A personalized virtual space automatic construction method based on user intention

Abstract

The present application relates to the technical field of virtual scene construction, and discloses a personalized virtual space automatic construction method based on user intention, which comprises the following steps: receiving a natural language query text input by a user, calling a preset model context protocol platform, and constructing a multi-element database corresponding to a virtual space construction intention vector; extracting key information entities corresponding to the virtual space construction intention vector from the multi-element database, and calculating the semantic similarity between the virtual space construction intention vector and the key information entities; establishing a virtual space configuration set corresponding to the multi-element database to generate an interactive preview view corresponding to the virtual space construction intention vector; and generating a personalized virtual space based on the multi-element database, the virtual space configuration set and the interactive preview view. The present application can make the virtual space construction process deeply match the real-time intention of the user, and improve the user adaptation flexibility of the virtual space.

Description

Technical Field

[0001] This invention relates to a method for automatically constructing personalized virtual spaces based on user intent, belonging to the field of virtual scene construction technology. Background Technology

[0002] Personalized virtual spaces are dynamic digital scene systems built around user intent using Unity engine technology. They are designed to precisely match users' core needs in different scenarios such as entertainment, work, and study, aiming to provide users with a highly customized, immersive, and interactive virtual experience.

[0003] However, traditional virtual spaces are mainly built based on predefined scripts and closed static resources. For example, the position and interaction rules of scene elements are set through fixed code logic. Although this method can ensure the stability of the scene in the early stage, its interaction mechanism relies entirely on the pre-set process. It cannot understand the user's real-time intentions, nor can it be integrated with external dynamic data (such as ecological monitoring information). Ultimately, this results in a serious lack of flexibility in user operation and adaptability of the scene.

[0004] Therefore, there is an urgent need for a solution that can deeply align the virtual space construction process with the user's real-time intentions and improve the user adaptability flexibility of the virtual space. Summary of the Invention

[0005] This invention provides a method for automatically constructing personalized virtual spaces based on user intent. Its main purpose is to make the virtual space construction process deeply match the user's real-time intent and improve the user adaptability and flexibility of the virtual space.

[0006] To achieve the above objectives, the present invention provides a method for automatically constructing a personalized virtual space based on user intent, comprising:

[0007] The system receives natural language query text input by the user to generate a virtual space construction intent vector for the user, calls a preset model context protocol platform, and constructs a multivariate database corresponding to the virtual space construction intent vector. The multivariate database includes an environmental element library, an object instance library, and an interaction logic library.

[0008] Configure the multidimensional attribute meta tags of the multidimensional database to establish the inverted index and feature vector index of the multidimensional database. Based on the inverted index and the feature vector index, extract the key information entities corresponding to the virtual space construction intent vector from the multidimensional database.

[0009] Calculate the semantic similarity between the virtual space construction intent vector and the key information entity, and based on the semantic similarity, generate the semantic embedding vector and spatial topological constraints corresponding to the virtual space construction intent vector;

[0010] Based on the semantic embedding vector and the spatial topology constraints, a virtual space configuration set corresponding to the multivariate database is established, and based on the virtual space configuration set, the corresponding entity data in the multivariate database is parsed.

[0011] By combining the entity data, the semantic embedding vector, and the spatial topology constraints, an interactive preview view corresponding to the virtual space construction intent vector is generated. Based on the multivariate database, the user's personalized virtual space is generated through the virtual space configuration set and the interactive preview view.

[0012] Optionally, receiving natural language query text input by the user to generate the user's virtual space construction intent vector includes:

[0013] The user's desired virtual space type is determined by the natural language query text.

[0014] Extract the set of spatial entity objects, spatial relation operators, and spatial attribute modifiers from the natural language query text;

[0015] Based on the set of spatial entity objects, the spatial relation operators, and the spatial attribute modifiers, construct a spatial topology relation graph of the required virtual space type;

[0016] Based on the spatial topology graph, the multi-granularity layout constraints of the required virtual space type are analyzed.

[0017] Based on the spatial topology graph and the multi-granularity layout constraints, the user's virtual space construction intent vector is generated.

[0018] Optionally, the step of invoking a preset model context protocol platform to construct a multivariate database corresponding to the virtual space construction intent vector includes:

[0019] The spatial environment elements, entity object elements, and interactive behavior elements in the virtual space construction intent vector are extracted.

[0020] Based on the data flow module of the model context protocol platform, the spatial environment elements, the entity object elements and the interactive behavior elements are converted into standardized data units to obtain environment data units, object data units and behavior data units.

[0021] Based on the tool invocation module of the model context protocol platform, environmental semantic tags, functional semantic tags, and behavioral semantic tags are extracted from the environmental data unit, the object data unit, and the behavioral data unit.

[0022] Based on the environmental semantic tags, the functional semantic tags, and the behavioral semantic tags, a cross-modal association set of the environmental data unit, the object data unit, and the behavioral data unit is constructed;

[0023] Based on the association strength value in the cross-modal association set, calculate the data call association weight for each data unit in the environmental data unit, the object data unit, and the behavior data unit;

[0024] Based on the data call association weights, a multi-source data scheduling priority strategy is generated for the cross-modal association set;

[0025] Based on the multi-source data scheduling priority strategy and the cross-modal association set, a multi-dimensional database corresponding to the virtual space construction intent vector is constructed.

[0026] Optionally, establishing the virtual space configuration set corresponding to the multivariate database based on the semantic embedding vector and the spatial topological constraints includes:

[0027] The environmental element library, object instance library, and interaction logic library in the multivariate database are parsed out.

[0028] Extract spatial style semantics and functional intent semantics from the semantic embedding vector;

[0029] Based on the spatial style semantics, candidate environmental elements in the environmental element library are selected;

[0030] Based on the functional intent semantics, candidate object instances are retrieved from the object instance library;

[0031] The spatial topology constraints are decomposed into environmental layout constraints and object relationship constraints.

[0032] Based on the environmental layout constraints, generate environmental spatial layout schemes and environmental element instances corresponding to the candidate environmental elements.

[0033] Based on the object relationship constraints and the environmental spatial layout scheme, output the set of object instances corresponding to the candidate object instances and their spatial pose parameters;

[0034] Based on the functional intent semantics, relevant candidate interaction logic is matched from the interaction logic library;

[0035] Calculate the compatibility coefficient between the candidate interaction logic and the object relationship constraints;

[0036] Based on the compatibility coefficient, the target interaction logic is selected from the candidate interaction logics;

[0037] By integrating the environmental spatial layout scheme, the environmental element instances, the object instance set, the spatial pose parameters, and the target interaction logic, a virtual space configuration set corresponding to the multivariate database is established.

[0038] Optionally, calculating the compatibility coefficient between the candidate interaction logic and the object relationship constraints includes:

[0039] Analyze the interaction space requirements of the candidate interaction logic, and calculate the spatial matching degree between the interaction space requirements and the object relationship constraints;

[0040] Identify the strictness level of the object relationship constraints and determine the constraint strength coefficient corresponding to the strictness level;

[0041] Parse the functional intent tags of the candidate interaction logic, and calculate the semantic matching degree between the functional intent tags and the current virtual space scene;

[0042] Identify the computational complexity characteristics corresponding to the candidate interaction logic and the urgency of real-time requirements corresponding to the object relationship constraints;

[0043] Based on the computational complexity characteristics and the urgency of the real-time requirements, the system load impact factor of the candidate interaction logic and the object relationship constraints is comprehensively extracted.

[0044] Combining the spatial matching degree, the constraint strength coefficient, the contextual semantic matching degree, and the system load impact factor, the compatibility coefficient between the candidate interaction logic and the object relationship constraints is calculated using the following formula:

[0045]

[0046] in, Indicates the compatibility factor. Indicates the constraint strength coefficient. The weighting coefficient represents the balance between spatial matching and contextual semantic matching. Indicates spatial matching degree, Indicates the semantic matching degree of the context. Indicates the system load impact factor. This represents the weight balance coefficient of the system load impact factors.

[0047] Optionally, configuring the multidimensional attribute meta tags of the multivariate database includes:

[0048] Extract the core data features of the environmental element library, object instance library, and interaction logic library from the multivariate database;

[0049] Based on the core data features, a set of general dimension labels for the multivariate database is defined;

[0050] The spatial location attributes, physical quantity types, and spatiotemporal granularity of the environmental element database are parsed to generate environment-specific meta tags for the multivariate database.

[0051] Extract the entity type, functional purpose, and geometric features from the object instance library to determine the object class-specific meta tags of the multivariate database;

[0052] Analyze the logic triggering mode, response time and scope of the interaction logic library to generate interaction-specific meta tags for the multi-dimensional database;

[0053] By integrating the general dimension tag set, the environment-specific meta tags, the object-specific meta tags, and the interaction-specific meta tags, the multidimensional attribute meta tags of the multivariate database are constructed.

[0054] Optionally, configuring the multidimensional attribute meta tags of the multivariate database to establish the inverted index and feature vector index of the multivariate database includes:

[0055] Based on the multidimensional attribute meta tags, the meta tag key-value pairs of the multivariate database are parsed out;

[0056] Extract the semantic features of the meta-tag key-value pairs;

[0057] Based on the semantic features of the meta-tags, a set of meta-tag feature vectors for the multivariate database is generated;

[0058] The feature vector index of the multivariate database is established using the meta-label feature vector set.

[0059] Identify high-frequency and rare keywords in the meta tag key-value pairs to calculate the keyword weight of the meta tag key-value pairs;

[0060] The weights of the inverted index items in the multivariate database are determined by the keyword weights.

[0061] Based on the weights of the inverted index items, a weighted inverted list of the multivariate database is generated;

[0062] Based on the weighted inverted list, an inverted index is established for the multivariate database.

[0063] Optionally, generating an interactive preview view corresponding to the virtual space construction intent vector by combining the entity data, the semantic embedding vector, and the spatial topological constraints includes:

[0064] The entity data is mapped to a preset Unity engine, and the environment element instance, object instance and interaction logic instance corresponding to the entity data are generated through the instantiation interface of the Unity engine.

[0065] Based on the semantic embedding vector, the semantic style preferences and functional interaction requirements of the virtual space construction intent vector are parsed out in order to adjust the multidimensional environment configuration parameters of the environmental element instances.

[0066] Based on the spatial topology constraints, generate the object pose configuration scheme for the object instance;

[0067] Extract the triggering conditions and response scripts of the interaction logic instance, and dynamically bind the triggering conditions, the response scripts, and specific events of the object instance through the Unity engine's event delegation system to obtain the interaction behavior logic chain;

[0068] By integrating the multi-dimensional environment configuration parameters, the object pose configuration scheme, and the interaction behavior logic chain, a Unity lightweight real-time rendering scene corresponding to the virtual space construction intent vector is constructed.

[0069] Based on the cluster center distance of the semantic embedding vector, the adaptive rendering optimization parameters of the Unity lightweight real-time rendering scene are defined.

[0070] Based on the lightweight real-time rendering scene and the adaptive rendering optimization parameters, an interactive preview view corresponding to the virtual space construction intent vector is generated.

[0071] Optionally, generating the semantic embedding vector corresponding to the virtual space construction intent vector based on the semantic similarity includes:

[0072] Based on the semantic similarity, candidate key information entities corresponding to the virtual space construction intent vector are selected;

[0073] Construct a multidimensional association matrix between the candidate key information entities and the virtual space construction intent vector;

[0074] Extract the principal component eigenvectors and their corresponding importance weights from the multidimensional correlation matrix;

[0075] Based on the importance weights, the principal component feature vectors are subjected to weighted fusion processing to obtain weighted fused feature vectors;

[0076] The weighted fusion feature vector is mapped to a low-dimensional semantic space to obtain the low-dimensional semantic embedding vector corresponding to the candidate key information entity.

[0077] Calculate the cluster center distance and distribution density of the low-dimensional semantic embedding vector in the low-dimensional semantic space;

[0078] Based on the cluster center distance and the distribution density, the low-dimensional semantic embedding vector is normalized to obtain the semantic embedding vector corresponding to the virtual space construction intent vector.

[0079] Optionally, generating the spatial topological constraints corresponding to the virtual space construction intent vector based on the semantic similarity includes:

[0080] Extract the spatial relation predicates from the virtual space construction intent vector, and determine the spatial constraint strength coefficients corresponding to the spatial relation predicates;

[0081] Based on the semantic similarity, calculate the semantic-feature coupling weights between different spatial relation predicates;

[0082] Based on the spatial constraint strength coefficient and the semantic-feature coupling weight, a priority ranking rule for the spatial relation predicate is established.

[0083] Based on the priority sorting rules, the constraint logic combination method and execution order corresponding to the virtual space construction intent vector are determined;

[0084] Based on the spatial constraint strength coefficient, define the constraint tolerance range corresponding to the virtual space construction intention vector;

[0085] By integrating the constraint logic combination method, the execution order, and the constraint tolerance range, spatial topological constraint conditions corresponding to the virtual space construction intent vector are generated.

[0086] Compared to the problems described in the background technology, this embodiment of the invention generates a virtual space construction intent vector by receiving natural language query text input by the user. This allows for targeted optimization of the intent-driven generation logic and dynamic data integration strategy for virtual spaces. Furthermore, by invoking a preset model context protocol platform to construct a multivariate database corresponding to the virtual space construction intent vector, this embodiment enhances the resource support and dynamic adaptation capabilities of the multivariate database for the entire virtual space construction process, ensuring the efficiency of personalized virtual space automatic construction and its matching degree with user needs. By configuring multidimensional attribute meta-tags in the multivariate database to establish its inverted index and feature vector index, this embodiment breaks the closed static attributes of virtual space data resources, forming a data management system that can be quickly retrieved and dynamically associated. This provides an efficient data retrieval framework for real-time capture of user intent and integration with external dynamic data (such as ecological monitoring information), ensuring the virtual space responds quickly to changes in user needs and updates to external data. Furthermore, by generating a semantic embedding vector corresponding to the virtual space construction intent vector based on the semantic similarity, this embodiment deeply integrates the semantic features of the user's real-time intent with the core features of key information entities. This invention accurately maps the dynamic adaptation relationship between user query needs and virtual space resources, significantly enhancing the virtual space's ability to understand user intent and integrate external dynamic data. Furthermore, by generating spatial topological constraints corresponding to the user query request based on semantic similarity, this invention deeply integrates the semantic features of the user's real-time intent with the core features of key information entities, accurately mapping the dynamic adaptation relationship between user query needs and virtual space resources, significantly enhancing the virtual space's ability to understand user intent and integrate external dynamic data. Further, by generating spatial topological constraints corresponding to the virtual space construction intent vector based on semantic similarity, this invention accurately anchors the association logic between user intent and virtual space elements, deeply coupling user demand orientation and scene resource distribution characteristics, significantly improving the virtual space's response accuracy to user real-time operation intent and its dynamic adaptation capability to scene topology. Finally, by combining the entity data, the semantic embedding vector, and the spatial topological constraints, this invention generates an interactive preview view corresponding to the virtual space construction intent vector, transforming the abstract configuration scheme of the virtual space into a visual and operable intuitive presentation, accurately associating user intent with the actual presentation effect of the virtual space.Finally, this embodiment of the invention generates personalized virtual spaces based on the multi-dimensional database, the virtual space configuration set, and the interactive preview view. This allows the virtual space construction process to deeply align with the dynamic expression needs of users' real-time intentions and the integration and adaptation requirements of external dynamic data. Simultaneously, it avoids user control rigidity caused by reliance on fixed code logic or external data integration obstacles caused by closed resource systems. This ensures the virtual space construction process deeply aligns with users' real-time intentions, improving the user adaptability and flexibility of the virtual space. Therefore, this embodiment of the invention provides a method for automatically constructing personalized virtual spaces based on user intentions, enabling the virtual space construction process to deeply align with users' real-time intentions and improving the user adaptability and flexibility of the virtual space. Attached Figure Description

[0087] Figure 1 This is a flowchart illustrating a method for automatically constructing a personalized virtual space based on user intent, as provided in an embodiment of the present invention.

[0088] Figure 2 This invention provides an embodiment of a method for automatically constructing personalized virtual spaces based on user intent.

[0089] User intent parsing flowchart;

[0090] Figure 3 A data interaction flowchart of a method for automatically constructing a personalized virtual space based on user intent, provided in an embodiment of the present invention;

[0091] Figure 4 This is a schematic diagram of a module for implementing a personalized virtual space automatic construction system based on user intent, provided as an embodiment of the present invention.

[0092] The objectives, features, and advantages of this invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0093] It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

[0094] This application provides a method for automatically constructing personalized virtual spaces based on user intent. The executing entity of this method includes, but is not limited to, at least one of the following electronic devices that can be configured to execute the method provided in this application: a server, a terminal, etc. In other words, the method for automatically constructing personalized virtual spaces based on user intent can be executed by software or hardware installed on a terminal device or a server device. The server includes, but is not limited to, a single server, a server cluster, a cloud server, or a cloud server cluster.

[0095] Reference Figure 1 The diagram shown is a flowchart illustrating a method for automatically constructing a personalized virtual space based on user intent, according to an embodiment of the present invention. In this embodiment, the method for automatically constructing a personalized virtual space based on user intent includes:

[0096] S1. Receive natural language query text input by the user to generate the user's virtual space construction intent vector, and call a preset model context protocol platform to construct a multivariate database corresponding to the virtual space construction intent vector, wherein the multivariate database includes an environmental element library, an object instance library, and an interaction logic library.

[0097] This invention, through receiving natural language query text input by a user, generates a virtual space construction intent vector for that user. This allows for targeted optimization of the intent-driven generation logic and dynamic data integration strategy for virtual spaces. The user refers to an entity that proposes personalized virtual space construction needs and conveys these needs through natural language input. This includes individual users (such as gamers and virtual office users) and institutional users (such as virtual exhibition hall design companies and educational content development institutions). Their core demand is to obtain a customized virtual space in the Unity engine that suits their specific usage scenarios (such as entertainment, office, and teaching). The natural language query text refers to the text content input by the user to express their virtual space construction needs, conforming to everyday human language habits. It is the direct linguistic carrier of the user's intent and must include the core functions of the virtual space (such as interaction and display), scene style (such as science fiction and retro), spatial composition (such as room layout and object placement), and special requirements (such as character interaction and lighting effects). The virtual space construction intent vector refers to vector data with a numerical expression form generated after semantic parsing and feature extraction of the user's input natural language query text using a large language model. It is the digital carrier of the user's virtual space construction needs.

[0098] As an embodiment of the present invention, receiving natural language query text input by the user to generate the user's virtual space construction intent vector includes:

[0099] The user's desired virtual space type is determined by the natural language query text.

[0100] Extract the set of spatial entity objects, spatial relation operators, and spatial attribute modifiers from the natural language query text;

[0101] Based on the set of spatial entity objects, the spatial relation operators, and the spatial attribute modifiers, construct a spatial topology relation graph of the required virtual space type;

[0102] Based on the spatial topology graph, the multi-granularity layout constraints of the required virtual space type are analyzed.

[0103] Based on the spatial topology graph and the multi-granularity layout constraints, the user's virtual space construction intent vector is generated.

[0104] The virtual space type refers to the functional category or style paradigm of the virtual space that the user intends to construct. It is defined by the core intent words in the natural language query text and serves as the overall constraint framework for subsequent space generation. For example, if a user inputs "I want a super-modern, technologically advanced conference room that can hold a meeting for 20 people," then "conference room" is the virtual space type. The space entity object set refers to the collection of all objects or structures that need to be instantiated in the virtual space, whether explicitly mentioned or implicitly included in the user's query text. For example, if a user inputs "Construct a home office space with 2 wooden desks, 1 floor lamp, and 3 interactive virtual assistants," the extracted space entity object set would be {desk (furniture category, 2 desks, 1 floor lamp, 3 floor lamps, 1 floor lamp, 1 floor lamp, 3 interactive virtual assistants)}. The list includes: floor lamp (decorative, 1 unit), virtual assistant (interactive, 3 units); the spatial relation operators refer to predicates or prepositional phrases that describe the relative position, orientation, or topological relationship between entities in the spatial entity object set, including positional relation operators (such as "next to," "above," "near"), interaction relation operators (such as "operable," "connected," "triggered"), and subordinate relation operators (such as "belongs to," "contains"); the spatial attribute modifiers refer to adjectives or quantitative terms used to describe the visual, physical, or stylistic characteristics of each entity or the entire space in the spatial entity object set. For example, if a user queries "place a large sofa and a white marble coffee table," then "large," "...", ... "White" and "marble" are spatial attribute modifiers; the spatial topology graph refers to an abstract description of the layout of a virtual space represented by a graph data structure. Its nodes are entity instances generated from a set of spatial entity objects, and the edges are defined by spatial relation operators. Both nodes and edges can be annotated with spatial attribute modifiers. For example, for the requirement that "the desk is located next to the floor-to-ceiling window, the desk is made of white glass, and the floor lamp illuminates and covers the desk," the constructed spatial topology graph has nodes "desk (ID1, attribute: white glass)", "floor-to-ceiling window (ID2)", and "floor lamp (ID3, attribute: illumination coverage)", and edges "ID1 - located next to... -> ID2 (weight 0.8)" and "ID...". 3-Light Coverage->ID1 (Weight 1.0)”; The multi-granularity layout constraints refer to the set of rules at different abstract levels derived from the spatial topology graph to guide the geometric generation of virtual space, including global constraints (rules that act on the entire space, such as space size, core functional area division, overall orientation, and main pedestrian channel settings), local constraints (rules that act on specific groups of entities or between entities, such as the center point of the TV should be on the same horizontal line as the center point of the sofa, and the coffee table must be within 0.5 meters in front of the sofa), and single entity constraints (rules that act on a single entity, mainly derived from spatial attribute modifiers, such as the size parameters of the sofa should be (length: 2m, width: 1m, height: 0.8m)).

[0105] Optionally, the user's required virtual space type can be determined by a large language model semantic parsing method, such as using global semantic features of natural language query text, employing intent classification or semantic clustering algorithms for space type matching, and generating space type identifiers to characterize the function and style category of the virtual space; the spatial topology relationship map of the required virtual space type can be generated by a graph structure construction method, such as using graph neural networks or knowledge graph technology for relation reasoning and fusion based on spatial entity object sets, spatial relation operators, and spatial attribute modifiers, and generating a topology map to characterize the structural associations and constraints between entities in the virtual space.

[0106] To clearly demonstrate the structured translation process from natural language to intent, please refer to [link / reference]. Figure 2 The diagram shown is a user intent parsing flowchart of a personalized virtual space automatic construction method based on user intent provided in an embodiment of the present invention. The flowchart clearly presents the entire process framework of user intent parsing and structured query generation, from "user input" to "text cleaning and word segmentation", then to "using large model recognition", "context processing", "structured organization", and finally "calling backend services". It can serve as the basic path for processing natural language query text to support subsequent operations such as determining virtual space type, extracting elements, and constructing graphs during the process of generating user virtual space construction intent vectors. By combining the correlation and influence weights of each link, the true correlation logic from text input to intent vector generation is accurately restored, thereby improving the efficiency and accuracy of virtual space construction intent vector generation.

[0107] Furthermore, by invoking a preset model context protocol platform, this embodiment of the invention constructs a multi-dimensional database corresponding to the virtual space construction intent vector. This enhances the multi-dimensional database's resource support and dynamic adaptation capabilities throughout the entire virtual space construction process, ensuring the efficiency of automatic construction of personalized virtual spaces and the degree of matching with user needs.

[0108] The Model Context Protocol Platform refers to an integrated platform with standardized communication protocols and modular functional components, designed to enable collaborative interaction between the large language model and the Unity engine during the automatic construction of personalized virtual spaces. Its core function is to manage message passing, tool invocation, state maintenance, data flow, and resource scheduling between the large language model and the Unity engine through a unified interface specification. This ensures efficient collaboration in the parsing of virtual space construction intent vectors, resource matching, and scene generation. It is a key middleware connecting natural language intent understanding and virtual space visualization construction. The platform ensures the orderly bidirectional transmission of messages through a request / response management module, supports the large language model in calling Unity's scene editing tools through a tool invocation module, synchronizes the construction progress and entity status of the Unity scene in real time through a state context module, processes the continuous interaction between intent vectors and resource data through a data flow module, and enables the large language model to accurately query and invoke the Unity resource library. The Multi-Dimensional Database refers to a multi-dimensional, computable, and associative virtual space-specific data management system built based on virtual space construction intent vectors. This system integrates spatial environment information, entity object information, and interaction behavior information, including an environment element library, an object instance library, and an interaction logic library.

[0109] As an embodiment of the present invention, the step of invoking a preset model context protocol platform to construct a multivariate database corresponding to the virtual space construction intent vector includes:

[0110] The spatial environment elements, entity object elements, and interactive behavior elements in the virtual space construction intent vector are extracted.

[0111] Based on the data flow module of the model context protocol platform, the spatial environment elements, the entity object elements and the interactive behavior elements are converted into standardized data units to obtain environment data units, object data units and behavior data units.

[0112] Based on the tool invocation module of the model context protocol platform, environmental semantic tags, functional semantic tags, and behavioral semantic tags are extracted from the environmental data unit, the object data unit, and the behavioral data unit.

[0113] Based on the environmental semantic tags, the functional semantic tags, and the behavioral semantic tags, a cross-modal association set of the environmental data unit, the object data unit, and the behavioral data unit is constructed;

[0114] Based on the association strength value in the cross-modal association set, calculate the data call association weight for each data unit in the environmental data unit, the object data unit, and the behavior data unit;

[0115] Based on the data call association weights, a multi-source data scheduling priority strategy is generated for the cross-modal association set;

[0116] Based on the multi-source data scheduling priority strategy and the cross-modal association set, a multi-dimensional database corresponding to the virtual space construction intent vector is constructed.

[0117] The spatial environment elements refer to the core elements extracted from the virtual space construction intent vector, used to define the overall environmental style and basic scene parameters of the virtual space. These include environmental style attributes (such as science fiction, retro, minimalist), spatial physical parameters (such as space size, terrain structure), and environmental atmosphere parameters (such as light intensity, color temperature, background sound effects, and weather effects). The entity object elements refer to the core attributes of interactive or static entities that constitute the "microscopic entity structure" of the virtual space, extracted from the virtual space construction intent vector. These include entity categories (such as furniture, equipment, decorations), basic entity attributes (such as material, size, and color), and entity functional attributes (such as whether they are interactive and the conditions for interaction triggering). The interactive behavior elements refer to the core elements parsed from the intent vector constructed in the virtual space, defining the interaction logic between the user and entity objects or the virtual space. These include interaction triggering methods (such as clicks, gestures, and voice), interaction execution actions (such as opening documents, playing videos, and switching scenes), and interaction feedback effects (such as pop-up prompts and sound effects). The data flow module refers to the core functional module in the model context protocol platform responsible for "data format conversion, multi-source data flow, and data transmission adaptation." It can convert spatial environment elements, entity object elements, and interactive behavior elements into standardized data units through preset protocols (such as JSON-RPC and WebSocket), realizing element data... It is compatible with bidirectional data transmission and formatting of multi-dimensional database sub-libraries and the Unity engine; the standardized data unit refers to a "structured data carrier" that conforms to the model context protocol platform transmission specifications and multi-dimensional database storage requirements after being processed by the data flow module. It is divided into environment data units, object data units, and behavior data units according to data type. Each unit contains a data identifier (such as ID), data type, core parameters, and associated indexes (identifiers associated with other units). The environment data unit refers to a sub-type of the corresponding spatial environment element within the standardized data unit, specifically storing structured data related to the virtual space environment, including environment style identifiers, spatial physical parameters (size, terrain), and environmental atmosphere parameters (…). The data includes lighting, sound effects, and environmental data association indexes (such as the associated entity object data unit ID); the object data unit refers to a sub-type of the corresponding entity object element in the standardized data unit, which specifically stores structured data related to virtual space entity objects, including entity category, basic attributes (material, size), functional attributes (interactivity), and object data association indexes (such as the associated triggered behavior data unit ID); the behavior data unit refers to a sub-type of the corresponding interactive behavior element in the standardized data unit, which specifically stores structured data related to virtual space interaction logic, including triggering method, execution action, feedback effect, and behavior data association indexes (such as the associated triggered object data unit ID).The tool invocation module refers to the core functional module in the model context protocol platform responsible for "invoking external tools (such as semantic analysis tools and resource query tools) and executing data processing logic (such as semantic tag extraction and association rule generation)". It calls tools through preset interfaces and feeds back the processing results to the data flow module or resource management module, supporting semantic association and data scheduling of multi-dimensional databases. The environmental semantic tags refer to the "concise semantic identifiers" extracted by the tool invocation module after semantic analysis of environmental data units, representing the core characteristics of the environmental data. These include environmental style tags (such as "sci-fi" and "retro") and environmental parameter tags (such as "high illumination - 500 lux" and "large space - 10m × 8m"). This system facilitates the rapid matching of environmental data with other types of data (such as entity object data). The functional semantic tags refer to the "concise semantic identifiers" extracted by the tool's calling module after semantic analysis of object data units, representing the core functions and attribute characteristics of entity objects. These include category tags (e.g., "furniture-desk," "equipment-electronic screen"), function tags (e.g., "static-non-interactive," "dynamic-click-interactive"), and attribute tags (e.g., "material-wood," "size-1.2m"), used to match the association between entity objects and their environment and interactive behaviors. The behavioral semantic tags refer to the "concise semantic identifiers" extracted by the tool's calling module after semantic analysis of behavioral data units, representing the core characteristics of interactive logic. The "recognition" includes trigger tags (such as "trigger-click O001"), action tags (such as "action-play video"), and feedback tags (such as "feedback-sound effect"), used to match the association between interactive behaviors and entity objects. The cross-modal association set refers to a structured set constructed based on environmental semantic tags, functional semantic tags, and behavioral semantic tags to describe the semantic associations between environmental data units, object data units, and behavioral data units. Each association contains four core fields: "source data unit ID - association type - target data unit ID - association strength value." The association strength value refers to the value used in the cross-modal association set to quantitatively describe the semantic relationship between the source data unit and the target data unit. The density index ranges from [0,1]. A value closer to 1 indicates a stronger association (e.g., an object-behavior trigger association strength of 0.92 indicates a strong binding), while a value below 0.5 indicates a weaker association (e.g., a sci-fi environment and a wooden desk association strength of 0.45 indicates low compatibility). The data call association weight refers to a weight index calculated based on the association strength value in the cross-modal association relationship set, combined with the "core contribution" of the data unit in the construction of the virtual space. This weight index is used to characterize the priority of data unit calls and ranges from [0,1]. For example, the weight of a core data unit (e.g., an electronic screen object that triggers interaction) is close to 1, while the weight of an auxiliary data unit (e.g., decorative background sound effects) is close to 0.3.The multi-source data scheduling priority strategy refers to a rule based on the call association weights of all data units, used to guide the order in which data is transmitted from a multi-source database to the Unity engine. This strategy guides the system to prioritize which data to process when resources are limited; for example, data units with a weight higher than a threshold of 0.7 are scheduled first; for data with the same weight, environmental data takes precedence over object data, and object data takes precedence over behavioral data.

[0118] Optionally, the conversion of the spatial environment elements, entity object elements, and interactive behavior elements into standardized data units can be achieved using a data flow module combined with a template mapping-based standardization conversion algorithm. For example, based on the type differences of the spatial environment elements, entity object elements, and interactive behavior elements, corresponding preset templates are called (environment templates contain a "style-physical parameter" field, object templates contain a "category-functional attribute" field, and behavior templates contain a "trigger method-execution action" field). The element parameters are mapped one by one according to the template fields, and dataID and relatedID are automatically generated, outputting standardized data in JSON format. The data units achieve a unified data format. The data retrieval association weight of each data unit in the environmental data unit, the object data unit, and the behavioral data unit can be calculated using a weighted fusion algorithm. For example, based on the core association strength value of the data unit (e.g., the association strength between object data unit O001 and behavioral data unit B001 is 0.92), the core contribution value (e.g., the core contribution value of an interactive electronic screen is 0.9), and other factors, a normalized weighting process is performed according to the formula "Data retrieval association weight = Core association strength value × 0.6 + Core contribution value × 0.4" to generate a weight value in the [0,1] interval to characterize the data unit's retrieval association weight. Priority is used; the multi-source data scheduling priority strategy for the cross-modal association set can be generated based on the weight threshold division algorithm of K-means clustering combined with the level-channel mapping rule. For example, according to the call association weight of all data units, the threshold is determined by K-means clustering (K=3) (Level 1 ≥ 0.85, Level 2 0.6-0.85, Level 3 < 0.6), and then the scheduling level is divided according to the threshold and the loading channel is matched (Level 1 corresponds to the cache channel, Level 2 corresponds to the regular high-speed channel), generating a structured strategy containing "data unit ID-scheduling level-loading channel" to guide the data loading order; the virtual space The multi-dimensional database corresponding to the intent vector can be constructed by integrating cross-modal association set, multi-source data scheduling priority strategy and Unity resource loading channel collaborative construction algorithm. For example, according to the multi-source data scheduling priority strategy, the corresponding channel is activated through the resource management module (first-level data goes through the high-speed cache channel) to retrieve Unity resources (electronic screen 3D model, interactive script), and the data unit is stored in the corresponding sub-database (object instance library, interactive logic library). At the same time, the association set (the adaptation relationship between E001 and O001) and scheduling strategy are stored to form a three-in-one database of "data storage-association index-scheduling configuration".

[0119] S2. Configure the multidimensional attribute meta tags of the multidimensional database to establish the inverted index and feature vector index of the multidimensional database. Based on the inverted index and the feature vector index, extract the key information entities corresponding to the virtual space construction intent vector from the multidimensional database.

[0120] This invention, through configuring multidimensional attribute meta tags of the multivariate database, establishes an inverted index and feature vector index for the multivariate database. This breaks the closed, static attributes of virtual space data resources, forming a data management system that can be quickly retrieved and dynamically associated. It provides an efficient data retrieval framework for real-time capture of user intent and integration with external dynamic data (such as ecological monitoring information), ensuring that the virtual space responds quickly to changes in user needs and updates to external data.

[0121] The multidimensional attribute meta-tags refer to a standardized set of tags that cover multidimensional attribute features and are configured for each data unit (such as environmental data unit, object data unit, and behavioral data unit) in the environment element library, object instance library, and interaction logic library of the multidimensional database. These tags include object name, theme category, style attribute, applicable scenario, and complexity index. The inverted index refers to an index structure that uses the attribute tags of data resources as index items to point in reverse to the storage location of the specific data resource containing the tag. The feature vector index refers to an index structure that abstracts the semantics, functions, and relationships of data into high-dimensional numerical vectors. Its core is to achieve fast similarity retrieval and association matching of data by measuring the similarity between high-dimensional vectors (such as distance and angle).

[0122] As an embodiment of the present invention, configuring the multidimensional attribute meta tags of the multivariate database includes:

[0123] Extract the core data features of the environmental element library, object instance library, and interaction logic library from the multivariate database;

[0124] Based on the core data features, a set of general dimension labels for the multivariate database is defined;

[0125] The spatial location attributes, physical quantity types, and spatiotemporal granularity of the environmental element database are parsed to generate environment-specific meta tags for the multivariate database.

[0126] Extract the entity type, functional purpose, and geometric features from the object instance library to determine the object class-specific meta tags of the multivariate database;

[0127] Analyze the logic triggering mode, response time and scope of the interaction logic library to generate interaction-specific meta tags for the multi-dimensional database;

[0128] By integrating the general dimension tag set, the environment-specific meta tags, the object-specific meta tags, and the interaction-specific meta tags, the multidimensional attribute meta tags of the multivariate database are constructed.

[0129] The environmental element library refers to a sub-database used to store and manage all elements constituting the macro background and static framework of the virtual space; the object instance library refers to a sub-database used to store and manage all identifiable and interactive discrete entities in the virtual space; the interaction logic library refers to a sub-database used to store and manage the rules, scripts, and logic governing how relationships occur between control objects, between objects and the environment, and between users and the virtual world; the core data features refer to the common attributes and key parameters extracted from the original data of the environmental element library, object instance library, and interaction logic library, reflecting their essence, including the essential attributes and key parameters of key information entities in the sub-databases, including unique identifiers (IDs), data names, data sources, creation times, version numbers, data types, etc.; the general dimension tag set This refers to a standardized set of tags, uniformly defined and applicable to all sub-databases, based on the common core data features of the environmental element library, object instance library, and interaction logic library in a multi-dimensional database. Its core function is to establish a unified classification benchmark across sub-databases, enabling universal identification and management of different data types. The spatial location attribute refers to the association information between environmental data in the environmental element library and virtual spatial geographic coordinates and regional divisions. For example, in a virtual urban traffic control scenario, the spatial location attributes of traffic flow data for a certain road segment include: absolute coordinates in the virtual scene (e.g., X: 120.5°, Y: 30.2°, Z: 0m, corresponding to real-world geographic coordinate conversion values), regional affiliation (e.g., eastern urban area - Zhongshan Road segment - from Zhongshan Road intersection to Renmin Road intersection), and spatial range (e.g., length 1.2km, lane width 3m).(5m / 4 lanes in total); The physical quantity type refers to the category of physical attributes represented by environmental data in the environmental element database. Its core function is to distinguish the measurement attributes and functional uses of environmental data, ensuring that the physical meaning of environmental data is clearly identifiable. This includes physical quantity types such as mechanical, thermal, electromagnetic, optical, and fluid dynamics. The spatiotemporal granularity refers to the set of granularities describing the precision and update frequency of environmental data in spatial and temporal details. It is mainly determined by the technical parameters of the data acquisition equipment or the configuration parameters of the generation algorithm. The environmental class-specific meta-tags refer to the set of meta-tags specifically defined to describe the characteristics of environmental elements in a multivariate database. The entity type refers to the category classification of entity objects in the object instance database. The function Purpose refers to the preset purpose or executable operation of an entity object; geometric morphological characteristics refer to the geometric shape, size parameters, and appearance of entity objects in the object instance library within the virtual scene; object class-specific meta tags refer to a set of meta tags specifically defined to describe the characteristics of object instances in a multivariate database; logical triggering mode refers to the method and conditions under which interactive behavior is activated in the interactive logic library; response timeliness refers to the maximum allowable delay time from the fulfillment of triggering conditions to the complete execution of interactive logic; scope refers to the range of target objects or spatial areas that a piece of interactive logic can affect; interactive class-specific meta tags refer to a set of meta tags specifically defined to describe the characteristics of interactive logic in a multivariate database.

[0130] Optionally, the core data features of the environmental element library, object instance library, and interaction logic library in the multivariate database can be extracted using principal component analysis. For example, by using the feature parameters of each element as variables and the data distribution characteristics as the analysis target, dimensionality reduction processing can be performed on the multidimensional data to extract key feature vectors and variance contribution rates, thus determining the core data feature set. The functional uses in the object instance library can be extracted using text mining techniques. For example, by using instance description text as corpus and functional keywords as extraction targets, word frequency statistics and semantic analysis can be used to construct a function-instance mapping relationship table and extraction accuracy indicators, thus clarifying the functional use classification system. The logical triggering modes of the interaction logic library can be analyzed using a state machine model. For example, by using logical triggering conditions as input and state transition paths as output, multiple sets of interaction records can be modeled to generate trigger rule sets and pattern matching parameters, thus determining typical triggering mode types. The scope of the interaction logic library can be determined using boundary value analysis. For example, by using the scope of objects involved in the interaction as the research object and boundary conditions as the judgment criteria, the context of logic execution can be defined, the scope coverage and boundary ambiguity can be calculated, and the effective scope of the scope can be clarified.

[0131] As another embodiment of the present invention, configuring the multidimensional attribute meta tags of the multivariate database to establish the inverted index and feature vector index of the multivariate database includes:

[0132] Based on the multidimensional attribute meta tags, the meta tag key-value pairs of the multivariate database are parsed out;

[0133] Extract the semantic features of the meta-tag key-value pairs;

[0134] Based on the semantic features of the meta-tags, a set of meta-tag feature vectors for the multivariate database is generated;

[0135] The feature vector index of the multivariate database is established using the meta-label feature vector set.

[0136] Identify high-frequency and rare keywords in the meta tag key-value pairs to calculate the keyword weight of the meta tag key-value pairs;

[0137] The weights of the inverted index items in the multivariate database are determined by the keyword weights.

[0138] Based on the weights of the inverted index items, a weighted inverted list of the multivariate database is generated;

[0139] Based on the weighted inverted list, an inverted index is established for the multivariate database.

[0140] The meta-tag key-value pairs refer to structured data units organized in a one-to-one correspondence between the tag dimensions and specific tag content of multidimensional attribute meta-tags in a multivariate database. Here, the key represents the attribute category, and the value represents the specific content of that attribute. The meta-tag semantic features refer to abstract features extracted from the values ​​of the meta-tag key-value pairs that represent their core meaning, including semantic keywords, semantic categories, and semantic relevance. The meta-tag feature vector set refers to a collection of high-dimensional numerical vectors formed by numerically transforming the meta-tag semantic features. The feature vector index refers to an index constructed based on the meta-tag feature vector set, used to quickly retrieve meta-tag feature vectors similar to the target vector and associate them with corresponding database data. The structure is as follows: High-frequency keywords refer to words that appear very frequently in the meta-tag values ​​of a multivariate database; rare keywords refer to words that appear very infrequently in the meta-tag values ​​of a multivariate database. For example, in a furniture database, "wood" may appear 1000 times (high-frequency), while "mahogany" may only appear 5 times (rare). Keyword weight refers to a quantitative value that measures the importance of a word in a document or dataset. For example, although mahogany may only appear twice in a document, because very few documents contain it, its inverse document frequency (IDF) value will be very high, and therefore its final weight will be very high. The calculation formula is: Inverted index item weight = term frequency. The formula is log(total number of documents / number of documents containing the word), where log(total number of documents / number of documents containing the word) represents the inverse document frequency; the inverted index item weight refers to the quantified value that maps the keyword weight in the meta tag key-value pair to the index item (keyword) of the inverted index, which can be directly used for relevance ranking during retrieval; the weighted inverted list refers to the structured list obtained by adding the inverted index item weight and data relevance score to each index item-data association pair on the basis of the traditional inverted list (such as index item → data storage address list).

[0141] Optionally, the meta tag key-value pairs of the multivariate database can be parsed using a structured tag parsing algorithm, such as a tag structured algorithm based on XML / JSON parsing (e.g., FastJSON parsing algorithm), which splits the text description of multidimensional attribute meta tags according to "dimension-content" and automatically generates key-value pair structures; the feature vector index of the multivariate database can be established using a weighted vector index construction method, such as grouping the meta tag feature vector set according to semantic categories, using a product quantization algorithm (PQ algorithm) to compress and encode the vectors, and combining a cosine similarity algorithm to set a vector matching threshold to construct a feature vector index that supports similarity retrieval; the weighted inverted list of the multivariate database can be generated using an ordered linked list construction algorithm, such as arranging the "index item-data" association pairs in descending order according to data relevance scores to generate a weighted inverted list.

[0142] Furthermore, this embodiment of the invention extracts key information entities corresponding to the virtual space construction intent vector from the multivariate database based on the inverted index and the feature vector index. This enables accurate matching and efficient retrieval of the data required for virtual space construction, ensuring that the extracted environment, object, and interaction-related key information entities are highly consistent with the user's intent. This avoids problems such as target data positioning delays, virtual space construction lag caused by redundant loading of irrelevant data, and user demand matching deviations caused by traditional full-database traversal retrieval. The key information entities refer to the set of data objects with the highest semantic and functional relevance to the virtual space construction intent vector, selected from the environment element library, object instance library, and interaction logic library of the multivariate database through joint query and semantic matching. Optionally, the inverted index can be used to filter out a set of candidate information entities based on the attribute requirements of the virtual space construction intent vector (such as "science fiction style" or "interactive electronic screen"). Then, the candidate set can be optimized by performing high-dimensional vector similarity calculation (such as cosine similarity) on the candidate set in combination with the feature vector index. Redundant and mismatched entities can be eliminated through association verification. This extracts the key information entities corresponding to the virtual space construction intent vector from the multivariate database.

[0143] S3. Calculate the semantic similarity between the virtual space construction intention vector and the key information entity, and based on the semantic similarity, generate the semantic embedding vector and spatial topological constraints corresponding to the virtual space construction intention vector.

[0144] This invention, through calculating the semantic similarity between the virtual space construction intent vector and the key information entities, can effectively establish a dynamic association mechanism between user real-time intent and scene elements, and achieve deep integration of the virtual space with external dynamic data. This reduces the scene rigidity problem caused by fixed code logic and preset interaction processes, and improves the flexibility of user operation and the adaptability of the scene. The semantic similarity refers to a numerical index obtained by quantifying the degree of semantic association between the virtual space construction intent vector and the key information entities (including environment-type, object-type, and interaction-type key information entities). The value range is usually [0,1]. The closer the value is to 1, the higher the semantic fit between the two in "user demand expression" and "data function representation". The closer the value is to 0, the greater the semantic deviation. Optionally, the semantic similarity between the virtual space construction intent vector and the key information entities can be calculated using the CLIP model. For example, the virtual space construction intent vector can be converted into a structured text description and input into CLIP. The model's text encoder (Transformer architecture) generates intent text semantic vectors; then, multimodal features are extracted for key information entity types. The scene rendering image of environment entities and the 3D model rendering image of object entities are input into CLIP's image encoder (ViT-L / 14 architecture) to generate image semantic vectors. The logical description of interactive entities is converted into text and then input into the text encoder to generate interactive text semantic vectors. Subsequently, based on the cosine similarity algorithm built into the CLIP model, the matching degree between the intent text semantic vector and the environment, object image semantic vectors, and interactive text semantic vectors is calculated to obtain similarity values ​​for function, style, and logic dimensions. Finally, a weighted fusion of the multi-dimensional similarities is performed using preset weights (function 0.3, style 0.3, logic 0.4) to generate semantic similarity values ​​in the range [0,1].

[0145] Furthermore, this embodiment of the invention generates a semantic embedding vector corresponding to the virtual space construction intent vector based on the semantic similarity. This can deeply integrate the semantic features of the user's real-time intent with the core features of key information entities, accurately map the dynamic adaptation relationship between user query needs and virtual space resources, and significantly enhance the virtual space's ability to understand user intent and integrate external dynamic data. The semantic embedding vector refers to a numerical vector with high-dimensional structured features generated by vector encoding technology based on the semantic similarity between the virtual space construction intent vector and the key information entities in the virtual space.

[0146] As an embodiment of the present invention, generating the semantic embedding vector corresponding to the virtual space construction intent vector based on the semantic similarity includes:

[0147] Based on the semantic similarity, candidate key information entities corresponding to the virtual space construction intent vector are selected;

[0148] Construct a multidimensional association matrix between the candidate key information entities and the virtual space construction intent vector;

[0149] Extract the principal component eigenvectors and their corresponding importance weights from the multidimensional correlation matrix;

[0150] Based on the importance weights, the principal component feature vectors are subjected to weighted fusion processing to obtain weighted fused feature vectors;

[0151] The weighted fusion feature vector is mapped to a low-dimensional semantic space to obtain the low-dimensional semantic embedding vector corresponding to the candidate key information entity.

[0152] Calculate the cluster center distance and distribution density of the low-dimensional semantic embedding vector in the low-dimensional semantic space;

[0153] Based on the cluster center distance and the distribution density, the low-dimensional semantic embedding vector is normalized to obtain the semantic embedding vector corresponding to the virtual space construction intent vector.

[0154] The candidate key information entities refer to the structured data unit combinations that initially conform to user intent, selected from a multivariate database based on the semantic similarity (meeting a preset threshold, such as ≥0.8) between the intent vector constructed in virtual space and the key information entities. The multidimensional association matrix is ​​a matrix structure constructed to quantify the dimensional semantic association between the candidate key information entities, with the semantic feature dimensions of the candidate key information entities as rows and the core requirement dimensions of the intent vector constructed in virtual space as columns. The matrix elements are the semantic similarity values ​​of the corresponding dimensions. The principal component feature vectors are orthogonal vectors obtained by dimensionality reduction of the multidimensional association matrix using Principal Component Analysis (PCA) algorithm, which best preserve the original information of the matrix (i.e., the dimensional association features between entities and intents). The importance weights are quantitative coefficients used to measure the importance of each principal component feature vector in representing the semantic association between entities and intents. The weighted fusion feature vector is a single, comprehensive feature vector obtained by weighted summation of each principal component feature vector using importance weights, such as a fusion vector = 0.85. PC1 vector +0.15 PC2 vector; the low-dimensional semantic space refers to a vector space with lower dimensionality (usually 2-10 dimensions) that can maintain the semantic relationships of data, obtained by mapping a high-dimensional feature space (such as the feature space corresponding to a multi-dimensional correlation matrix) through a dimensionality reduction algorithm (such as t-SNE, UMAP); the low-dimensional semantic embedding vector refers to the coordinate vector corresponding to the weighted fused feature vector after being mapped to the low-dimensional semantic space; the cluster center distance refers to the Euclidean distance from the low-dimensional semantic embedding vector to the cluster center of its semantic category; the distribution density refers to the distribution density in the low-dimensional semantic space with the low-dimensional semantic embedding vector as the center and a preset radius (such as 0.1). The normalization process is the ratio of the number of other low-dimensional vectors contained in the neighborhood to the neighborhood volume. For example, in a 2D low-dimensional semantic space, the preset neighborhood radius of the low-dimensional semantic embedding vector (0.45, 0.62) is 0.1. The neighborhood contains 5 other low-dimensional vectors, and the neighborhood volume (area of ​​a circle) is π×0.1²≈0.0314. Then the distribution density is 5 / 0.0314≈159.24 (vectors / unit area). The normalization process refers to scaling and translating the value of the low-dimensional semantic embedding vector according to the cluster center distance and distribution density, so that it falls into a stable and standard numerical range (such as [0,1]).

[0155] Optionally, the multidimensional association matrix between the candidate key information entities and the virtual space construction intent vector can be generated using cross-modal tensor construction algorithms, such as high-order tensor decomposition based on semantic alignment features, spatial co-occurrence probability, and functional coupling strength, to generate a joint representation matrix to quantify the multidimensional association between intent and entities; manifold learning dimensionality reduction algorithms can be used to map the weighted fusion feature vector to a low-dimensional semantic space, such as neighborhood structure-preserving dimensionality reduction of the high-dimensional fusion vector based on isomap or local linear embedding (LLE), to generate compact semantic embeddings to enhance the robustness and computational efficiency of vector representation.

[0156] This invention generates spatial topological constraints corresponding to the virtual space construction intent vector based on the semantic similarity. This can accurately anchor the association logic between user intent and virtual space elements, deeply couple user demand orientation and scene resource distribution characteristics, and significantly improve the response accuracy of the virtual space to the user's real-time operation intent and the dynamic adaptation capability of the scene topology. The spatial topological constraints refer to the set of constraint rules extracted and defined based on the semantic similarity between the virtual space construction intent vector and the candidate key information entities, which are used to regulate the spatial position relationship, hierarchical structure relationship and interaction range relationship of the candidate key information entities in the virtual space.

[0157] As an embodiment of the present invention, generating spatial topological constraints corresponding to the virtual space construction intent vector based on the semantic similarity includes:

[0158] Extract the spatial relation predicates from the virtual space construction intent vector, and determine the spatial constraint strength coefficients corresponding to the spatial relation predicates;

[0159] Based on the semantic similarity, calculate the semantic-feature coupling weights between different spatial relation predicates;

[0160] Based on the spatial constraint strength coefficient and the semantic-feature coupling weight, a priority ranking rule for the spatial relation predicate is established.

[0161] Based on the priority sorting rules, the constraint logic combination method and execution order corresponding to the virtual space construction intent vector are determined;

[0162] Based on the spatial constraint strength coefficient, define the constraint tolerance range corresponding to the virtual space construction intention vector;

[0163] By integrating the constraint logic combination method, the execution order, and the constraint tolerance range, spatial topological constraint conditions corresponding to the virtual space construction intent vector are generated.

[0164] The spatial relationship predicate refers to a standardized logical expression used to accurately describe the spatial positional and relative orientation relationships of candidate key information entities in virtual space; the spatial constraint strength coefficient refers to a numerical index used to quantify the strictness of spatial topological constraints (position, hierarchy, interaction range constraints), with a value range of [0,1]. The closer the value is to 1, the stricter the constraint (e.g., interaction range constraint strength 0.97), and the closer it is to 0, the looser the constraint (e.g., auxiliary entity position constraint strength 0.3); the semantic-feature coupling weight refers to a weight index used to quantify the contribution ratio of "semantic similarity" and "entity spatial features" in the generation of spatial topological constraints, with a value range of [0,1], including semantic weight (W1) and feature weight (W2), and W1+W2=1. The semantic weight is determined based on the semantic similarity between the intent vector and the entity (the higher the similarity, the larger W1), and the feature weight is determined based on the entity's own spatial features (e.g., size, functional type) (for core functional entities, W2). The priority ordering rule refers to the set of rules used to determine the execution order of multiple spatial topological constraints. Core rules include: constraint type priority (interaction range constraint > hierarchical structure constraint > spatial location constraint, as interaction logic directly affects user experience), and constraint of the same type sorted in descending order of constraint strength coefficient (interaction constraint with strength 0.97 > hierarchical constraint with strength 0.86). The constraint logic combination method refers to the logical operation method that integrates multiple independent spatial topological constraints (location, hierarchy, interaction range constraints) into an executable constraint set. Common methods include "AND", "OR", and "NOT". The constraint combination of core entities uses "AND" logic (all constraints must be satisfied simultaneously, such as an electronic screen needing to satisfy location AND hierarchy AND interaction range constraints), auxiliary entities use "OR" logic (satisfying any constraint is sufficient, such as greenery being able to be located in a corner OR against a wall), and conflicting constraints use "NOT". The logic (excluding non-compliant constraints, such as "non-occluded interaction areas"); the execution order refers to the complete execution process steps of spatial topology constraints from rule parsing to entity layout adjustment, which usually follows the order of constraint parsing → priority sorting → conflict detection → constraint execution → compliance verification. Each step corresponds to a specific operation (such as constraint parsing: extracting spatial relationship predicates and strength coefficients; conflict detection: determining whether there are contradictions between constraints); the constraint tolerance range refers to the reasonable range within which the layout parameters (location coordinates, hierarchical priority, interaction range size) of candidate key information entities are allowed to deviate from the constraint standard value during the execution of spatial topology constraints.

[0165] Optionally, the spatial constraint strength coefficient corresponding to the spatial relation predicate can be determined by an SVM machine learning model; the constraint logic combination method and execution order corresponding to the virtual space construction intention vector can be determined by a constraint priority scheduling algorithm, such as dynamic programming based on constraint conflict detection results, user intent weights, and constraint satisfaction costs, to generate the optimal constraint execution sequence to ensure the efficiency and reliability of virtual space construction; the constraint tolerance interval corresponding to the virtual space construction intention vector can be defined by a fuzzy membership function, such as Gaussian function fitting based on constraint type characteristics, user historical adjustment behavior data, and spatial function requirements, to generate an elastic constraint boundary to enhance the flexibility and fault tolerance of virtual space construction.

[0166] S4. Based on the semantic embedding vector and the spatial topology constraints, establish a virtual space configuration set corresponding to the multivariate database, and parse the corresponding entity data in the multivariate database based on the virtual space configuration set.

[0167] This invention establishes a virtual space configuration set corresponding to the multivariate database based on the semantic embedding vector and the spatial topology constraints. This deeply connects the user's real-time intent with the adaptation logic of virtual space resources, accurately maps user needs at the semantic level with scene construction rules at the topology level, and ensures that the automatic construction process of virtual space can specifically match the user's personalized operation requirements. The virtual space configuration set refers to a set of structured configuration schemes that can be directly used for automatic deployment of virtual space, based on the semantic embedding vector and spatial topology constraints as the core basis, by selecting suitable resources from the multivariate database and integrating the suitable resources.

[0168] As an embodiment of the present invention, the step of establishing a virtual space configuration set corresponding to the multivariate database based on the semantic embedding vector and the spatial topological constraints includes:

[0169] The environmental element library, object instance library, and interaction logic library in the multivariate database are parsed out.

[0170] Extract spatial style semantics and functional intent semantics from the semantic embedding vector;

[0171] Based on the spatial style semantics, candidate environmental elements in the environmental element library are selected;

[0172] Based on the functional intent semantics, candidate object instances are retrieved from the object instance library;

[0173] The spatial topology constraints are decomposed into environmental layout constraints and object relationship constraints.

[0174] Based on the environmental layout constraints, generate environmental spatial layout schemes and environmental element instances corresponding to the candidate environmental elements.

[0175] Based on the object relationship constraints and the environmental spatial layout scheme, output the set of object instances corresponding to the candidate object instances and their spatial pose parameters;

[0176] Based on the functional intent semantics, relevant candidate interaction logic is matched from the interaction logic library;

[0177] Calculate the compatibility coefficient between the candidate interaction logic and the object relationship constraints;

[0178] Based on the compatibility coefficient, the target interaction logic is selected from the candidate interaction logics;

[0179] By integrating the environmental spatial layout scheme, the environmental element instances, the object instance set, the spatial pose parameters, and the target interaction logic, a virtual space configuration set corresponding to the multivariate database is established.

[0180] The spatial style semantics refers to the semantic information extracted from the semantic embedding vector that characterizes the user's non-functional needs such as visual style, environmental atmosphere, and scene tone of the virtual space; the functional intent semantics refers to the semantic information extracted from the semantic embedding vector that characterizes the user's functional needs such as core functions, operational requirements, and data interaction goals of the virtual space; the candidate environmental elements refer to the basic elements for building the environment that are adapted to the visual style of the virtual space, selected from the environmental element library of a multivariate database based on the spatial style semantics, including background elements (such as forest background images, farmland background images), atmosphere elements (such as raindrop effects, sunlight and shadow), and scene decoration elements (such as tree models, crop models); the candidate object instances refer to the functional entity objects that are adapted to the core functions of the virtual space, retrieved from the object instance library of a multivariate database based on the functional intent semantics, including data... Display objects (such as PM2.5 data panels, humidity dashboards), control operation objects (such as irrigation switch buttons, brightness adjustment sliders), and analysis tool objects (such as trend chart generators, anomaly warning pop-ups), etc.; the environmental layout constraints refer to the constraints obtained after decomposing the spatial topology constraints, which are used to regulate the layout position, hierarchical relationship, and display rules of candidate environmental elements in the virtual space, including position constraints (such as background elements filling the entire virtual space canvas, tree models distributed in the left area of ​​the scene), hierarchical constraints (such as background elements being at the bottom layer, and atmosphere effects being in the middle layer), and display constraints (such as displaying sunlight and shadows in daytime scenes, and displaying moonlight and shadows in nighttime scenes), etc.; the object relationship constraints refer to the constraints obtained after decomposing the spatial topology constraints, which are used to regulate the association relationship, interaction dependency, and data flow logic between candidate object instances, including association constraints (such as PM2.5).5. The data panel needs to be associated with the trend chart generator to share data; dependency constraints (e.g., the irrigation switch button needs to depend on the parameters of the irrigation duration input box to perform control operations); data constraints (e.g., the abnormal warning pop-up needs to receive abnormal data from the data panel to trigger display); the environmental spatial layout scheme refers to the specific layout planning scheme formulated for candidate environmental elements based on environmental layout constraints, including element location coordinates (e.g., forest background map: coordinates (0,0) to (1920,1080); tree model: coordinates (200,300), (400,350)), element size parameters, etc. (e.g., tree model: 100px wide, 200px high; background image: 1920px wide, 1080px high), element display sequence (e.g., displaying moonlight and shadow from 0-6 o'clock, and sunlight and shadow from 6-18 o'clock), etc.); the environmental element instance refers to a directly deployable environmental element instance generated after parameterizing the candidate environmental elements according to the environmental spatial layout scheme; the object instance set refers to a collaborative object instance combination formed after associating and configuring the candidate object instances according to object relationship constraints and the environmental spatial layout scheme; the spatial pose parameter refers to the parameter describing the object. The quantitative parameters of the instance's spatial position, attitude, and size in the virtual space include 2D pose (such as planar coordinates (X,Y), rotation angle, and planar dimensions (width × height)) and 3D pose (such as three-dimensional coordinates (X,Y,Z), pitch angle, yaw angle, roll angle, and three-dimensional dimensions (length × width × height)). The candidate interaction logic refers to the logical rule templates for user operations and system responses matched from the interaction logic library of a multi-dimensional database based on functional intent semantics, and adapted to the core functions of the virtual space. These include data viewing logic (such as clicking the data panel → displaying detailed data) and control operation logic (such as sliding to adjust the slider). The interaction logic includes: (1) sending control commands to update device status; (2) analyzing triggering logic (e.g., clicking the generate button to call the algorithm to display a trend chart); the compatibility coefficient is a numerical parameter that quantifies the degree of adaptation and conflict-free nature between candidate interaction logic and object relationship constraints. Its value ranges from [0,1]. A larger value indicates greater compatibility between the candidate interaction logic and object relationship constraints, while a smaller value indicates poorer compatibility. The target interaction logic refers to the final interaction logic selected from the candidate interaction logics, whose compatibility coefficient with object relationship constraints is ≥ a preset threshold (e.g., 0.7), and can be directly used for virtual space interaction implementation.

[0181] Optionally, the spatial topological constraints can be decomposed into environmental layout constraints and object relationship constraints using a topological constraint decomposition algorithm. For example, using the environmental-related semantic roles, object-related semantic roles, and relationship-related semantic roles in the spatial topological constraints as identification indicators, a mapping model between constraints and semantic roles can be constructed. The specific rules for decomposing constraints into environmental layout constraints and object relationship constraints can be determined by solving the model. The spatial pose parameters corresponding to the candidate object instances can be obtained through a pose parameter mapping algorithm of the environmental spatial layout scheme. For example, using the object position description, size specifications, and posture requirements in the environmental spatial layout scheme as input indicators, a conversion model between layout description and quantified pose parameters (coordinates, size, angle) can be constructed. The natural language layout description can be converted into the spatial pose parameters of the candidate object instances by solving the model.

[0182] Specifically, the establishment of the virtual space configuration set relies on semantic retrieval of a multivariate database and model context management, see [link / reference]. Figure 3 The diagram shown is a data interaction flowchart of a method for automatically constructing a personalized virtual space based on user intent, provided by an embodiment of the present invention. The diagram presents the complete logic from user-initiated query, transmission from the client to the large model, and then connection of data collection and database through the model context protocol to complete the interaction of multi-database resources. The multi-module data flow simulates the pre-construction data interaction process of virtual space configuration set construction, which can help to analyze the transformation state of user intent into semantic retrieval instructions of multi-database, the model context's guarantee logic for cross-module data consistency, and the retrieval path of database resources (environmental elements, object instances, interaction logic), etc. It should be noted that this flowchart focuses on the data interaction logic before the establishment of the virtual space configuration set. In actual applications, it can be adapted and adjusted according to the storage structure of the multi-database and the deployment form of the large model, and does not limit the specific execution form of the process.

[0183] As another embodiment of the present invention, calculating the compatibility coefficient between the candidate interaction logic and the object relationship constraints includes:

[0184] Analyze the interaction space requirements of the candidate interaction logic, and calculate the spatial matching degree between the interaction space requirements and the object relationship constraints;

[0185] Identify the strictness level of the object relationship constraints and determine the constraint strength coefficient corresponding to the strictness level;

[0186] Parse the functional intent tags of the candidate interaction logic, and calculate the semantic matching degree between the functional intent tags and the current virtual space scene;

[0187] Identify the computational complexity characteristics corresponding to the candidate interaction logic and the urgency of real-time requirements corresponding to the object relationship constraints;

[0188] Based on the computational complexity characteristics and the urgency of the real-time requirements, the system load impact factor of the candidate interaction logic and the object relationship constraints is comprehensively extracted.

[0189] By combining the spatial matching degree, the constraint strength coefficient, the context semantic matching degree, and the system load impact factor, the compatibility coefficient between the candidate interaction logic and the object relationship constraint is calculated.

[0190] The interaction space requirement refers to the range and shape requirements of the spatial area necessary for the candidate interaction logic to be correctly triggered and executed. For example, the interaction space requirement for a door opening interaction logic includes: triggering area: a hemisphere with a radius of 0.5 meters in front of the door; action area: the rotating envelope of the door leaf. The spatial matching degree refers to the quantitative value of the degree of overlap between the interaction space requirement and the spatial allowable range specified by the object relationship constraints, usually represented by a value between 0 and 1, where 1 represents a complete match. The strictness level refers to the qualitative classification of the degree of mandatory compliance of the rules specified in the object relationship constraints, such as level 1: advisory, level 2: recommended, and level 3: mandatory. The constraint strength coefficient refers to the coefficient determined by the object relationship constraints. The stringency level of the condition is converted into a quantitative numerical parameter, with a value range of [0,1]. This parameter reflects the weight of the constraint in the compatibility calculation; the higher the stringency level, the larger the coefficient (strong constraints correspond to 0.8-1.0, medium constraints to 0.5-0.7, and weak constraints to 0.1-0.4). The functional intent label refers to a machine-readable semantic identifier used to describe the core purpose or function of the candidate interaction logic, including function type labels (such as data viewing, parameter control, analysis generation), operation type labels (such as click, drag, input), and target object labels (such as data panel, irrigation slider). The current virtual space scene refers to a comprehensive description of the virtual space functional scene and user intent scene in which the candidate interaction logic is located, including... This includes scene functional objectives (such as ecological monitoring data visualization, farm irrigation control), object instances contained in the scene (such as data panels, map component irrigation sliders, buttons), and scene user intent (such as users needing to view real-time data, users needing to adjust irrigation parameters), etc.; the contextual semantic matching degree refers to the quantitative value of the degree of correlation between the semantics of the functional intent label and the contextual semantics of the current virtual space scene; the computational complexity feature refers to the characteristic indicators used to describe the system computing resources and computation time consumed by the candidate interaction logic during execution, including the number of computation steps (such as simple click interaction: 3 computation steps; trend chart generation interaction: 50 computation steps), data processing volume (such as processing 10 data points, processing 1000 data points), algorithm, etc. Complexity level (e.g., constant level, quadratic level); The urgency of real-time requirements refers to an indicator that quantifies the time sensitivity and priority of real-time requirements (e.g., response latency, data update speed) based on the scenario function priority, user operation goal urgency, and data timeliness requirements corresponding to object relationship constraints. The value range is usually [0,1]. The larger the value, the more urgent the real-time requirement (fast response must be prioritized, and delay will lead to serious consequences), and the smaller the value, the lower the urgency (a certain delay is allowed, and the impact on the function is small); The system load impact factor refers to the load pressure index expected to be caused to the system when the computational complexity requirements and the urgency of real-time requirements of the interaction logic are met simultaneously.

[0191] Optionally, the strictness level of the object relationship constraints can be obtained through a level mapping algorithm, such as pre-constructing a strictness level-keyword correspondence library (strong constraint keywords: "must", "indispensable", "otherwise invalid"; medium constraint keywords: "suggested", "should", "priority"; weak constraint keywords: "may", "allowed", "optional"). Then, the semantic keywords in the object relationship constraints are extracted by the BERT semantic recognition model, and the level is determined by matching the keyword library. The computational complexity features corresponding to the candidate interaction logic can be identified using abstract syntax tree (AST) analysis technology, such as parsing the execution code of the candidate interaction logic, statistically analyzing the number of code statement executions, the number of loop nesting levels, the function call depth, the data processing volume, and other feature parameters, and mapping them to computational complexity indicators (e.g., a statement count > 100 and loop nesting ≥ 3 levels are judged as high complexity).

[0192] In another embodiment of the present invention, the compatibility coefficient between the candidate interaction logic and the object relationship constraint is calculated by the following formula:

[0193]

[0194] in, Indicates the compatibility factor. Indicates the constraint strength coefficient. The weighting coefficient represents the balance between spatial matching and contextual semantic matching. Indicates spatial matching degree, Indicates the semantic matching degree of the context. Indicates the system load impact factor. This represents the weight balance coefficient of the system load impact factors.

[0195] In detail, the constraint strength coefficient ( This coefficient is used to dynamically adjust the binding force of object relationship constraints on the overall compatibility assessment results. It can be assigned a specific value by analyzing the type of constraint, historical violation records, and the safety threshold of the current scenario. For example, for rigid constraints involving virtual human safety or equipment operation safety (such as prohibiting clipping), the constraint strength coefficient can be set to 1.0 to strictly ensure the absolute satisfaction of the constraint; while for flexible constraints involving aesthetics or comfort (such as the ideal distance between objects being 0.5 meters), the coefficient can be set to 0.6 to ensure a certain degree of flexibility. The weighting balance coefficient between spatial matching degree and contextual semantic matching degree (…) ) used to dynamically adjust spatial matching degree ( ) and context semantic matching degree ( The relative contribution of the weighting factor to the compatibility coefficient can be assigned a specific value by analyzing historical interaction data and the accuracy requirements of the current virtual scene. For example, in high-precision scenarios such as surgical simulation training, accurate spatial and semantic matching is crucial, and in this case, the weighting factor... It can be set to 0.8 to significantly increase the weight of spatial and semantic matching; while for lightweight interactive scenarios such as virtual tours, the accuracy requirements can be appropriately relaxed, and the coefficient can be adjusted accordingly. It can be set to 0.4 to reduce the contribution ratio of spatial and semantic matching, thereby enhancing interaction flexibility. The weighting balance coefficient of the system load influence factor ( ) is used to dynamically adjust the system load influence factor ( The degree of suppression of the compatibility coefficient can be assigned a specific value through real-time system performance monitoring and the scene's requirements for smoothness. For example, in scenarios with extremely high real-time requirements, such as multi-person online virtual reality conferences, system smoothness should be prioritized. In this case, the weighting balance coefficient... It can be set to 0.8 to significantly increase the influence weight of the system load factor, ensuring automatic degradation or rejection of low-priority interactions under high load; while for offline rendered virtual building roaming scenes, the real-time requirements are lower, and the coefficient... It can be set to 0.2 to weaken the inhibitory effect of system load and prioritize the integrity of interactive functions.

[0196] It should be noted that by introducing spatial matching degree and context semantic matching degree This ensures that the calculated compatibility coefficient R can simultaneously meet the system's functional requirements for interactive logic (e.g., completing interactive responses within a specific time window to guarantee system real-time performance) and structural requirements (e.g., maintaining logical consistency even when system load increases, avoiding logical conflicts or performance degradation caused by resource contention), thus enabling the system to maintain high reliability and adaptability even in complex interactive environments; by introducing the hyperbolic tangent function... Smoothing the risk parameter λ can suppress the nonlinearity of the system load influence factors and enhance the system's robustness.

[0197] Furthermore, by parsing the corresponding entity data in the multivariate database based on the virtual space configuration set, this embodiment of the invention can accurately connect the abstract configuration rules and concrete construction resources of the virtual space, deeply couple the user intent-oriented configuration scheme with the scene support data in the multivariate database, and ensure that the personalized virtual space automatic construction process can specifically call the appropriate environment, objects and interaction resources. The entity data refers to the original data assets with clear format and semantics stored in the multivariate database (environment database, object database, interaction logic database) used to instantiate specific elements of the virtual space. The data types of the entity data include geometric model data, texture mapping data, material parameter data, audio sample data, etc.

[0198] S5. Combining the entity data, the semantic embedding vector, and the spatial topology constraints, generate an interactive preview view corresponding to the virtual space construction intent vector. Based on the multivariate database, generate the user's personalized virtual space through the virtual space configuration set and the interactive preview view.

[0199] This invention, by combining the entity data, the semantic embedding vector, and the spatial topology constraints, generates an interactive preview view corresponding to the virtual space construction intent vector. This transforms the abstract configuration scheme of the virtual space into a visual and operable intuitive presentation, accurately linking user intent with the actual presentation effect of the virtual space. The interactive preview view refers to a visual interface generated by virtual rendering technology based on the extracted entity data, used to preview the scene effect before the final construction of the virtual space and to support user interaction.

[0200] As an embodiment of the present invention, the step of generating an interactive preview view corresponding to the virtual space construction intent vector by combining the entity data, the semantic embedding vector, and the spatial topological constraints includes:

[0201] The entity data is mapped to a preset Unity engine, and the environment element instance, object instance and interaction logic instance corresponding to the entity data are generated through the instantiation interface of the Unity engine.

[0202] Based on the semantic embedding vector, the semantic style preferences and functional interaction requirements corresponding to the virtual space construction intent vector are parsed out in order to adjust the multi-dimensional environment configuration parameters of the environmental element instance;

[0203] Based on the spatial topology constraints, generate the object pose configuration scheme for the object instance;

[0204] Extract the triggering conditions and response scripts of the interaction logic instance, and dynamically bind the triggering conditions, the response scripts, and specific events of the object instance through the Unity engine's event delegation system to obtain the interaction behavior logic chain;

[0205] By integrating the multi-dimensional environment configuration parameters, the object pose configuration scheme, and the interaction behavior logic chain, a Unity lightweight real-time rendering scene corresponding to the virtual space construction intent vector is constructed.

[0206] Based on the cluster center distance of the semantic embedding vector, the adaptive rendering optimization parameters of the Unity lightweight real-time rendering scene are defined.

[0207] Based on the lightweight real-time rendering scene and the adaptive rendering optimization parameters, an interactive preview view corresponding to the virtual space construction intent vector is generated.

[0208] The Unity engine, developed by Unity Technologies, is a real-time 3D interactive content development engine with cross-platform capabilities. Its core functions include, but are not limited to, 3D / 2D scene rendering, entity object instantiation, physics engine simulation, interactive logic editing, script compilation and execution, and multi-platform (e.g., PC, mobile devices, web pages) output. It can be combined with a visual editor and programmatic interfaces (such as the C# script API) to achieve a complete development process from virtual scene construction to interactive content generation. The instantiation interface refers to the application programming interface in the Unity engine used to create editable and interactive instance objects from resource templates (such as prefabs and model files) within the scene. API); The environment element instance refers to a scene object instance generated in the Unity engine based on environment element entity data (such as 3D model files, texture maps, audio resources, lighting presets, etc.) through an instantiation interface (such as Instantiate) to build the basic environment framework of the virtual space; the object instance refers to a scene instance object generated in the Unity engine based on object instance entity data (such as UI control prefabs, functional module prefabs) through an instantiation interface (such as Instantiate) to carry the core functions of the virtual space and user interaction operations; the interaction logic instance refers to an instance created by adding a component to an object through Unity's AddComponent interface. The instance is a script component instance used to define user operation and system response logic; the semantic style preference refers to the non-functional requirements of users regarding the visual style, atmosphere, and design tone of the virtual space, as parsed through semantic embedding vectors; the functional interaction requirements refer to the functional requirements of users regarding the core functions, interactive operations, and data processing goals of the virtual space, as parsed through semantic embedding vectors; the multi-dimensional environment configuration parameters refer to a set of multi-dimensional quantitative parameters used to define the visual presentation, atmosphere creation, and functional association of environmental element instances, including visual attribute parameters (such as material color, model size, texture resolution), atmosphere rendering parameters (such as light intensity, sound effect volume, particle effect density), and functional... The associated parameters (such as scene switching labels, time period display rules, etc.); the object pose configuration scheme refers to the structured configuration scheme generated based on spatial topological constraints to define the position, pose, size, and level of an object instance in virtual space; the trigger condition refers to the pre-set objective judgment rules or event thresholds used to activate the execution of response scripts in the interactive behavior logic chain in an interactive virtual scene built by the Unity engine; the response script refers to the code snippets or function sets defined in the interactive logic instance to execute the response actions corresponding to the trigger condition; the event delegation system refers to the event handling mechanism in the Unity engine used to decouple and associate object instances, trigger conditions, and response scripts.The specific event refers to a pre-defined event type in the event component of an object instance that corresponds to the user operation type or scene state type; the interactive behavior logic chain refers to a closed-loop execution link formed by connecting the specific event, trigger condition, and response script of an object instance according to trigger-response logic through the event delegation system; the Unity lightweight real-time rendering scene refers to a virtual scene instance built in the Unity engine based on lightweight resources and optimized rendering strategies, supporting real-time interactive rendering with low hardware overhead; the cluster center distance refers to the Euclidean or cosine distance between the semantic embedding vector of the user query request and the pre-defined semantic cluster center vector, calculated by a clustering algorithm; the adaptive rendering optimization parameters refer to a set of quantized parameters dynamically generated based on the cluster center distance of the semantic embedding vector and the terminal hardware performance detection results, used to adjust the rendering strategy of the Unity lightweight real-time rendering scene.

[0209] Optionally, the object pose configuration scheme of the object instance can be implemented using a spatial topology constraint parsing method. For example, a mapping model between topology rules and pose parameters can be constructed using position parameters (3D coordinates, relative spacing), posture parameters (rotation angle, orientation), and hierarchy parameters (display priority, occlusion relationship) as constraint indicators. The object pose configuration scheme that makes all indicators compliant can be determined by solving the model. The Unity lightweight real-time rendering scene corresponding to the virtual space construction intent vector can be constructed using a lightweight resource integration method, such as using resource complexity (model face count, texture resolution), rendering performance (frame rate, video memory usage), and interaction compatibility. Using response latency and event trigger success rate as evaluation metrics, an adaptation model between element parameters and scene performance is constructed. Through model optimization, a lightweight real-time rendering scene that balances all metrics is determined. The interactive preview view corresponding to the intent vector of the virtual space can be generated by adaptive tuning of rendering parameters. For example, using visual fidelity (style matching degree, detail reproduction rate), interaction smoothness (operation response speed, logic execution accuracy) and terminal adaptability (multi-platform compatibility, resource utilization rate) as output metrics, a control model between optimization parameters and preview effect is constructed. By solving the model, an interactive preview view that optimizes all metrics is determined.

[0210] Furthermore, this embodiment of the invention generates a personalized virtual space based on the multi-dimensional database, the virtual space configuration set, and the interactive preview view. This allows the virtual space construction process to deeply match the user's real-time intention dynamic expression needs and the integration and adaptation requirements of external dynamic data. At the same time, it can avoid the user's rigid operation caused by relying on fixed code logic or the external data integration obstacles caused by closed resource systems, greatly improving the user adaptation flexibility and scene dynamic adaptability of the virtual space.

[0211] The personalized virtual space refers to a virtual scene dynamically constructed based on entity resources (such as environment models, functional components, and interaction scripts) from a multi-dimensional database. This is achieved through a virtual space configuration set (structured parameters carrying user intent) and an interactive preview view (a user feedback verification carrier), resulting in a virtual scene with user intent adaptability, functional customizability, and data integrability. The core logic of constructing the personalized virtual space includes: first, extracting entity resources such as environmental elements and functional components from the multi-dimensional database that match the semantic embedding vector of the user query; then, the virtual space configuration set completes the structured organization of these resources (e.g., defining environment style parameters, object pose rules, and interaction logic relationships); next, generating an initial interactive preview view based on the configuration set; and obtaining real-time intent data through user operation feedback (e.g., dragging to adjust object positions, switching functional modules), which feeds back into the configuration set for parameter iteration (e.g., updating object coordinate parameters, adding new interaction logic rules); simultaneously, the configuration set continuously synchronizes external dynamic data (e.g., ecological monitoring information) from the multi-dimensional database, driving real-time updates to the preview view (e.g., refreshing data panel content, status warning prompts), forming a closed loop of database resource invocation, configuration set parameter optimization, and preview view interactive feedback.

[0212] Compared to the problems described in the background technology, this embodiment of the invention, by receiving natural language query text input by the user to generate the user's virtual space construction intent vector, can specifically optimize the intent-driven generation logic and dynamic data integration strategy of the virtual space. Furthermore, by calling a preset model context protocol platform to construct a multivariate database corresponding to the virtual space construction intent vector, this embodiment of the invention can enhance the resource support and dynamic adaptation capabilities of the multivariate database for the entire virtual space construction process, ensuring the efficiency of personalized virtual space automatic construction and the matching degree of user needs. By configuring multidimensional attribute meta-tags of the multivariate database to establish an inverted index and feature vector index of the multivariate database, this embodiment of the invention can break the closed static attributes of virtual space data resources, forming a data management system that can be quickly retrieved and dynamically associated. This provides an efficient data calling framework for real-time capture of user intent and integration with external dynamic data (such as ecological monitoring information), ensuring the rapid response of the virtual space to changes in user needs and updates to external data. Furthermore, by generating a semantic embedding vector corresponding to the virtual space construction intent vector based on the semantic similarity, this embodiment of the invention can deeply integrate the semantic features of the user's real-time intent with the core features of key information entities. This invention accurately maps the dynamic adaptation relationship between user query needs and virtual space resources, significantly enhancing the virtual space's ability to understand user intent and integrate external dynamic data. Furthermore, by generating spatial topological constraints corresponding to the user query request based on semantic similarity, this invention deeply integrates the semantic features of the user's real-time intent with the core features of key information entities, accurately mapping the dynamic adaptation relationship between user query needs and virtual space resources, significantly enhancing the virtual space's ability to understand user intent and integrate external dynamic data. Further, by generating spatial topological constraints corresponding to the virtual space construction intent vector based on semantic similarity, this invention accurately anchors the association logic between user intent and virtual space elements, deeply coupling user demand orientation and scene resource distribution characteristics, significantly improving the virtual space's response accuracy to user real-time operation intent and its dynamic adaptation capability to scene topology. Finally, by combining the entity data, the semantic embedding vector, and the spatial topological constraints, this invention generates an interactive preview view corresponding to the virtual space construction intent vector, transforming the abstract configuration scheme of the virtual space into a visual and operable intuitive presentation, accurately associating user intent with the actual presentation effect of the virtual space.Finally, this embodiment of the invention generates personalized virtual spaces based on the multi-dimensional database, the virtual space configuration set, and the interactive preview view. This allows the virtual space construction process to deeply align with the dynamic expression needs of users' real-time intentions and the integration and adaptation requirements of external dynamic data. Simultaneously, it avoids user control rigidity caused by reliance on fixed code logic or external data integration obstacles caused by closed resource systems. This ensures the virtual space construction process deeply aligns with users' real-time intentions, improving the user adaptability and flexibility of the virtual space. Therefore, this embodiment of the invention provides a method for automatically constructing personalized virtual spaces based on user intentions, enabling the virtual space construction process to deeply align with users' real-time intentions and improving the user adaptability and flexibility of the virtual space.

[0213] like Figure 4 The diagram shown is a functional block diagram of a personalized virtual space automatic construction system based on user intent according to the present invention.

[0214] The user-intent-based personalized virtual space automatic construction system 200 described in this invention can be installed in an electronic device. Depending on the functions implemented, the user-intent-based personalized virtual space automatic construction system may include an intent parsing module 201, a database indexing module 202, a query matching module 203, a space configuration generation module 204, and a space visualization module 205. The modules described in this invention can also be referred to as units, which are a series of computer program segments that can be executed by the processor of an electronic device and perform a fixed function, and are stored in the memory of the electronic device.

[0215] In this embodiment of the invention, the functions of each module / unit are as follows:

[0216] The intent parsing module 201 is used to receive natural language query text input by the user, generate the user's virtual space construction intent vector, call a preset model context protocol platform, and construct a multi-dimensional database corresponding to the virtual space construction intent vector, wherein the multi-dimensional database includes an environmental element library, an object instance library, and an interaction logic library.

[0217] The database index module 202 is used to configure the multidimensional attribute meta tags of the multivariate database to establish the inverted index and feature vector index of the multivariate database, and extract key information entities corresponding to the virtual space construction intent vector from the multivariate database according to the inverted index and the feature vector index.

[0218] The query matching module 203 is used to calculate the semantic similarity between the virtual space construction intent vector and the key information entity, and based on the semantic similarity, generate the semantic embedding vector and spatial topological constraints corresponding to the virtual space construction intent vector.

[0219] The spatial configuration generation module 204 is used to establish a virtual spatial configuration set corresponding to the multivariate database based on the semantic embedding vector and the spatial topology constraints, and to parse the corresponding entity data in the multivariate database based on the virtual spatial configuration set.

[0220] The spatial visualization module 205 is used to combine the entity data, the semantic embedding vector, and the spatial topology constraints to generate an interactive preview view corresponding to the virtual space construction intent vector. Based on the multivariate database, the user's personalized virtual space is generated through the virtual space configuration set and the interactive preview view.

[0221] In detail, the modules in the user-intent-based personalized virtual space automatic construction system 200 described in this embodiment of the invention employ the same methods as described above when in use. Figure 1 This method employs the same technical means as the one described above for automatically constructing personalized virtual spaces based on user intent, and can produce the same technical effects, so it will not be elaborated here.

[0222] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention.

[0223] Finally, it should be noted that in the above embodiments, each embodiment can be combined with each other or independent. Deleting any one of them will not affect the technical implementation of other embodiments. The above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A method for automatically constructing personalized virtual spaces based on user intent, characterized in that, The method includes: The system receives natural language query text input by the user to generate a virtual space construction intent vector for the user. It then invokes a preset model context protocol platform to construct a multivariate database corresponding to the virtual space construction intent vector. This multivariate database includes an environment element library, an object instance library, and an interaction logic library. The invocation of the preset model context protocol platform to construct the multivariate database corresponding to the virtual space construction intent vector includes: The spatial environment elements, entity object elements, and interactive behavior elements in the virtual space construction intent vector are extracted. Based on the data flow module of the model context protocol platform, the spatial environment elements, the entity object elements and the interactive behavior elements are converted into standardized data units to obtain environment data units, object data units and behavior data units. Based on the tool invocation module of the model context protocol platform, environmental semantic tags, functional semantic tags, and behavioral semantic tags are extracted from the environmental data unit, the object data unit, and the behavioral data unit. Based on the environmental semantic tags, the functional semantic tags, and the behavioral semantic tags, a cross-modal association set of the environmental data unit, the object data unit, and the behavioral data unit is constructed; Based on the association strength value in the cross-modal association set, calculate the data call association weight for each data unit in the environmental data unit, the object data unit, and the behavior data unit; Based on the data call association weights, a multi-source data scheduling priority strategy is generated for the cross-modal association set; Based on the multi-source data scheduling priority strategy and the cross-modal association set, a multi-dimensional database corresponding to the virtual space construction intent vector is constructed. Configure the multidimensional attribute meta tags of the multidimensional database to establish the inverted index and feature vector index of the multidimensional database. Based on the inverted index and the feature vector index, extract the key information entities corresponding to the virtual space construction intent vector from the multidimensional database. Calculate the semantic similarity between the virtual space construction intent vector and the key information entity, and based on the semantic similarity, generate the semantic embedding vector and spatial topological constraints corresponding to the virtual space construction intent vector; Based on the semantic embedding vector and the spatial topology constraints, a virtual space configuration set corresponding to the multivariate database is established, and based on the virtual space configuration set, the corresponding entity data in the multivariate database is parsed. By combining the entity data, the semantic embedding vector, and the spatial topology constraints, an interactive preview view corresponding to the virtual space construction intent vector is generated. Based on the multivariate database, the user's personalized virtual space is generated through the virtual space configuration set and the interactive preview view.

2. The method for automatically constructing a personalized virtual space based on user intent as described in claim 1, characterized in that, The process of receiving natural language query text input by the user to generate the user's virtual space construction intent vector includes: The user's desired virtual space type is determined by the natural language query text. Extract the set of spatial entity objects, spatial relation operators, and spatial attribute modifiers from the natural language query text; Based on the set of spatial entity objects, the spatial relation operators, and the spatial attribute modifiers, construct a spatial topology relation graph of the required virtual space type; Based on the spatial topology graph, the multi-granularity layout constraints of the required virtual space type are analyzed. Based on the spatial topology graph and the multi-granularity layout constraints, the user's virtual space construction intent vector is generated.

3. The method for automatically constructing a personalized virtual space based on user intent as described in claim 1, characterized in that, The step of establishing a virtual space configuration set corresponding to the multivariate database based on the semantic embedding vector and the spatial topological constraints includes: The environmental element library, object instance library, and interaction logic library in the multivariate database are parsed out. Extract spatial style semantics and functional intent semantics from the semantic embedding vector; Based on the spatial style semantics, candidate environmental elements in the environmental element library are selected; Based on the functional intent semantics, candidate object instances are retrieved from the object instance library; The spatial topology constraints are decomposed into environmental layout constraints and object relationship constraints. Based on the environmental layout constraints, generate environmental spatial layout schemes and environmental element instances corresponding to the candidate environmental elements. Based on the object relationship constraints and the environmental spatial layout scheme, output the set of object instances corresponding to the candidate object instances and their spatial pose parameters; Based on the functional intent semantics, relevant candidate interaction logic is matched from the interaction logic library; Calculate the compatibility coefficient between the candidate interaction logic and the object relationship constraints; Based on the compatibility coefficient, the target interaction logic is selected from the candidate interaction logics; By integrating the environmental spatial layout scheme, the environmental element instances, the object instance set, the spatial pose parameters, and the target interaction logic, a virtual space configuration set corresponding to the multivariate database is established.

4. The method for automatically constructing a personalized virtual space based on user intent as described in claim 3, characterized in that, The calculation of the compatibility coefficient between the candidate interaction logic and the object relationship constraints includes: Analyze the interaction space requirements of the candidate interaction logic, and calculate the spatial matching degree between the interaction space requirements and the object relationship constraints; Identify the strictness level of the object relationship constraints and determine the constraint strength coefficient corresponding to the strictness level; Parse the functional intent tags of the candidate interaction logic, and calculate the semantic matching degree between the functional intent tags and the current virtual space scene; Identify the computational complexity characteristics corresponding to the candidate interaction logic and the urgency of real-time requirements corresponding to the object relationship constraints; Based on the computational complexity characteristics and the urgency of the real-time requirements, the system load impact factor of the candidate interaction logic and the object relationship constraints is comprehensively extracted. Combining the spatial matching degree, the constraint strength coefficient, the contextual semantic matching degree, and the system load impact factor, the compatibility coefficient between the candidate interaction logic and the object relationship constraints is calculated using the following formula: in, Indicates the compatibility factor. Indicates the constraint strength coefficient. The weighting coefficient represents the balance between spatial matching and contextual semantic matching. Indicates spatial matching degree, Indicates the semantic matching degree of the context. Indicates the system load impact factor. This represents the weight balance coefficient of the system load impact factors.

5. The method for automatically constructing a personalized virtual space based on user intent as described in claim 1, characterized in that, The configuration of the multidimensional attribute meta tags of the multivariate database includes: Extract the core data features of the environmental element library, object instance library, and interaction logic library from the multivariate database; Based on the core data features, a set of general dimension labels for the multivariate database is defined; The spatial location attributes, physical quantity types, and spatiotemporal granularity of the environmental element database are parsed to generate environment-specific meta tags for the multivariate database. Extract the entity type, functional purpose, and geometric features from the object instance library to determine the object class-specific meta tags of the multivariate database; Analyze the logic triggering mode, response time and scope of the interaction logic library to generate interaction-specific meta tags for the multi-dimensional database; By integrating the general dimension tag set, the environment-specific meta tags, the object-specific meta tags, and the interaction-specific meta tags, the multidimensional attribute meta tags of the multivariate database are constructed.

6. The method for automatically constructing a personalized virtual space based on user intent as described in claim 1, characterized in that, The configuration of the multidimensional attribute meta-labels of the multivariate database to establish the inverted index and feature vector index of the multivariate database includes: Based on the multidimensional attribute meta tags, the meta tag key-value pairs of the multivariate database are parsed out; Extract the semantic features of the meta-tag key-value pairs; Based on the semantic features of the meta-tags, a set of meta-tag feature vectors for the multivariate database is generated; The feature vector index of the multivariate database is established using the meta-label feature vector set. Identify high-frequency and rare keywords in the meta tag key-value pairs to calculate the keyword weight of the meta tag key-value pairs; The weights of the inverted index items in the multivariate database are determined by the keyword weights. Based on the weights of the inverted index items, a weighted inverted list of the multivariate database is generated; Based on the weighted inverted list, an inverted index is established for the multivariate database.

7. The method for automatically constructing a personalized virtual space based on user intent as described in claim 1, characterized in that, The step of generating an interactive preview view corresponding to the virtual space construction intent vector by combining the entity data, the semantic embedding vector, and the spatial topological constraints includes: The entity data is mapped to a preset Unity engine, and the environment element instance, object instance and interaction logic instance corresponding to the entity data are generated through the instantiation interface of the Unity engine. Based on the semantic embedding vector, the semantic style preferences and functional interaction requirements of the virtual space construction intent vector are parsed out in order to adjust the multidimensional environment configuration parameters of the environmental element instance. Based on the spatial topology constraints, generate the object pose configuration scheme for the object instance; Extract the triggering conditions and response scripts of the interaction logic instance, and dynamically bind the triggering conditions, the response scripts, and specific events of the object instance through the Unity engine's event delegation system to obtain the interaction behavior logic chain; By integrating the multi-dimensional environment configuration parameters, the object pose configuration scheme, and the interaction behavior logic chain, a Unity lightweight real-time rendering scene corresponding to the virtual space construction intent vector is constructed. Based on the cluster center distance of the semantic embedding vector, the adaptive rendering optimization parameters of the Unity lightweight real-time rendering scene are defined. Based on the lightweight real-time rendering scene and the adaptive rendering optimization parameters, an interactive preview view corresponding to the virtual space construction intent vector is generated.

8. The method for automatically constructing a personalized virtual space based on user intent as described in claim 1, characterized in that, The step of generating a semantic embedding vector corresponding to the virtual space construction intent vector based on the semantic similarity includes: Based on the semantic similarity, candidate key information entities corresponding to the virtual space construction intent vector are selected; Construct a multidimensional association matrix between the candidate key information entities and the virtual space construction intent vector; Extract the principal component eigenvectors and their corresponding importance weights from the multidimensional correlation matrix; Based on the importance weights, the principal component feature vectors are subjected to weighted fusion processing to obtain weighted fused feature vectors; The weighted fusion feature vector is mapped to a low-dimensional semantic space to obtain the low-dimensional semantic embedding vector corresponding to the candidate key information entity. Calculate the cluster center distance and distribution density of the low-dimensional semantic embedding vector in the low-dimensional semantic space; Based on the cluster center distance and the distribution density, the low-dimensional semantic embedding vector is normalized to obtain the semantic embedding vector corresponding to the virtual space construction intent vector.

9. The method for automatically constructing a personalized virtual space based on user intent as described in claim 1, characterized in that, The step of generating spatial topological constraints corresponding to the virtual space construction intent vector based on the semantic similarity includes: Extract the spatial relation predicates from the virtual space construction intent vector, and determine the spatial constraint strength coefficients corresponding to the spatial relation predicates; Based on the semantic similarity, calculate the semantic-feature coupling weights between different spatial relation predicates; Based on the spatial constraint strength coefficient and the semantic-feature coupling weight, a priority ranking rule for the spatial relation predicate is established. Based on the priority sorting rules, the constraint logic combination method and execution order corresponding to the virtual space construction intent vector are determined; Based on the spatial constraint strength coefficient, define the constraint tolerance range corresponding to the virtual space construction intention vector; By integrating the constraint logic combination method, the execution order, and the constraint tolerance range, spatial topological constraint conditions corresponding to the virtual space construction intent vector are generated.

Images