AI mediation system based on large data model

Abstract

The invention discloses an AI mediation system based on a large data model, and the system comprises an interaction system which is used for receiving the multi-modal information of a party, and feeding back the multi-modal dialogue information outputted by a mediation engine system to the party, so as to complete the multi-modal dialogue interaction between an AI mediator and the party; the mediation engine system is used for initiating one or more rounds of AI mediation of one or more AI mediators, and generating multi-modal dialogue information by combining a multi-modal mediation template of the mediation large model system and a context of a dialogue in the interaction system, so as to send the multi-modal dialogue information to the interaction system; and the mediation large model system is used for generating the multi-modal mediation template according to the multi-modal mediation data and the multi-dimensional intelligent labeling result of the AI mediator so as to send the multi-modal mediation template to the mediation engine system. According to the scheme, the mediation efficiency can be improved, and the mediation quality can be enhanced.

Description

Technical Field

[0001] The present application relates to the field of artificial intelligence, and in particular to an AI mediation system based on a large data model. Background Art

[0002] Case mediation plays an important role in the modern judicial system. It aims to resolve disputes through non-litigation means, reduce the burden on courts, and improve judicial efficiency. With the continuous development and complexity of society, the number of cases has increased dramatically. Traditional litigation methods are time-consuming, labor-intensive, and costly. Case mediation provides an efficient and economical alternative that can quickly resolve disputes, reduce the time and financial burden on the parties, and promote social harmony. The existing case mediation system technology is mainly based on the experience and judgment of manual mediators. Mediators are usually professionals with legal knowledge and mediation skills. They communicate with the parties, analyze the case, and propose solutions to facilitate the two parties to reach an agreement. This method is effective to a certain extent, but there are also obvious problems: First, the quality and ability of mediators vary, and the quality of mediation is difficult to guarantee. Second, manual mediation is inefficient, especially when the case volume is large, it is difficult for mediators to cope with it. Finally, the manual mediation process is easily affected by subjective factors, which may lead to unfair results. At the same time, the number of various types of dispute cases is large and growing rapidly, and judicial resources are insufficient. The problems faced are "difficult to remind, difficult to collect, and difficult to litigate". For example: due to the high time cost, mediation organizations and mediators represented by lawyers lack motivation to participate and are inefficient. Grassroots people's courts have too many cases and too few people, and are overwhelmed.

[0003] At present, artificial intelligence technologies based on commercial big models such as judicial big models and Pangu big models have been well utilized in many fields and have driven the rapid development of the industry. However, it is still a wasteland in the field of mediation. With the rapid development of the Internet, social conflicts and disputes in the digital age show a development trend of "many, fast, wide, complex and difficult". The mediation industry has the following problems: mediation is not standardized and no industry standards have been formed. The number of mediators is limited and their experience is limited, resulting in insufficient professionalism. The number of mediation dispute cases is large and the growth rate is fast, so they cannot be effectively resolved in time. The most basic intelligent outbound calls can no longer meet the needs of mediation.

[0004] Therefore, there is an urgent need for a technical solution that can improve mediation efficiency and enhance mediation quality. Summary of the invention

[0005] In order to solve the deficiencies of the prior art, the embodiment of the present application provides an AI mediation system based on a large data model. The present application solves the problems of insufficient number of mediators and long training cycle.

[0006] An embodiment of the present application provides an AI mediation system based on a data big model, comprising: an interactive system, a mediation engine system and a mediation big model system; wherein the interactive system is used to receive multimodal information of the parties, and feed back the multimodal dialogue information output by the mediation engine system to the parties, so as to complete the multimodal dialogue interaction between the AI ​​mediator and the parties; the mediation engine system is used to initiate one or more rounds of AI mediation by one or more AI mediators, and generate the multimodal dialogue information in combination with the multimodal mediation template of the mediation big model system and the context of the dialogue in the interactive system, so as to send it to the interactive system; the mediation big model system is used to generate the multimodal mediation template according to the multimodal mediation data of the AI ​​mediator and the multi-dimensional intelligent labeling results, so as to send it to the mediation engine system.

[0007] In a possible implementation, the interactive system includes: an interactive subsystem, a digital human subsystem and a manual mediation subsystem; wherein, when receiving the multimodal information of the parties and feeding back the multimodal dialogue information output by the mediation engine system to the parties to complete the multimodal dialogue interaction between the AI ​​mediator and the parties: the interactive subsystem is used to respond to the multimodal information of the parties to drive the AI ​​mediator to conduct multimodal dialogue interaction operations with the parties; the digital human subsystem is used to receive the multimodal dialogue information of the interactive subsystem to conduct multimodal dialogue interaction with the parties; the manual mediation subsystem is used to receive instructions for converting AI mediation to manual mediation to assign the mediation case to the corresponding manual mediator.

[0008] In a possible implementation, the mediation engine system includes: a mediation initiation subsystem, a mediation dialogue subsystem, an intelligent annotation subsystem and an intelligent monitoring subsystem; wherein, when initiating one or more rounds of AI mediation by one or more AI mediators, and combining the multimodal mediation template of the mediation large model system and the context of the dialogue in the interactive system, the multimodal dialogue information is generated to be sent to the interactive system: the mediation initiation subsystem is used to import the mediation case into the system, and select the target AI mediator to transmit the AI ​​mediation parameters to the mediation dialogue subsystem; the mediation dialogue subsystem is used to generate multimodal dialogue information according to the AI ​​mediation parameters of the mediation initiation subsystem, the multimodal mediation template of the mediation library subsystem in the mediation large model system, the multidimensional intelligent annotation of the intelligent annotation subsystem and the context content of this mediation dialogue, so as to transmit it to the interactive subsystem in the interactive system; the intelligent annotation subsystem is used to receive the text elements, audio text elements and video text elements of the interactive subsystem in the interactive system, and generate corresponding data tags to transmit the results of the multidimensional intelligent annotation to the mediation dialogue subsystem and the mediation library subsystem in the mediation large model system respectively; the intelligent monitoring subsystem is used to monitor the manual mediation process and check the mediation results.

[0009] In a possible implementation, the mediation big model system includes: a mediation library subsystem, a mediation training subsystem, a mediation modeling subsystem and a third-party resource library subsystem; wherein, the mediation big model system is used to generate the multimodal mediation template according to the multimodal mediation data and multi-dimensional intelligent labeling results of the AI ​​mediator, and send it to the mediation engine system: the mediation library subsystem is used to construct core mediation data according to the output of one or more subsystems; the mediation training subsystem is used to optimize the multimodal mediation template using the AI ​​strategy feature set and the target machine learning model; the mediation modeling subsystem is used to integrate the target algorithm through a multi-core algorithm structure, and generate a multimodal mediation algorithm model for the mediation training subsystem to call; the third-party resource library subsystem is used to obtain the target database and the target big model.

[0010] In a possible implementation, the response to the multimodal information of the parties is used to drive the AI ​​mediator to conduct a multimodal dialogue interaction operation with the parties, including: collecting the audio information of the parties, and transmitting the target audio clip and the corresponding text element information to the intelligent annotation subsystem in the mediation engine system; collecting the video information of the parties, and identifying the movements and micro-expressions of the parties, so as to transmit the target video clip and the corresponding emotional text to the intelligent annotation subsystem in the mediation engine system; collecting the instant messages of the parties during the interaction, and transmitting the corresponding text elements to the intelligent annotation subsystem in the mediation engine system; according to the multimodal dialogue information received from the mediation dialogue subsystem in the mediation engine system, driving the digital human subsystem to conduct a multimodal dialogue with the parties, and transferring to the manual mediator based on preset conditions; based on the text records, voice records, video records of the parties and the identified and sorted text elements, audio elements, video elements, etc., constructing the multimodal mediation data of the parties, and transmitting it to the mediation library subsystem in the mediation large model system for multimodal storage.

[0011] In a possible implementation, the monitoring of the manual mediation process and checking of the mediation results include: retrieving multimodal mediation supervision data from the mediation library subsystem in the mediation large model system, and acquiring multimodal mediation dialogue data of the manual mediator in real time; identifying whether the mediation behavior of the manual mediator complies with the mediation process specifications, and prompting the mediator to make real-time improvements; generating a mediation analysis report for the manual mediator, and combining the mediation supervision process data to form multimodal mediation supervision data for transmission to the mediation library subsystem in the mediation large model system.

[0012] In a possible implementation, the core mediation data is constructed based on the output of one or more subsystems, including: obtaining target legal information in a third-party resource library subsystem; obtaining multimodal mediation data in a manual mediation subsystem in an interactive system; obtaining multimodal mediation data in an interactive subsystem in an interactive system; obtaining multimodal intelligent labeling results in an intelligent labeling subsystem in a mediation engine system; generating multimodal mediation data based on multimodal mediation process data for transmission to a mediation training subsystem; and receiving multimodal mediation templates of trained mediators in the mediation training subsystem.

[0013] In a possible implementation, the use of AI strategy feature sets and target machine learning models to optimize multimodal mediation templates includes: retrieving multimodal mediation data from a mediation library subsystem, and retrieving a multimodal mediation algorithm model generated in a mediation modeling subsystem; and training the multimodal mediation algorithm model based on existing multimodal mediation templates and multimodal mediation data to optimize the multimodal mediation templates.

[0014] In the AI ​​mediation system based on the data big model provided above, the embodiment of the present application, based on the big model, combines the current advanced technologies such as natural language, voice conversion, semantic recognition, etc. to enable mediation, and can gather different causes of action, relevant laws and regulations, dispute focus, precedents, successful mediation cases, mediation skills, mediation speech, etc. to train AI mediators, and enable them to have the mediation ability of gold medal mediators, thereby improving mediation efficiency. Furthermore, in some embodiments, by training AI mediation quality inspectors to monitor the mediation process in real time, the quality of mediation can be enhanced, and the efficiency, compliance, and security of mediation can be ensured. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0016] Figure 1 A system architecture diagram of an AI mediation system provided in an embodiment of the present application; Figure 2 An AI mediation flow chart of an AI mediation system provided in an embodiment of the present application; Figure 3 An AI mediation training flow chart of an AI mediation system provided in an embodiment of the present application; Figure 4 An AI mediation supervision flowchart of an AI mediation system provided in an embodiment of the present application. DETAILED DESCRIPTION

[0017] Various exemplary embodiments of the present application will now be described in detail with reference to the accompanying drawings. It should be noted that unless otherwise specifically stated, the relative arrangement of components and steps, numerical expressions and numerical values ​​set forth in these embodiments do not limit the scope of the present application.

[0018] Those skilled in the art will appreciate that the terms "first", "second" and the like in the embodiments of the present application are only used to distinguish different steps, devices or subsystems, etc., and do not represent any specific technical meaning, nor do they represent the necessary logical order between them. It should also be understood that in the embodiments of the present application, "multiple" may refer to two or more, and "at least one" may refer to one, two or more. It should also be understood that for any component, data or structure mentioned in the embodiments of the present application, in the absence of explicit limitation or contrary revelation given in the context, it can generally be understood as one or more. In addition, the term "and / or" in the present application is only a description of the association relationship of the associated objects, indicating that there can be three relationships, for example, A and / or B, which can represent: A exists alone, A and B exist at the same time, and B exists alone. In addition, the character " / " in the present application generally indicates that the associated objects before and after are an "or" relationship. It should also be understood that the description of each embodiment in the present application emphasizes the differences between the embodiments, and the same or similar aspects can be referenced to each other. For the sake of brevity, they will not be repeated one by one.

[0019] At the same time, it should be understood that, for ease of description, the sizes of the various parts shown in the drawings are not drawn according to the actual proportional relationship. The following description of at least one exemplary embodiment is actually only illustrative and is by no means intended to limit the present application and its application or use. The techniques, methods and devices known to ordinary technicians in the relevant fields may not be discussed in detail, but where appropriate, the techniques, methods and devices should be considered as part of the specification. It should be noted that similar numbers and letters represent similar items in the following drawings, so once an item is defined in one drawing, it does not need to be further discussed in subsequent drawings.

[0020] In order to make the purpose, technical solution and advantages of the embodiments of the present application clearer, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0021] Figure 1A system architecture diagram of an AI mediation system provided for an embodiment of the present application. It should be understood that the system shown in the figure is exemplary and not restrictive. This means that the system architecture involved is not limited to a specific form or design, but is presented as an example. In other words, the architecture shown in the figure can be regarded as a way of expression to clearly describe related concepts and relationships, and does not exclude other forms of architecture. Therefore, when interpreting the architecture in the picture, it should be understood that the model is flexible and diverse, and its purpose is to provide an exemplary description rather than a restrictive provision on a specific form.

[0022] like Figure 1 As shown, the interactive system 101 receives the multimodal information of the parties, and feeds back the multimodal dialogue information output by the mediation engine system to the parties, so as to complete the multimodal dialogue interaction between the AI ​​mediator and the parties. It should be noted that the interactive system includes: an interactive subsystem, a digital human subsystem and an artificial mediation subsystem. The AI ​​interactive system is responsible for the interaction between the AI ​​mediator and the parties. On the one hand, it receives the multimodal dialogue (including text, voice, tone, action, expression and other information) information of the AI ​​mediation engine system and completes a multimodal dialogue with the parties; on the other hand, it collects the multimodal information (text, audio and video) of the parties in a multimodal dialogue.

[0023] Further (such as Figure 2 As shown in the figure, first, the case that needs to be mediated is imported into the AI ​​mediation system through API or files (such as forms). Then, the AI ​​mediation system conducts case analysis and calls the AI ​​mediation robot trained by the mediation model. Secondly, the AI ​​mediation robot uses ICT, AR / VR, big data, blockchain and other technologies to mediate with the parties based on the content of the mediation application. Next, the AI ​​mediation robot reaches the parties through phone calls, text messages, videos, AR / VR, digital people, etc. Subsequently, the AI ​​mediation robot profiles the parties, analyzes the communication intentions of the parties, conducts continuous mediation communication with the parties, conducts interactive and superimposed communication and negotiation, guides the parties to mediate, reaches a mediation agreement, signs legal documents such as the mediation agreement / judicial confirmation application, and performs the legal obligations agreed in the mediation agreement. Finally, the mediation supervisor monitors the operation of the AI ​​mediator in real time, discovers business problems in the mediation process in real time, adjusts the mediation strategy, and maintains the stable operation of the mediation robot.

[0024] Specifically, the interactive subsystem is used to respond to the multimodal information of the parties to drive the AI ​​mediator to conduct multimodal dialogue interaction operations with the parties. This includes: collecting the audio information of the parties, and transmitting the target audio clip and the corresponding text element information to the intelligent annotation subsystem in the mediation engine system; collecting the video information of the parties, and identifying the actions and micro-expressions of the parties, so as to transmit the target video clip and the corresponding emotional text to the intelligent annotation subsystem in the mediation engine system; collecting the instant messages of the parties during the interaction, and transmitting the corresponding text elements to the intelligent annotation subsystem in the mediation engine system; according to the multimodal dialogue information of the mediation dialogue subsystem in the mediation engine system received, the digital human subsystem is driven to conduct a multimodal dialogue with the parties, and transferred to the manual mediator based on preset conditions; based on the text records, voice records, video records and the identified and sorted text elements, audio elements, video elements, etc. of the parties, the multimodal mediation data of the parties is constructed, and transmitted to the mediation library subsystem in the mediation large model system for multimodal storage.

[0025] In one embodiment, the AI ​​interaction subsystem is responsible for the interaction between the AI ​​mediator and the parties via voice, video, and text (such as SMS, instant messaging, etc.).

[0026] First, the audio information of the parties is collected and extracted. The continuous analog audio signal is converted into a discrete digital signal, framed, and MFCC features are extracted. The optimal probability path is found, the acoustic model recognizes phonemes, and the vocabulary model and language model recognize words and sentences. The mediator and the mediation assistant recognize the audio signals of the parties during the mediation process and organize them into small audio segments and corresponding text element information (keywords of legal and regulatory cognition such as understanding the law and not understanding the law, keywords of life status such as income, work, food, clothing, housing, transportation, consumption, family, social activities, etc., keywords of mental state such as positive, pessimistic, and depressed, keywords of performance such as willingness to perform, delaying performance, and refusing to perform, emotional sensitive words such as joy, anger, sorrow, intelligence, rudeness, disgust, hate, and gratitude, negative keywords such as pornography, terrorism, violence, and insults, the tone of the audio corresponding to the keywords, etc.), and transmit them to the AI ​​intelligent labeling subsystem.

[0027] Then, the video information of the parties is collected and extracted. Computer vision technology is used to detect the position and outline of the face; deep learning models are used to identify key points on the face. Machine learning models, convolutional neural networks (CNN) and recurrent neural networks (RNN) are used to identify micro-expressions and movements to form video element slices. Local video element slices such as micro-movement videos of eyes, ears, mouth, nose, eyebrows, hair, wrinkles, cheeks, facial muscles, etc., and overall video element slices such as action videos of head, shoulders, neck, hands, elbows, arms, torso, hips, limbs, etc., identify and organize the emotional text elements of the parties in local video element slices and overall video element segments: happy, surprised, afraid, angry, sad, disgusted, contemptuous, interested, etc. The identified emotions are transmitted to the AI ​​intelligent annotation subsystem to form video text element information.

[0028] Secondly, the parties' instant messages are collected and extracted. The system collects and extracts the instant messages between the parties and the mediators (such as text messages, WeChat, Fetion, etc.); analyzes the text, punctuation marks, emoticons, etc. sent by the parties in the instant messages, identifies network terms, analyzes the context, and organizes the text elements of the parties (keywords of legal and regulatory cognition such as understanding the law, not understanding the law, etc., keywords of life status such as income, work, food, clothing, housing, transportation, consumption, family, social activities, etc., keywords of mental state such as positive, pessimistic, and depressed, keywords of performance such as willingness to perform, delaying performance, and refusing to perform, emotionally sensitive words such as joy, anger, sorrow, and happiness, intellectuality, rudeness, disgust, hate, and gratitude, and negative keywords such as pornography, terrorism, violence, and insults). The text elements of the parties identified and organized are transmitted to the AI ​​intelligent labeling subsystem.

[0029] Next, the multimodal dialogue (including text, voice, tone, action, expression, etc.) of the AI ​​mediation dialogue subsystem is received, and the AI ​​digital human subsystem is called to complete a mediation dialogue with the parties. Subsequently, the multimodal dialogue (including text, voice, tone, action, expression, etc.) of the AI ​​mediation dialogue subsystem is received. When the parties have no intention to mediate or the parties clearly need manual mediation, this AI mediation ends, and the AI ​​mediation record is transferred to the AI ​​manual mediation subsystem, and this mediation is transferred to the manual mediator. Next, based on the original text, voice, and video records of the parties, as well as the identified and sorted text elements, audio elements, video elements, etc., the multimodal mediation data of the parties is formed and transmitted to the AI ​​mediation library subsystem of the AI ​​mediation large model system for multimodal storage.

[0030] The digital human subsystem is used to receive multimodal dialogue information from the interactive subsystem to conduct multimodal dialogue interaction with the parties. Specifically, the AI ​​digital human subsystem receives multimodal dialogue information from the AI ​​interactive subsystem, and uses multiple AI core technologies such as speech synthesis, speech recognition, semantic understanding, image processing, machine translation, and virtual image drive to realize the digital mediator's legal publicity, legal consultation, and interactive communication with the parties, so as to guide the parties to have the willingness to mediate and reach mediation.

[0031] The manual mediation subsystem is used to receive instructions for AI mediation to be transferred to manual mediation, so as to assign the mediation case to the corresponding manual mediator. Specifically, the AI ​​manual mediation subsystem receives instructions for AI mediation to be transferred to manual mediation, and assigns the mediation case to the corresponding manual mediator. The manual mediator uses the ICT technology integrated by the AI ​​manual mediation subsystem and digital intelligence technologies such as cloud computing, blockchain, and big data to conduct multimodal mediation with the parties through telephone, text messages, and videos. The manual multimodal mediation process and results (no willingness to mediate, no mediation agreement reached, mediation agreement reached, case closed, etc.) are fed back to the AI ​​mediation library subsystem.

[0032] The mediation engine system 102 initiates one or more rounds of AI mediation by one or more AI mediators, and combines the multimodal mediation template of the mediation large model system and the context of the dialogue in the interactive system to generate the multimodal dialogue information to be sent to the interactive system. It should be noted that the mediation engine system includes: a mediation initiation subsystem, a mediation dialogue subsystem, an intelligent labeling subsystem and an intelligent monitoring subsystem. The AI ​​mediation engine system is responsible for initiating one or more rounds of AI mediation by one or more AI mediators, and outputs the intelligently generated words to the AI ​​interactive system according to the trained AI multimodal mediation template received in the mediation large model system, combined with the multi-dimensional intelligent labeling results formed in the interaction between the AI ​​interactive system and the parties, and combined with the context before this dialogue, to drive the AI ​​interactive system to mediate with the parties.

[0033] Specifically, the mediation initiation subsystem is used to import mediation cases into the system and select a target AI mediator to transmit AI mediation parameters to the mediation dialogue subsystem. Furthermore, the case is imported into the AI ​​mediation system AI mediation subsystem through an API or a file (such as a form, etc.). The AI ​​mediation initiation subsystem automatically analyzes the case, and automatically (manually) selects the AI ​​mediation robot corresponding to the mediator with a high mediation success rate according to the different causes of the case and the household registration, age, occupation and other information of the parties; sets AI mediation parameters: the number of mediation robots, mediation robots, start time, termination conditions such as time, keywords, manual active intervention, etc. The AI ​​mediation initiation subsystem starts AI mediation and transmits AI mediation parameters to the AI ​​mediation dialogue subsystem.

[0034] The mediation dialogue subsystem is used to generate multimodal dialogue information based on the AI ​​mediation parameters of the mediation start subsystem, the multimodal mediation template of the mediation library subsystem in the mediation large model system, the multi-dimensional intelligent annotation of the intelligent annotation subsystem and the context content of this mediation dialogue, so as to transmit it to the interactive subsystem in the interactive system. In one embodiment, the dialogue content between the AI ​​mediator and the parties is generated in real time. The AI ​​mediation parameters of the AI ​​mediation start subsystem are received, and the multimodal mediation template of the mediator in the AI ​​mediation library subsystem is called. A multimodal dialogue (text, voice, tone, action, expression, etc.) to be communicated with the parties is generated and fed back to the AI ​​interactive subsystem. The multi-dimensional intelligent annotation of the AI ​​intelligent annotation subsystem is received. According to the multi-dimensional intelligent annotation of the AI ​​intelligent annotation subsystem and the multimodal mediation template of the mediator of the AI ​​mediation library subsystem, combined with the context content of this mediation dialogue, a multimodal dialogue (text, voice, tone, action, expression, etc.) to be communicated with the parties is generated and fed back to the AI ​​interactive subsystem.

[0035] The intelligent annotation subsystem is used to receive the text elements, audio text elements and video text elements of the interactive subsystem in the interactive system, and generate corresponding data tags to transmit the results of multi-dimensional intelligent annotation to the mediation dialogue subsystem and the mediation library subsystem in the mediation large model system respectively. Specifically, multi-modal intelligent annotation is performed on the parties. The AI ​​intelligent annotation subsystem receives the message text elements, audio text elements and video text elements of the AI ​​interactive subsystem. The message text elements, audio text elements and video text elements of the parties are combined and analyzed. When keywords are matched in the elements, the parties are intelligently annotated in multiple dimensions (living status, character, honesty coefficient, repayment ability, mediation willingness, mediation success rate, etc.): preset data tags are added. The multi-dimensional intelligent annotation includes the original files, audio and video information of the parties, the corresponding message / audio / video text elements analyzed and sorted, and the corresponding data tags. The relationship between the example elements and annotations. The multi-dimensional intelligent annotation of the parties is fed back to the AI ​​mediation dialogue subsystem. The multi-modal intelligent annotation and analysis results of the parties are transmitted to the AI ​​mediation library subsystem for multi-modal storage.

[0036] The intelligent monitoring subsystem is used to monitor the manual mediation process and check the mediation results. Specifically, it retrieves the multimodal mediation supervision data of the mediation library subsystem in the mediation large model system, and obtains the multimodal mediation dialogue data of the manual mediator in real time; identifies whether the mediation behavior of the manual mediator complies with the mediation process specifications, and prompts the mediator to make real-time improvements; generates a mediation analysis report for the manual mediator, and combines the process data of the mediation supervision to form multimodal mediation supervision data for transmission to the mediation library subsystem in the mediation large model system.

[0037] In one embodiment, the mediation process is monitored and the mediation results (e.g. Figure 4 As shown). Retrieve multimodal mediation supervision data such as mediation process specification data, mediation compliance data, mediation supervision cases, etc. from the AI ​​mediation library subsystem. Obtain the multimodal mediation dialogue data of the manual mediator in real time, sort out the text, voice, video elements, and the multimodal mediation elements of the mediator such as tone, expression, etc. in real time, analyze whether the manual mediator follows the mediation process norms of "indicating the identity of the mediator, verifying the identity of the parties, explaining the reasons for mediation, listening to the voices of the parties, etc.", and analyze whether the manual mediator is compliant in mediation, such as whether sensitive words are hit, whether the neutrality of mediation is maintained, whether the mediation intention of the parties is respected, whether the emotions are out of control, etc. The analysis finds that the mediation behavior of the manual mediator does not meet the mediation process norms, and prompts the mediator to make real-time improvements. The analysis finds that the mediation of the manual mediator is not compliant, and takes real-time supervision actions, such as blocking the call when the content of the sensitive word library is detected, and recording and reminding the mediator; when the dialogue is detected to be excessive and the emotions are out of control, remind the manual mediation administrator or AI mediation supervisor to take over the mediation, or immediately end the mediation. Generate a mediation analysis report for the manual mediator, and analyze the good points and areas that need improvement of the mediator. The process data and analysis reports of mediation supervision form multimodal mediation supervision data and are output to the AI ​​mediation library subsystem.

[0038] The mediation big model system 103 generates the multimodal mediation template based on the multimodal mediation data and multidimensional intelligent annotation results of the AI ​​mediator, and sends it to the mediation engine system. It should be noted that the mediation big model system includes: a mediation library subsystem, a mediation training subsystem, a mediation modeling subsystem and a third-party resource library subsystem. The mediation big model system accepts the multimodal mediation data, multidimensional intelligent annotation results and the third-party resource library laws, regulations and case library of the mediator from the AI ​​mediation engine system, performs AI training according to certain modeling and algorithms, and outputs the mediator's AI multimodal mediation template for the AI ​​mediation engine system to call when initiating AI mediation.

[0039] Specifically, the mediation library subsystem is used to construct core mediation data based on the output of one or more subsystems. In one embodiment, the target legal information in the third-party resource library subsystem is obtained; the multimodal mediation data in the manual mediation subsystem in the interactive system is obtained; the multimodal mediation data in the interactive subsystem in the interactive system is obtained; the multimodal intelligent annotation results of the intelligent annotation subsystem in the mediation engine system are obtained; multimodal mediation data is generated based on the multimodal mediation process data to be transmitted to the mediation training subsystem; and the multimodal mediation template of the trained mediator in the mediation training subsystem is received.

[0040] In an implementation scenario, obtain information on different causes of action and related laws, regulations, precedents, etc. from a third-party resource library. Receive and store multimodal mediation data of all mediators who mediate disputes of different causes of action from the manual mediation subsystem: mediation process data, desensitized case materials. Receive and store multimodal mediation data of the parties collected, analyzed, and organized during AI mediation from the AI ​​interactive subsystem. Receive and store multimodal intelligent annotations and analysis results of the parties collected, analyzed, and organized by the mediator in a mediation from the AI ​​intelligent annotation subsystem. Analyze and organize multimodal mediation process data of dispute cases that have been mediated or not mediated by the mediator, including mediation words, mediation skills, mediation dialogues, and other multimodal mediation data with high mediation success rates for different causes of action, and output them to the AI ​​mediation training subsystem. The multimodal mediation template of the trained mediator is received from the AI ​​mediation training subsystem. The multimodal mediation template includes the mediator's mediation experience, mediation language, mediation skills, mediation cases in various types of dispute cases, the laws and regulations involved in such disputes, successful mediation cases of third parties, litigation precedents, etc., the mediator's thinking logic, speaking logic, etc., the mediator's score of the parties' willingness to mediate, as well as the mediator's usual voice tone, facial expressions and other data, information and processes.

[0041] The mediation training subsystem is used to optimize the multimodal mediation template using the AI ​​strategy feature set and the target machine learning model. Specifically, the multimodal mediation data in the mediation library subsystem is retrieved, and the multimodal mediation algorithm model generated in the mediation modeling subsystem is retrieved; based on the existing multimodal mediation template and multimodal mediation data, the multimodal mediation algorithm model is trained to optimize the multimodal mediation template.

[0042] In one embodiment (such as Figure 3As shown in the figure), the AI ​​mediation training subsystem retrieves the multimodal mediation data of mediators mediating different dispute cases stored in the AI ​​mediation library subsystem. The AI ​​mediation training subsystem retrieves the multimodal mediation algorithm model generated and released by the AI ​​mediation modeling subsystem; the AI ​​mediation training subsystem calls, debugs, and runs the multimodal mediation algorithm model of the AI ​​mediation modeling subsystem based on the existing multimodal mediation template and multimodal mediation data of the designated mediator, and builds a complete set of model training tools such as entity recognition models, intent classification models, sentiment classification models, ASR-related models, TTS-related models, etc., to achieve technological innovation of AI and efficient training process. Build AI strategy feature set, machine learning model engine, self-iteration, continuously update and improve the AI ​​strategy engine, and output AI strategy. Using AI strategy engines and model training tools, the mediator's multimodal mediation template is continuously iterated and optimized. The multimodal mediation template includes the mediator's mediation process map in various dispute cases, including: mediation context and path, mediation nodes (multi-dimensional intelligent labeling of the parties, mediation language, etc.), mediation skills, successful mediation cases, failed mediation cases, laws and regulations related to such disputes, successful mediation cases of third parties, litigation precedents, etc., the mediator's thinking logic, speaking logic, etc., as well as the mediator's usual voice intonation, facial expressions and other data and information; the multimodal mediation template is output to the AI ​​mediation library subsystem.

[0043] The mediation modeling subsystem is used to integrate the target algorithm through a multi-core algorithm structure and generate a multi-modal mediation algorithm model for the mediation training subsystem to call. Specifically, the AI ​​mediation modeling subsystem uses a multi-core algorithm structure to effectively integrate mainstream and professional field algorithm forms (LR-Linear Regression linear regression, Log-LR, LDA-Linear Discriminant Analysis linear discriminant analysis, Cost-sensitive learning cost sensitive learning, DAG directed acyclic graph, Adaboost iterative algorithm, GBDT-Gradient Boosting Decision Tree gradient boosting decision tree algorithm, XGBoost-eXtreme Gradient Boosting extreme gradient boosting tree, KNN (K-Nearest Neighbor) K nearest neighbor algorithm, RNN-Recurrent Neural Network recurrent neural network, AGNES-AGglomerative NESting hierarchical clustering algorithm, etc.), forming a multi-modal mediation algorithm model, ensuring accuracy while also providing mutual verification for the AI ​​mediation training subsystem to call.

[0044] The third-party resource library subsystem is used to obtain target databases and target big models, including but not limited to legal and regulatory databases, judicial big models, government affairs big models, consumer big models, etc.

[0045] It should be noted that the entire mediation process of this application is recorded in real time, and pictures, texts, audio and video are recorded throughout the process, so that there is a traceable mediation to avoid unnecessary disputes. The AI ​​mediator has the mediation ability of a gold medal mediator, which can greatly alleviate the current problem of insufficient number and professionalism of the mediator team. The AI ​​quality inspector can effectively ensure the compliance of the mediation by detecting the mediation process of the mediator in real time. The mediation big model analyzes the information of the case in multiple dimensions, and combines the mediation experience database to provide character portrait analysis, mediation reports, mediation plans, etc., effectively improving the success rate of mediation.

[0046] In addition, the design of the AI ​​mediation system framework, through the efficient operation and coordination of the AI ​​mediation subsystems, realizes the automatic mediation, self-learning, self-iterative optimization evolution of AI mediators, and interaction with human mediators. The mediation big model realizes the analysis, organization, and modeling of the mediator's multimodal mediation data, and trains to form the mediator's multimodal mediation template. The AI ​​interactive system collects, analyzes, and organizes to form the real-time multimodal mediation data of the parties. The AI ​​mediation engine realizes multimodal and multi-dimensional intelligent labeling of the parties in a mediation dialogue, calls the multimodal mediation template, and conducts mediation independently. The AI ​​mediation engine calls the mediation big model, self-learns, iterates and optimizes to improve the mediator's multimodal mediation template. The mediation big model continuously introduces laws and regulations, and legal data related to mediation, and can continuously optimize the mediator's multimodal mediation template through modeling and training.

[0047] Furthermore, an embodiment of the present application also provides a computer program product, which, when executed on a terminal device, enables the terminal device to execute any of the above methods.

[0048] It can be known from the description of the above implementation mode that those skilled in the art can clearly understand that all or part of the steps in the above-mentioned embodiment method can be implemented by means of software plus a necessary general hardware platform. Based on such an understanding, the technical solution of the present application can be essentially or partly embodied in the form of a software product that contributes to the prior art. The computer software product can be stored in a storage medium such as ROM / RAM, a disk, an optical disk, etc., including several instructions for enabling a computer device (which can be a personal computer, a server, or a network communication device such as a media gateway, etc.) to execute the methods described in the various embodiments of the present application or certain parts of the embodiments.

[0049] It should be noted that the various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments, and the same or similar parts between the various embodiments can be referred to each other. For the device disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the method part description.

[0050] It should also be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "comprise a ..." do not exclude the presence of other identical elements in the process, method, article or device including the elements.

[0051] The above description of the disclosed embodiments enables those skilled in the art to implement or use the present application. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present application. Therefore, the present application will not be limited to the embodiments shown herein, but will conform to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. An AI mediation system based on a big data model, characterized in that: include: Interaction system, mediation engine system and mediation large model system; among them, The interactive system is used to receive multimodal information from the parties and feed back the multimodal dialogue information output by the mediation engine system to the parties, so as to complete the multimodal dialogue interaction between the AI ​​mediator and the parties; The mediation engine system is used to initiate one or more rounds of AI mediation by one or more AI mediators, and generate the multimodal dialogue information in combination with the multimodal mediation template of the mediation large model system and the context of the dialogue in the interactive system, so as to be sent to the interactive system; The mediation big model system is used to generate the multimodal mediation template based on the multimodal mediation data and multi-dimensional intelligent labeling results of the AI ​​mediator, and send it to the mediation engine system.

2. The AI ​​mediation system according to claim 1, characterized in that: The interactive system includes: an interactive subsystem, a digital human subsystem and an artificial mediation subsystem; wherein, when receiving the multimodal information of the parties, the multimodal dialogue information output by the mediation engine system is fed back to the parties to complete the multimodal dialogue interaction between the AI ​​mediator and the parties: The interactive subsystem is used to respond to the multimodal information of the parties to drive the AI ​​mediator to conduct multimodal dialogue interaction operations with the parties; The digital human subsystem is used to receive multimodal dialogue information from the interactive subsystem to conduct multimodal dialogue interaction with the parties concerned; The manual mediation subsystem is used to receive instructions for converting AI mediation to manual mediation so as to assign the mediation case to the corresponding manual mediator.

3. The AI ​​mediation system according to claim 1, characterized in that: in, The mediation engine system includes: a mediation initiation subsystem, a mediation dialogue subsystem, an intelligent labeling subsystem and an intelligent monitoring subsystem; wherein, when initiating one or more rounds of AI mediation by one or more AI mediators, and combining the multimodal mediation template of the mediation large model system and the context of the dialogue in the interactive system, the multimodal dialogue information is generated and sent to the interactive system: The mediation initiation subsystem is used to import mediation cases into the system and select a target AI mediator to transmit AI mediation parameters to the mediation dialogue subsystem; The mediation dialogue subsystem is used to generate multimodal dialogue information based on the AI ​​mediation parameters of the mediation initiation subsystem, the multimodal mediation template of the mediation library subsystem in the mediation large model system, the multi-dimensional intelligent annotation of the intelligent annotation subsystem and the context content of this mediation dialogue, so as to transmit it to the interactive subsystem in the interactive system; The intelligent annotation subsystem is used to receive text elements, audio text elements and video text elements of the interactive subsystem in the interactive system, and generate corresponding data tags to transmit the results of the multi-dimensional intelligent annotation to the mediation dialogue subsystem and the mediation library subsystem in the mediation large model system respectively; The intelligent monitoring subsystem is used to monitor the manual mediation process and check the mediation results.

4. The AI ​​mediation system according to claim 1, characterized in that: in, The mediation big model system includes: a mediation library subsystem, a mediation training subsystem, a mediation modeling subsystem and a third-party resource library subsystem; wherein, the mediation big model system is used to generate the multimodal mediation template according to the multimodal mediation data and multi-dimensional intelligent annotation results of the AI ​​mediator, and send it to the mediation engine system: The mediation library subsystem is used to construct core mediation data according to the output of one or more subsystems; The mediation training subsystem is used to optimize the multimodal mediation template using the AI ​​strategy feature set and the target machine learning model; The mediation modeling subsystem is used to integrate the target algorithm through a multi-core algorithm structure and generate a multi-modal mediation algorithm model for the mediation training subsystem to call; The third-party resource library subsystem is used to obtain the target database and the target large model.

5. The AI ​​mediation system according to claim 2, characterized in that: in, The responding to the multimodal information of the parties to drive the AI ​​mediator to conduct multimodal dialogue interaction operations with the parties includes: Collect the audio information of the parties, and transmit the target audio clip and the corresponding text element information to the intelligent annotation subsystem in the mediation engine system; Collect the video information of the parties involved, and identify their movements and micro-expressions, so as to transmit the target video clips and corresponding emotional texts to the intelligent annotation subsystem in the mediation engine system; Collect the instant messages of the parties during the interaction and transmit the corresponding text elements to the intelligent annotation subsystem in the mediation engine system; According to the received multimodal dialogue information of the mediation dialogue subsystem in the mediation engine system, the digital human subsystem is driven to conduct a multimodal dialogue with the parties, and the dialogue is transferred to the human mediator based on preset conditions; Based on the parties' written records, voice records, video records and the identified and organized text elements, audio elements, video elements, etc., the parties' multimodal mediation data is constructed and transmitted to the mediation library subsystem in the mediation big model system for multimodal storage.

6. The AI ​​mediation system according to claim 3, characterized in that: in, The monitoring of the manual mediation process and checking of the mediation results include: Retrieve the multimodal mediation supervision data of the mediation library subsystem in the mediation large model system, and obtain the multimodal mediation dialogue data of the manual mediator in real time; Identify whether the mediation behavior of human mediators conforms to the mediation process norms and prompt the mediators to make improvements in real time; Generate a mediation analysis report for the manual mediator, and combine it with the process data of mediation supervision to form multimodal mediation supervision data for transmission to the mediation library subsystem in the mediation large model system.

7. The AI ​​mediation system according to claim 4, characterized in that: in, The core mediation data is constructed according to the output of one or more subsystems, including: Obtain target legal information from the third-party repository subsystem; Obtain multimodal mediation data in the human mediation subsystem of the interactive system; Obtain multimodal mediation data in interactive subsystems of an interactive system; Obtain the multimodal intelligent annotation results of the intelligent annotation subsystem in the mediation engine system; generating multimodal mediation data based on the multimodal mediation process data for transmission to a mediation training subsystem; Receive the multimodal mediation templates of the trained mediators in the mediation training subsystem.

8. The AI ​​mediation system according to claim 4, characterized in that: in, The method of optimizing the multimodal mediation template using the AI ​​strategy feature set and the target machine learning model includes: Retrieve the multimodal mediation data in the mediation library subsystem, and retrieve the multimodal mediation algorithm model generated in the mediation modeling subsystem; Based on the existing multimodal mediation templates and multimodal mediation data, the multimodal mediation algorithm model is trained to optimize the multimodal mediation templates.

Images