CN-121998242-A - Flood event construction method driven by cooperation of geographic scene and social data

CN121998242ACN 121998242 ACN121998242 ACN 121998242ACN-121998242-A

Abstract

The invention belongs to the technical field of flood control and flood prevention, in particular to a flood event construction method driven by cooperation of a geographic scene and social data, which comprises the following steps of S1, collecting and preprocessing social media data; S2, geographic scene body structural modeling oriented to disasters, S3, flood disaster situation knowledge map construction driven by a large language model, S4, a multi-mode space-time narrative expression mechanism of flood disasters, and S5, disaster space-time evolution visual expression driven by semantic knowledge reasoning. Through the technology closed loop of 'semantic analysis-knowledge generation-space-time deduction-visual narrative', the full chain path of systematic cognition from fragmented perception of disaster information is reconstructed, the spanning from static display to dynamic narrative and from macroscopic description to causal reasoning of disaster information expression is realized, the real-time cognition and autonomous response capability of the public to disaster risks are enhanced, and the scientificity and social cooperativity of disaster risk management are obviously enhanced.

Inventors

ZHAO ZIHENG
Ran Qingqing
LI JIAHAO

Assignees

西南交通大学

Dates

Publication Date: 20260508
Application Date: 20260116

Claims (9)

1. The flood event construction method driven by the cooperation of the geographic scene and the social data is characterized by comprising the following steps: s1, collecting and preprocessing social media data; s2, performing disaster-oriented geographic scene body structural modeling; S3, constructing a flood disaster situation knowledge graph driven by the large language model; s4, a flood disaster multi-mode space-time narrative expression mechanism; s5, disaster time-space evolution visual expression driven by semantic knowledge reasoning; The specific implementation flow of the step S1 comprises the following steps: Preliminarily collecting data information of users of each platform, and carrying out social media data Scoring the quality of (2): ; Wherein, the Scoring the time accuracy; Scoring spatial accuracy; scoring the content credibility; 、、 Is a weight coefficient; When the quality is scored When the following conditions are met, the piece of social media data is recorded Recording into an original social media dataset In (a): ; Wherein, the Is a scoring threshold; forming an original social media dataset : ; Wherein each record Representing a separate piece of social media data containing text, time, location information.
2. The method for constructing the flood event driven by the cooperation of the geographic scene and the social data according to claim 1, wherein the implementation flow of the step S1 further comprises the following steps: For original social media data set Preprocessing at least including noise filtering, format correction, unified time stamping, spatial alignment and text normalization to obtain cleaned social media data set : 。
3. The method for constructing the flood event driven by the cooperation of the geographic scene and the social data according to claim 2, wherein the specific implementation flow of the step S2 comprises the following steps: Constructing a geographical scene ontology model oriented to flood disasters as a semantic framework: ; Wherein, the Is an entity set; is an attribute set; is a set of relationships; is a hierarchy of classification levels; is a constraint rule set.
4. The method for constructing the flood event driven by the cooperation of the geographic scene and the social data according to claim 3, wherein the specific implementation flow of the step S3 comprises the following steps: In a geographic scene ontology Under the constraint of (1) from a social media dataset using a large language model Extracting structured knowledge : Knowledge obtained by extraction Expressed as triples: ; Wherein, the Is a disaster condition entity set; is an attribute set; Is a semantic relation set; Will extract the result Mapping to ontologies In the process, a flood disaster situation knowledge graph is constructed : ; ; Wherein, the For node collection, corresponding entity instance ; For edge collection, corresponding relation instance ; For node attribute set, corresponding attribute instance 。
5. The method for constructing a flood event driven by a combination of a geographic scene and social data according to claim 4, wherein in step S3, the large language model is LLM, and the social media data set is extracted by using the following method Structured knowledge of (a): ; Wherein, the Is a decimation function.
6. The method for constructing a flood event driven by cooperation of a geographical scene and social data according to claim 4, wherein the flood disaster situation knowledge graph in step S3 Evolution with time update: ; Wherein, the Updating operators for the maps; Is the increment of the map; Is newly added data; Is a time decay factor.
7. The method for constructing the flood event driven by the cooperation of the geographic scene and the social data according to claim 4, wherein the specific implementation flow of the step S4 comprises the following steps: Introduction of visual modalities And interaction modality Forming a flood narrative expression model: ; Wherein, the Expressing results for the narrative; Is a fusion function; At the semantic mapping layer, disaster entities are subjected to semantic hierarchy and space constraint in the ontology Binding with the geographic space position to realize semantic projection from a knowledge layer to a space visualization layer: ; Wherein, the Is the disaster situation occurrence position; is a corresponding time segment; is a semantic attribute vector; At the visual layer, a multi-mode expression strategy is adopted, semantic nodes, space tracks and disaster images are combined, and the spreading and influencing process of flood is displayed in a color coding, dynamic symbolization and time sequence animation mode; At the interaction level, users may be based on semantic relationships And carrying out dynamic query to form semantic-driven interactive narrative experience.
8. The method for constructing the flood event driven by the cooperation of the geographic scene and the social data according to claim 7, wherein the specific implementation procedure of the step S5 comprises the following steps: calculating causal link strength between events based on semantic reasoning: ; Wherein, the And (3) with Respectively represent knowledge patterns Two event entities in (a); Representing events And (3) with Semantic relationships between; is a contextual association between events; representing knowledge graph Global structural constraints of (a); is a semantic reasoning function based on a large language model; is a logical reasoning function based on ontology rules; representing semantic intensity and causal direction between events; Each event is processed Mapping into a triplet And (3) forming a space-time constraint set: ; Wherein, the Is the time at which the event occurred; is the spatial location of the event occurrence; Causal relationships between events are defined by semantic reasoning functions Determining, as an event And events When the following formula is satisfied, then both are considered to satisfy the conditions that make up the causal chain: ; Wherein, the Is a filtering threshold.
9. The method for constructing the flood event driven by the cooperation of the geographic scene and the social data according to claim 8, wherein the implementation process of the step S5 further comprises the following steps: at the visual level, the inference results are mapped into the narrative visual framework to form a semantic-driven visual expression mechanism, which specifically comprises three mapping modes: The time evolution mapping is used for displaying the forming, spreading and recovering processes of disasters through time axis and stage layering animation; Space diffusion mapping, namely, based on a three-dimensional terrain scene, showing a propagation path and an influence range of flood in a geographic space; and (3) mapping the semantic narrative, namely embedding the disaster event chain obtained by reasoning into a visual interface in the form of text narrative and semantic nodes, and realizing a closed-loop process from the semantic reasoning to the visual narrative.

Description

Flood event construction method driven by cooperation of geographic scene and social data Technical Field The invention belongs to the technical field of flood control and flood prevention, and particularly relates to a flood event construction method driven by cooperation of a geographic scene and social data. Background Flood disasters are one of the most frequent and most destructive natural disasters worldwide, and sudden, diffuse and space-time uncertainties cause disaster information propagation delay, so that public risk cognition and emergency decision efficiency are severely restricted. The traditional flood visualization method depends on remote sensing images and numerical models, and can show macroscopic inundation range, but has the problems of missing disaster details, low public participation, poor information timeliness and the like, and is difficult to meet urgent demands of the public on disaster information visualization. The rise of social media data provides a new dimension for flood disaster dynamic perception, millions of disaster related push messages are generated daily by platforms such as Twitter and microblog, and the social media data has the advantages of being strong in instantaneity, wide in coverage range, high in public participation and the like. However, the existing research mostly adopts the traditional NLP technology to process social media data, and faces the problems of low information extraction efficiency, serious semantic ambiguity, prominent noise interference and the like, so that extracted disaster entities and relations are inaccurate, and further the reliability and reliability of a visual result are affected. Therefore, the invention provides a flood event construction method driven by the cooperation of a geographic scene and social data. Disclosure of Invention In order to overcome the deficiencies of the prior art, at least one technical problem presented in the background art is solved. The technical scheme adopted for solving the technical problems is that the flood event construction method driven by the cooperation of the geographic scene and the social data comprises the following steps: s1, collecting and preprocessing social media data; s2, performing disaster-oriented geographic scene body structural modeling; S3, constructing a flood disaster situation knowledge graph driven by the large language model; s4, a flood disaster multi-mode space-time narrative expression mechanism; S5, disaster time-space evolution visual expression driven by semantic knowledge reasoning. Preferably, the specific implementation procedure of step S1 includes: Preliminarily collecting data information of users of each platform to form an original social media data set : ; For original social media data setPreprocessing at least including noise filtering, format correction, unified time stamping, spatial alignment and text normalization to obtain cleaned social media data set: ; Preferably, step S1 is performed on social media dataBefore collection, the quality of the materials is scored: ; When the quality is scored The piece of social media data is only selected when the following is satisfiedRecording into an original social media datasetIn (a):; preferably, the specific implementation procedure of step S2 includes: Constructing a geographical scene ontology model oriented to flood disasters as a semantic framework: ; Preferably, the specific implementation procedure of step S3 includes: In a geographic scene ontology Under the constraint of (1) from a social media dataset using a large language modelExtracting structured knowledge: Knowledge obtained by extractionExpressed as triples: Will extract the result Mapping to ontologiesIn the process, a flood disaster situation knowledge graph is constructed: Preferably, the large language model in step S3 is LLM, and the social media data set is extracted using the following methodStructured knowledge of (a): preferably, the flood disaster situation knowledge graph in step S3 Evolution with time update: Preferably, the specific implementation procedure of step S4 includes: Introduction of visual modalities And interaction modalityForming a flood narrative expression model: At the semantic mapping layer, disaster entities are subjected to semantic hierarchy and space constraint in the ontology Binding with the geographic space position to realize semantic projection from a knowledge layer to a space visualization layer: At the visual layer, a multi-mode expression strategy is adopted, semantic nodes, space tracks and disaster images are combined, and the spreading and influencing process of flood is displayed in a color coding, dynamic symbolization and time sequence animation mode; At the interaction level, users may be based on semantic relationships And carrying out dynamic query to form semantic-driven interactive narrative experience. Preferably, the specific implementation procedure of step S5 includes: calculating ca