AI organization information system and intelligent assistant thereof

Abstract

The invention relates to the technical field of organization management informatization, in particular to an AI organization information system and an intelligent assistant thereof, and the AI organization information system comprises an intelligent assistant module, a multi-mode interaction module, an AIGC modeling engine module and a data center module. Through deep integration of artificial intelligence, multi-modal interaction, meta-universe technology and big data management, the problems of technology islanding, single form, low efficiency, weak interactivity, difficulty in model generation and the like existing in a traditional organization informatization tool are solved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of organization management information technology, and in particular to an AI organization information system and an intelligent assistant thereof. BACKGROUND

[0002] With the rapid development of information technology, the digitalization and intelligentization transformation of organization management work has become a key path to strengthen grassroots team building in the new era. However, the existing organization information system generally has the following technical bottlenecks:

[0003] (1) Weak intelligent interaction capability: traditional systems rely on fixed knowledge base and rule engine, and it is difficult to dynamically process open management consultation. General large models lack professional knowledge in the field of organization management, and are prone to generate content that does not conform to the organization's charter and specifications, while professional organization management models are difficult to cover general scenarios, and it is urgent to establish a dual-model collaborative decision-making mechanism and real-time specification database linkage capability;

[0004] (2) Limited virtual meeting experience: online organization activities based on video conferencing have problems such as poor sense of presence and single interaction. The network latency of mainstream VR platforms often exceeds 100ms when multiple people interact in real time, resulting in strong dizziness for participants, and structured meeting records cannot be automatically aggregated from speech data;

[0005] (3) Low modeling efficiency: professional modelers spend weeks to complete 3D display space and other digital scenes, and existing AIGC tools have difficulty in precisely controlling organization identifiers and symbolic elements with strict specifications when generating building models, and cannot automatically adapt to specific cultural style parameters based on text descriptions;

[0006] (4) Data security and processing bottlenecks: sensitive data such as member information and meeting records are growing exponentially, and traditional centralized storage architecture is difficult to support high concurrency and real-time synchronization, and lacks hierarchical processing mechanisms for edge computing nodes. Speech and gesture data in multi-modal interaction are prone to compliance risks, and a real-time sensitive content interception system needs to be established.

[0007] In the current disclosed technology: an organization question and answer system based on NLP, but the problem of dynamic scheduling of multiple large models has not been solved; a general VR conference system, which does not design parallel space segmentation and content review layers for organization management scenarios; and an AIGC generated 3D model, but lacks a generation constraint mechanism for specific identifiers.

[0008] In summary, it is urgent to build an AI-assisted organization system that integrates large model decision-making, metaverse interaction, and secure modeling engine, breaking through the limitations of traditional technology in multi-modal intelligent response, real-time immersive meetings, compliant content generation, and distributed data governance. SUMMARY

[0009] To address the aforementioned technical problems, this invention provides an AI-powered organizational information system and its intelligent assistant.

[0010] In a first aspect, the present invention provides an AI-based organizational information system, comprising:

[0011] The intelligent assistant module includes an integrated open API interface and a large model gateway for dynamically accessing organization-specific large models and general large models;

[0012] A multimodal interaction module provides a 2D / 3D digital human avatar editor for users to customize the digital human's name, appearance, and response logic;

[0013] The VR meeting management module is based on the WebGL+WebRTC architecture to build a metaverse meeting space for real-time multi-user interaction, wherein the network transmission latency achieved by the metaverse meeting space is ≤30ms.

[0014] AIGC modeling engine module, which receives text, voice or sketch input and generates compliant 3D models of organizations;

[0015] The data platform module adopts a distributed storage architecture to manage the member database, policy and regulation database, and organizational activity records, and is used to achieve real-time synchronization of relevant data.

[0016] Preferably, the intelligent assistant module further includes:

[0017] A knowledge base management module based on the RAG enhanced architecture is used to realize multi-level permission knowledge retrieval and dynamic updates.

[0018] The dual-model collaborative routing module is used to automatically call the organization's dedicated large model or general large model based on the problem type.

[0019] Preferably, the VR meeting management module includes:

[0020] The parallel space dynamic segmentation module is used to automatically generate subspace copies of the metaverse meeting space based on the number of participants;

[0021] The four-dimensional feature extraction module integrates speech-to-text, speech time sequence graphs, document keyword frequency, and voting behavior data to automatically generate structured meeting minutes.

[0022] Preferably, the AIGC modeling engine module includes:

[0023] The diffusion model's element generator module is trained using training data containing specific cultural characteristics and presets prohibited areas in the generation model to avoid generating specific identifiers or charter content; and

[0024] The parametric model generation interface module receives text description input and automatically adjusts the building scale and material texture parameters of the generated organization 3D model.

[0025] Preferably, the multimodal interaction module further includes: a voice command recognition module; and a gesture-controlled digital human module.

[0026] Preferably, the data platform module adopts a three-cloud collaborative architecture, dynamically schedules the AI ​​service group through the Kubernetes container orchestration platform, and realizes hierarchical processing of data streams based on edge computing nodes.

[0027] Secondly, the present invention provides a smart assistant method for an AI organizational information system as described in the first aspect, comprising the following steps:

[0028] S71. Receive user query requests through the API gateway, and call the large model gateway to dispatch the query requests to the organization-specific model or general model.

[0029] S72. Combining the RAG architecture, information is retrieved in real time from the policy and regulation database of the data platform module to generate compliant answers;

[0030] S73. The compliant answer is output in a multimodal form as a response through the digital human image of the multimodal interaction module.

[0031] Preferably, the intelligent assistant method for AI organizational information systems further includes the following steps:

[0032] S74. Perform real-time sensitive content detection on user input content, wherein the sensitive content detection is filtered in accordance with relevant regulations;

[0033] S75. When sensitive or suspected illegal content is detected, a manual review mechanism is triggered to block the content.

[0034] S76. The SM4 encryption algorithm is used to encrypt and store the interactive data.

[0035] This invention provides a method for generating organizational metaverse space, employing an AI organizational information system as described in the first aspect and / or an intelligent assistant method as described in the second aspect, characterized by comprising the following steps:

[0036] S91, Receive input commands in the form of text, voice or sketch;

[0037] S92. Based on the AIGC modeling engine module, generate a 3D model of the organization according to the input instructions;

[0038] S93. Call the model review layer to perform topology verification and content compliance verification on the generated organization 3D model;

[0039] S94. Use the modular drag-and-drop editor to combine the verified 3D organizational models to generate standardized spatial templates;

[0040] S95 outputs a lightweight 3D metaverse scene.

[0041] Preferably, in step S92, the process of generating the 3D model of the tissue includes:

[0042] The model details are optimized based on the GAN network;

[0043] An attention mechanism is applied to compress model data to reduce model size;

[0044] By applying dynamic perception algorithms, the layout of the metaverse space is dynamically adjusted based on real-time collected user behavior data.

[0045] Compared with the prior art, the beneficial effects of the present invention are:

[0046] (1) This invention overcomes the problems of technological silos, single form, low efficiency, weak interactivity and difficulty in model generation of traditional organizational information tools by deeply integrating artificial intelligence, multimodal interaction, metaverse technology and big data management.

[0047] (2) This invention uses the dual-model collaborative routing mechanism of the intelligent assistant module to combine the advantages of the organization-specific large model and the general large model to accurately and efficiently handle different types of member inquiries and business requests, and significantly improve the professionalism, accuracy and coverage of the answers.

[0048] (3) The knowledge base management module of this invention based on the RAG architecture realizes real-time, accurate and permission-controlled knowledge retrieval and updating. Combined with dynamic retrieval of the central policy and regulation database, it ensures that the answers output by the intelligent assistant strictly comply with organizational norms and policy requirements, effectively avoiding the risk of information misleading.

[0049] (4) This invention combines a multimodal interaction module to provide a highly human-like, natural and intuitive user interface, enhancing the willingness to use and the learning experience.

[0050] (5) The VR meeting management module based on the WebGL+WebRTC architecture of this invention successfully realizes a multi-person real-time interactive metaverse meeting space with network transmission latency ≤30ms, providing an immersive collaborative experience with near-zero latency, and effectively overcoming the pain points of weak presence and low participation in remote meetings.

[0051] (6) Based on the parallel space dynamic segmentation module, this invention can flexibly generate subspace copies according to the number of participants, effectively improving the organizational efficiency and user experience of large-scale meetings. The four-dimensional feature extraction module integrates voice, time sequence, text keywords and behavioral data to automatically generate high-quality, structured meeting minutes, greatly reducing the burden of manual recording and improving the accuracy of information retention.

[0052] (7) The present invention supports multiple input forms such as text, voice or sketches based on the AIGC modeling engine module, and automatically generates 3D models that meet the requirements of the organization (such as organizational activity scenes, miniature models of cultural landmarks), which greatly reduces the technical threshold and time cost of professional modeling.

[0053] (8) The present invention uses data training containing specific cultural characteristics in the diffusion model cultural element generator module and presets a restricted area mechanism to ensure the compliance of the model generation process from the root. The parameterized model generation interface module allows users to quickly adjust key parameters such as building proportions and material textures of the model through simple descriptions. Combined with GAN network detail optimization, attention mechanism compression and other technologies, it ensures that the generated model is both high-quality and aesthetically pleasing as well as lightweight. The data platform module adopts a distributed storage architecture to uniformly integrate and manage multi-source heterogeneous data such as member databases, policy and regulation databases, and organizational activity records, so as to achieve unified data standards, centralized storage and real-time synchronization, and break down data silos. Based on the three-cloud collaborative architecture and Kubernetes container orchestration platform, it realizes efficient elastic scheduling and dynamic load balancing of background AI computing resources, and improves the overall processing capacity and resource utilization of the system. The edge computing nodes process data streams in a hierarchical manner to reduce the pressure on the central node and improve the response speed. Combined with real-time sensitive content detection (based on the Central Cyberspace Administration's thesaurus), manual review and interception mechanism and SM4 data encryption storage, it constructs a multi-layered and in-depth data security system to ensure that sensitive information is not leaked and illegal content is effectively intercepted.

[0054] (9) This invention is based on a method for generating metaverse space by combining intelligent assistant and AIGC engine, which realizes a rapid generation process from concept input to lightweight 3D metaverse scene output; the topology verification and content compliance verification of the model review layer ensure the standardization and security of the final space content; the dynamic perception algorithm adjusts the metaverse space layout in real time according to user behavior data, improving the adaptability and comfort of user experience. Attached Figure Description

[0055] Figure 1 A flowchart illustrating the information system process for AI organization;

[0056] Figure 2 This is a flowchart illustrating the process of using a smart assistant.

[0057] Figure 3A schematic diagram of the process for generating the metaverse space; Detailed Implementation

[0058] The technical solution of the present invention will be clearly and completely described below with reference to specific embodiments. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the protection scope of the present invention.

[0059] Example 1 Figure 1 As shown, the AI-powered organizational management information system includes the collaborative operation of the following modules:

[0060] The intelligent assistant module includes a dynamic model access module, a dual-model collaborative routing module, and a RAG knowledge base management module;

[0061] The dynamic model access module uses the KongAPI gateway to convert between multiple protocols such as HTTP / gRPC / WebSocket and access the Star Organization Management Large Model (40B parameters) and general large models (such as ChatGLM);

[0062] The dual-model collaborative routing module uses an SVM classifier to identify question types. Policy-related questions (such as "organizing meeting procedures") are routed to the dedicated management model, while routine inquiries (such as "meeting equipment usage") are routed to the general model. The routing accuracy is 98.7%, and the module is trained based on more than 2,000 labeled data.

[0063] The RAG knowledge base management module searches the ElasticSearch policy and regulation database in real time and automatically adds policy basis when generating answers; in the multi-level access control, provincial level can view data for the entire province, while branch level can only view data for its own organization.

[0064] The multimodal interaction module includes a digital human editor module and an interaction control module;

[0065] The Unity3D editor in the Digital Human Editor module supports uploading images to generate 2D / 3D avatars; the response logic settings adopt a strict mode / interactive mode.

[0066] The interactive control module uses an ASR engine for speech recognition (word error rate <5%), and a regular camera for skeletal joint point recognition (accuracy 92.7%). Raising a hand triggers the oath-taking process.

[0067] The VR meeting management module includes an underlying architecture module, a parallel space segmentation module, and an intelligent meeting minutes module.

[0068] The underlying architecture module adopts a dual-engine approach of WebGL + WebRTC, with dynamic bandwidth scheduling for edge computing nodes (500M bandwidth allocated per 100 participants); in the parallel space segmentation module, when the number of participants exceeds 300, subspace copies are automatically generated, and each subspace independently transmits audio and video streams; in the intelligent meeting minutes module, four-dimensional feature fusion includes speech-to-text transcription, speaking time sequence graphs, document keyword frequency, and voting behavior data, outputting structured minutes;

[0069] The AIGC modeling engine module includes a diffusion model generator module and a parameterized generation interface;

[0070] The diffusion model generator module contains 136 standard organizational identification models (badges / flags, etc.) in its training data and also presets prohibited areas such as prohibiting modification of the charter shape;

[0071] The parametric generation interface includes input text commands (such as "retro-style meeting room that can accommodate 50 people") and automatically adjusts the space size and material texture;

[0072] The data platform module includes a storage architecture module and a three-cloud collaboration module;

[0073] Storage architecture module: MySQL cluster stores member information, MongoDB stores activity records, and Apache Spark processes behavioral data in real time;

[0074] The three-cloud collaboration includes private cloud, public cloud, and edge nodes, and also covers Kubernetes dynamic scheduling of GPU resources.

[0075] like Figure 2 As shown, the implementation of the intelligent assistant method includes:

[0076] Request processing: The user inputs "How to conduct an organizational review of members?" via voice; the API gateway receives the input and transmits it to the large model gateway;

[0077] Model routing and response: The dual-model collaborative module identifies policy-related issues, retrieves relevant Article 20 of the management regulations using the RAG architecture, and broadcasts the information via a 3D digital human after SM4 encryption;

[0078] Security controls: Real-time filtering of sensitive words (response time <200ms), manual review triggered when illegal content is detected;

[0079] like Figure 3 As shown, the implementation of the metaverse space generation method includes:

[0080] Model generation: Input a hand-drawn sketch of a "circular oath platform", and generate a 3D draft by spreading the model;

[0081] During compliance reviews, topology verification detects issues such as floating / clipping; logo proportion reviews ensure that badge proportions comply with relevant national standards.

[0082] Space construction: Modular drag-and-drop includes a combination of a swearing-in platform and a "historical bookshelf", and uses a lightweight processing attention mechanism to compress the model to 3MB.

[0083] The specific embodiments further describe the present invention. However, it should be understood that the specific description herein should not be construed as limiting the nature and scope of the present invention. Various modifications made to the above embodiments by those skilled in the art after reading this specification are all within the scope of protection of the present invention.

Claims

1. An AI-based organizational information system, characterized in that, include: The intelligent assistant module includes an integrated open API interface and a large model gateway for dynamically accessing organization-specific large models and general large models; A multimodal interaction module provides a 2D / 3D digital human avatar editor for users to customize the digital human's name, appearance, and response logic; The VR meeting management module is based on the WebGL+WebRTC architecture to build a metaverse meeting space for real-time multi-user interaction, wherein the network transmission latency achieved by the metaverse meeting space is ≤30ms. AIGC modeling engine module, which receives text, voice or sketch input and generates compliant 3D models of organizations; The data platform module adopts a distributed storage architecture to manage the member database, policy and regulation database, and organizational activity records, and is used to achieve real-time synchronization of relevant data.

2. The AI ​​organizational information system according to claim 1, characterized in that, The intelligent assistant module also includes: A knowledge base management module based on the RAG enhanced architecture is used to realize multi-level permission knowledge retrieval and dynamic updates. The dual-model collaborative routing module is used to automatically call the organization's dedicated large model or general large model based on the problem type.

3. The AI ​​organizational information system according to claim 1, characterized in that, The VR conference management module includes: The parallel space dynamic segmentation module is used to automatically generate subspace copies of the metaverse meeting space based on the number of participants; The four-dimensional feature extraction module integrates speech-to-text, speech time sequence graphs, document keyword frequency, and voting behavior data to automatically generate structured meeting minutes.

4. The AI ​​organizational information system according to claim 1, characterized in that, The AIGC modeling engine module includes: The diffusion model organization element generator module is trained using training data containing specific cultural characteristics and presets prohibited areas in the generated model to avoid generating specific logos or charter content; and the parametric model generation interface module receives text description input and automatically adjusts the building scale and material texture parameters of the generated organization 3D model.

5. The AI ​​organizational information system according to claim 1, characterized in that, The multimodal interaction module also includes: a voice command recognition module; and a gesture-controlled digital human module.

6. The AI ​​organizational information system according to claim 1, characterized in that, The data platform module adopts a three-cloud collaborative architecture, dynamically schedules AI service groups through the Kubernetes container orchestration platform, and realizes hierarchical processing of data streams based on edge computing nodes.

7. A method for an intelligent assistant applied to an AI organizational information system as described in any one of claims 1-6, characterized in that, Includes the following steps: S71. Receive user query requests through the API gateway, and call the large model gateway to dispatch the query requests to the organization-specific model or general model. S72. Combining the RAG architecture, information is retrieved in real time from the policy and regulation database of the data platform module to generate compliant answers; S73. The compliant answer is output in a multimodal form as a response through the digital human image of the multimodal interaction module.

8. The intelligent assistant method according to claim 7, characterized in that, It also includes the following steps: S74. Perform real-time sensitive content detection on user input content, wherein the sensitive content detection is filtered in accordance with relevant regulations; S75. When sensitive or suspected illegal content is detected, a manual review mechanism is triggered to block the content. S76. The SM4 encryption algorithm is used to encrypt and store the interactive data.

9. A method for generating an organizational metaverse space, employing an AI organizational information system as described in any one of claims 1-6 and / or an intelligent assistant method as described in claims 7-8, characterized in that, Includes the following steps: S91, Receive input commands in the form of text, voice or sketch; S92. Based on the AIGC modeling engine module, generate a 3D model of the organization according to the input instructions; S93. Call the model review layer to perform topology verification and content compliance verification on the generated organization 3D model; S94. Use the modular drag-and-drop editor to combine the verified 3D organizational models to generate standardized spatial templates; S95 outputs a lightweight 3D metaverse scene.

10. The method for generating organizational metaverse space according to claim 9, characterized in that, In step S92, the process of generating the 3D model of the tissue includes: The model details are optimized based on the GAN network; An attention mechanism is applied to compress model data to reduce model size; By applying dynamic perception algorithms, the layout of the metaverse space is dynamically adjusted based on real-time collected user behavior data.

Images