CN117077631A - Knowledge graph-based engineering emergency plan generation method - Google Patents

Knowledge graph-based engineering emergency plan generation method Download PDF

Info

Publication number
CN117077631A
CN117077631A CN202311330094.4A CN202311330094A CN117077631A CN 117077631 A CN117077631 A CN 117077631A CN 202311330094 A CN202311330094 A CN 202311330094A CN 117077631 A CN117077631 A CN 117077631A
Authority
CN
China
Prior art keywords
knowledge
model
entity
emergency
knowledge graph
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.)
Pending
Application number
CN202311330094.4A
Other languages
Chinese (zh)
Inventor
刘婷
张群
王明疆
高焕焕
李俊
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
PowerChina Northwest Engineering Corp Ltd
Original Assignee
PowerChina Northwest Engineering Corp Ltd
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by PowerChina Northwest Engineering Corp Ltd filed Critical PowerChina Northwest Engineering Corp Ltd
Priority to CN202311330094.4A priority Critical patent/CN117077631A/en
Publication of CN117077631A publication Critical patent/CN117077631A/en
Pending legal-status Critical Current

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F40/00—Handling natural language data
    • G06F40/10—Text processing
    • G06F40/166—Editing, e.g. inserting or deleting
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/045—Combinations of networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/0464—Convolutional networks [CNN, ConvNet]
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N5/00—Computing arrangements using knowledge-based models
    • G06N5/02—Knowledge representation; Symbolic representation
    • G06N5/022—Knowledge engineering; Knowledge acquisition
    • G06N5/025—Extracting rules from data
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q10/00—Administration; Management
    • G06Q10/10—Office automation; Time management
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
    • G06Q50/10—Services
    • G06Q50/26—Government or public services

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Business, Economics & Management (AREA)
  • General Physics & Mathematics (AREA)
  • General Engineering & Computer Science (AREA)
  • Data Mining & Analysis (AREA)
  • Artificial Intelligence (AREA)
  • Health & Medical Sciences (AREA)
  • Computational Linguistics (AREA)
  • General Health & Medical Sciences (AREA)
  • Evolutionary Computation (AREA)
  • Mathematical Physics (AREA)
  • Human Resources & Organizations (AREA)
  • Strategic Management (AREA)
  • Computing Systems (AREA)
  • Software Systems (AREA)
  • Tourism & Hospitality (AREA)
  • Molecular Biology (AREA)
  • Entrepreneurship & Innovation (AREA)
  • General Business, Economics & Management (AREA)
  • Biophysics (AREA)
  • Biomedical Technology (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Marketing (AREA)
  • Economics (AREA)
  • Operations Research (AREA)
  • Audiology, Speech & Language Pathology (AREA)
  • Primary Health Care (AREA)
  • Educational Administration (AREA)
  • Development Economics (AREA)
  • Quality & Reliability (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)

Abstract

The application discloses a knowledge graph-based engineering emergency plan generation method, belongs to the technical field of engineering emergency, and can solve the problems of low intelligent level and poor emergency disposal efficiency of the existing emergency management mode. The method comprises the following steps: s1, acquiring multi-source security management data of an engineering to construct a text data set; s2, carrying out knowledge extraction on the text data set to construct a knowledge graph; s3, acquiring a test problem, and generating an emergency plan according to the test problem and the knowledge graph. The method is used for generating the engineering emergency plan.

Description

Knowledge graph-based engineering emergency plan generation method
Technical Field
The application relates to a knowledge-graph-based engineering emergency plan generation method, and belongs to the technical field of engineering emergency.
Background
Dam engineering is large in scale and complex in geological conditions, risk factors for triggering emergency events in engineering construction and operation processes are various, and once dangerous events occur, serious influence can be caused on surrounding ecological environments, life and property of surrounding residents can be threatened, so that timely and effective emergency disposal work is a serious issue for safety management of the dam.
The emergency management mode of the dam engineering at present is to store an emergency plan in a paper or electronic text form, emergency disposal work mainly relies on a security manager to manually search an emergency plan text file, although the content search of the plan text can be roughly realized through a keyword search method, the search result is usually more in content and larger in range, the emergency disposal scheme is difficult to quickly find, and the emergency disposal mode based on the static text is usually limited by experience and capability level of disposal staff and cannot quickly search and acquire the plan information, so that the emergency time is long, the efficiency is low and omission is easy to generate, and the security management efficiency is reduced.
Disclosure of Invention
The application provides a knowledge graph-based engineering emergency plan generation method, which can solve the problems of low intelligent level and poor emergency treatment efficiency of the existing emergency management mode.
The application provides a knowledge graph-based engineering emergency plan generation method, which comprises the following steps:
s1, acquiring multi-source security management data of an engineering to construct a text data set;
s2, carrying out knowledge extraction on the text data set to construct a knowledge graph;
s3, acquiring a test problem, and generating an emergency plan according to the test problem and the knowledge graph.
Optionally, the S2 specifically includes:
s21, entity extraction is carried out on the text data set by adopting an entity extraction model, and an entity tag sequence is obtained;
s22, performing relation extraction on the entity tag sequence by adopting a segmented convolutional neural network to obtain a knowledge triplet;
s23, constructing a knowledge graph according to the knowledge triplet.
Optionally, the step S21 specifically includes:
s211, performing entity labeling on the text data set to obtain labeled text;
s212, inputting the marked text into an entity extraction model to obtain an entity tag sequence.
Optionally, the step S211 specifically includes:
and performing BIO entity labeling of type expansion on the text data set to obtain labeled text.
Optionally, the entity extraction model is a BERT-BiLSTM-CRF fusion model constructed by a BERT model, a BiLSTM model and a CRF model;
the BERT model is input as a text after labeling; the character embedding layer of the BERT model is an input layer of the BiLSTM model; the input layer of the CRF model is a BiLSTM layer of the BiLSTM model; the output of the CRF model is the physical tag sequence.
Optionally, the step S22 specifically includes:
s221, determining relation classification among entities according to the entity tag sequence;
s222, carrying out relation extraction on the entity tag sequence by adopting a segmented convolution neural network according to the relation classification to obtain a knowledge triplet.
Optionally, a pooling layer in the segmented convolutional neural network adopts a maximum pooling mode.
Optionally, after S1, the method further includes:
s4, preprocessing the text data set to obtain a processed text data set;
correspondingly, the step S2 specifically comprises the following steps:
and carrying out knowledge extraction on the processed text data set to construct a knowledge graph.
Optionally, the preprocessing includes: text washing, disabling word removal, and character segmentation.
Optionally, the step S23 specifically includes:
and storing the knowledge triples into a graph database, and constructing a knowledge graph corresponding to the knowledge triples in the graph database.
The application has the beneficial effects that:
(1) According to the engineering emergency plan generation method, the standardized, structured and semantic engineering emergency plan knowledge base is constructed based on the knowledge graph technology, and the intelligent level and emergency disposal efficiency of on-site emergency management are improved through efficient storage, visual display and quick retrieval of the plan knowledge.
(2) The engineering emergency plan generation method provided by the application establishes a knowledge graph of the emergency plan for the field of hydroelectric engineering. The fusion model provided in the entity extraction process of the knowledge graph solves the problems of word ambiguity, lack of a corpus in the field of hydroelectric engineering, low recognition accuracy and the like. The knowledge graph comprises various elements and related knowledge of the hydropower engineering, such as monitoring equipment, early warning information, emergency plan steps and the like. Knowledge in the hydropower field can be fused by constructing a knowledge graph, and an association relationship between the knowledge is established, so that a foundation is provided for generating an emergency plan.
(3) The engineering emergency plan generating method provided by the application fuses information from different data sources, including field emergency specifications, historical event data, real-time monitoring data and the like. The knowledge graph can structure various concepts, attributes and relationships so that related knowledge can be stored and retrieved in a more flexible and extensible manner. Compared with the traditional document or database storage mode, the knowledge graph can provide richer and more accurate association information. By comprehensively utilizing the information of different data sources, the emergency plan can be generated more comprehensively and accurately.
(4) According to the engineering emergency plan generation method provided by the application, the plan generation process is automatic and intelligent, and the plan is automatically analyzed, processed and generated through an artificial intelligent algorithm and a model, so that the requirement of manual intervention is reduced, and the efficiency and consistency of plan generation are improved. Compared with the traditional static plan, the method can be used for adjusting and improving the plan more flexibly, and the accuracy and operability of emergency response are improved.
Drawings
FIG. 1 is a flow chart of a method for generating an engineering emergency plan according to an embodiment of the present application;
FIG. 2 is a schematic diagram of an overall construction flow of an engineering emergency knowledge graph provided by an embodiment of the application;
FIG. 3 is a schematic diagram of a physical extraction model architecture according to an embodiment of the present application;
FIG. 4 is a schematic diagram of a BERT model structure according to an embodiment of the present application;
FIG. 5 is a schematic diagram of a BiLSTM model network structure according to an embodiment of the present application;
FIG. 6 is a schematic diagram of a segmented convolutional neural network according to an embodiment of the present application;
fig. 7 is a schematic diagram of a functional flow for querying and displaying an emergency plan of a knowledge graph according to an embodiment of the present application.
Detailed Description
The present application is described in detail below with reference to examples, but the present application is not limited to these examples.
The knowledge graph is developed on the basis of a traditional knowledge base and is generally composed of an entity-relation-entity or entity-relation-attribute triplet, and compared with the traditional knowledge base, the knowledge graph contains fewer semantics and rich examples, is flexible in form and is higher in expandability.
In order to cope with the emergency disposal work of complex hydroelectric engineering (such as dam engineering), the operation bottleneck of the traditional emergency disposal depending on manual experience is broken through, the intelligent level of safety management is improved, the scattered text of an emergency plan is converted into the mutually associated knowledge by means of a knowledge graph technology, and timely and effective auxiliary information is provided for the emergency decision making.
The application defines the knowledge graph as a setWherein->Representing a set of entities in a knowledge base, comprising +.>Seed entity (L.)>Representing a set of relationships in a knowledge base, comprising +.>A relationship. />Represents the fact set, one fact is triad +.>Representation of-> ，/>，Representing the head, relationship and tail entities of the triplet, respectively, indicating that the two entities are connected by a specific relationship. The overall construction flow of the hydropower engineering emergency knowledge graph is shown in fig. 2, and the core of the overall construction flow is knowledge extraction, and is divided into entity extraction and relation attribute extraction, then the extracted triples are stored in a graph database to form the knowledge graph, and knowledge application is realized through system development.
Specifically, an embodiment of the present application provides a method for generating an engineering emergency plan based on a knowledge graph, as shown in fig. 1 to fig. 7, where the method includes:
s1, acquiring multi-source security management data of engineering to construct a text data set.
In the embodiment of the application, the text data set can be constructed by selecting multi-source data such as monitoring reports, patrol data, accident descriptions, emergency plans and the like.
Since raw text data often contains various noise information that affects the accuracy of entity recognition. Thus, after S1, the method may further comprise:
s4, preprocessing the text data set to obtain a processed text data set;
specifically, the text preprocessing may be performed according to the following three steps: (1) Text cleaning is performed to remove useless data such as spaces, punctuation marks, letters and the like which interfere with the recognition effect in the training process. (2) Disabling word removal, removing individual chinese characters, e.g., have been used, individual, etc., that are commonly used in the dataset, can reduce the efficiency of training. (3) Character segmentation, dividing input text data into individual characters (E 1 ， E 2 ，…，E N ) The individual characters are then stored as a vocabulary, i.e. each character has a corresponding unique identifier (denoted as character: ID (identity)]）。
S2, carrying out knowledge extraction on the text data set to construct a knowledge graph.
The method specifically comprises the following steps:
s21, entity extraction is carried out on the text data set by adopting an entity extraction model, and an entity tag sequence is obtained.
S21 specifically comprises:
s211, entity labeling is carried out on the text data set, and labeled text is obtained.
In practical application, BIO entity labeling of type expansion can be performed on the text data set, and labeled text is obtained.
The method is characterized in that BIO is adopted for entity marking aiming at text data in different forms, wherein B-X represents the beginning of an entity X, I-X represents the end of the entity X, O does not belong to any entity type, and the traditional entity identification category is expanded from three categories (organization name, person name and place name) to ten categories (incident time, incident type, place, personnel, supplies, equipment, organization, incident grade, incident cause and disposal measure) by combining with a text set in the field of dam engineering, and the concrete is shown in the table 1.
TABLE 1 text naming entity labeling for dam engineering Emergency plans
S212, inputting the marked text into the entity extraction model to obtain an entity tag sequence.
The entity extraction model can be a BERT-BiLSTM-CRF fusion model constructed by a BERT (Bidirectional Encoder Representation from Transformers) model, a BiLSTM (Bi-directional Long Short-Term Memory, two-way long and short Term Memory network) model and a CRF (Conditional Random Fields, conditional random field) model;
the BERT model is input as a text after labeling; the character embedding layer of the BERT model is an input layer of the BiLSTM model; the input layer of the CRF model is a BiLSTM layer of the BiLSTM model; the output of the CRF model is the physical tag sequence.
The application provides a BERT-BiLSTM-CRF fusion model based on machine learning and deep learning for entity extraction of marked text of a dam, which is shown by referring to FIG. 3 and mainly comprises the following steps:
(1) Character vectorization: a pretrained language model based on BERT is constructed, in which text is represented by a single character as a model input. After extracting features by the bi-directional transducer encoder, a character-level vectorized representation (T 1 ，T 2 ，…，T N ). Referring to fig. 4, the BERT model core structure is two bidirectional transducer encoders, and each transducer block is composed of two sublayers, namely a multi-head self-attention mechanism and a fully-connected feedforward network. Each sub-layer is added with residual connection and normalization, and the output of the sub-layer is that。
In the transducer block, the data first passes through the multi-head attention module to obtain a weighted feature vector. In the attention mechanism, each character has 3 different vectors, namely a Query vector (Q), a Key vector (K) and a Value vector (V), and the length is 64. The present application uses scaling dot products to calculate vectors in the attention mechanism to prevent it from being too large in the inner product, where +.>Representation->And->Is calculated as the dimension of。
The multi-head attention mechanism is realized byA different pair of linear transformations->、/>、/>And (3) projecting, and finally connecting different attention results in series:
，
。
wherein:for the output layer->，/>，/>Vectors respectively->、/>And->Linear layer weights of (2); />Is the number of characters.
The data is then sent to a fully connected feed forward network module comprising a two-layer structure: the first layer is an activation function ReLU, the second layer is a linear activation function, and the output character vectorization result is:
；
wherein:and->Represents the weight and deviation of the first layer, respectively, ">And->Representing the weight and bias of the second layer, respectively. After extracting features by the bi-directional transducer encoder, a character-level vectorized representation (T 1 ，T 2 ，…，T N ）。
(2) Semantic feature extraction: the application is thatThe dynamic character vector constructed in the last step is used as a character embedding layer to improve an input layer of a BiLSTM model, the improved BiLSTM model is used as a feature extractor to model to obtain complex semantic features, the improved BiLSTM is a combination of forward LSTM and backward LSTM, the improved BiLSTM network structure is shown in figure 5, and the LSTM model is composed ofInput character +.>Cell state->Temporary cell status->Hidden layer status->Forgetful doorMemory door->And an output door->Composition is prepared. It delivers useful information for subsequent computations by discarding garbage and memorizing new information. First of all, the forgetting gate is calculated and the information to be forgotten is selected +.>Then calculate the memory gate and temporary cell states, < +.>，/>Then calculate the current cell stateFinally, the current states of the output layer and the hidden layer are calculated +.>，/>。
Wherein:and->Weights and deviations in neurons, respectively; subscript->、/>And->A forget gate, an input gate and an output gate respectively; />Representing the scalar product of the two vectors. Finally, the hidden layer states output by the BiLSTM layer are connected to form a sentence-level feature vector +.>。
(3) Entity tag sequence output: by modifying the output layer of BiLSTM, the output of BiLSTM layer is used as the input of CRF layer during training, the application adopts linear chain member random field, whereinFor the input variables, the sequence to be marked is indicated, < >>Representing input text +.>Vectors of individual characters; />For outputting the sequence, express AND->One-to-one tag sequence. The goal of the model is to maximize +.>Obtaining conditional probability by maximum likelihood estimation>The likelihood function during training is +.>And outputting an output sequence with the maximum overall probability according to the following formula in final prediction:
；
in the method, in the process of the application,represents a weight vector, consisting of->And->Determining; />Represents global feature vector, by->And->Determining; />Representing all tag sequences that meet the BIO marking rules.
S22, performing relation extraction on the entity tag sequence by adopting a segmented convolutional neural network to obtain a knowledge triplet.
Based on the entity tag sequence, a segmented convolutional neural network (PCNN) is adopted, and the word segmentation technology is combined to realize the relation extraction of the emergency plan, wherein the extraction content is a head entityRelation->And tail entity->Triads of three elements。
S22 specifically comprises the following steps:
s221, determining the relation classification among the entities according to the entity tag sequence.
Defining relationship classifications among entities according to entity classifications in the dataset, wherein 9 entity relationships are defined in total: < emergency organization, management, emergency materials >, < accident type, where they occur >, < place of occurrence, where they are located >, < emergency personnel, which are classified as emergency organization >, < accident type, reporting, emergency organization >, < accident type, classification, accident level >, < accident cause, resulting in accident type >, < accident cause, time of occurrence >, < accident type, take action, dispose action >.
S222, according to the relation classification, carrying out relation extraction on the entity tag sequence by adopting a segmented convolution neural network to obtain a knowledge triplet.
Wherein, the pooling layer in the segmented convolutional neural network (PCNN) adopts a maximum pooling mode.
The relation extraction model PCNN is generated by modifying the pooling layer of the Convolutional Neural Network (CNN) model by changing the pooling layer from local pooling to maximum pooling, as shown in FIG. 6As shown. A sentence is divided into three sections, and each section is subjected to maximum pooling after convolution to obtain three values, so that compared with the traditional maximum pooling, only one value can be obtained from each convolution kernel, and the local maximum pooling can be used for extracting the characteristic information of the sentence more fully. For a length ofThe word vector dimension is->Is +.>The convolution kernel is dimensioned +.>When the remaining length is insufficient in the convolution process>When it is filled with 0, the length of the convolution result is +.>The convolution operation has the formula of
，/>。
In order to capture the different features, the application will useThe convolution kernels, the output of the convolution operation is，/>Wherein->The value range is related to the sentence length. The output of each convolution kernel is divided into three parts by two entities, and a vector with a segment maximum pooling output length of 3 can be expressed as:
；
to three-dimensional vectorCan get->After being activated by a nonlinear function, the output is as follows:
；
then integrating the sentence characteristics obtained by the convolution layer and the pooling layer to realize the extraction of the relation types, predicting the conditional probability by a softmax classifier and generating the relation with the highest probability, as shown in the following formula,yrepresenting the total category of the relationship classification, representing the final output result.
；
。
S23, constructing a knowledge graph according to the knowledge triplet.
The method comprises the following steps: and storing the knowledge triples into a graph database, and constructing a knowledge graph corresponding to the knowledge triples in the graph database.
And constructing a knowledge graph according to the entities and the relations in the extracted knowledge triples, wherein the entities correspond to the nodes in the knowledge graph, and the relations correspond to the edges in the knowledge graph. The process of constructing the emergency plan knowledge graph in the Neo4j graph database mainly comprises three steps, namely the following specific steps: (1) Importing an entity file, creating entity nodes, and adding a UNIQUE index (UNIQUE) for each entity; (2) Importing a relation file, matching the relation among the nodes, and generating a relation edge; (3) creating a knowledge graph.
S3, acquiring a test problem, and generating an emergency plan according to the test problem and the knowledge graph.
The extracted knowledge triples are firstly stored in a Neo4j graph database to realize the visualization of key data in an emergency plan, structural knowledge such as accident reasons, accident types, emergency measures and the like in the running process of the hydroelectric engineering can be intuitively known through a knowledge chain in the graph, and the emergency decision efficiency is improved. The emergency knowledge graph application system comprises three functional pages of entity identification, relation inquiry and plan retrieval. The entity identification page can automatically generate and check an entity identification result after inputting a dam engineering emergency plan text and clicking and submitting the text; the relation inquiry page can check a relation link map and a relation list associated with the relation inquiry page through inputting a keyword entity; the plan query page can generate a map demonstration of a corresponding emergency plan when a specific test problem, such as what is the emergency treatment method of a flood disaster, is input, and can obtain sudden accidents such as flood disaster on the flood dike, local landslide of a slope, and unstable slope toe, which are caused by the flood disaster, different accidents correspond to different treatment measures, and when the unstable slope toe occurs, measures such as stone throwing and pressure foot of a geomembrane bag can be adopted.
The constructed hydropower engineering emergency knowledge graph application system is based on a Windows 10 operating system, and is developed by adopting a Python 3.6 programming language, a Django development framework and a Neo4j graph database. Wherein Django is a Web application framework of an open source code written by Python, and adopts MVT (Model-View-Template-View) development mode to realize retrieval and visual display of entity and relation related sub-graphs. Fig. 7 shows a functional flow of the Django framework for implementing knowledge graph query display, and the query request sent by the browser is transmitted to the view layer through the HTTP transmission protocol, the Web server interface and the route, and the view layer encapsulates the operation of the application page. The model layer encapsulates the query operation on the Neo4j graph database, the template layer encapsulates the page display function written by the HTML/CSS/JS files, and the sub-graph can be displayed after the browser receives the query result.
The engineering emergency plan generation method based on the knowledge graph improves the emergency plan generation efficiency: the plan generating method based on the knowledge graph can automatically extract and organize information from a large amount of knowledge in the hydropower field, reduces the workload of manual collection and retrieval, and greatly improves the efficiency of plan generation. Along with the continuous evolution and update of the knowledge in the field of hydroelectric engineering, the method can quickly add the latest knowledge to the knowledge map and be timely applied to the generation of emergency plans, so that the capability of the method is maintained.
The knowledge-graph-based engineering emergency plan generation method improves the accuracy and the comprehensiveness of the plan: by abstracting and modeling knowledge and rules in the field of hydroelectric engineering and introducing real-time monitoring data, accident data, early warning data and the like, information sharing, communication and auxiliary decision making in a decision making process are supported, and the rationality and flexibility of a plan are improved. The plan generating method based on the knowledge graph can make full use of information in the knowledge graph to carry out reasoning and deducing, and the generated plan is based on comprehensive and accurate professional knowledge, so that artificial omission and mistakes are effectively avoided, and the quality and accuracy of the plan are improved.
The engineering emergency plan generating method based on the knowledge graph improves emergency response capability: the generated emergency plan can better adapt to different emergency situations and requirements through verification and optimization links. The proposal contains various emergency treatment measures and coping strategies, which can help emergency personnel to make correct decisions and actions under emergency conditions and improve emergency response capability. The method can generate personalized plans according to the specific requirements and site conditions of users, considers the differences and the particularities of various emergency scenes, and provides more flexible and highly-adaptive plan generating capability.
The engineering emergency plan generating method based on the knowledge graph supports emergency drilling and training: the generated plan can be used as the basis of emergency exercises and training, and can be customized and adjusted according to actual conditions and requirements. The plan contains detailed operation steps and emergency procedures, so that emergency personnel can be helped to be familiar with and master the requirements and methods of emergency work.
The engineering emergency plan generating method based on the knowledge graph provides decision support: based on the requirements of management personnel and the characteristics of hydropower engineering, an intelligent plan recommendation system is designed. By analyzing the problem or scene description input by the user, the applicable emergency plan or similar cases are automatically recommended by using the expertise and experience of the related fields, and quick and personalized plan generation suggestions are provided. The generated plans can be used as part of a decision support system to provide decision references and basis for the manager. The plan contains information in the aspects of risk assessment, resource allocation and the like, so that a manager can be helped to make a reasonable decision, and the emergency management level is improved.
While the application has been described in terms of preferred embodiments, it will be understood by those skilled in the art that various changes and modifications can be made without departing from the scope of the application, and it is intended that the application is not limited to the specific embodiments disclosed.

Claims (10)

1. The method for generating the engineering emergency plan based on the knowledge graph is characterized by comprising the following steps of:
s1, acquiring multi-source security management data of an engineering to construct a text data set;
s2, carrying out knowledge extraction on the text data set to construct a knowledge graph;
s3, acquiring a test problem, and generating an emergency plan according to the test problem and the knowledge graph.
2. The method according to claim 1, wherein S2 specifically comprises:
s21, entity extraction is carried out on the text data set by adopting an entity extraction model, and an entity tag sequence is obtained;
s22, performing relation extraction on the entity tag sequence by adopting a segmented convolutional neural network to obtain a knowledge triplet;
s23, constructing a knowledge graph according to the knowledge triplet.
3. The method according to claim 2, wherein S21 specifically comprises:
s211, performing entity labeling on the text data set to obtain labeled text;
s212, inputting the marked text into the entity extraction model to obtain an entity tag sequence.
4. A method according to claim 3, wherein S211 is specifically:
and performing BIO entity labeling of type expansion on the text data set to obtain labeled text.
5. A method according to claim 2 or 3, wherein the entity extraction model is a BERT-BiLSTM-CRF fusion model constructed from a BERT model, a BiLSTM model and a CRF model;
the BERT model is input as a text after labeling; the character embedding layer of the BERT model is an input layer of the BiLSTM model; the input layer of the CRF model is a BiLSTM layer of the BiLSTM model; the output of the CRF model is the physical tag sequence.
6. The method according to claim 2, wherein S22 specifically comprises:
s221, determining relation classification among entities according to the entity tag sequence;
s222, carrying out relation extraction on the entity tag sequence by adopting the segmented convolution neural network according to the relation classification to obtain a knowledge triplet.
7. The method of claim 6, wherein the pooling layer in the segmented convolutional neural network employs a maximum pooling approach.
8. The method according to claim 1, characterized in that after said S1, the method further comprises:
s4, preprocessing the text data set to obtain a processed text data set;
correspondingly, the step S2 specifically comprises the following steps:
and carrying out knowledge extraction on the processed text data set to construct a knowledge graph.
9. The method of claim 8, wherein the preprocessing comprises: text washing, disabling word removal, and character segmentation.
10. The method according to claim 2, wherein S23 is specifically:
and storing the knowledge triples into a graph database, and constructing a knowledge graph corresponding to the knowledge triples in the graph database.
CN202311330094.4A 2023-10-16 2023-10-16 Knowledge graph-based engineering emergency plan generation method Pending CN117077631A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202311330094.4A CN117077631A (en) 2023-10-16 2023-10-16 Knowledge graph-based engineering emergency plan generation method

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202311330094.4A CN117077631A (en) 2023-10-16 2023-10-16 Knowledge graph-based engineering emergency plan generation method

Publications (1)

Publication Number Publication Date
CN117077631A true CN117077631A (en) 2023-11-17

Family

ID=88708364

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202311330094.4A Pending CN117077631A (en) 2023-10-16 2023-10-16 Knowledge graph-based engineering emergency plan generation method

Country Status (1)

Country Link
CN (1) CN117077631A (en)

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117993499A (en) * 2024-04-03 2024-05-07 江西省水利科学院(江西省大坝安全管理中心、江西省水资源管理中心) Multi-mode knowledge graph construction method for four pre-platforms for flood control in drainage basin
CN117993499B (en) * 2024-04-03 2024-06-04 江西省水利科学院(江西省大坝安全管理中心、江西省水资源管理中心) Multi-mode knowledge graph construction method for four pre-platforms for flood control in drainage basin

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115438195A (en) * 2022-10-19 2022-12-06 中国农业银行股份有限公司 Construction method and device of knowledge graph in financial standardization field
CN115952297A (en) * 2022-12-15 2023-04-11 中车工业研究院有限公司 Subway fire scene-oriented emergency plan knowledge graph construction method and device
CN116304070A (en) * 2022-09-09 2023-06-23 清华大学 Automatic generation method of power grid emergency treatment scheme based on knowledge graph construction

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116304070A (en) * 2022-09-09 2023-06-23 清华大学 Automatic generation method of power grid emergency treatment scheme based on knowledge graph construction
CN115438195A (en) * 2022-10-19 2022-12-06 中国农业银行股份有限公司 Construction method and device of knowledge graph in financial standardization field
CN115952297A (en) * 2022-12-15 2023-04-11 中车工业研究院有限公司 Subway fire scene-oriented emergency plan knowledge graph construction method and device

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117993499A (en) * 2024-04-03 2024-05-07 江西省水利科学院(江西省大坝安全管理中心、江西省水资源管理中心) Multi-mode knowledge graph construction method for four pre-platforms for flood control in drainage basin
CN117993499B (en) * 2024-04-03 2024-06-04 江西省水利科学院(江西省大坝安全管理中心、江西省水资源管理中心) Multi-mode knowledge graph construction method for four pre-platforms for flood control in drainage basin

Similar Documents

Publication Publication Date Title
CN108897857B (en) Chinese text subject sentence generating method facing field
CN104318340B (en) Information visualization methods and intelligent visible analysis system based on text resume information
CN113723632A (en) Industrial equipment fault diagnosis method based on knowledge graph
CN113392986B (en) Highway bridge information extraction method based on big data and management maintenance system
CN110990590A (en) Dynamic financial knowledge map construction method based on reinforcement learning and transfer learning
CN109918671A (en) Electronic health record entity relation extraction method based on convolution loop neural network
CN110188192B (en) Multi-task network construction and multi-scale criminal name law enforcement combined prediction method
CN111858940B (en) Multi-head attention-based legal case similarity calculation method and system
CN107368521B (en) Knowledge recommendation method and system based on big data and deep learning
CN117453921B (en) Data information label processing method of large language model
CN117271767B (en) Operation and maintenance knowledge base establishing method based on multiple intelligent agents
CN114647741A (en) Process automatic decision and reasoning method, device, computer equipment and storage medium
CN113343690B (en) Text readability automatic evaluation method and device
CN110750635A (en) Joint deep learning model-based law enforcement recommendation method
CN116610816A (en) Personnel portrait knowledge graph analysis method and system based on graph convolution neural network
CN116822625A (en) Divergent-type associated fan equipment operation and detection knowledge graph construction and retrieval method
CN114331122A (en) Key person risk level assessment method and related equipment
Qu et al. Knowledge-driven recognition methodology for electricity safety hazard scenarios
CN117574898A (en) Domain knowledge graph updating method and system based on power grid equipment
CN117151222A (en) Domain knowledge guided emergency case entity attribute and relation extraction method thereof, electronic equipment and storage medium
CN116881689A (en) Knowledge-enhanced user multi-mode online comment quality evaluation method and system
CN117236676A (en) RPA process mining method and device based on multi-mode event extraction
CN117216617A (en) Text classification model training method, device, computer equipment and storage medium
CN116069951A (en) Construction worker safety knowledge extraction and knowledge graph construction method
CN117077631A (en) Knowledge graph-based engineering emergency plan generation method

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