CN-121996678-A - Information visualization and interactive management method and management system thereof

CN121996678ACN 121996678 ACN121996678 ACN 121996678ACN-121996678-A

Abstract

The application relates to the technical field of computers, discloses an information visualization and interactive management method and system, and aims to solve the technical problems of static information expression, single interactive mode, high cognitive load and weak logic association of the conventional visualization system. The method comprises the steps of constructing a five-dimensional information node model fusing space-time coordinates and task contexts, performing visual arrangement by adopting a dynamic layout algorithm mixed by force guidance and grid constraint, and realizing focus area detail expansion, path highlighting and view zooming by fusing three paths of eye movement, touch control and gesture, wherein the system comprises modules of multi-source data access, semantic map construction, context enhancement, a dynamic layout engine, multi-mode interaction perception, causal chain deduction and the like. According to the scheme, the collaborative optimization of the situation awareness, the layout stability and the interaction naturalness of the information expression is realized, and the cognition efficiency and the decision support capability under the complex service scene are remarkably improved.

Inventors

HU YINGYING
ZHANG FENGJIAO
LI XINLAI

Assignees

湖北经济学院

Dates

Publication Date: 20260508
Application Date: 20260123

Claims (10)

1. An information visualization and interactive management method, comprising: acquiring original business data from a plurality of heterogeneous data sources, wherein the original business data comprises a structured database record, a semi-structured log file and unstructured text or image data; Carrying out semantic analysis and entity extraction on the original business data, identifying key information entities and attributes thereof, and establishing a logic association relationship between the key information entities to form an initial information map; Performing context enhancement processing on the initial information map, and constructing an information node with five-dimensional feature vectors by introducing four metadata such as a time stamp, a geographic position, a user role and a task context, wherein the five-dimensional feature vectors are composed of entity identifications, attribute sets, associated edge sets, space-time coordinates and context labels; Inputting the information nodes with the five-dimensional feature vectors into a dynamic layout engine based on a force guiding and grid constraint mixing mechanism, wherein the dynamic layout engine calculates optimal space coordinates of each information node in a visual canvas according to a preset visual priority weight matrix, and the visual priority weight matrix is jointly determined by importance scores of the nodes, a user attention historical value and current task correlation; rendering the information nodes and the associated edges thereof on a visual canvas to generate an initial visual view, wherein the initial visual view adopts a layered coloring strategy, the core nodes are presented in high-saturation colors, and the edge nodes are displayed in low-transparency gray scales; Synchronously acquiring a fixation point coordinate sequence, a finger contact area and hand space pose data of a user by deploying a multi-mode interaction sensing unit; Calculating a current visual focus area of a user based on the gaze point coordinate sequence, triggering detail unfolding operation of information nodes in the area if the duration time of the visual focus area exceeds a preset threshold, and superposing and displaying sub-level attributes or related events of the nodes in a pop-up micro-view mode; based on the finger contact area, identifying zooming, translation or frame selection operation executed by a user, if a continuous double-finger pinch gesture is detected, executing global zooming on the current view, wherein the zooming scale is in a linear relation with the gesture interval change rate; based on the hand space gesture data, analyzing an air gesture instruction executed by a user, when a specific gesture that a palm faces a screen and five fingers are opened is identified, starting a panoramic overview mode, switching a current view to a bird's eye view angle and reducing detail levels of all nodes; in a deep analysis state, a context reasoning engine is called, causal chain deduction is carried out on information nodes in a locking area, a forward influence path and a backward tracing path are generated, arrows with different colors are respectively marked, and meanwhile, confidence scores on each path are calculated, and the confidence scores are weighted by association strength, data freshness and source credibility; And superposing the forward influence path and the backward tracing path to the current visual view, supporting a user to trigger a drill-down operation by clicking any path node, and expanding the bottom data source and the processing logic layer by layer until the original record level.
2. The method for visualizing and interactively managing information as in claim 1, wherein performing semantic parsing and entity extraction on the raw business data comprises: Directly extracting the field name of the table as an attribute key and the field value as an attribute value for the structured database record; Matching key fields with a semi-structured log file by adopting a regular expression template, wherein the regular expression template is pre-configured with a time format, an IP address mode and an error code dictionary; Calling a named entity recognition model for unstructured text data, wherein the named entity recognition model is trained based on a two-way long-short-term memory network and conditional random field joint architecture and supports recognition of four types of entities including organization, personnel, equipment numbers and event types; And extracting embedded text from unstructured image data through an optical character recognition engine, and recognizing a dashboard, an indicator light or a flow chart symbol in the image by combining with a target detection algorithm to convert the dashboard, the indicator light or the flow chart symbol into a structured state description.
3. The method for visualizing and interactively managing information as in claim 2, wherein the constructing the visual priority weight matrix comprises: Assigning a base importance score to each information node, the base importance score being equal to a logarithmic value of a sum of an ingress and an egress of the node; Inquiring a user behavior log library, acquiring the access frequency of the user to the node within the past thirty days, and normalizing the access frequency to be used as a user attention historical value; comparing task identifiers transmitted by a current task planning interface, if the service domain of the node is matched with the current task, giving a task correlation degree of one, otherwise, giving zero; and carrying out weighted summation on the three numerical values according to weight coefficients of 0.5, 0.3 and 0.2 to obtain the final vision priority weight.
4. The method of claim 3, wherein the dynamic layout engine uses a hybrid constraint mechanism, wherein the force directing part simulates a spring-repulsive force model, attractive forces are applied between adjacent nodes, repulsive forces are applied between non-adjacent nodes with a distance smaller than a preset threshold, the grid constraint part divides the visual canvas into a plurality of logical cells with equal size, each cell accommodates at most one core node, if a plurality of core nodes are distributed to the same cell, the core nodes are ranked according to their visual priority weights, only the highest weight is reserved, and the rest nodes are automatically degraded to be subordinate marks and offset to adjacent idle cells.
5. The method for visualizing and interactively managing information as in claim 4, wherein the data fusion process of the multi-modal interactive awareness unit comprises: performing sliding window smooth filtering on the gaze point coordinate sequence output by the eye movement tracking sensor, wherein the window length is ten frames; Carrying out space-time alignment on the coordinates of the contact point reported by the touch panel and hand gesture data output by the gesture recognition module, wherein the time deviation of the alignment accuracy is not more than twenty milliseconds, and the space deviation is not more than five pixels; When the three types of perception data are uniformly directed to the same area in time and space, the effective interaction intention is judged, otherwise, the three types of perception data are regarded as noise to be discarded.
6. The method for visualizing and interactively managing information as in claim 5, wherein calculating a current visual focus area of the user based on the gaze point coordinate sequence, and triggering a detailed expansion operation of the information nodes in the area if the duration of the visual focus area exceeds a preset threshold, comprises: Retrieving all direct sub-attributes of the node and associated events; filtering irrelevant information according to the context label; And ejecting the card-type micro view at the upper right of the node, wherein the micro view background is a semitransparent frosted glass effect, and the content adopts compact typesetting to support rolling viewing.
7. The method of information visualization and interactive management according to claim 6, wherein identifying a zoom, pan, or box operation performed by a user based on the finger contact area comprises: If a continuous double-finger pinch gesture is detected, global scaling is performed on the current view, and the scaling is in a linear relationship with the gesture pitch change rate; And if the single-finger long-press dragging action is detected, activating a path following mode, highlighting all the traversed information nodes along a dragging track, and generating a temporary association chain.
8. The method of claim 7, wherein invoking a context inference engine to perform causal chain deduction on information nodes within a locked area in a deep analysis state comprises: performing breadth-first search along the forward direction of the associated edge from the focus node, and multiplying the path confidence coefficient by the associated strength coefficient of the edge every time one layer is expanded; And performing depth-first search along the reverse direction of the associated edge, and multiplying the path confidence by a data freshness attenuation factor every step of backtracking, wherein the data freshness attenuation factor is an exponential function, the base number of the data freshness attenuation factor is 0.98, and the exponent is the difference between the current time and the data generation time divided by eight hours.
9. An information visualization and interactive management system, comprising: the multi-source data access module is used for acquiring original service data from a plurality of heterogeneous data sources, wherein the original service data comprises a structured database record, a semi-structured log file and unstructured text or image data; The semantic graph construction module is used for carrying out semantic analysis and entity extraction on the original business data, identifying key information entities and attributes thereof, and establishing a logic association relationship between the key information entities to form an initial information graph; The context enhancement module is used for performing context enhancement processing on the initial information map, and constructing an information node with a five-dimensional feature vector by introducing four metadata including a time stamp, a geographic position, a user role and a task context; The dynamic layout engine module is used for inputting the information nodes with the five-dimensional feature vectors into a dynamic layout engine based on a force guiding and grid constraint mixing mechanism, and calculating the optimal space coordinates of each information node in the visual canvas according to a preset visual priority weight matrix; the visual rendering module is used for rendering the information nodes and the associated edges thereof on a visual canvas, generating an initial visual view and distinguishing the importance of the nodes by adopting a layered coloring strategy; the multi-mode interaction sensing module is used for deploying an eye tracking sensor, a touch panel and a six-degree-of-freedom gesture recognition module and synchronously collecting the gaze point coordinate sequence, the finger contact area and the hand space pose data of a user; the focus response control module is used for calculating the current visual focus area of the user based on the gaze point coordinate sequence and triggering detail unfolding operation when the preset stay condition is met; The gesture instruction analysis module is used for identifying zoom, translation, frame selection, panoramic overview and depth analysis instructions executed by a user based on the finger contact area and the hand space pose data and generating corresponding view control signals; the causal chain deduction module is used for performing causal chain deduction on the information nodes in the locking area under the state of deep analysis, generating a forward influence path and a backward tracing path, and calculating the confidence score of each path; And the drill-down navigation module is used for supporting the user to trigger drill-down operation by clicking the path node, and expanding the bottom data source and the processing logic layer by layer until the original record level.
10. The information visualization and interactive management system of claim 9, wherein the dynamic layout engine module is configured to: simulating a spring-repulsive force model, applying attractive force between adjacent nodes, and applying repulsive force between non-adjacent nodes with a distance smaller than a preset threshold value; The visual canvas is divided into a plurality of logic cells with equal size, each cell accommodates at most one core node, if a plurality of core nodes are distributed to the same cell, the visual priority weight of the core nodes is sorted according to the visual priority weight of the core nodes, only the highest weight of the core nodes is reserved to occupy the center of the cell, and the rest nodes are automatically degraded to be attached marks and offset to the edge positions of adjacent idle cells.

Description

Information visualization and interactive management method and management system thereof Technical Field The invention belongs to the technical field of computers, and particularly relates to an information visualization and interactive management method and an information visualization and interactive management system. Background With the deep advancement of big data, artificial intelligence and enterprise digital transformation, information management has gradually evolved from a traditional static storage and retrieval mode to a high-order form of dynamically organizing, association mining and real-time response to massive heterogeneous data. In this context, the information visualization serves as a key bridge for connecting data and decisions, and has a core value in visually presenting complex data relationships in a graphical manner, so as to assist users in quickly understanding information structures and making decisions. However, currently mainstream information management systems generally display data by adopting static charts or preset views, and lack dynamic modeling capability for deep semantic association between information, which makes it difficult for users to explore, reconstruct or adjust an information network in real time in an interaction process. Among other things, interactive information management techniques aim to drive the dynamic evolution of data views through user operations (such as drag, screening, focusing, or path tracking), thereby enabling a "what you see is what you control" management experience. The basic principle of the technology is to map the user intent to an operation instruction to the underlying data model and update the visual interface synchronously to reflect the change of information state. However, existing systems face multiple challenges in achieving this goal, on the one hand, the visualization engine is highly coupled with the data management logic, resulting in view update delay, interaction response delay, and on the other hand, most schemes only support unidirectional operations (such as query only or edit only), and cannot support multi-granularity, multi-dimensional coordinated edits while maintaining information consistency. In the prior art, information visualization often depends on a fixed template to generate a chart, and is difficult to adapt to different data topological structures, interactive operation is often limited to surface layer UI events and is not deeply bound with semantic relation of information entities, and more importantly, a unified mechanism is lacked, so that the integrity, timeliness and traceability of information association can be synchronously maintained in the user interaction process. Especially in complex business scenes (such as knowledge graph operation and maintenance, project dependence management or supply chain cooperation), the defects cause that a user cannot efficiently complete dynamic recombination of information, causal tracing or influence range deduction, and severely restrict the agility and accuracy of man-machine cooperation decision. Therefore, an information processing method of deep fusion visual expression and interactive management logic is needed to realize dynamic disclosure, flexible control and closed-loop verification of information association. Disclosure of Invention The invention provides an information visualization and interactive management method, which comprises the steps of constructing a unified semantic mapping model of multi-source heterogeneous data, carrying out structural analysis and context association modeling on original data from different service systems to form an information map with a hierarchical topological relation, automatically arranging nodes and edges in the information map in a two-dimensional or three-dimensional visualization canvas according to a preset visual priority rule by adopting a dynamic layout algorithm based on space coordinate constraint, simultaneously, introducing a multi-channel man-machine interaction perception mechanism, capturing operation intention, gazing focus and gesture track of a user in real time, driving local re-rendering of visual elements and self-adaptive adjustment of information granularity to realize conversion from static display to context perception interaction, and finally achieving collaborative optimization of information expression definition, user cognitive load minimization and operation response accuracy. According to one aspect of the present invention, there is provided an information visualization and interactive management method, including: acquiring original business data from a plurality of heterogeneous data sources, wherein the original business data comprises a structured database record, a semi-structured log file and unstructured text or image data; Carrying out semantic analysis and entity extraction on the original business data, identifying key information entities and attributes thereof, an