CN115544275B - Natural disaster knowledge hypergraph construction method considering space-time process and disaster mechanism - Google Patents
Natural disaster knowledge hypergraph construction method considering space-time process and disaster mechanism Download PDFInfo
- Publication number
- CN115544275B CN115544275B CN202211502906.4A CN202211502906A CN115544275B CN 115544275 B CN115544275 B CN 115544275B CN 202211502906 A CN202211502906 A CN 202211502906A CN 115544275 B CN115544275 B CN 115544275B
- Authority
- CN
- China
- Prior art keywords
- disaster
- knowledge
- natural disaster
- layer
- natural
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
- G06F16/36—Creation of semantic tools, e.g. ontology or thesauri
- G06F16/367—Ontology
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
- G06F16/38—Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually
- G06F16/387—Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually using geographical or spatial information, e.g. location
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02A—TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE
- Y02A10/00—TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE at coastal zones; at river basins
- Y02A10/40—Controlling or monitoring, e.g. of flood or hurricane; Forecasting, e.g. risk assessment or mapping
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Databases & Information Systems (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Animal Behavior & Ethology (AREA)
- Computational Linguistics (AREA)
- Library & Information Science (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses a natural disaster knowledge hypergraph construction method considering space-time process and disaster mechanism, comprising the following steps: constructing a knowledge system of natural disasters; constructing a hierarchical natural disaster knowledge conceptual model, and establishing mapping between the conceptual model and an instance model; based on the multi-source data, natural disaster knowledge extraction is realized, and entity alignment and knowledge fusion are carried out; based on a knowledge hypergraph theory method, the expression of the hypergraph-based natural disaster entity and the hyperrelationship knowledge and the expression of the natural disaster instance are realized; and (3) analyzing the specific natural disaster event by adopting the steps 1 to 4, and forming a natural disaster knowledge hypergraph based on the TypeDB knowledge hypergraph visualization platform. The method not only inherits the advantage of simple structure of the existing knowledge graph, but also realizes the expression of the natural disaster space-time process knowledge and the complex disaster mechanism knowledge, and can provide knowledge support for natural disaster management and decision in the actual application scene.
Description
Technical Field
The invention relates to the technical field of disaster big data mining and geochemical knowledge modeling, in particular to a natural disaster knowledge hypergraph construction method considering space-time process and disaster mechanism.
Background
At present, the natural disaster monitoring method mainly comprises geological monitoring, remote sensing monitoring, air-space-ground integrated monitoring and the like, provides high-precision multi-source monitoring big data for natural disaster hidden danger monitoring, prediction and early warning and the like, and plays an important role in human disaster prevention, disaster reduction and emergency management. However, natural disasters have the complex characteristics of concealment, burstiness, uncertainty of space-time evolution, unclear disaster causing conditions, insufficient internal structure cognition, unclear deformation and cracking process and disaster forming mechanism and the like. Moreover, the daily and monthly mass natural disaster monitoring data has large data volume, wide sources and complex structure, but only a small part of the data can provide useful knowledge for natural disaster monitoring and predictive analysis. Therefore, a knowledge system and a knowledge base in the natural disaster field are constructed, and the mechanism knowledge and the interrelation of disaster factors, disaster-tolerant environments, disaster-bearing bodies and the like are checked, so that the hidden danger base of the natural disaster is mastered, and the early monitoring and forecasting and early warning capability of the natural disaster is improved.
At present, a relational database is generally adopted for expressing, storing and inquiring natural disaster information. The method is convenient for users to understand, but the data query is only an execution code for limiting certain areas in the database, and natural disaster knowledge cannot be obtained. And part of scholars explore the natural disaster knowledge expression based on the knowledge graph, and according to the knowledge representation and organization management mode of the knowledge graph, the nodes represent natural disaster entities, the sides represent the attributes of the natural disaster entities and the relations between the natural disaster entities, and a natural disaster knowledge net structure is constructed, so that a fact type natural disaster knowledge base which can be traced back, understood, interpreted and calculated by the machine is formed. However, knowledge such as geological environments, external induction factors, unique space-time evolution processes, disaster mechanisms and the like with complex and various natural disasters, particularly complex multi-element relations among disaster causing factors, disaster pregnancy environments and disaster bearing bodies, cannot be expressed by adopting a general knowledge graph expression model. In addition, the knowledge graph stores data in the form of triples, so that the complexity of geographical space data and natural disaster knowledge in the real world and the complex relationship between knowledge units are excessively simplified, and the disaster mechanism and the time-space evolution process of the natural disaster are difficult to express.
Disclosure of Invention
The invention aims to provide a natural disaster knowledge hypergraph construction method, which comprises the steps of firstly researching a natural disaster knowledge expression model considering a space-time process and a disaster mechanism, modeling a conceptual entity and a relationship of a natural disaster event, and realizing the expression of basic attribute characteristics, procedural knowledge and mechanistic knowledge of the natural disaster; then, constructing a layering natural disaster knowledge representation model which takes space and time as a framework and takes an object as a core and reveals a state-process space and time evolution process and a rule characteristic-disaster causing mode disaster mechanism; and finally, considering different entities and hierarchical structures forming the natural disasters, taking dynamic characteristics of continuous evolution and development of the natural disasters and mechanism knowledge formed in development change into consideration, constructing a natural disaster field knowledge hypergraph model oriented to hypergraph theory, and describing a natural disaster knowledge unit and semantic relation thereof in three aspects of depth, breadth and granularity.
In order to achieve the above purpose, the method for constructing the natural disaster knowledge hypergraph taking into account the space-time process and the disaster mechanism of the invention specifically comprises the following steps:
step 1: analyzing the characteristics and association relation of three elements of a disaster causing factor, a disaster-tolerant environment and a disaster-bearing body of a natural disaster, and the mechanism action of the three elements in each stage of the natural disaster space-time development process, condensing the space-time frame of the natural disaster, and constructing a natural disaster knowledge system;
step 2: based on a natural disaster knowledge system, a top-down and bottom-up combined mode is adopted to analyze the expression of the relationship between natural disaster entities on a disaster mechanism and a space-time development process, a layered natural disaster knowledge expression model is abstracted, and the model comprises a concept layer, an element layer, a rule layer, a mode layer and an instance layer, and mapping between the concept layer and the instance layer is established;
step 3: based on structured, semi-structured and unstructured data sources, a deep learning model is adopted, manual correction is assisted to realize extraction of natural disaster entities and relations, and entity alignment and knowledge fusion are carried out;
step 4: based on a hierarchical natural disaster knowledge expression model and a knowledge hypergraph theory method, carrying out hyperedge representation of natural disaster entities, attribute characteristics of the entities and connection and mapping among layers of a concept layer, an element layer, a rule layer, a mode layer and an instance layer, and realizing hypergraph-based natural disaster entity and hyperrelationship knowledge expression and natural disaster instance expression;
step 5: and (3) analyzing the specific natural disaster event by adopting the steps 1 to 4, and forming a natural disaster knowledge hypergraph based on the TypeDB knowledge hypergraph visualization platform.
Furthermore, in the step 3, the natural disaster entity is extracted by using CRF, word2vec-BiLSTM-CRF and BERT-BiLSTM-CRF models, the natural disaster relation is extracted by using CNN, attention-BiLSTM, transformer models, and a manual auxiliary mode is adopted to obtain a natural disaster knowledge extraction result.
Further, in step 3, similarity among object names, attributes and space-time development processes among natural disaster entities is judged by using word vector similarity, and then entity alignment and knowledge fusion are carried out according to the similarity.
Further, in the step 4, the natural disaster entity knowledge expression based on hypergraph specifically comprises modeling of a concept layer, an element layer, a rule layer, a mode layer and an instance layer of the natural disaster;
formalized representation of conceptual layer modeling of natural disasters:
in the method, in the process of the invention,Va set of vertices representing a natural disaster event,
vertex set representing external disaster causing factors, +.>
Vertex set representing gestational disaster environment, ++>
Representing a vertex set of the disaster-bearing body;Esuperlimit set representing natural disaster event, +.>
Is the beyond of the external disaster factor +.>
For the pregnant disaster environment, the people are beyond the limit and are added with the drugs>
Is the disaster-bearing body overtlimit; />
Indicating that is at->
The superside of the concept level consists of entities contained in the superside, and represents three levels of superside concept knowledge of a natural disaster knowledge system; />
Representing entities contained in a natural disaster knowledge system;
formalized representation of element-level modeling of natural disasters is:
in the superb
Represents the damage of the external disaster factor suffered by the natural disaster under a certain specific space-time state +.>
Variation of the intrinsic gestational disaster Environment->
And threat to disaster-bearing body->
；/>
Refers to the state of the natural disaster space-time development process, which comprises a sprouting stage, an initial stage, a development stage and a destruction stage;
the formalization of the rule layer modeling of natural disasters is expressed as:
in the method, in the process of the invention,
is the superside of the single element layer entity rule, < ->
First step of representing natural disaster factoriA rule;
formalized representation of model layer modeling of natural disasters is:
in the superb
Disaster causing mode representing natural disaster, +.>
Rule indicating disaster causing mode->
Refers to the state of the space-time development process of natural disasters;
the formal expression of the example layer modeling of natural disasters is:
in the method, in the process of the invention,
represent the firstjNatural disaster event of the strip.
Further, in the step 4, the hypergraph-based natural disaster hyperrelationship expression specifically comprises Sub relationships, change relationships, inclusion relationships, reasoning relationships, attribute relationships and instance layer semantic relationships;
formalized representation of Sub relationships is:
in the method, in the process of the invention,
and->
Is the upper and lower level of the overtlimit in the natural disaster conceptual layer; />
Representing +.>
The first part included by the supersidenThe sliver overtlimit.
Formalizing the change relation is expressed as:
in the method, in the process of the invention,
refers to the change relation beyond the limit, natural disaster entity in a certain state +.>
Natural disaster entity +.>
Which occurred before;
formalized representation of inclusion relationships is:
in the method, in the process of the invention,
representing a relationship overrun;
the reasoning relation is developed between the rule layer and the mode layer, and a unique disaster causing mode is determined by a plurality of rule nodes;
formalized representation of attribute relationships is:
in the method, in the process of the invention,
attribute override representing an instance comprising event content belonging to the instance in natural disaster event E>
；
Formalized representation of instance layer semantic relationships is:
in the method, in the process of the invention,
representing two natural disaster events->
The semantic relationship between them is superb.
In summary, the method provided by the invention has the following advantages: 1) The model takes different entities and hierarchical structures forming natural disasters into consideration, considers dynamic characteristics of continuous evolution and development of the natural disasters and mechanism knowledge formed in development change, takes space and time as a framework, takes objects as cores, reveals layered natural disaster knowledge of 'state-process' space and time evolution and 'rule feature-disaster-causing mode' disaster mechanism, and builds a bridge of mapping natural disaster data to natural disaster knowledge hypergraph. 2) The natural disaster knowledge hypergraph disclosed by the invention not only can inherit the advantage of simple structure of the existing knowledge graph, but also can express the multi-element complex relation problem among complex knowledge such as natural disaster factors, disaster-tolerant environments, disaster-bearing bodies and the like, has better universality in the disaster field, and can also provide knowledge support for natural disaster management and decision in actual application scenes.
Drawings
FIG. 1 is an overall flow of the present invention;
FIG. 2 is a landslide knowledge hypergraph construction technical flow in the invention;
FIG. 3 is a landslide knowledge system in accordance with the present invention;
FIG. 4 is a conceptual model of landslide knowledge in accordance with the present invention;
FIG. 5 is a graph based landslide concept knowledge representation in the present invention;
FIG. 6 is a graph of the present invention showing knowledge representation of landslide examples based on hypergraph;
FIG. 7 is a mapping of a landslide conceptual layer and an example layer based on hypergraph in the present invention.
Detailed Description
The technical scheme of the invention is further described below with reference to the accompanying drawings and examples.
The overall flow of the present invention is shown in figure 1. The method of the invention is explained in detail by landslide hazard, the technological process of landslide knowledge hypergraph construction is shown in figure 2, and the specific analysis process is as follows.
Step 1: as shown in fig. 3, the characteristics and association relation of three elements of a disaster causing factor, a disaster-tolerant environment and a disaster-bearing body of the landslide are analyzed, and the mechanism action of the three elements in each stage of the space-time development process of the landslide is exerted, so that a space-time framework of the landslide is condensed, and a landslide knowledge system is constructed.
Step 2: as shown in fig. 4, based on a landslide knowledge system, a top-down and bottom-up combined mode is adopted to analyze the expression of the relation between entities on the disaster mechanism and the process state, a landslide knowledge conceptual model is constructed, a layered landslide knowledge representation model, namely a conceptual layer, an element layer, a rule layer, a mode layer and an instance layer is abstracted, mapping between the conceptual model (conceptual layer) and the instance model (instance layer) is established, and basic units and association relations of landslide knowledge are depicted in three aspects of depth, breadth and granularity.
Step 3: based on structured, semi-structured and unstructured data sources such as landslide databases, monographs, documents and web page texts, a mode of combining rule models, machine learning and deep learning is adopted, and manual correction is assisted to realize landslide knowledge (entity+relation) extraction, so that a structured landslide knowledge extraction result is obtained, and entity alignment and knowledge fusion are carried out.
Step 4: as shown in FIG. 5, based on a hierarchical landslide knowledge representation model and a knowledge hypergraph theory method, a concept layer, an element layer, a rule layer, a mode layer landslide entity, attribute characteristics (such as basic attribute characteristics and rule characteristics) of the landslide entity and the relationship and mapping hyperedge representation among the layers are carried out, so that the hypergraph-based landslide entity and hyperrelationship knowledge expression is realized. As shown in FIG. 6, a hierarchical landslide knowledge representation model and a knowledge hypergraph theory method are based, so that the landslide entity of an instance layer, the attribute characteristics of the entity and the relationship and mapping hyperedge representation among the entity are realized, and the landslide instance expression based on hypergraph is realized.
Step 5: as shown in fig. 7, the landslide event is specifically analyzed by adopting the steps 1 to 4, and a landslide knowledge hypergraph is formed based on a TypeDB knowledge hypergraph visualization platform.
As an optimal technical scheme, landslide event composition objects in the step 1 are divided into disaster causing factors, disaster pregnancy environments and disaster bearing bodies. Disaster factors are various external factors causing landslide, including natural factors and human factors. The disaster-tolerant environment is a deep geological condition for generating landslide and the surface topography and land form, and is an inherent control factor of the landslide. The disaster-bearing body is an object acted by various landslide disaster-causing factors and disaster-pregnant environments, and is a collection of human society and ecological environments. Landslide is the result of the coupling synergistic effect of internal and external power, and can be classified into geological deformation, stratum lithology, landform characteristics and the like according to the difference of the objects. It should be noted that different objects can refer to related field classification methods, and perform more detailed type classification according to actual needs.
As a preferable technical scheme of the invention, the landslide space-time development process in the step 1 comprises 4 state stages, which are respectively: an initial deformation stage, a constant-speed deformation stage, an acceleration deformation stage and a breaking and stacking stage.
Initial deformation stage: the initial deformation stage of the slope body is that the deformation is from no to no, the trailing edge starts to generate intermittent cracks, the deformation curve shows relatively large slope, but the slope of the deformation curve is slowed down along with the time, and the characteristic of deceleration deformation is shown.
Constant velocity deformation stage: based on initial deformation, under the action of gravity and other factors, the rear edge cracks are communicated, the loess slope continues to deform at a similar rate basically, and the monitoring curve fluctuates due to the interference of external factors, but the macroscopic deformation rate at the stage basically remains unchanged.
Acceleration of deformation phase: when the deformation of the slope body is developed to a certain stage, the sliding surface is basically communicated, the local collapse is realized, the deformation speed is accelerated, the deformation speed can show a continuous acceleration and growth trend, the deformation curve is nearly steep before the whole slope body is unstable, and at the stage, the tangent angle of the loess bedrock landslide is smaller than that of the loess inner landslide.
Breaking up the accumulation phase: when the deformation of the slope is accelerated to a certain degree, the whole body rapidly slides down, the sliding body moves towards the slope toe against the friction resistance of the substrate under the action of gravitational potential energy, the sliding speed is slowed down, and the sliding body is piled up in a certain range to form a new landform.
As a preferable technical scheme, the expression of the relation between landslide entities in the step 2 on the disaster mechanism and the process state comprises three types, namely fact type knowledge, space-time process knowledge, rules and mechanism knowledge.
Facts type knowledge: including conceptual knowledge and instance knowledge of the different objects making up the landslide and its hierarchical structure.
Knowledge of spatiotemporal processes: through the evolution process of element-layer landslide factors and different states of the element-layer landslide factors in the whole landslide evolution process, the expression of the landslide space-time evolution process knowledge based on the state-process is realized, the basic conditions of state division are adopted by four evolution stages, and the space-time characteristics of the landslide factors such as loess landslide disaster factors, disaster-tolerant environments and disaster-bearing bodies in different stages and the process knowledge of landslide easily occur are described.
Rule and mechanism knowledge: by means of abstract knowledge expression of the condition-result of a rule layer (condition node) and a mode layer (result node), connection between a plurality of rule features of the landslide and a unique disaster causing mode is established, and complete expression of mechanism knowledge is achieved in landslide development change. As shown in fig. 4, the abstract knowledge representation based on the "state-process" and the "condition-result" is that the innermost side is a landslide conceptual layer, the landslide conceptual layer is transferred to an element layer through a relation, the element layer shows the dynamic evolution process of landslide factors under different states, and a series of rule features of landslide are obtained from the element layer to a rule layer. The outermost side is a mode layer, and the connection between a plurality of regularity features of the landslide and a unique disaster causing mode is established through a rule knowledge representation method based on a condition-result.
As a preferred technical solution, the data sources for landslide knowledge extraction in step 3 include three types, which are classified into structured data, semi-structured data and unstructured data.
Structured data: the method mainly comprises a landslide relation database, an image interpretation database and a public landslide database which are interpreted by a 'sky-earth-interior' monitoring means, and structural data which can be converted into a CSV format are directly stored in a knowledge hypergraph. The relational database represents and stores quantitative data such as numbers, symbols and the like through a two-dimensional table, each row of the table represents a landslide event, each column represents attribute characteristics of the landslide event, and the same type of entities have the same number and type of attributes; the image interpretation database mainly comprises landslide point position data, wherein the landslide event number of the data is large, but the attribute information is not detailed enough, and only comprises space-time information (occurrence time and occurrence position) of landslide and basic attributes (length, width, volume and the like) of the landslide; the disclosed landslide database includes not only global, nationwide, and wide-range data on government websites, but also regional landslide data sets constructed by students.
Semi-structured data: the method mainly refers to website disaster data such as news websites, weather websites, social media and the like, and the website disaster data is usually expressed as XML documents, JSON documents and the like, and structural data which can be converted into a CSV format by analyzing the XML documents and the JSON documents is directly stored in a knowledge hypergraph.
Unstructured data: the method mainly uses the geological environment of landslide, weather changes affecting landslide, artificial loading and unloading slope changes and other character records, landslide live photos, map data, satellite images and other pictures, and landslide point monitoring video and other audio and video expression forms. For the type of data, text data with higher quality is preferentially selected, and landslide knowledge in the text is extracted in a mode of adding manual auditing through an automatic extraction deep learning method, so that unstructured data is converted into structured data.
The specific steps of landslide knowledge extraction of unstructured data include:
s1, constructing a landslide event information annotation corpus of Chinese texts (journal documents, disaster reports and the like), wherein the annotated content comprises names, time, positions, attributes and behavior information elements for describing landslide objects.
S2, according to the labeling corpus, adopting a knowledge extraction model of two main streams of CRF and BiLSTM-CRF and a word2vec and BERT pre-training vector model to realize automatic extraction of landslide entities and attribute information in the text, and utilizing the advantages of CRF learning transfer characteristics to conduct label prediction of output information so as to complete extraction of landslide entities and attribute characteristic values.
S3, according to the labeling corpus, selecting CRF, word2vec-BiLSTM-CRF and BERT-BiLSTM-CRF models to extract entities, and selecting CNN, attention-BiLSTM, transformer models to extract relations, wherein the evaluation results are shown in the table 1.
Table 1 landslide knowledge extraction evaluation results
Note that: the accuracy P represents the proportion of the predicted correct case data to the predicted case data; the recall rate R represents the proportion of the predicted correct case data to the actual case data; the metric f1=2×pxr/(p+r) is used to measure the accuracy and recall.
S4, aligning landslide entities by calculating cosine similarity among word vectors, and carrying out landslide knowledge fusion. And generating word vectors of landslide entities by using a word2vec pre-training vector model, fine-tuning the model, defining the dimension of the word group to be 100-dimensional, and measuring the cosine similarity between any two entities under the same type of marks according to the following formula, wherein the cosine similarity has a value between 0 and 1, and the closer the result is to 1, the higher the similarity between the two entities is. When the cosine similarity is larger than the threshold value, the cosine similarity is considered to be the same landslide entity, the merging operation is carried out, and the merging precision is controlled by a manual auxiliary means.
In the method, in the process of the invention,
is a vector representation of two landslide entities processed by a pre-trained vector model.
As a preferable technical scheme, in the step 4, the landslide entity modeling based on the concept layer, the element layer, the rule layer, the mode layer and the example layer of the hypergraph theory is performed, and the modeling process of each layer is specifically as follows.
Modeling of a concept layer: the concept knowledge of landslide is expressed in a concept layer, and factors of a landslide knowledge system are divided into an overtone and an entity, wherein the overtone consists of the entity belonging to the overtone. Formalized representation:
in the method, in the process of the invention,Va set of vertices representing a landslide event,
vertex set representing disaster causing factor H, +.>
Vertex set representing disaster-tolerant environment P, +.>
Representing a top point set of the disaster-bearing body S;Esuperlimit set representing landslide event, +.>
Is a landslide disaster factor overrun, +.>
Is a landslide pregnancy disaster environment overtravel, +.>
Is a landslide disaster-bearing body overtlimit；/>
Indicating that is at->
The overtlimit of the concept level consists of entities contained in the overtlimit and represents three levels of overtlimit concept knowledge of a landslide knowledge system; />
Representing the entities contained in the landslide knowledge system.
Modeling of element layers: the element layer emphasizes the spatiotemporal process and the interaction mechanism generated by landslide factors in the process, and focuses on spatiotemporal knowledge representation based on 'state-process'. At the finest particle size
In concept, state knowledge is superimposed, and element layers of state-process are expressed, and rich space-time state information is contained, and formalized representation is as follows:
in the superb
Represents the damage of the external disaster factor to landslide in a specific space-time state +.>
Variation of the intrinsic gestational disaster Environment->
And threat to disaster-bearing body->
；/>
The state of the space-time development process of landslide event is used for revealing dynamic changes of different stages of the same landslide factor and landslide factors required for identifying landslide under a specific state, and the landslide factors comprise a sprouting stage, an initial stage, a development stage and a destruction stage.
Rule layer modeling: the rule layer entity carries out constraint expression on the value range, the type and the combination mode of the element layer entity, and the rule layer formalized representation is as follows:
in the method, in the process of the invention,
is the superside of the single element layer entity rule, < ->
Item number representing landslide event elementiRules are provided.
Modeling a mode layer: and the mode layer entity obtains the disaster-causing mode of the landslide by reasoning based on the result of the combined action of the rule layer entities. The uniquely determined disaster-causing mode can be obtained by reasoning according to different rules of different landslide factors under a plurality of states, and the rules are conditions and are defined bynThe individual conditions result in a unique disaster mode.
In the superb
Disaster-causing mode representing landslide event, +.>
Rule indicating disaster causing mode->
Refers to the state of the landslide event in the space-time development process.
Example layer modeling: the instance layer refers to a landslide event, and represents the attribute and the event content of the specific landslide event, and not only comprises disaster causing factors, disaster-tolerant environments and stage rule characteristic attributes of disaster-bearing bodies, but also comprises self attribute characteristics such as time attributes, space attributes, disaster intensity, disaster frequency and the like. The disaster event formalized representation of the instance layer is:
in the method, in the process of the invention,
represent the firstjA landslide event.
For a landslide event, its attribute features may be formally represented as:
in the method, in the process of the invention,
for regularity feature>
Is a basic attribute characteristic of landslide, and is->
Representing the overrun consisting of the regularity feature and the landslide basic attribute feature.
As a preferable technical scheme, landslide super relation modeling based on super graph theory in step 4 comprises Sub relation, change relation, inclusion relation, reasoning relation, attribute relation and instance layer semantic relation.
Sub relation: refers to the hierarchical classification between different supersides having common attributes, representing the logical hierarchical classification structure between upper and lower supersides. The Sub relationship formalized is expressed as:
in the method, in the process of the invention,
and->
Is the upper and lower stages of the overtlimit in the landslide concept layer; />
Representing +.>
The first part included by the supersidenThe sliver overtlimit.
Change relation: the relation among a plurality of elements in a continuous state in the element layer is used for describing the space-time evolution process of the landslide, and can be used for revealing the knowledge of the mechanism node representation process of the occurrence of the landslide. The change relation is formalized as:
in the method, in the process of the invention,
refers to the landslide entity ++under a certain state with the change relation exceeding the limit>
Landslide entity +.>
Which occurred before.
The inclusion relationship is as follows: the landslide element layer and the rule layer are contained relations, and the rule layer is a rule knowledge summary of element layer entities. For a certain element layer entity, a rule superside may be formed, containing rule layer entities belonging to the entity. The inclusion relationship is formally expressed as:
in the method, in the process of the invention,
representing inclusion relationships.
Inference relation: a unique disaster causing mode is determined by a plurality of rule nodes by expanding between a rule layer and a mode layer. The scheme provides a rule knowledge representation method based on a rule-mode, and the complete expression of rule knowledge in a landslide knowledge hypergraph is realized by means of a condition compound node.
Attribute relationship: landslide event
Attribute features of landslide event and content of each entity containing the event
There is an attribute relationship between them, including data attributes and object attributes. The attribute relationships are used to connect the element layer and the instance layer, formally expressed as:
in the method, in the process of the invention,
attribute overrun representing an instance, including event content belonging to the instance in landslide event E
；/>
And the same name as the corresponding entity node in the element layer.
Instance layer semantic relationship: in the example layer, after many landslide events, especially landslide with high grade and high intensity occur, a series of other disasters are often induced, and semantic relations such as initiation, mass-sending and the like exist between primary disaster events and secondary disaster events. The instance layer semantic relationship formalized representation is:
in the method, in the process of the invention,
representing two landslide events +.>
The semantic relationship between them is superb.
As a preferred technical solution, step 5 takes "party Chuan 2#" landslide as an example, and the landslide knowledge hypergraph conceptual layer realizes knowledge representation modeling of landslide conceptual entities, different hyperedges are composed of the conceptual entities contained therein, and low-level hyperedges are contained by high-level hyperedges, which is a high summary of landslide knowledge (fig. 5). The instance layer and concept layer models add a systematic "party Sichuan 2#" landslide instance (FIG. 6). The mapping of the concept knowledge and the instance knowledge comprises the steps of establishing a mapping relation between a party Chuan 2# "landslide event entity and a mode layer disaster-causing mode entity, establishing a mapping relation between a party Chuan 2#" landslide event characteristic value and a state entity of an element layer, mapping each characteristic value node of the party Chuan 2# "landslide event to a corresponding element layer state node corresponding to a unique landslide state, and mapping from scattered specific knowledge to system abstract knowledge so as to express the dynamic evolution process and disaster-causing mechanism of the party Chuan 2#" landslide instance (figure 7).
The foregoing is a specific embodiment of the present invention, but the scope of the present invention should not be limited thereto. Any changes or substitutions that would be obvious to one skilled in the art are deemed to be within the scope of the present invention, and the scope is defined by the appended claims.
Claims (5)
1. The natural disaster knowledge hypergraph construction method taking the space-time process and the disaster mechanism into consideration is characterized by comprising the following steps:
step 1: analyzing the characteristics and association relation of three elements of a disaster causing factor, a disaster-tolerant environment and a disaster-bearing body of a natural disaster, and the mechanism action of the three elements in each stage of the natural disaster space-time development process, condensing the space-time frame of the natural disaster, and constructing a natural disaster knowledge system;
step 2: based on a natural disaster knowledge system, a top-down and bottom-up combination mode is adopted, the expressions of relation among natural disaster entities on a disaster mechanism and a space-time development process are respectively analyzed through fact knowledge, space-time process knowledge, rules and mechanism knowledge, the rules and mechanism knowledge are abstract knowledge expressions by means of condition-result of a rule layer and a mode layer, connection between a plurality of regularity features of the natural disaster and a unique disaster causing mode is established, and complete expression of the mechanism knowledge is realized in the development and change of the natural disaster;
abstract the knowledge representation based on the status-process and the condition-result, abstract the layered natural disaster knowledge expression model, the innermost side is a conceptual layer, the conceptual layer is transferred to an element layer through the relation, the dynamic evolution process of natural disaster factors under different states is shown in the element layer, a series of rule features of the natural disaster are obtained from the element layer to a rule layer, the outermost side is a mode layer and an instance layer, and the mapping between the conceptual layer and the instance layer is established;
step 3: based on structured, semi-structured and unstructured data sources, a deep learning model is adopted, manual correction is assisted to realize extraction of natural disaster entities and relations, and entity alignment and knowledge fusion are carried out;
step 4: based on a hierarchical natural disaster knowledge expression model and a knowledge hypergraph theory method, carrying out hyperedge representation of natural disaster entities, attribute characteristics of the entities and connection and mapping among layers of a concept layer, an element layer, a rule layer, a mode layer and an instance layer, and realizing hypergraph-based natural disaster entity and hyperrelationship knowledge expression and natural disaster instance expression;
step 5: and (3) analyzing the specific natural disaster event by adopting the steps 1 to 4, and forming a natural disaster knowledge hypergraph based on the TypeDB knowledge hypergraph visualization platform.
2. The natural disaster knowledge hypergraph construction method considering space-time process and disaster mechanism as claimed in claim 1, wherein the method comprises the following steps: in the step 3, CRF, word2vec-BiLSTM-CRF and BERT-BiLSTM-CRF models are selected to extract natural disaster entities, CNN, attention-BiLSTM, transformer models are selected to extract natural disaster relations, and a manual auxiliary mode is adopted to obtain a natural disaster knowledge extraction result.
3. The natural disaster knowledge hypergraph construction method considering space-time process and disaster mechanism as claimed in claim 2, wherein the method comprises the following steps: and 3, judging the similarity among object names, attributes and space-time development processes among natural disaster entities by using word vector similarity, and then carrying out entity alignment and knowledge fusion according to the similarity.
4. The natural disaster knowledge hypergraph construction method considering space-time process and disaster mechanism as claimed in claim 3, wherein: the natural disaster entity knowledge expression based on hypergraph in step 4 specifically comprises modeling of a concept layer, an element layer, a rule layer, a mode layer and an instance layer of natural disasters;
formalized representation of conceptual layer modeling of natural disasters:
V＝{V H ,V P ，V S }
wherein V represents the vertex set of the natural disaster event, V H Vertex set representing external disaster factor, V P Vertex set representing pregnant disaster environment, V S Representing a vertex set of the disaster-bearing body; e represents a hyperedge set of natural disaster events,
is the beyond of the factor caused by the external disaster,
for the pregnant disaster environment, the people are beyond the limit and are added with the drugs>
Is the disaster-bearing body overtlimit; e (E) i Representing supersides in the concept level of i, consisting of entities contained in the supersides, representing three levels of superside concept knowledge of a natural disaster knowledge system; />
Representing entities contained in a natural disaster knowledge system;
formalized representation of element-level modeling of natural disasters is:
in the superside S c Represents the damage V of the external disaster factor suffered by the natural disaster under a certain specific space-time state Hc Variation of intrinsic pregnancy and disaster environments V Pc And to disaster-bearing bodyThreat V caused Sc ；S i Refers to the state of the natural disaster space-time development process, which comprises a sprouting stage, an initial stage, a development stage and a destruction stage;
the formalization of the rule layer modeling of natural disasters is expressed as:
F E ＝{S c1 ，S c2 ，…，S ci ，…，S cn }
wherein F is E Is the superside of the entity rule of a single element layer, S ci An ith rule indicating a natural disaster element;
formalized representation of model layer modeling of natural disasters is:
in the superside M E Represents a disaster causing mode of a natural disaster,
rules representing disaster causing modes S i Refers to the state of the space-time development process of natural disasters;
the formal expression of the example layer modeling of natural disasters is:
E＝{E 1 ，E 2 ，…，E j ，…，E n }
wherein E is j Represents the natural disaster event of the j th item.
5. The natural disaster knowledge hypergraph construction method considering space-time process and disaster mechanism as claimed in claim 4, wherein the method comprises the following steps: the natural disaster super-relationship expression based on the super-graph in the step 4 specifically comprises Sub relationships, change relationships, inclusion relationships, reasoning relationships, attribute relationships and instance layer semantic relationships;
formalized representation of Sub relationships is:
wherein E is i And E is i+1 Is the upper and lower level of the overtlimit in the natural disaster conceptual layer;
representing E in a natural disaster conceptual layer i The n-th sub superside contained in the superside;
formalizing the change relation is expressed as:
wherein R is v Refers to a natural disaster entity under a certain state with a change relation exceeding the limit
Natural disaster entity +.>
Which occurred before;
formalized representation of inclusion relationships is:
R contain ＝{S C ，V Hc ，V Pc ，V Sc ，S i }，i∈(1，4)
wherein R is contain Representing a relationship overrun;
the reasoning relation is developed between the rule layer and the mode layer, and a unique disaster causing mode is determined by a plurality of rule nodes;
formalized representation of attribute relationships is:
wherein R is property Attribute overrun representing an instance, including event content belonging to the instance in natural disaster event E
Formalized representation of instance layer semantic relationships is:
R instance ＝{E 1 ，E 2 }
wherein R is instance Representing two natural disaster events E 1 ，E 2 The semantic relationship between them is superb.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211502906.4A CN115544275B (en) | 2022-11-29 | 2022-11-29 | Natural disaster knowledge hypergraph construction method considering space-time process and disaster mechanism |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211502906.4A CN115544275B (en) | 2022-11-29 | 2022-11-29 | Natural disaster knowledge hypergraph construction method considering space-time process and disaster mechanism |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN115544275A CN115544275A (en) | 2022-12-30 |
| CN115544275B true CN115544275B (en) | 2023-06-16 |
Family
ID=84722771
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211502906.4A Active CN115544275B (en) | 2022-11-29 | 2022-11-29 | Natural disaster knowledge hypergraph construction method considering space-time process and disaster mechanism |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115544275B (en) |
Families Citing this family (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116186294B (en) * | 2023-04-28 | 2023-08-18 | 浙江大学 | Hypergraph-based intelligent factory element universe knowledge representation method and application thereof |
| CN117252255B (en) * | 2023-07-31 | 2024-06-18 | 中国矿业大学(北京) | Disaster emergency knowledge graph construction method oriented to auxiliary decision |
| CN117992617A (en) * | 2023-12-26 | 2024-05-07 | 中国科学院空天信息创新研究院 | Method, device, equipment and storage medium for representing earth resource environment tuple |
| CN117951238A (en) * | 2024-02-29 | 2024-04-30 | 江西师范大学 | Space-time knowledge graph construction method based on geographic scene |
| CN118313451B (en) * | 2024-06-11 | 2024-09-17 | 中国科学院沈阳应用生态研究所 | Ecological protection red line risk knowledge graph structure optimization method based on space-time hypergraph |
Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109992672A (en) * | 2019-04-11 | 2019-07-09 | 华北科技学院 | Knowledge mapping construction method based on disaster scene |
Family Cites Families (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2015184221A1 (en) * | 2014-05-30 | 2015-12-03 | Georgetown University | A process and framework for facilitating information sharing using a distributed hypergraph |
| CN110874833B (en) * | 2019-10-15 | 2022-03-22 | 合肥工业大学 | SAR image change detection method based on hypergraph matching |
| CN115098696A (en) * | 2022-06-20 | 2022-09-23 | 广州市阿尔法软件信息技术有限公司 | Method and device for constructing urban physical examination knowledge graph and storage medium |
-
2022
- 2022-11-29 CN CN202211502906.4A patent/CN115544275B/en active Active
Patent Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109992672A (en) * | 2019-04-11 | 2019-07-09 | 华北科技学院 | Knowledge mapping construction method based on disaster scene |
Also Published As
| Publication number | Publication date |
|---|---|
| CN115544275A (en) | 2022-12-30 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN115544275B (en) | Natural disaster knowledge hypergraph construction method considering space-time process and disaster mechanism | |
| Rubin et al. | Veracity roadmap: Is big data objective, truthful and credible? | |
| Qiu et al. | Automatic spatiotemporal and semantic information extraction from unstructured geoscience reports using text mining techniques | |
| Demir et al. | Predicting occurrence of liquefaction-induced lateral spreading using gradient boosting algorithms integrated with particle swarm optimization: PSO-XGBoost, PSO-LightGBM, and PSO-CatBoost | |
| Yue et al. | Intelligent services for discovery of complex geospatial features from remote sensing imagery | |
| CN112328794B (en) | Typhoon event information aggregation method | |
| US9594755B2 (en) | Electronic document repository system | |
| Masmoudi et al. | Memon: Modular environmental monitoring ontology to link heterogeneous earth observed data | |
| Seedah et al. | Ontology for querying heterogeneous data sources in freight transportation | |
| Kumar et al. | BS-LSTM: an ensemble recurrent approach to forecasting soil movements in the real world | |
| Morkner et al. | Distilling data to drive carbon storage insights | |
| Shbita et al. | Building spatio-temporal knowledge graphs from vectorized topographic historical maps | |
| Cui et al. | Inferring twitters’ socio-demographics to correct sampling bias of social media data for augmenting travel behavior analysis | |
| Dorodnykh et al. | Conceptual model engineering for industrial safety inspection based on spreadsheet data analysis | |
| Ostermann et al. | Reproducible research and GIScience: An evaluation using GIScience conference papers | |
| Wen et al. | Construction and application of a multilevel geohazard domain ontology: A case study of landslide geohazards | |
| Cao et al. | Geographic process modeling based on geographic ontology | |
| Bimonte et al. | From volunteered geographic information to volunteered geographic OLAP: A VGI data quality-based approach | |
| Brodaric et al. | DOLCE ROCKS: Integrating Geoscience Ontologies with DOLCE. | |
| Xu et al. | Carbonate Ontology and Its Application for Integrating Microfacies Data | |
| Manoj et al. | OntoINT: a framework for ontology integration based on entity linking from heterogeneous knowledge sources | |
| Umuhoza et al. | Trustworthiness assessment of knowledge on the semantic sensor web by provenance integration | |
| Li et al. | Event cube–a conceptual framework for event modeling and analysis | |
| Zhang | Text Complexity Classification Data Mining Model Based on Dynamic Quantitative Relationship between Modality and English Context | |
| Gong et al. | A term extraction algorithm based on machine learning and comprehensive feature strategy |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |