feature set. This represents the m-th feature component in the k-th emotion reference feature. The feature dimensions that represent the emotional reference features. The emotional reference feature belongs to 3D real number characteristic space.

[0086] The formula for the emotion fusion weight of the target emotion reference feature is:

[0087] in:

[0088] in, This represents the emotion fusion weight corresponding to the first target emotion reference feature generated for the i-th emotion feature. This represents the emotion fusion weight corresponding to the second target emotion reference feature generated for the i-th emotion feature. This represents the first emotional feature component in the i-th audio segment. This represents the second emotional feature component in the i-th audio segment. i represents the audio segment number corresponding to the current emotional feature.

[0089] The weighted fusion formula for multiple target sentiment reference features is as follows:

[0090] in, This represents the fused emotional feature generated for the i-th emotional feature. This represents the emotion fusion weight corresponding to the first target emotion reference feature. This represents the emotion fusion weight corresponding to the second target emotion reference feature. This represents the first target emotion reference feature selected for the i-th emotion feature. This represents the second target emotion reference feature selected for the i-th emotion feature.

[0091] This embodiment determines the basic characteristics of a character by using object attribute data and historical interaction preference data, which can keep the basic attributes of the character corresponding to the object data. By filtering and fusing multiple target emotion reference features for each emotion feature, the continuity of the emotion state representation can be improved. By unifying the feature dimensions and generating emotion transition features between adjacent character emotion features, the abrupt changes caused by direct switching of emotion states can be reduced, thereby improving the consistency and continuity between the character-driven feature sequence and the audio emotion changes.

[0092] In one embodiment, such as Figure 6 As shown, step S40 above includes: S401, Based on the rhythm features in the audio feature sequence, determine the action time nodes and pause time intervals in the communication audio, and generate the character action sequence; S402, the character-driven feature sequence and the character action timing are mapped according to the time sequence of the communication audio to generate animation-driven data; S403, generate multiple animation clips based on the character's basic characteristics, the animation-driven data, and the communication audio; S404, for each group of adjacent animation segments in the plurality of animation segments, extract the ending action state of the previous animation segment and the starting action state of the next animation segment, and extract the emotional transition features corresponding to each group of adjacent animation segments from the character driving feature sequence, and generate a transition frame sequence based on the ending action state, the starting action state and the emotional transition features. S405, the transition frame sequence is inserted between corresponding adjacent animation segments to connect the multiple animation segments and generate a connecting animation.

[0093] In this embodiment, the rhythmic features in the audio feature sequence record changes in speech rate, duration of pronunciation, duration of pauses, and distribution of stress according to the temporal position of the communication audio. Based on the rhythmic changes, suitable time positions for action changes are determined. Sentence stress, speech rate changes, or changes in pronunciation intensity are identified as action time nodes, and continuous low-energy intervals or intervals without effective pronunciation are identified as pause time intervals. Action time nodes are used to mark the moments when character posture, head movements, or localized movements change, and pause time intervals are used to define the duration of action maintenance, deceleration, or stillness. The action time nodes and pause time intervals are organized according to the temporal order of the communication audio to form a character action sequence, giving action changes a clear temporal position.

[0094] The character-driven feature sequence records the changes in a character's emotional characteristics and emotional transitions over time, while the character's action sequence records the changes in action changes and pauses over time. Following a unified timeline for the communication audio, character-driven features within the same time position or overlapping time intervals are mapped to action states. Data with different time resolutions are interpolated, resampled, or aligned with time windows to form animation-driven data. This animation-driven data consists of time stamps, emotional state parameters, and action state parameters, allowing for the simultaneous recall of emotional and action changes at the same time position.

[0095] The character's basic features provide stable parameters such as appearance, facial structure, skeletal structure, and basic posture. Animation-driven data provides emotional and action states that change over time, while communication audio provides the duration and sound content of the animated segments. Segment ranges are divided based on sentence boundaries, pauses, or duration in the communication audio. Within each segment range, the character's basic features remain consistent, and facial expressions, postures, and action parameters are updated based on the animation-driven data to generate multiple animated segments corresponding to the respective communication audio intervals.

[0096] For each pair of adjacent animation segments, the character's posture, joint position, and facial expression are read from the last frame of the preceding animation segment to form the ending action state. The corresponding state is then read from the starting frame of the following animation segment to form the starting action state. Based on the temporal position of the adjacent animation segments in the character-driven feature sequence, corresponding emotional transition features are extracted. The ending action state, starting action state, and emotional transition features are used together to determine the amplitude of action change, the amplitude of facial expression change, and the duration of the transition in the intermediate frames. The transition frame sequence can be generated using linear interpolation, curve interpolation, or piecewise interpolation to ensure that the action state gradually changes from the ending action state to the starting action state, and that the facial expression change is consistent with the emotional transition features.

[0097] Each transition frame sequence is inserted between corresponding adjacent animation segments, and the animation frames and audio time markers are reorganized according to the time sequence of the communicating audio. During the insertion process, the frame rate, duration, and boundary time positions of adjacent segments and transition frame sequences are corrected to avoid duplicate frames, missing frames, or audio-visual time offsets, thus forming a continuous communicating animation.

[0098] For example, the formula for generating emotional attention in animation clips is:

[0099] in, This indicates the output of emotional attention during the animation clip generation process. The denoting element represents the latent video features of the animation generation network in the t-th generation iteration. In an embodiment employing a diffusion-based animation generation network, t can represent the denoising iteration position. This represents the emotional state data corresponding to the i-th time position in the animation-driven data. This represents a projection function that maps latent video features to query features. This represents a projection function that maps emotional state data to key features. This represents a projection function that maps emotional state data to value features. Dimensions representing key features in emotional attention. This indicates the transpose operation. Softmax represents normalization of attention relevance.

[0100] The interpolation formula for the transition frame position between the end action state and the start action state is:

[0101] in, This indicates the coordinates of the target action position in the final action state of the previous animation segment. This indicates the coordinates of the target action position in the initial motion state of the next animation segment. This represents the coordinates of the target action position in the t-th transition frame. N represents the number of transition frames inserted between adjacent animation segments. t represents the sequential number of the transition frame.

[0102] This embodiment determines the action time nodes and pause time intervals through rhythm features, and maps the character-driven feature sequence to the character action sequence according to the time order of the communication audio, which can improve the temporal consistency between emotional changes, action changes and sound rhythm. By combining the character's basic features and animation-driven data to generate multiple animation segments, and using the end action state, start action state and emotional transition features to generate a transition frame sequence, it can reduce abrupt action changes and emotional jumps between adjacent animation segments, thereby improving the continuity and naturalness of the communication animation.

[0103] In one embodiment, such as Figure 7 As shown, step S50 above includes: S501, extract factual information and basis identifiers from the structured objection data, and locate the target text content from the basis text based on the basis identifiers; S502, perform semantic matching between the factual information of the matter and the target text content to determine the location of key text, and generate key annotation content based on the location of key text; S503, generate text display content based on the target text content, generate an interactive interpretation entry based on the highlighted content, and combine the text display content, the highlighted content, and the interactive interpretation entry to generate display content based on the text content; S504, determine the reference content corresponding to the target text content from the communication script, determine the playback time interval corresponding to the reference content in the communication audio, and determine the display time interval of the reference content based on the playback time interval; S505, according to the stated display time interval, synchronize the displayed content with the communication animation in time and merge the images to generate objection communication content.

[0104] In this embodiment, factual information and basis identifiers are read from structured objection data. The factual information may include the object, action, time, status, and result related to the current objection. The basis identifier points to a specific text range within the basis text. Based on the basis identifier, text retrieval conditions are established. Corresponding content can be located within the basis text by entry number, paragraph number, heading level, text index, or character position, and the text within the located range is determined as the target text content. When multiple versions of the basis text exist, the version information stored in the structured objection data can be combined to determine the matching text, avoiding the mixing of different versions.

[0105] When semantically matching factual information with target text content, terminology normalization, semantic segmentation, and key information extraction are performed on both sides, and the degree of correspondence between objects, behaviors, conditions, time, and states is compared. Based on the matching results, the positions of sentences, phrases, or fields with a high degree of relevance to the factual information are determined, and the character range, paragraph position, or text node position is recorded to form key text positions. When generating highlighted content based on key text positions, the annotation range, annotation type, and display priority can be saved to ensure that the text to be highlighted corresponds to the specific position in the target text content.

[0106] The target text content is organized into paragraphs, its hierarchy preserved, and its display format converted to form the displayed text content. Highlighted content is used to configure prominent areas within the text, and interactive interpretation entry points are generated based on the location of the highlighted text and bound to the corresponding text area. These interactive interpretation entry points can be represented by control icons, trigger areas, or event icons, and are positionally associated with the corresponding highlighted content. The displayed text content, highlighted content, and interactive interpretation entry points are combined according to a unified layout data structure to form the displayed content, ensuring that text content, highlighted locations, and interactive objects maintain a correspondence within the same display data.

[0107] The communication script retrieves text segments that are semantically consistent with the target text or have a clear referencing relationship to determine the reference content. Using the time alignment between the communication script and the communication audio, the start and end times corresponding to the reference content are determined, forming a playback time interval. Time alignment can be achieved based on sentence boundaries, word boundaries, phoneme time stamps, or speech synthesis timestamps. The display time interval for the reference content is determined based on the playback time interval, ensuring that the reference content is displayed within the time frame during which the relevant reference content is played in the communication audio.

[0108] The display time range is mapped to the timeline of the communication animation. Within the corresponding time range, the display status based on the displayed content is configured, and the image fusion is completed using image overlay, area juxtaposition, or layered compositing methods. During the fusion process, the time markers, image size, and display area of ​​the communication animation and the displayed content are unified, and the start position, duration, and end position of the display are corrected to form a communication message that simultaneously includes both animation content and the displayed content.

[0109] This embodiment improves the accuracy of the correspondence between disputed facts and presented evidence by locating target text content based on identifiers and determining the location of key text using factual information. By associating and organizing the text display content, highlighted content, and interactive interpretation entry points, and determining the display time interval based on the communication script and audio, the evidence content can be presented within the corresponding communication time range. By merging the evidence display content with the communication animation according to a unified timeline, the temporal and content consistency between the evidence description and the animated communication content can be improved.

[0110] In one embodiment, such as Figure 8 As shown, step S60 above includes: S601, extract output channel data, interaction scenario data and terminal parameter data from the interaction configuration data, generate content output parameters based on the output channel data, the interaction scenario data and the terminal parameter data, and output the objection communication content based on the content output parameters; S602, receive an interaction request for the content of the objection communication, extract the interaction type and interaction content from the interaction request, and generate interaction task data based on the interaction type and the interaction content; S603, determine the target processing node based on the interaction type, establish an association between the interaction task data and the target matter and the objection communication content, and send the interaction task data to the target processing node for interaction processing; S604, receive the processing status data and processing result data returned by the target processing node, and generate interactive processing data based on the interactive task data, the processing status data and the processing result data; S605, Generate a data summary based on the content of the objection communication and the interaction processing data, and organize the content of the objection communication, the interaction processing data and the data summary according to the target matter to generate evidence storage data.

[0111] In this embodiment, output channel data, interaction scenario data, and terminal parameter data are read from the interaction configuration data. Output channel data identifies the web page, mobile device, messaging interface, or other available output channels. Interaction scenario data records the current content presentation environment and interaction state. Terminal parameter data records screen size, media format support capabilities, audio playback capabilities, and interaction interface capabilities. Based on these three types of data, parameters are combined for the output format, screen size, media encoding, playback method, and interaction entry point of the objection communication content to generate content output parameters. The objection communication content is then format-converted and sent according to these parameters, ensuring that the output content matches the current channel, scenario, and terminal capabilities.

[0112] Interaction requests carry request information related to the previously output objection communication content, from which the interaction type and interaction content are extracted. The interaction type can be used to distinguish different request categories such as content viewing, information feedback, material submission, application processing, or status inquiry. The interaction content stores the submitted text, data items, file identifiers, or operation information. Based on the interaction type, the task category and data structure are determined, and the interaction content is written into the corresponding fields to generate interaction task data, thus converting different forms of interaction requests into data objects that can be recognized by subsequent nodes.

[0113] Based on the interaction type, the target processing node is determined in the preset node configuration. The target processing node can consist of a service interface, a business module, or a data processing unit. When establishing a relationship between the interaction task data and the target matter and objection communication content, a matter identifier, a content identifier, and a task identifier can be written, enabling the interaction request to locate the corresponding matter and corresponding communication content. After the association is completed, the interaction task data is sent to the target processing node, and the sending time, receiving status, and task identifier are retained.

[0114] The target processing node returns processing status data and processing result data. The processing status data can record states such as pending, processing, completed, or exception, while the processing result data can record the returned content, result identifier, update time, and exception information. The interactive task data, processing status data, and processing result data are associated according to the task identifier to form interactive processing data, ensuring consistent data ownership across request content, processing status, and processing result.

[0115] When generating a data summary based on the content of objection communication and interaction processing data, the two types of data can be sorted and formatted first, and then a summary generation function can be used to form a fixed-length data summary. The data summary, objection communication content, and interaction processing data are organized and associated with the target matter, and the matter identifier, content version, interaction record, and time information are recorded to generate evidence data. Evidence data can be stored using structured records, serialized data objects, or hierarchical data sets to ensure that the communication content and interaction results under the same target matter remain completely linked.

[0116] This embodiment generates content output parameters by using output channel data, interaction scenario data, and terminal parameter data, which can improve the matching degree between the objection communication content and the output environment. By converting interaction requests into interaction task data and associating them with target matters, objection communication content, and target processing nodes, it can maintain a consistent correspondence between the request source, processing process, and processing result. By generating data summaries and associating objection communication content and interaction processing data according to target matters, it can improve the integrity, consistency, and traceability of communication records and interaction records.

[0117] In one embodiment, an insurance claims platform processes a medical insurance claim denial objection. A customer appeals because facial wrinkle removal surgery was not covered, arguing that the treatment was medically necessary. The platform retrieves the reason for denial, medical records, expense list, submitted materials, and appeal request based on the claim case number. It also retrieves the effective insurance contract terms, exclusions, and appeal material requirements at the time of the incident. Furthermore, it reads the customer's account information, including age, language preference, historical service channels, and content presentation preferences, as well as the current mobile application, terminal resolution, subtitle support, and interface configuration.

[0118] The platform identifies claim case numbers and customer identifiers from case data, breaking down customer objections into reasons for objection, treatment facts, existing materials, and requests for re-review. The insurance contract is divided into multiple basis units according to clause numbers and paragraphs. Clause 5.1 states that cosmetic procedures are not covered and includes a definition of cosmetic procedures. Based on the customer's objection regarding treatment purpose, the platform filters clauses related to cosmetic procedures, treatment necessity, and coverage. It then semantically matches facial wrinkle removal surgery, diagnostic results, and treatment purpose with the definitions of each clause, determining Clause 5.1 as the target basis. Regarding the customer's request for re-review, the platform further verifies medical necessity certificates, medical records, and expense materials, identifying any unsubmitted medical necessity certificates as supplementary materials, and generating structured objection data that includes case facts, target clauses, the status of existing materials, and the appeal request.

[0119] The platform generates customer-facing communication content based on structured objection data. The emotional expression section addresses customer questions regarding the denied claim result. The explanation section clarifies the definition of beauty procedures in Article 5.1 and the correspondence between facial wrinkle removal surgery and the current clause. The follow-up guidance section explains the need for supplementary medical necessity proof during the re-evaluation process. The platform re-verifies the clause numbers, meanings, and material requirements in the candidate scripts to avoid interpretations exceeding the insurance contract's scope or adding unrecorded material requirements, thus forming the final communication script.

[0120] The communication script is converted into gentle, professional audio, and divided into different content sections according to empathy explanations, clause interpretations, and appeal guidance. The platform extracts information on tone changes, vocal energy, speech rate, pauses, and stress from each audio section. For example, it retains obvious pauses before and after sections such as Article 5.1, beauty treatments, and proof of medical necessity, so that the emotional state and rhythm at different times form a continuous audio feature sequence.

[0121] The customer is 58 years old and has primarily checked their claim progress via mobile application over the past three months. Historical preferences show that the customer is more receptive to realistic figures, clear lip movements, and gentle gestures. Based on this, the platform selected realistic and gentle character traits, professional attire, and moderate gestures. For different emotional states in empathic explanations, clause interpretations, and appeal guidance, matching content is selected and integrated from emotional reference features such as empathy, professionalism, and patience. For example, empathic explanations emphasize empathy and patience, while clause interpretations emphasize professionalism and patience; when there are significant emotional shifts between adjacent content, gradual emotional transition features are generated between the preceding and following states, forming a continuous sequence of character-driven features.

[0122] The platform determines the timing and position of nods, gestures, and movements based on the speech rate, emphasis, and pauses in the audio. For example, it reduces the frequency of hand gestures when explaining section 5.1, maintains eye contact when explaining the definition of beauty treatments, and adds subtle guiding gestures when explaining medical necessity. The character's driving state and the timing of the actions are correlated with the same audio timeline, forming animation-driven data, and multiple animation clips are generated by combining the character's appearance and the audio.

[0123] For communication content exceeding the duration of a single segment, the platform reads the hand position, head posture, and facial expression at the end of the previous segment and the corresponding state at the beginning of the next segment, and combines these with emotional changes such as empathy towards professionalism and professionalism towards patience to generate transition frames. These transition frames are inserted between adjacent animation segments, allowing the character's actions and expressions to change continuously, forming a complete communication animation for claims objections.

[0124] The platform locates Article 5.1 of the medical insurance contract based on the case facts and clause identifiers in the structured objection data. It matches case facts such as facial wrinkle removal surgery and treatment purposes with the clause content to determine the location of key texts such as the definition of beauty treatment procedures. The clause display area retains the clause number, title, and related text, highlighting content directly relevant to the current case, and also provides an interactive interpretation entry. When the communication audio plays to Article 5.1 and the definition of beauty treatment procedures, the corresponding content is simultaneously displayed in the right-hand clause display window, forming the objection communication content together with the character animation on the left.

[0125] After viewing the objection communication content on the claims details page of the mobile application, the customer clicks the appeal application entry and submits a request for re-review and supporting medical necessity documents. The platform identifies the interaction type as an appeal application, links the appeal statement, supporting documents, and the current claims case, and sends it to the appeal review node. The appeal review node returns processing statuses such as accepted or pending review, as well as the appeal task number, document receipt results, and estimated processing time. Based on this, the platform generates interaction processing data.

[0126] Communication animations, communication scripts, terms and conditions display records, customer appeal content, material submission records, processing status, and review results are linked according to the same claim case number, and a data summary is generated to form the corresponding case's archival data. When customers subsequently supplement materials, check review progress, or submit new objections, the relevant records continue to be grouped into the same case, ensuring a continuous correspondence between claim objection communication, appeal processing, and process records.

[0127] In one embodiment, a role-driven objection communication device is provided, which corresponds one-to-one with the role-driven objection communication method described in the above embodiments. (Refer to...) Figure 9 , Figure 9 This is a functional module diagram of a preferred embodiment of the objection communication device based on role-driven features of the present invention. The objection communication device based on role-driven features includes: an objection information verification module 10, an audio feature generation module 20, a role-driven generation module 30, a communication animation generation module 40, a data display and fusion module 50, and an interactive evidence storage module 60. Detailed descriptions of each functional module are as follows: The objection information verification module 10 is used to obtain objection information, basis text, object profile data and interaction configuration data of the target matter, verify the correspondence between the objection information and the basis text, and generate structured objection data. The audio feature generation module 20 is used to generate a communication script based on the structured objection data, generate communication audio based on the communication script, and extract emotional features and rhythm features from the communication audio to generate an audio feature sequence. The character-driven generation module 30 is used to determine the basic characteristics of the character based on the object profile data, and to generate a character-driven feature sequence based on the emotional features in the audio feature sequence and a preset set of emotional reference features. The communication animation generation module 40 is used to generate multiple animation segments based on the character's basic features, the character's driving feature sequence, the audio feature sequence, and the communication audio, and to connect the multiple animation segments to generate a communication animation. According to the display fusion module 50, it is used to generate display content based on the structured objection data and the objection text, and merge the display content with the communication animation to generate objection communication content; The interactive evidence storage module 60 is used to output the objection communication content based on the interactive configuration data, receive interactive requests for the objection communication content and perform interactive processing, generate interactive processing data, and generate evidence storage data based on the objection communication content and the interactive processing data.

[0128] In one embodiment, the objection information verification module 10 is specifically used for: Obtain objection information for the target matter, extract the matter identifier and object identifier from the objection information, obtain the supporting text based on the matter identifier, and obtain object profile data and interaction configuration data based on the object identifier; Extract the objection reason information, the factual information of the matter, the information of the submitted materials and the information of subsequent processing requirements from the objection information, and extract multiple basis units and material constraint content from the basis text. Each basis unit includes a corresponding basis identifier, basis content and definition content. Based on the objection reason information, candidate basis units are selected from the multiple basis units, and the factual information of the matter is semantically matched with the definition content in each candidate basis unit to determine the target basis unit; Based on the subsequent processing requirements information, candidate material requirements are selected from the material constraints. The candidate material requirements are matched with the submitted material information to determine the material requirements to be supplemented. Based on the target basis unit and the material requirements to be supplemented, the objection information and the basis text are verified to generate corresponding verification results. Based on the corresponding verification results, the matter identifier, the objection reason information, the matter fact information, the submitted material information, the subsequent processing requirement information, the target basis unit, and the supplementary material requirements are organized according to the field correspondence to generate structured objection data.

[0129] In one embodiment, the audio feature generation module 20 is specifically used for: Extract the objection reason information, factual information, target basis unit, required supplementary materials, and subsequent processing information from the structured objection data; Based on the objection reason information and the factual information of the matter, generate emotional expression content; based on the target basis unit and the factual information of the matter, generate basis explanation content; and based on the supplementary material requirements and the subsequent processing needs information, generate subsequent guidance content. Organize the emotional expression content, the basis explanation content, and the subsequent guidance content in the order of content to generate a candidate communication script. Extract the reference content and material guidance content from the candidate communication script, match the reference content with the target reference unit, and match the material guidance content with the requirements of the supplementary materials. Based on the matching results, modify the candidate communication script and generate a communication script. Based on the communication script, a communication audio is generated. The communication script is divided into multiple script content units, and the audio time interval corresponding to each script content unit in the communication audio is determined. The communication audio is divided into multiple audio segments according to the multiple audio time intervals. The emotional and rhythmic features of each audio segment are extracted, and the corresponding emotional and rhythmic features are organized according to the time sequence of each audio segment to generate an audio feature sequence.

[0130] In one embodiment, the role-driven generation module 30 is specifically used for: Extract object attribute data and historical interaction preference data from the object profile data, determine target character features from a preset character feature library based on the object attribute data and the historical interaction preference data, and generate basic character features based on the target character features; For each emotion feature in the audio feature sequence, multiple target emotion reference features are selected from a preset emotion reference feature set, and emotion fusion weights corresponding to each target emotion reference feature are generated. For each emotion feature, based on the emotion fusion weights corresponding to the emotion features, multiple target emotion reference features corresponding to the emotion features are weighted and fused to generate fused emotion features. The fused emotion features are then mapped to the same feature dimension as the character's basic features to generate character emotion features corresponding to the emotion features. Emotional transition features are generated based on the character emotional features corresponding to adjacent emotional features in the audio feature sequence. The character's emotional features are organized according to the time order of each emotional feature in the audio feature sequence, and corresponding emotional transition features are inserted between adjacent character emotional features to generate a character-driven feature sequence.

[0131] In one embodiment, the communication animation generation module 40 is specifically used for: Based on the rhythm features in the audio feature sequence, determine the action time nodes and pause time intervals in the communication audio, and generate the character action sequence. The character-driven feature sequence and the character action timing are mapped to the time sequence of the communication audio to generate animation-driven data; Multiple animation clips are generated based on the character's basic characteristics, the animation-driven data, and the communication audio. For each group of adjacent animated segments in the plurality of animated segments, the ending action state of the previous animated segment and the starting action state of the next animated segment are extracted, and the emotional transition features corresponding to each group of adjacent animated segments are extracted from the character driving feature sequence. A transition frame sequence is generated based on the ending action state, the starting action state and the emotional transition features. The transition frame sequence is inserted between corresponding adjacent animation segments to connect the multiple animation segments and generate a connecting animation.

[0132] In one embodiment, the display fusion module 50 is specifically used for: Extract factual information and evidence identifiers from the structured objection data, and locate the target text content from the evidence text based on the evidence identifiers; The factual information of the matter is semantically matched with the target text content to determine the location of key text, and key annotation content is generated based on the location of key text. Based on the target text content, generate text display content; based on the highlighted content, generate an interactive interpretation entry point; and combine the text display content, the highlighted content, and the interactive interpretation entry point to generate display content based on the text content. The reference content corresponding to the target text content is determined from the communication script, and the playback time interval corresponding to the reference content is determined from the communication audio. The display time interval of the reference content is determined based on the playback time interval. According to the stated display time interval, the displayed content and the communication animation are synchronized in time and the images are merged to generate objection communication content.

[0133] In one embodiment, the interactive evidence storage module 60 is specifically used for: Extract output channel data, interaction scenario data, and terminal parameter data from the interaction configuration data; generate content output parameters based on the output channel data, interaction scenario data, and terminal parameter data; and output the objection communication content based on the content output parameters. Receive an interaction request for the content of the objection communication, extract the interaction type and interaction content from the interaction request, and generate interaction task data based on the interaction type and interaction content; Based on the interaction type, a target processing node is determined, and the interaction task data is associated with the target matter and the content of the objection communication. The interaction task data is then sent to the target processing node for interaction processing. Receive the processing status data and processing result data returned by the target processing node, and generate interactive processing data based on the interactive task data, the processing status data and the processing result data; A data summary is generated based on the content of the objection communication and the interaction processing data. The objection communication content, the interaction processing data, and the data summary are then associated and organized according to the target matter to generate evidence storage data.

[0134] In one embodiment, a computer device is provided, which may be a server, and its internal structure diagram may be as follows: Figure 10As shown. The computer device includes a processor, memory, network interface, and database connected via a system bus. The processor provides determination and control capabilities. The memory includes non-volatile and / or volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and database. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage media. The network interface is used to communicate with external clients via a network connection. When executed by the processor, the computer program implements server-side functions or steps of a role-driven objection communication method.

[0135] In one embodiment, a computer device is provided, which may be a client, and its internal structure diagram may be as follows: Figure 11 As shown, the computer device includes a processor, memory, network interface, display screen, and input devices connected via a system bus. The processor provides determination and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage media. The network interface is used to communicate with an external server via a network connection. When executed by the processor, the computer program implements client-side functions or steps of a role-driven objection communication method.

[0136] In one embodiment, a computer device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to perform the following steps: Obtain objection information, supporting text, object profile data, and interaction configuration data for the target matter; verify the correspondence between the objection information and the supporting text; and generate structured objection data. A communication script is generated based on the structured objection data, a communication audio is generated based on the communication script, and emotional and rhythmic features are extracted from the communication audio to generate an audio feature sequence. Based on the object profile data, the basic characteristics of the character are determined, and based on the emotional characteristics in the audio feature sequence and the preset emotional reference feature set, a character-driven feature sequence is generated. Based on the character's basic features, the character's driving feature sequence, the audio feature sequence, and the communication audio, multiple animation clips are generated, and the multiple animation clips are connected to generate a communication animation; Based on the structured objection data and the basis text, the basis display content is generated, and the basis display content is integrated with the communication animation to generate objection communication content; The objection communication content is output based on the interaction configuration data, the interaction request for the objection communication content is received and the interaction is processed, the interaction processing data is generated, and the evidence storage data is generated based on the objection communication content and the interaction processing data.

[0137] In one embodiment, a computer-readable storage medium is provided, which may be non-volatile or volatile, and a computer program is stored thereon, which, when executed by a processor, performs the following steps: Obtain objection information, supporting text, object profile data, and interaction configuration data for the target matter; verify the correspondence between the objection information and the supporting text; and generate structured objection data. A communication script is generated based on the structured objection data, a communication audio is generated based on the communication script, and emotional and rhythmic features are extracted from the communication audio to generate an audio feature sequence. Based on the object profile data, the basic characteristics of the character are determined, and based on the emotional characteristics in the audio feature sequence and the preset emotional reference feature set, a character-driven feature sequence is generated. Based on the character's basic features, the character's driving feature sequence, the audio feature sequence, and the communication audio, multiple animation clips are generated, and the multiple animation clips are connected to generate a communication animation; Based on the structured objection data and the basis text, the basis display content is generated, and the basis display content is integrated with the communication animation to generate objection communication content; The objection communication content is output based on the interaction configuration data, the interaction request for the objection communication content is received and the interaction is processed, the interaction processing data is generated, and the evidence storage data is generated based on the objection communication content and the interaction processing data.

[0138] It should be noted that the functions or steps that can be implemented by the computer-readable storage medium or computer device described above can be referred to the relevant descriptions on the server side and client side in the foregoing method embodiments. To avoid repetition, they will not be described one by one here.

[0139] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), Rambus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0140] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional units and modules is used as an example. In practical applications, the above functions can be assigned to different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above.

[0141] It should be noted that any AI models, software tools, or components not belonging to this company appearing in the embodiments of this application are merely illustrative examples and do not represent actual use. All user personal information involved in the embodiments of this application has been authorized (with the knowledge and consent) by the relevant parties or has been fully authorized by all parties, and the executing entity may obtain it through various legal and compliant means. The collection, storage, use, processing, transmission, provision, and disclosure of the information, data, and signals involved all comply with relevant laws and regulations and do not violate public order and good morals.

[0142] The above-described embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included within the protection scope of the present invention.

Claims

1. A method for objection communication based on role-driven characteristics, characterized in that, Includes the following steps: Obtain objection information, supporting text, object profile data, and interaction configuration data for the target matter; verify the correspondence between the objection information and the supporting text; and generate structured objection data. A communication script is generated based on the structured objection data, a communication audio is generated based on the communication script, and emotional and rhythmic features are extracted from the communication audio to generate an audio feature sequence. Based on the object profile data, the basic characteristics of the character are determined, and based on the emotional characteristics in the audio feature sequence and the preset emotional reference feature set, a character-driven feature sequence is generated. Based on the character's basic features, the character's driving feature sequence, the audio feature sequence, and the communication audio, multiple animation clips are generated, and the multiple animation clips are connected to generate a communication animation; Based on the structured objection data and the basis text, the basis display content is generated, and the basis display content is integrated with the communication animation to generate objection communication content; The objection communication content is output based on the interaction configuration data, the interaction request for the objection communication content is received and the interaction is processed, the interaction processing data is generated, and the evidence storage data is generated based on the objection communication content and the interaction processing data.

2. The objection communication method based on role-driven characteristics as described in claim 1, characterized in that, Obtain objection information, supporting text, object profile data, and interaction configuration data for the target matter; verify the correspondence between the objection information and the supporting text; and generate structured objection data, including: Obtain objection information for the target matter, extract the matter identifier and object identifier from the objection information, obtain the supporting text based on the matter identifier, and obtain object profile data and interaction configuration data based on the object identifier; Extract the objection reason information, the factual information of the matter, the information of the submitted materials and the information of subsequent processing requirements from the objection information, and extract multiple basis units and material constraint content from the basis text. Each basis unit includes a corresponding basis identifier, basis content and definition content. Based on the objection reason information, candidate basis units are selected from the multiple basis units, and the factual information of the matter is semantically matched with the definition content in each candidate basis unit to determine the target basis unit; Based on the subsequent processing requirements information, candidate material requirements are selected from the material constraints. The candidate material requirements are matched with the submitted material information to determine the material requirements to be supplemented. Based on the target basis unit and the material requirements to be supplemented, the objection information and the basis text are verified to generate corresponding verification results. Based on the corresponding verification results, the matter identifier, the objection reason information, the matter fact information, the submitted material information, the subsequent processing requirement information, the target basis unit, and the supplementary material requirements are organized according to the field correspondence to generate structured objection data.

3. The objection communication method based on role-driven characteristics as described in claim 1, characterized in that, A communication script is generated based on the structured objection data, and communication audio is generated based on the communication script. Emotional and rhythmic features are extracted from the communication audio to generate an audio feature sequence, including: Extract the objection reason information, factual information, target basis unit, required supplementary materials, and subsequent processing information from the structured objection data; Based on the objection reason information and the factual information of the matter, generate emotional expression content; based on the target basis unit and the factual information of the matter, generate basis explanation content; and based on the supplementary material requirements and the subsequent processing needs information, generate subsequent guidance content. Organize the emotional expression content, the basis explanation content, and the subsequent guidance content in the order of content to generate a candidate communication script. Extract the reference content and material guidance content from the candidate communication script, match the reference content with the target reference unit, and match the material guidance content with the requirements of the supplementary materials. Based on the matching results, modify the candidate communication script and generate a communication script. Based on the communication script, a communication audio is generated. The communication script is divided into multiple script content units, and the audio time interval corresponding to each script content unit in the communication audio is determined. The communication audio is divided into multiple audio segments according to the multiple audio time intervals. The emotional and rhythmic features of each audio segment are extracted, and the corresponding emotional and rhythmic features are organized according to the time sequence of each audio segment to generate an audio feature sequence.

4. The objection communication method based on role-driven characteristics as described in claim 1, characterized in that, Based on the object profile data, basic character features are determined, and a character-driven feature sequence is generated based on the emotional features in the audio feature sequence and a preset emotional reference feature set, including: Extract object attribute data and historical interaction preference data from the object profile data, determine target character features from a preset character feature library based on the object attribute data and the historical interaction preference data, and generate basic character features based on the target character features; For each emotion feature in the audio feature sequence, multiple target emotion reference features are selected from a preset emotion reference feature set, and emotion fusion weights corresponding to each target emotion reference feature are generated. For each emotion feature, based on the emotion fusion weights corresponding to the emotion features, multiple target emotion reference features corresponding to the emotion features are weighted and fused to generate fused emotion features. The fused emotion features are then mapped to the same feature dimension as the character's basic features to generate character emotion features corresponding to the emotion features. Emotional transition features are generated based on the character emotional features corresponding to adjacent emotional features in the audio feature sequence. The character's emotional features are organized according to the time order of each emotional feature in the audio feature sequence, and corresponding emotional transition features are inserted between adjacent character emotional features to generate a character-driven feature sequence.

5. The objection communication method based on role-driven characteristics as described in claim 1, characterized in that, Based on the character's basic features, the character's driving feature sequence, the audio feature sequence, and the communication audio, multiple animation clips are generated, and the multiple animation clips are linked together to generate a communication animation, including: Based on the rhythm features in the audio feature sequence, determine the action time nodes and pause time intervals in the communication audio, and generate the character action sequence. The character-driven feature sequence and the character action timing are mapped to the time sequence of the communication audio to generate animation-driven data; Multiple animation clips are generated based on the character's basic characteristics, the animation-driven data, and the communication audio. For each group of adjacent animated segments in the plurality of animated segments, the ending action state of the previous animated segment and the starting action state of the next animated segment are extracted, and the emotional transition features corresponding to each group of adjacent animated segments are extracted from the character driving feature sequence. A transition frame sequence is generated based on the ending action state, the starting action state and the emotional transition features. The transition frame sequence is inserted between corresponding adjacent animation segments to connect the multiple animation segments and generate a connecting animation.

6. The objection communication method based on role-driven characteristics as described in claim 1, characterized in that, Based on the structured objection data and the supporting text, supporting content is generated and then integrated with the communication animation to generate objection communication content, including: Extract factual information and evidence identifiers from the structured objection data, and locate the target text content from the evidence text based on the evidence identifiers; The factual information of the matter is semantically matched with the target text content to determine the location of key text, and key annotation content is generated based on the location of key text. Based on the target text content, generate text display content; based on the highlighted content, generate an interactive interpretation entry point; and combine the text display content, the highlighted content, and the interactive interpretation entry point to generate display content based on the text content. The reference content corresponding to the target text content is determined from the communication script, and the playback time interval corresponding to the reference content is determined from the communication audio. The display time interval of the reference content is determined based on the playback time interval. According to the stated display time interval, the displayed content and the communication animation are synchronized in time and the images are merged to generate objection communication content.

7. The objection communication method based on role-driven characteristics as described in claim 1, characterized in that, Based on the interactive configuration data, the objection communication content is output; interactive requests for the objection communication content are received and processed; interactive processing data is generated; and evidence storage data is generated based on the objection communication content and the interactive processing data, including: Extract output channel data, interaction scenario data, and terminal parameter data from the interaction configuration data; generate content output parameters based on the output channel data, interaction scenario data, and terminal parameter data; and output the objection communication content based on the content output parameters. Receive an interaction request for the content of the objection communication, extract the interaction type and interaction content from the interaction request, and generate interaction task data based on the interaction type and interaction content; Based on the interaction type, a target processing node is determined, and the interaction task data is associated with the target matter and the content of the objection communication. The interaction task data is then sent to the target processing node for interaction processing. Receive the processing status data and processing result data returned by the target processing node, and generate interactive processing data based on the interactive task data, the processing status data and the processing result data; A data summary is generated based on the content of the objection communication and the interaction processing data. The objection communication content, the interaction processing data, and the data summary are then associated and organized according to the target matter to generate evidence storage data.

8. A role-driven objection communication device, characterized in that, The objection communication device based on role-driven characteristics includes: The objection information verification module is used to obtain objection information, basis text, object profile data and interaction configuration data of the target matter, verify the correspondence between the objection information and the basis text, and generate structured objection data. The audio feature generation module is used to generate a communication script based on the structured objection data, generate communication audio based on the communication script, and extract emotional features and rhythm features from the communication audio to generate an audio feature sequence. The character-driven generation module is used to determine the basic characteristics of the character based on the object profile data, and to generate a character-driven feature sequence based on the emotional features in the audio feature sequence and a preset set of emotional reference features. The communication animation generation module is used to generate multiple animation clips based on the character's basic features, the character's driving feature sequence, the audio feature sequence, and the communication audio, and to connect the multiple animation clips to generate a communication animation. The display fusion module is used to generate display content based on the structured objection data and the objection text, and to fuse the display content with the communication animation to generate objection communication content. The interactive evidence storage module is used to output the objection communication content based on the interactive configuration data, receive interactive requests for the objection communication content and perform interactive processing, generate interactive processing data, and generate evidence storage data based on the objection communication content and the interactive processing data.

9. A computer device, characterized in that, The computer device includes a memory, a processor, and a role-driven objection communication program stored in the memory and executable on the processor, wherein the role-driven objection communication program, when executed by the processor, implements the steps of the role-driven objection communication method as described in any one of claims 1-7.

10. A computer-readable storage medium, characterized in that, The storage medium stores a role-driven objection communication program, which, when executed by a processor, implements the steps of the role-driven objection communication method as described in any one of claims 1-7.

Images