Interactive artificial intelligence system based on multi-dimensional spatio-temporal information vector graphics

Abstract

To provide a mutual generation type artificial intelligence system which completely releases an engineering engineer and supports an intelligent system for automatic and quick mutual processing of voice, characters, multi-dimensional vector graphics, and a specific theme graphics text document in a related field.SOLUTION: A multi-modal spatio-temporal large-scale model is pre-trained for multi-dimensional spatio-temporal information such as a two-dimension, a three dimensional vector, and a time phase, to implement mutual representation and generation of engineering multi-dimensional vector graphics and a theme specific graphics and text document, so as to produce an intelligent engineering graphics and text data processing application. computer-aided design.SELECTED DRAWING: Figure 1

Description

[Technical Field]

[0001] The present invention relates to the field of artificial intelligence technology and engineering applications, and in particular to a mutually generative artificial intelligence system based on multidimensional spatiotemporal information vector graphics. [Background technology]

[0002] In recent years, artificial intelligence technologies, including large-scale models, have developed rapidly. Large-scale model technology is based on training on massive datasets and has demonstrated powerful semantic understanding and reasoning capabilities, promising for solving complex problems in fields such as speech, translation, medicine, education, finance, and the arts. Multimodal large-scale models combine large-scale models' natural language processing capabilities with their ability to understand and generate other modal data (e.g., visual, audio, etc.) to integrate multiple types of input and output, such as text, images, and voice, thereby mimicking the multimodal information processing characteristics of human cognition and providing a richer, more natural interaction experience. However, currently, AI technologies, including multimodal large-scale models, still process data types such as text, voice, images, and video, which are essentially text-based and raster-formatted. These technologies primarily serve fields such as literary and artistic creation, question-and-answer systems, speech translation, and data summarization analysis, but have yet to reach the human factors application field, which is characterized by vector-formatted data. From the moment we are born, humans have been confronted with a wealth of multimedia spatiotemporal data in the real world. These data types include text, audio, video, graphics, and images, as well as complex features such as numbers, text, attributes, metadata, interaction processes, and spatiotemporal relationships. This gives humans intelligence and even wisdom. Language is merely an organic component of multimedia spatiotemporal data, and humans' representation and interpretation of data are primarily text, vector, and raster graphics image formats. Artificial intelligence (AI) is not just about current text and statistical intelligence or machine vision-based spatial intelligence; it is also about spatiotemporal intelligence (STAI). Large-scale models will evolve into spatiotemporal large models (STLMs) in the future, and large language models (LLMs) are merely a subset of STLMs. Vector models are important and widely used in spatiotemporal intelligence processing systems.

[0003] A geographic information system (GIS) is a computer system whose task is to collect, store, manage, calculate, analyze, display, and describe the location and distribution of geospatial objects and solve user problems. Computer-aided design (CAD) is a computer system that uses computers and their graphics facilities to provide engineering designers with functions such as geometric modeling, feature calculation, and computer drafting. GIS and CAD are widely used in fields such as spatial data management, engineering design, and drafting, and have accumulated large amounts of graphics-related data such as electronic maps, engineering blueprints, and fabrication and construction drawings, all of which are primarily vector spatial data. Vector data is an inherently abstract data format, an abstract representation of spatial objects and engineering processes. Whether in GIS or CAD, vector graphics are drawn interactively by humans, and the creation of specific thematic graphics and text documents is carried out by humans based on their knowledge base and actual engineering needs. GIS allows for partial automatic processing due to its topological and spatial relationships, but CAD lacks the ability to fully describe spatiotemporal relationships and attributes, making it difficult to achieve intelligent processing. In general, current GIS and CAD-related vector graphics drawing and the processing of specific subject graphics and textual engineering documents are one-way rather than two-way intelligent mutual generation, making it impossible to achieve intelligent automatic processing.

[0004] While large-scale modeling technology enables the mutual generation of vector graphics and graphics / text documents, current research and applications of large-scale models still focus on text or raster-type data, including applications such as converting text to raster video, and rarely involve the modeling of vector data such as points, lines, surfaces, bodies, and networks. Therefore, to address the need to process a large number of multidimensional vector graphics and specific-theme graphics / text documents in the field of ergonomics, it is hoped that a system can be proposed that can understand, analyze, and process multidimensional spatiotemporal information vector graphics and mutually generate specific-theme graphics / text documents. Summary of the Invention [Problem to be solved by the invention]

[0005] In view of the above problems, the present invention proposes a mutually generative artificial intelligence system based on multidimensional spatiotemporal information vector graphics. [Means for solving the problem]

[0006] An embodiment of the present invention provides an interactive artificial intelligence system based on multidimensional spatiotemporal information vector graphics, which includes a multidimensional vector spatiotemporal large-scale model side, a multidimensional spatiotemporal information processing agent side, and an intelligent information system application side.

[0007] The multidimensional vector spatiotemporal large-scale model side uniformly models and represents multiple types of spatiotemporal data to construct a multimodal vector spatiotemporal large-scale model, and the vector spatiotemporal large-scale model is used for mutual representation, understanding, and analysis of multidimensional vector spatiotemporal data and natural language descriptions including voice and text, and is used to realize vertical domain knowledge reinforcement learning of the engineering expert knowledge system and enhance the processing ability of the engineering field knowledge and data processing flow, and the multiple types of spatiotemporal data include text, voice, image, video, and multidimensional vector graphics, The multidimensional spatiotemporal information processing agent side, based on the learning and inference capabilities of the vector spatiotemporal large-scale model, in combination with the processing and analysis capabilities of a geographic information system (GIS) or computer-aided design (CAD) platform, learns and refers to the task execution process of typical business flows, automatically analyzes the interaction business flow tasks of the multiple types of spatiotemporal data processing, decomposes the business flow tasks into simple subtasks, and autonomously executes, transmits, and solves the subtasks; The intelligent information system application side converts various multidimensional spatio-temporal data into corresponding natural language descriptions based on the inference and understanding capabilities of the vector spatio-temporal large-scale model, realizes artificial intelligence understanding of the multidimensional vector graphics and various multidimensional spatio-temporal data, and completes the interaction, analysis, and mutual generation between the agent-driven multidimensional vector graphics and the specific theme graphics and text documents in the engineering field through simple input methods including voice and text. Based on the inference of the vector spatio-temporal large-scale model and the processing capabilities of the agent, the parameters in the multidimensional spatio-temporal data are accurately obtained, and the related multidimensional spatio-temporal data is automatically updated, thereby realizing the automatic generation and update of the multidimensional vector graphics or the specific theme graphics and text documents.

[0008] Optionally, the domain knowledge of the engineering expert knowledge system is embedded in the multi-dimensional vector spatio-temporal large-scale model side to realize vertical domain knowledge reinforcement learning for the engineering expert knowledge system, and enhance the understanding and execution ability of the knowledge and data processing flow of the engineering field; The process of constructing the vector spatiotemporal large-scale model on the multidimensional vector spatiotemporal large-scale model side includes: natural language description of the multidimensional vector spatiotemporal data; pre-training and fine-tuning of the vector spatiotemporal large-scale model; and understanding and analysis of the multidimensional vector spatiotemporal data; Embedding the domain knowledge of the engineering expert knowledge system into the multi-dimensional vector spatio-temporal large-scale model and realizing vertical domain knowledge reinforcement learning for the engineering expert knowledge system includes: fine-tuning the general-purpose large-scale model using the data of the engineering field as a pre-training dataset, and expanding the lexicon of the engineering vertical domain, thereby pre-training the general-purpose large-scale model using the data of the engineering field, and realizing embedding the domain knowledge of the engineering expert knowledge system into the general-purpose large-scale model; Embedding domain knowledge of the processing flow of engineering graphics and text data into the multidimensional vector spatiotemporal large-scale model to enhance the understanding and execution ability of the engineering knowledge and data processing flow specifically includes describing a processing flow with clear engineering feature data, defining a language for the processing flow of engineering data, and training the general-purpose large-scale model with the uniformly expressed processing flow of engineering data as a fine-tuned data set, thereby realizing knowledge embedding of the data processing flow, and enhancing the understanding and execution ability of the general-purpose large-scale model for the processing flow of engineering data; The natural language description of the multidimensional vector spatio-temporal data is a natural language description and conversion of vector spatio-temporal data of various types, including but not limited to geographic information systems (GIS) and computer-aided drawing (CAD), which have continuous space-time (x, y, z, t) information and attribute information of points, lines, surfaces, and bodies, as needed for training, understanding, and processing of vector spatio-temporal large-scale models. The converted natural language includes all information of the original vector spatio-temporal data, including but not limited to the geometric type, geometric features, display style, reference point coordinates, relative geometric data based on the reference point, temporal information, and attribute information of the vector object. pre-training and fine-tuning the vector spatio-temporal large-scale model includes creating a dataset including natural language descriptions of the multidimensional vector spatio-temporal data and features of vector objects, and continuously pre-training and fine-tuning the general-purpose large-scale model based on the pre-training dataset and the fine-tuning dataset to form a vector spatio-temporal large-scale model capable of understanding the vector spatio-temporal data; The understanding and analysis of the multidimensional vector spatio-temporal data includes: inputting vector spatio-temporal data described in natural language into the vector spatio-temporal large-scale model, and outputting features of the vector object understood by the vector spatio-temporal large-scale model; inputting part of the features of the vector object into the vector spatio-temporal large-scale model, and outputting complete vector spatio-temporal data through understanding or analysis by the vector spatio-temporal large-scale model; and further completing the understanding and processing of geometric features of the vector spatio-temporal data, the understanding and processing of attribute features, the determination and processing of spatial relationships, and the determination and processing of spatio-temporal relationships by the analytical capabilities of the vector spatio-temporal large-scale model based on the spatio-temporal features or data of the output vector object; understanding and processing the geometric features of the vector spatiotemporal data includes processing the geometric coordinates of the vector objects themselves, including, but not limited to, changing and editing the shape, size, and position; The understanding and processing of the attribute features of the vector spatiotemporal data includes processing the attribute information of the vector object itself, including, but not limited to, modifying, querying, analyzing, and statistical processing; The determination and processing of the spatial relationships of the vector spatiotemporal data includes determining and processing the topological spatial relationships, ordinal spatial relationships, and metric spatial relationships between vector objects, where topological spatial relationships refer to the relationships of association, adjacency, inclusion, intersection, overlap, and separation between spatial objects, ordinal spatial relationships refer to the spatial arrangement order of spatial objects or events, including directional relationships such as front-back, left-right, up-down, east-west, north-south, and so on, and metric spatial relationships refer to the distance or perspective relationship between spatial objects.

[0009] Optionally, the multidimensional spatiotemporal information processing agent side includes a task planning module, a task storage module, and a task behavior module, wherein the task planning module, the task storage module, and the task behavior module each represent a corresponding processor.

[0010] The task planning module decomposes and plans subtasks of the vector graphics processing interaction workflow included in various multidimensional spatiotemporal data processing based on the domain knowledge, workflow, data comprehension, and text generation capabilities of the vector spatiotemporal large-scale model, where each subtask is a general logic processing or geographic information space analysis operation that is executed independently, has clear input and output, and executes the subtasks in conjunction with each other to complete the complex workflow processing; The task memory module allows each task to perceive information from the environment or obtain information from storage, provide data required for each task, and store process data during execution; the task memory module allows the multidimensional spatiotemporal information vector processing agent to accumulate data and experience, gradually undergo self-evolution, and provide the vector spatiotemporal large-scale model with iterative support; The task behavior module executes each subtask planned by the vector spatiotemporal large-scale model as a specific result, and the process of task execution depends on, but is not limited to, the inference capability of the vector spatiotemporal large-scale model, the spatial analysis and processing capability of a geographic information system, and the processing capability of a computer-aided drafting system, and the objects of interaction in the execution process include, but are not limited to, sensors, model libraries, controllers, and databases.

[0011] Optionally, the intelligent information system application side realizes artificial intelligence understanding of the multidimensional vector graphics and various multidimensional spatiotemporal data, and the steps of realizing the interaction, analysis and mutual generation between the multidimensional vector graphics and specific theme graphics and text documents include processing and understanding of multidimensional spatiotemporal data, generating a multidimensional spatiotemporal data processing agent, and mutual generation between the multidimensional vector graphics and specific theme graphics and text documents, specifically including the following steps:

[0012] Step S1: Multidimensional spatio-temporal data is input to the intelligent information system application side, and the multidimensional spatio-temporal data is converted into a natural language description that can be processed by the vector spatio-temporal large-scale model. Based on the inference and understanding capabilities of the vector spatio-temporal large-scale model, the multidimensional spatio-temporal information vector data and temporal data in the multidimensional spatio-temporal data are artificially analyzed and understood.

[0013] Step S2: Input user instructions using an interaction method including voice input and text input, which are converted into text by voice recognition. Based on the understanding, analysis, and processing capabilities of the vector spatiotemporal large-scale model, input the user instructions written in natural language and convert them into editing, querying, analyzing, and output instructions for formatted information system data, thereby generating a multidimensional spatiotemporal data processing agent.

[0014] Step S3: Based on the multidimensional spatiotemporal information processing agent, a mutually generating artificial intelligence application capable of automatic generation and two-way updating of automated and intelligent multidimensional vector graphics, specific theme graphics and text documents is realized by integrating and calling the multidimensional spatiotemporal information processing agent and executing it multiple times, thereby realizing the mutual generation of the multidimensional vector graphics and various multidimensional spatiotemporal data.

[0015] Optionally, creating a dataset including natural language descriptions of the multi-dimensional vector spatio-temporal data and features of vector objects, and continuously pre-training and fine-tuning the general-purpose large-scale model based on the pre-training dataset and the fine-tuning dataset, specifically includes: Step T1: collecting a large amount of vector spatiotemporal data and corresponding text natural language descriptions of the vector spatiotemporal data and feature descriptions of vector objects to form a vector spatiotemporal training dataset, wherein the vector spatiotemporal data includes predefined geometric structure information of points, lines, surfaces, and bodies, and attribute text information corresponding to the geometric bodies; Step T2: using the vector spatiotemporal training data set to pre-train the general-purpose large-scale model, sampling vector text as training samples by random walking, and inputting the vector information knowledge into the general-purpose large-scale model by GPT generation to fine-tune the model; Step T3 includes collecting multiple question-and-answer mode data including vector data-topological relationship, vector data-attribute information, and attribute description-vector data, forming a fine-tuning dataset for a vector spatiotemporal model, and fine-tuning the general-purpose large-scale model pre-trained in step T2 to effectively align natural language and vector data.

[0016] Optionally, the task planning module searches for similar tasks in an engineering flow library according to the task description, and the task planning method of the task planning module includes planning without feedback and planning with feedback, and the feedback may come from the environment, the user, or the execution result of the vector space-time large-scale model, and the business flow is decomposed into multiple subtasks and executed, and the multiple subtasks are connected in a cascade or tree shape, and subsequent subtasks are determined based on the task execution result after each subtask is completed.

[0017] Optionally, the task memory module stores information perceived from the environment, task execution records, and task execution results, and utilizes the recorded memories to facilitate future actions, the task memory module including stored data and operations; The stored data includes a short-term memory of input information within a context window and a long-term memory of an external vector store searched by a quick query.

[0018] Optionally, the task behavior module specifically uses tools including the vector spatiotemporal large-scale model itself and external tools including algorithm models, program compilations, databases, and APIs to perform each task, and also uses a spatial database to perform subspace analysis, uses a model in a model library to complete a specific task, and uses an API to obtain real-time or historical data, where the specific task refers to a task corresponding to each model in the model library.

[0019] Optionally, realizing the automatic generation of the multidimensional vector graphics or theme-specific graphics and text document comprises: According to the content requirements of the multidimensional vector graphics and the specific theme graphics and text document, manually create a multidimensional vector graphics or a specific theme graphics and text document template, or automatically generate a specific theme graphics and text document template by the vector-space-time large-scale model according to the content requirements, wherein the graphics in the specific theme graphics and text document are in a multidimensional vector graphics format or a raster format converted from the multidimensional vector graphics; Extracting and generating each type of parameter in the multidimensional vector graphics or the specific theme graphics and text document template through query analysis of an information system or understanding and inference of the vector spatiotemporal large-scale model, accurately obtaining the parameters in the multidimensional vector spatiotemporal data, and fusing the parameters with the template to generate the multidimensional vector graphics or the specific theme graphics and text document.

[0020] Optionally, the mutual generation of the multidimensional vector graphics and the specific theme graphics text document is performed by inputting the changes in the relevant spatiotemporal information of the engineering data or the specific theme graphics text document based on the understanding and analysis capabilities of the vector spatiotemporal large-scale model, and then generating parameters, descriptions or agents for updating the multidimensional vector graphics using the vector spatiotemporal large-scale model, and automatically updating the data of the multidimensional vector graphics; After the multidimensional vector graphics are drawn or modified to change the multidimensional vector graphics, the multidimensional vector graphics modified by the vector spatiotemporal large-scale model are obtained, the corresponding engineering parameters or document description are updated, and the related content of the specific theme graphics text document is further updated, thereby realizing the mutual generation of the multidimensional vector graphics and the specific theme graphics text document. [Effects of the Invention]

[0021] This invention builds a multidimensional vector spatiotemporal large-scale model, a multidimensional spatiotemporal information processing agent, and an intelligent information system application based on a geographic information system (GIS) or computer-aided design (CAD) platform and data source. For multidimensional spatiotemporal data such as two-dimensional and three-dimensional vectors and time phases, a large-scale model with the ability to understand engineering expert knowledge systems, data processing flow, multidimensional vector graphics, and the structure of topic-specific documents is pre-trained. This realizes the mutual representation and generation of multidimensional vector graphics, topic-specific spatiotemporal information, and specialized document descriptions, thereby creating an intelligent data processing system application in the engineering field. By creating an artificial intelligence system for the engineering field of human society, this invention fully liberates engineers and provides intelligent system support for the automatic and rapid mutual processing of voice, text, multidimensional vector graphics, topic-specific graphics, and text documents in related fields, making it highly practical. [Brief explanation of the drawings]

[0022] Various other benefits and advantages will become apparent to those skilled in the art upon reading the following detailed description of the preferred embodiments. The drawings are only for purposes of illustrating the preferred embodiments and are not to be construed as limiting the invention. Furthermore, like elements are designated by like reference numerals throughout the drawings.

[0023] [Figure 1] 1 is a block diagram showing the configuration of a multidimensional spatiotemporal information vector graphics interactive generation type artificial intelligence system according to an embodiment of the present invention; [Figure 2] 1 is a flowchart of an application based on a vector spatiotemporal large-scale model at the application side of an intelligent information system in an embodiment of the present invention; [Figure 3] 1 is a flowchart illustrating the interaction, analysis, and mutual generation between vector graphics and graphics-text documents based on a vector-space-time large-scale model in an embodiment of the present invention. [Figure 4] FIG. 10 is a schematic diagram of yet another embodiment of a multidimensional spatiotemporal information vector graphics interactive generation artificial intelligence system according to an embodiment of the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0024] In order to make the above-mentioned objects, features and advantages of the present invention more clearly understandable, the present invention will be described in more detail below with reference to the drawings and specific embodiments. It should be understood that the specific examples described in this specification are for the purpose of explaining the present invention, and are only some examples of the present invention, but are not all examples, and do not limit the present invention.

[0025] 1 shows a block diagram of an interactive artificial intelligence system based on multidimensional spatiotemporal information vector graphics according to this embodiment. This interactive artificial intelligence system specifically includes a multidimensional vector spatiotemporal large-scale model side, a multidimensional spatiotemporal information processing agent side, and an intelligent information system application side. An agent refers to an independent small module or application that can complete a single task, and an information processing agent refers to a small module for information processing.

[0026] The multidimensional vector spatiotemporal large-scale model is used to unify and represent multiple types of spatiotemporal data, including text, audio, images, videos, and multidimensional vector graphics.

[0027] The multidimensional vector space-time large-scale model side is also used to construct a multimodal vector space-time large-scale model that is used for the mutual representation, understanding, and analysis of multidimensional vector space-time data and natural language descriptions including speech and text, thereby realizing knowledge reinforcement learning in the vertical domain of the engineering expert knowledge system and enhancing the processing capabilities for the processing flow of knowledge and data in the engineering field.

[0028] Preferably, the domain knowledge of the engineering expert knowledge system is embedded in the multi-dimensional vector spatio-temporal large-scale model side, so as to realize vertical domain knowledge reinforcement learning for the engineering expert knowledge system, and to enhance the understanding and execution ability of the knowledge and data processing flow in the engineering field.

[0029] The process by which the multidimensional vector space-time large-scale model side constructs a vector space-time large-scale model includes: a natural language description of the multidimensional vector space-time data; pre-training and fine-tuning of the vector space-time large-scale model; and understanding and analysis of the multidimensional vector space-time data. While FIG. 1 exemplarily illustrates elements by which the multidimensional vector space-time large-scale model side constructs a vector space-time large-scale model to better represent the structure of the mutually generative artificial intelligence system, it does not indicate that the multidimensional vector space-time large-scale model side is used only for constructing a vector space-time large-scale model. The elements by which the multidimensional vector space-time large-scale model side constructs a vector space-time large-scale model include a natural language description of the vector space-time data (i.e., a natural language description of the multidimensional vector space-time data), pre-training and fine-tuning of the vector space-time large-scale model (i.e., pre-training and fine-tuning of the vector space-time large-scale model), and understanding and analysis of the vector space-time data (i.e., understanding and analysis of the multidimensional vector space-time data).

[0030] Among these, embedding domain knowledge of an engineering expert knowledge system into a multidimensional vector spatiotemporal large-scale model to realize vertical domain knowledge reinforcement learning for the engineering expert knowledge system specifically involves fine-tuning a general-purpose large-scale model using engineering data as a pre-training dataset and expanding the lexicon of the engineering vertical domain, thereby pre-training the general-purpose large-scale model using an engineering dataset and embedding domain knowledge of the engineering expert knowledge system into the general-purpose large-scale model. Here, the general-purpose large-scale model, also known as a "base large-scale model," refers to a large-scale language model across domains or tasks, typically trained by a large organization, with emphasis on the model's generalization and generalization capabilities. Meanwhile, the spatiotemporal large-scale model in this specification is a domain / vertical large-scale model that enhances specific capabilities based on a general-purpose large-scale model.

[0031] Embedding domain knowledge of engineering data processing flows into the multidimensional vector spatiotemporal large-scale model and enhancing the engineering knowledge and understanding and execution capabilities of the data processing flows specifically includes describing a processing flow with clear engineering feature data, defining a language for the engineering data processing flow, and training a general-purpose large-scale model using the uniformly expressed engineering data processing flow as a fine-tuned dataset to realize knowledge embedding of the data processing flow and enhance the general-purpose large-scale model's understanding and execution capabilities of the engineering data processing flow.

[0032] The natural language description of multidimensional vector spatiotemporal data is a natural language description and conversion of vector spatiotemporal data of various types, including points, lines, surfaces, and bodies, which have continuous space-time (x, y, z, t) information and attribute information of each domain, including but not limited to geographic information systems (GIS) and computer-aided drawing (CAD), in a natural language manner as needed for training, understanding, and processing of vector spatiotemporal large-scale models. The converted natural language contains all information of the original vector spatiotemporal data, including but not limited to the geometric type, geometric features, display style, reference point coordinates, relative geometric data based on the reference point, temporal information, and attribute information of the vector object.

[0033] Specifically, the pre-training and fine-tuning of the vector spatiotemporal large-scale model includes creating a dataset containing natural language descriptions of multidimensional vector spatiotemporal data and features of vector objects, and continuously pre-training and fine-tuning the general-purpose large-scale model based on the pre-training dataset and the fine-tuning dataset to form a vector spatiotemporal large-scale model capable of understanding the vector spatiotemporal data.

[0034] Preferably, creating a dataset including natural language descriptions of multi-dimensional vector spatio-temporal data and features of vector objects, and continuously pre-training and fine-tuning a general-purpose large-scale model based on the pre-training dataset and the fine-tuning dataset, specifically: Step T1: collecting a large amount of vector spatiotemporal data and corresponding text natural language descriptions of the vector spatiotemporal data and feature descriptions of vector objects to form a vector spatiotemporal training dataset, where the vector spatiotemporal data includes predefined geometric structure information of points, lines, surfaces, and bodies, and attribute text information corresponding to the geometric bodies; Step T2: using the vector spatiotemporal training dataset to pre-train a general-purpose large-scale model, sample vector text as training samples by random walking, and input the vector information knowledge into the general-purpose large-scale model by GPT generation to fine-tune the model; Step T3 includes collecting multiple question-and-answer mode data including vector data-topological relationship, vector data-attribute information, and attribute description-vector data, forming a fine-tuning dataset for a vector spatiotemporal model, and fine-tuning the general-purpose large-scale model pre-trained in step T2 to effectively align natural language and vector data.

[0035] By the above three-step method, a vector space-time large-scale model is constructed on the multidimensional vector space-time large-scale model side.

[0036] The understanding and analysis of multidimensional vector spatio-temporal data includes inputting vector spatio-temporal data described in natural language into a vector spatio-temporal large-scale model and outputting the features of the vector object understood by the vector spatio-temporal large-scale model; inputting part of the features of the vector object into the vector spatio-temporal large-scale model and outputting the complete vector spatio-temporal data through understanding or analysis by the vector spatio-temporal large-scale model; and further completing the understanding and processing of the geometric features, understanding and processing of attribute features, determination and processing of spatial relationships, and determination and processing of spatio-temporal relationships of the vector spatio-temporal data using the analytical capabilities of the vector spatio-temporal large-scale model based on the spatio-temporal features or data of the output vector object.

[0037] Understanding and processing the geometric features of vector spatiotemporal data includes processing the geometric coordinates of the vector objects themselves, including, but not limited to, modifying and editing the shape, size, and position; understanding and processing the attribute features of vector spatiotemporal data includes processing the attribute information of the vector objects themselves, including, but not limited to, modifying, querying, analyzing, and statistical processing; determining and processing the spatial relationships of vector spatiotemporal data includes determining and processing the topological spatial relationships, ordinal spatial relationships, and metric spatial relationships between vector objects, where topological spatial relationships refer to the relationships of association, adjacency, containment, intersection, overlap, and separation between spatial objects; ordinal spatial relationships refer to the spatial arrangement order of spatial objects or events, including directional relationships such as front-back, left-right, up-down, east-west, north-south, and so on; and metric spatial relationships refer to the distance or perspective relationship between spatial objects.

[0038] The multidimensional spatiotemporal information processing agent, based on the learning and inference capabilities of vector spatiotemporal large-scale models and combined with the processing and analysis capabilities of a geographic information system (GIS) or computer-aided design (CAD) platform, learns and references the task execution process of typical workflows, automatically analyzes multiple types of spatiotemporal data processing interaction workflow tasks, decomposes the workflow tasks into simple subtasks, and autonomously executes, communicates, and resolves the subtasks. Typical workflows refer to the steps and processes for processing spatiotemporal information data described herein, including GIS- and CAD-related data processing methods such as drawing graphics, editing, adding attributes, and spatial analysis and processing.

[0039] Preferably, the multidimensional spatiotemporal information processing agent side includes a task planning module, a task storage module, and a task behavior module.

[0040] The task planning module decomposes and plans subtasks of the vector graphics processing interaction workflow involved in various multidimensional spatiotemporal data processing based on the domain knowledge, workflow, data understanding, and text generation capabilities of the vector spatiotemporal large-scale model. Each subtask is an independently executed general logic process or geospatial analysis operation with distinct inputs and outputs. The subtasks are executed in conjunction with each other to complete complex workflow processing. Specifically, the task planning module first searches for similar tasks in an engineering workflow library according to the task description. The task planning method of the task planning module includes feedback-free planning and feedback-based planning, where feedback can come from the environment, the user, or the execution results of the vector spatiotemporal large-scale model. The workflow is decomposed into multiple subtasks and executed. The multiple subtasks are connected in a cascade or tree structure, and subsequent subtasks are determined based on the task execution results after each subtask is completed.

[0041] The task memory module is used for each task to perceive information from the environment or obtain information from storage, provide the data required for each task, and store process data during execution. The multidimensional spatiotemporal information vector processing agent uses the task memory module to accumulate data and experience, gradually self-evolve, and provide iterative support for the vector spatiotemporal large-scale model. Specifically, the task memory module stores information perceived from the environment, task execution records, and task execution results, and may use the recorded memories to promote future actions. The task memory module includes stored data and operations, and the stored data includes short-term memory of input information in the context window and long-term memory of external vector stores searched by quick query.

[0042] The task behavior module executes each subtask planned by the vector spatiotemporal large-scale model as a specific result, and the process of task execution depends on, but is not limited to, the inference capability of the vector spatiotemporal large-scale model, the spatial analysis and processing capability of the geographic information system, and the processing capability of the computer-aided cartography system, and the objects of interaction in the execution process include, but are not limited to, sensors, model libraries, controllers, and databases. Specifically, the task behavior module executes each task using tools including the vector spatiotemporal large-scale model itself and external tools including algorithm models, program compilations, databases, and APIs, and also performs subspace analysis using a spatial database, completes specific tasks using models in the model library, and obtains real-time or historical data using APIs, and a specific task refers to a task corresponding to each model in the model library.

[0043] In Fig. 1, three modules on the multidimensional spatiotemporal information processing agent side are shown as an example to better represent the structure of the mutually generating artificial intelligence system, but this does not mean that the multidimensional spatiotemporal information processing agent side includes only these three modules. The three modules on the multidimensional spatiotemporal information processing agent side include task planning (i.e., task planning module), task memory (i.e., task memory module), and task behavior (i.e., task behavior module).

[0044] The intelligent information system application uses the inference and understanding capabilities of the vector-space-time large-scale model to convert various multidimensional spatio-temporal data into corresponding natural language descriptions, thereby enabling artificial intelligence to understand multidimensional vector graphics and various multidimensional spatio-temporal data. Simple input methods, including voice and text, are used to complete the interaction, analysis, and mutual generation between agent-driven engineering multidimensional vector graphics and specific-theme graphics and text documents. The inference capabilities of the vector-space-time large-scale model and the agent's processing capabilities are used to accurately obtain parameters in the multidimensional spatio-temporal data, automatically update the related multidimensional spatio-temporal data, and automatically generate and update multidimensional vector graphics or specific-theme graphics and text documents. In this embodiment of the present invention, multidimensional vector graphics refer to two-dimensional, three-dimensional, and temporal-dimensional vector data used in engineering, and specific-theme graphics and text documents refer to document reports that combine graphics and text and have specific engineering meanings or uses.

[0045] More preferably, the intelligent information system application side realizes artificial intelligence understanding of multidimensional vector graphics and various multidimensional spatio-temporal data, and the steps of realizing the interaction, analysis, and mutual generation between multidimensional vector graphics and specific theme graphics and text documents include processing and understanding multidimensional spatio-temporal data, generating a multidimensional spatio-temporal data processing agent, and mutual generation between multidimensional vector graphics and specific theme graphics and text documents, where the multidimensional spatio-temporal data processing agent refers to an agent that processes reading, editing, analysis, etc. of multidimensional spatio-temporal data, and specifically includes the following steps:

[0046] Step S1: Input the multidimensional spatiotemporal data into the intelligent information system application, convert the multidimensional spatiotemporal data into a natural language description that can be processed by the vector spatiotemporal large-scale model, and based on the inference and understanding capabilities of the vector spatiotemporal large-scale model, artificially analyze and understand the multidimensional spatiotemporal information vector data and temporal data in the multidimensional spatiotemporal data.

[0047] Step S2: User instructions are input using an interaction method including voice input and text input, which are converted into text by voice recognition. Based on the understanding, analysis, and processing capabilities of vector spatiotemporal large-scale models, user instructions written in natural language are input and converted into editing, querying, analyzing, and output instructions for formatted information system data, thereby generating a multidimensional spatiotemporal data processing agent.

[0048] Step S3: Based on the multidimensional spatiotemporal information processing agent, by utilizing the fusion call and multiple execution of the multidimensional spatiotemporal information processing agent, an automated and intelligent mutual generation artificial intelligence application capable of automatic generation and bidirectional update of multidimensional vector graphics and specific theme graphics and text documents is realized, and the mutual generation of multidimensional vector graphics and specific theme graphics and text documents is realized.

[0049] To better illustrate the structure of the interactive AI system, Figure 1 exemplifies the intelligent information system application's AI understanding of multidimensional vector graphics and various multidimensional spatiotemporal data, and the interaction, analysis, and interactive generation of multidimensional vector graphics and specific-theme graphics and text documents. However, this does not imply that the intelligent information system application is limited to these technical capabilities. These capabilities include processing and understanding multidimensional spatiotemporal data (i.e., AI analysis and understanding of multidimensional spatiotemporal information vector data and temporal data in multidimensional spatiotemporal data), generating multidimensional spatiotemporal information processing agents (i.e., generating multidimensional spatiotemporal data processing agents), and interactive generation of multidimensional spatiotemporal information vector graphics and specific-theme graphics and text documents (i.e., interactive generation of multidimensional vector graphics and specific-theme graphics and text documents).

[0050] To realize the automatic generation of multidimensional vector graphics or specific theme graphics and text documents, specifically, First, according to the content requirements of the multidimensional vector graphics or the specific theme graphics and text document, manually create a multidimensional vector graphics or a specific theme graphics and text document template, or automatically generate a specific theme graphics and text document template according to the content requirements using a vector-space-time large-scale model, where the graphics in the specific theme graphics and text document are in a multidimensional vector graphics format or a raster format converted from multidimensional vector graphics; then, through query analysis of the information system or understanding and inference of the vector-space-time large-scale model, extract and generate each type of parameter in the multidimensional vector graphics or the specific theme graphics and text document template, accurately obtain the parameters in the multidimensional vector-space-time data; and finally, combine the parameters with the template to generate the multidimensional vector graphics or the specific theme graphics and text document.

[0051] The mutual generation of multidimensional vector graphics and specific theme graphics and text documents is based on the understanding and analysis capabilities of the vector spatiotemporal large-scale model. After inputting the changes in related spatiotemporal information from engineering data or specific theme graphics and text documents, the vector spatiotemporal large-scale model generates parameters, descriptions, or agents for updating multidimensional vector graphics, and automatically updates the data of multidimensional vector graphics. After the multidimensional vector graphics are drawn or modified to change the multidimensional vector graphics, the modified multidimensional vector graphics are obtained through the vector space-time large-scale model, the corresponding engineering parameters or document description are updated, and the related content of the specific theme graphics text document is further updated, thereby realizing the mutual generation of the multidimensional vector graphics and the specific theme graphics text document.

[0052] 2 is a flowchart of an application based on the vector-space-time large-scale model on the intelligent information system application side in this embodiment, which realizes the mutual representation and generation of natural language such as voice input and text input, and multidimensional vector graphics, specific theme graphics, text documents, etc. The "understanding and analysis of vector images and time phases," "vector image and time phase application commands," and "information processing agent" in FIG. 2 can be understood to correspond to the above-mentioned steps S1 to S3, and therefore will not be described in detail.

[0053] Figure 3 shows a flowchart for implementing the interaction, analysis, and mutual creation between multidimensional vector graphics and specific theme-specific graphics and text documents based on the vector-space-time large-scale model in this embodiment. Based on the understanding and analysis capabilities of the vector-space-time large-scale model, the system realizes the mutual creation and updating of various types of multidimensional spatiotemporal data, such as engineering data, graphics and text documents, and other spatiotemporal data sources, with various vector graphics (such as two-dimensional graphics, three-dimensional graphics, and other vector types) and data (vector graphics parameters, vector temporal parameters, vector graphics descriptions, and vector graphics agents). Here, the vector graphics agent refers to a small module that handles tasks related to vector graphics (e.g., reading, editing, saving, etc.).

[0054] As described above, this invention utilizes a geographic information system (GIS) or computer-aided design (CAD) platform and data source to build a multidimensional vector-space-time large-scale model side, a multidimensional space-time information processing agent side, and an intelligent information system application side. For multidimensional space-time data such as two-dimensional and three-dimensional vectors and time phases, a large-scale model capable of understanding engineering expert knowledge systems, data processing flow, multidimensional vector graphics, and the structure of topic-specific documents is pre-trained, enabling the mutual representation and generation of multidimensional vector graphics, topic-specific space-time information, and specialized document descriptions, thereby creating an intelligent data processing system application in the engineering field. By creating an artificial intelligence system for the engineering field of human society, this invention fully liberates engineers and provides intelligent system support for the automatic and rapid interaction of speech, text, and engineering multidimensional vector graphics, topic-specific graphics, and text documents in related fields, making it highly practical.

[0055] The above-described embodiment of the interactive artificial intelligence system based on multidimensional spatiotemporal information vector graphics is merely illustrative. The units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, i.e., they may be located in one place or distributed across multiple network units. Depending on actual needs, some or all of these modules may be selected to achieve the objectives of this embodiment. Those skilled in the art can understand and implement this embodiment without any creative effort.

[0056] Each component of the present disclosure may be implemented in hardware, software modules running on one or more processors, or a combination thereof. Those skilled in the art will appreciate that a microprocessor or digital signal processor (DSP) may be used to implement some or all of the functions of some or all of the components of an electronic device according to an embodiment of the present disclosure. The present disclosure may also be implemented as a device or apparatus program (e.g., a computer program and computer program product) for executing some or all of the methods described herein. Such a program implementing the present disclosure may be stored on a computer-readable medium or may take the form of one or more signals. Such signals may be downloaded from an Internet website, provided on a carrier signal, or provided in any other form.

[0057] For example, FIG. 4 illustrates an interactive artificial intelligence system based on multidimensional spatiotemporal information vector graphics, in which the present disclosure can be implemented. This interactive artificial intelligence system based on multidimensional spatiotemporal information vector graphics conventionally includes a processor 1010 and a computer program product or computer-readable medium in the form of a memory 1020. The memory 1020 may be an electronic memory such as a flash memory, an EEPROM (electrically erasable programmable read-only memory), an EPROM, a hard disk, or a ROM. The memory 1020 has a storage space 1030 for program code 1031 for executing the method steps of any of the above methods. For example, the program code storage space 1030 may include each program code 1031 used to implement the various steps of the above methods. These program codes may be read from or written to one or more computer program products. These computer program products include program code carriers such as a hard disk, a compact disc (CD), a memory card, or a floppy disk. Such computer program products are typically portable or fixed storage units. The storage unit may have storage segments, storage spaces, etc. arranged similarly to the memory 1020 in the interactively generative artificial intelligence system based on multidimensional spatiotemporal information vector graphics of FIG. 4. Program code compression may be performed, for example, in an appropriate manner. Typically, the storage unit includes computer-readable code, i.e., code that can be read by a processor, such as 1010, which, when executed by an electronic device, causes the electronic device to perform the steps of the above-described method.

[0058] Regarding the content of the mutually generating artificial intelligence system based on multidimensional spatiotemporal information vector graphics in this embodiment, the above embodiment can be referred to, so a detailed description will not be given here.

[0059] Although preferred embodiments of the present invention have been described, additional changes and modifications can be made to these embodiments by those skilled in the art once they have acquired the basic creative concept. Therefore, it is intended that the appended claims be interpreted as including all changes and modifications that fall within the scope of the preferred embodiments and embodiments of the present invention.

[0060] It should be noted that, in this specification, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply that such an actual relationship or order exists between these entities or operations. Furthermore, the terms "comprise," "include," or any other variant thereof are intended to encompass a non-exclusive inclusion, such that a process, method, article, or terminal device that includes a set of elements includes not only those elements but also other elements not expressly listed, or elements inherent in such process, method, article, or terminal device. Unless further limited, an element defined by the phrase "comprises..." does not exclude the presence of other identical elements in the process, method, article, or terminal device that includes that element.

[0061] Although the embodiments of the present invention have been described above with reference to the drawings, the present invention is not limited to the above-mentioned specific embodiments, which are schematic and not limiting. Those skilled in the art can implement many more forms under the teachings of the present invention without departing from the spirit of the present invention and the scope of the claims, and all of these fall within the scope of protection of the present invention.

Claims

1. An interactive artificial intelligence system based on multidimensional spatiotemporal information vector graphics, comprising a processor and a memory, wherein instructions are stored in the memory, and when the processor executes the instructions, A multi-modal vector spatio-temporal large-scale model is constructed by unifying the representation of multiple types of spatio-temporal data, and the vector spatio-temporal large-scale model is used to understand and analyze natural language descriptions of multi-dimensional vector spatio-temporal data, thereby enhancing the processing capabilities of engineering knowledge and data processing flow, where the types of spatio-temporal data include text, audio, image, video, and multi-dimensional vector graphics, and the natural language descriptions of the multi-dimensional vector spatio-temporal data include audio and text; Based on the learning and inference capabilities of the vector spatiotemporal large-scale model, combined with the processing and analysis capabilities of a geographic information system (GIS) or computer-aided design (CAD) platform, the system learns and references the task execution process of typical business flows, automatically analyzes the interaction business flow tasks of the multiple types of spatiotemporal data processing, decomposes the business flow tasks into simple subtasks, and autonomously executes, communicates, and solves the subtasks; Based on the inference and understanding capabilities of the vector space-time large-scale model, various multidimensional space-time data are converted into corresponding natural language descriptions, and through simple input methods including voice and text, the interaction, analysis, and mutual generation of agent-driven multidimensional vector graphics in the engineering field and specific theme graphics and text documents are completed; based on the inference of the vector space-time large-scale model and the processing capabilities of the agent, the parameters in the multidimensional space-time data are accurately obtained, and the related multidimensional space-time data are automatically updated, and further the automatic generation and update of the multidimensional vector graphics or the specific theme graphics and text documents is realized; The process of constructing the vector spatio-temporal large-scale model includes: a natural language description of the multi-dimensional vector spatio-temporal data; pre-training and fine-tuning the vector spatio-temporal large-scale model; and understanding and analyzing the multi-dimensional vector spatio-temporal data; The multidimensional vector spatiotemporal large-scale model uses engineering data as a pre-training dataset to fine-tune the general-purpose large-scale model and expands the vocabulary list of the engineering field, thereby pre-training the general-purpose large-scale model using the pre-training dataset to enhance the general-purpose large-scale model's knowledge of the engineering field and its understanding and execution ability of the data processing flow; The processor further comprises: The method is configured to describe a processing flow of data in the engineering field, define a language for the processing flow of data in the engineering field, train the general-purpose large-scale model using the defined processing flow of data in the engineering field as a fine-tuning dataset, and enhance the general-purpose large-scale model's ability to understand and execute the processing flow of data in the engineering field; The natural language description of the multidimensional vector spatio-temporal data includes vector spatio-temporal data of various types, such as points, lines, surfaces, and bodies, having continuous space-time (x, y, z, t) information and attribute information for each domain, including geographic information systems (GIS) and computer-aided cartography (CAD), and is described and converted in a natural language manner according to the needs of training, understanding, and processing the vector spatio-temporal large-scale model. The converted natural language description includes all information of the original vector spatio-temporal data, including the geometric type, geometric features, display style, reference point coordinates, relative geometric data based on the reference point, temporal information, and attribute information of the vector object. pre-training and fine-tuning the vector spatio-temporal large-scale model includes creating a dataset including natural language descriptions of the multidimensional vector spatio-temporal data and features of vector objects, and continuously pre-training and fine-tuning the general-purpose large-scale model based on the pre-training dataset and the fine-tuning dataset to form a vector spatio-temporal large-scale model capable of understanding the vector spatio-temporal data; The understanding and analysis of the multidimensional vector spatio-temporal data includes: inputting vector spatio-temporal data described in natural language into the vector spatio-temporal large-scale model, and outputting features of the vector object understood by the vector spatio-temporal large-scale model; inputting part of the features of the vector object into the vector spatio-temporal large-scale model, and outputting complete vector spatio-temporal data through understanding or analysis by the vector spatio-temporal large-scale model; and further completing the understanding and processing of geometric features of the vector spatio-temporal data, the understanding and processing of attribute features, the determination and processing of spatial relationships, and the determination and processing of spatio-temporal relationships by the analytical capabilities of the vector spatio-temporal large-scale model based on the spatio-temporal features or data of the output vector object; understanding and processing the geometric features of the vector spatiotemporal data includes processing geometric coordinates of the vector objects themselves; understanding and processing the attribute features of the vector spatiotemporal data includes processing attribute information of the vector objects themselves; determining and processing spatial relationships of the vector spatiotemporal data includes determining and processing topological spatial relationships, ordinal spatial relationships, and metric spatial relationships between the vector objects; creating a dataset including natural language descriptions of the multi-dimensional vector spatio-temporal data and features of vector objects, and continuously pre-training and fine-tuning the general-purpose large-scale model based on the pre-training dataset and the fine-tuning dataset; Step T1: collecting a large amount of vector spatiotemporal data and corresponding textual natural language descriptions of the vector spatiotemporal data and feature descriptions of the vector objects to form a vector spatiotemporal training dataset, wherein the vector spatiotemporal data includes predefined geometric structure information of points, lines, surfaces, and bodies, and attribute text information corresponding to the geometric bodies; Step T2: using the vector spatiotemporal training dataset to perform pre-training based on the general-purpose large-scale model, sampling vector texts by random walking, and inputting the sampled vector texts as training samples into the general-purpose large-scale model to fine-tune the model; Step T3: collecting a plurality of question-and-answer mode data, including vector data-topological relationship, vector data-attribute information, and attribute description-vector data, to form a fine-tuning dataset for a vector spatiotemporal model, and fine-tuning the general-purpose large-scale model pre-trained in step T2; The processor is further configured to search for a similar task in an engineering flow library according to the task description, and the task planning method includes planning without feedback and planning with feedback, and the processor is further configured to feed back execution results from an environment, a user, or a vector space-time large-scale model, and to decompose the business flow into multiple subtasks and execute them; The processor is further configured to interact, analyze, and mutually generate the multidimensional vector graphics and the specific theme graphics and text document, which steps include processing and understanding the multidimensional spatiotemporal data, generating the multidimensional spatiotemporal data processing agent, and mutually generating the multidimensional vector graphics and the specific theme graphics and text document, and further Step S1: inputting the multidimensional spatio-temporal data, converting the multidimensional spatio-temporal data into a natural language description that can be processed by the vector spatio-temporal large-scale model, and analyzing and understanding the multidimensional spatio-temporal information vector data and temporal data in the multidimensional spatio-temporal data in an artificial intelligence manner based on the inference and understanding capabilities of the vector spatio-temporal large-scale model; Step S2: inputting a user command through an interaction method including voice input and text input, and converting the user command written in natural language into a formatted information system data editing, query, analysis and output command based on the vector spatiotemporal large-scale model understanding, analysis and processing capability, thereby generating a multidimensional spatiotemporal data processing agent; Step S3: based on the multidimensional spatiotemporal information processing agent, by utilizing the fusion call and multiple execution of the multidimensional spatiotemporal information processing agent, to realize an interactive generation type artificial intelligence application capable of automatic generation and two-way update of automated and intelligent multidimensional vector graphics, specific theme graphics and text documents, and to realize the interactive generation of the multidimensional vector graphics and various multidimensional spatiotemporal data; The generating of the multidimensional vector graphics or the specific theme graphics and text document includes accurately obtaining parameters in the multidimensional vector spatiotemporal data, and then fusing the parameters with a template to generate the multidimensional vector graphics or the specific theme graphics and text document, in an interactive artificial intelligence system.

2. The interactive artificial intelligence system of claim 1, wherein processing of the geometric coordinates of the vector object itself includes changing and editing the shape, size, and position, and processing of the attribute information of the vector object itself includes modification, query, analysis, and statistics, the topological spatial relationship means the relationship of association, adjacency, containment, intersection, overlap, and separation between spatial objects, the order spatial relationship means the spatial arrangement order of the spatial objects or events, including directional relationships of front and back, left and right, up and down, east and west, north and south, and the distance spatial relationship means the distance or perspective relationship between the spatial objects.

3. The processor further comprises: Based on the domain knowledge, workflow, data comprehension, and text generation capabilities of the vector spatiotemporal large-scale model, tasks of the vector graphics processing interaction workflow included in various multidimensional spatiotemporal data processing are decomposed and planned, and each subtask is a general logic processing or geographic information spatial analysis operation that is executed independently and has clear input and output; Link subtasks together to complete complex workflows. When each task perceives information from the environment or acquires information from storage, provide the data required for each task and store the process data during execution. The multi-dimensional spatio-temporal information vector processing agent accumulates data and experience, gradually undergoes self-evolution, and provides the vector spatio-temporal large-scale model with support for repetitive ability; The interactive artificial intelligence system of claim 1, characterized in that each subtask planned by the vector-space-time large-scale model is executed as a concrete result, the dependencies of the task execution process include the inference capability of the vector-space-time large-scale model, the spatial analysis and processing capability of a geographic information system, and the processing capability of a computer-aided drafting system, and the interaction targets of the execution process are configured to include sensors, model libraries, controllers, and databases.

4. 2. The interactive artificial intelligence system according to claim 1, wherein the plurality of subtasks are connected in a cascade or tree configuration, and subsequent subtasks are determined based on the task execution results after each subtask is completed.

5. The processor is further configured to store information perceived from the environment, task performance records, and task performance results, and to utilize the recorded memory, including the stored data and operations, to facilitate future actions; 4. The interactive artificial intelligence system of claim 3, wherein the stored data includes a short-term memory of input information in the context window and a long-term memory of an external vector store searched by a quick query.

6. The processor is further configured to perform each task using tools including the vector spatiotemporal large-scale model itself and external tools including algorithmic models, program compilations, databases, and APIs, and to perform subspace analysis using a spatial database, complete specific tasks using models in a model library, and obtain real-time or historical data using APIs, wherein the specific tasks refer to tasks corresponding to each model in the model library.

7. The automatic generation of the multidimensional vector graphics or the specific theme graphics and text document is realized by: According to the content requirements of the multidimensional vector graphics and the specific theme graphics and text document, manually create a multidimensional vector graphics or a specific theme graphics and text document template, or automatically generate a specific theme graphics and text document template according to the content requirements by the vector-space-time large-scale model, wherein the graphics in the specific theme graphics and text document are in a multidimensional vector graphics format or a raster format converted from the multidimensional vector graphics; The interactive artificial intelligence system of claim 1, further comprising: extracting and generating each type of parameter in the multidimensional vector graphics or the specific theme graphics and text document template through query analysis of an information system or understanding and inference of the vector spatiotemporal large-scale model, accurately obtaining the parameters in the multidimensional vector spatiotemporal data, and fusing the parameters with the template to generate the multidimensional vector graphics or the specific theme graphics and text document.

8. The mutual generation of the multidimensional vector graphics and the specific theme graphics / text document is carried out by inputting the changes in the related spatiotemporal information due to the engineering data or the specific theme graphics / text document based on the understanding and analysis capabilities of the vector spatiotemporal large-scale model, and then generating parameters, descriptions or agents for updating the multidimensional vector graphics using the vector spatiotemporal large-scale model, and automatically updating the data of the multidimensional vector graphics; The interactive artificial intelligence system of claim 1, characterized in that after drawing or modifying the multidimensional vector graphics to change the multidimensional vector graphics, the modified multidimensional vector graphics are obtained through the vector-space-time large-scale model, the corresponding engineering parameters or document descriptions are updated, and the related content of the specific theme graphics / text document is further updated.

Images