Domain expert knowledge driven crisis event evolution context generation method and device

Abstract

The application discloses a crisis event evolution context generation method and device driven by domain expert knowledge, and belongs to the field of artificial intelligence, comprising the following steps: based on the experience of domain experts, a basic data model is constructed by using a program, and the model is run based on a computer processor, which is used for supporting the whole event evolution context analysis; event detection analysis is performed based on internet data by using a computer, an event sentence is formed, and a model is run by using the computer to mine basic crisis clues and important crisis clues; meanwhile, the development trend of the crisis event is tracked; time sequence, coordination and causality analysis are performed based on the basic crisis clues, the important crisis clues and the development trend by using the computer, the connection between crisis events is strengthened, and the crisis event evolution context is generated in different dimensions of event types and target types by using the computer through crisis event classification. The application can generate a complete and rich crisis event evolution context.

Description

Technical Field

[0001] This invention relates to the field of artificial intelligence, and more specifically, to a method and apparatus for generating crisis event evolution context driven by domain expert knowledge. Background Technology

[0002] A crisis refers to a dangerous and calamous moment, a test of decision-making and problem-solving abilities, and a turning point for individuals, groups, and society, involving life-or-death situations, transfers of interests, and many other impacts. Crisis events are typically sudden or major social events that may directly threaten the fundamental goals and values ​​of individuals, organizations, or society, such as life, property, safety, and order. In today's globalized world, the world faces numerous problems such as rising temperatures, increased extreme weather events, environmental pollution, resource shortages, and economic instability, all of which present significant crisis risks. Therefore, it is necessary to strengthen the ability and level of crisis identification, understanding, and management, and to take effective and specific measures to address them.

[0003] Crises are generally characterized by their magnitude, complexity, danger, and difficulty in prediction, exhibiting a high degree of suddenness and immediacy. With the rapid development of internet technology and the explosive growth of data in cyberspace, crises are often obscured by the boundless information. Simply relying on human observation and analysis is no longer sufficient for effective and rapid comprehensive analysis and understanding of crisis events, hindering their prevention and control. Therefore, with advancements in science and technology and the accumulation of extensive domain knowledge and experience, there is an urgent need to effectively combine technology and knowledge to conduct evolutionary analysis of the development path of crisis events, generating information such as the preceding background and current state of the crisis, thereby improving the comprehensive understanding and control of crisis events and reducing the damage and losses caused by them. Summary of the Invention

[0004] The purpose of this invention is to overcome the shortcomings of the prior art and provide a method and apparatus for generating crisis event evolution context driven by domain expert knowledge. It utilizes computer technology to provide a method and apparatus that can generate a complete and rich crisis event evolution context.

[0005] The objective of this invention is achieved through the following solution:

[0006] A domain expert knowledge-driven method for generating crisis event evolution timelines includes the following steps:

[0007] S1, based on the knowledge and experience of domain experts, uses a program to build a basic data model containing the characteristics and indicators of crisis events in the domain, and runs the model on a computer processor to support the analysis of the entire event evolution.

[0008] S2, using computers to perform event detection and analysis based on Internet data, forming event sentences, and using computers to run the model to uncover basic and important crisis clues; at the same time, tracking the development trend of crisis events;

[0009] S3. Using computers, based on the basic crisis clues, important crisis clues and development trends, perform time-series, coordination and causal relationship analysis to strengthen the connection between crisis events. Also, using computers, crisis event classification is used to generate crisis event evolution patterns in different dimensions such as event type and target type.

[0010] Furthermore, the basic crisis clues include relevant meetings, statements, and holidays; the important crisis clues include behaviors and states; and the development trend of the crisis event includes the occurrence, intensification, weakening, and resolution of the crisis event.

[0011] Furthermore, in step S1, the step of constructing a basic data model containing crisis event characteristics and indicators within the domain, based on the knowledge and experience of domain experts, specifically includes the following sub-steps:

[0012] Using computer programming, a data model is constructed from two dimensions: crisis event characteristics and crisis event indicators, to establish a unified crisis event model template. Then, crisis event models in different fields are based on this template and are visualized and instantiated using domain expert knowledge.

[0013] Furthermore, in step S2, the use of a computer to perform event detection and analysis based on Internet data to form event sentences, and the use of a computer to run the model to mine basic crisis clues and important crisis clues, specifically includes text data analysis and spatiotemporal data analysis;

[0014] The text data analysis specifically includes the following sub-steps:

[0015] Guided by the crisis event indicators of the crisis event model, we first use computers to mine background signs, then use the BERT pre-trained language model to detect events in text data to obtain event sentences, then classify events to obtain major conference events and speech events, and analyze the emotional polarity of the speech. At the same time, we build an important holiday database based on domain expert knowledge and obtain current important holiday information in real time.

[0016] Then, using computers to mine behavioral signs, semantic calculations are performed on text and behavioral indicators based on the Bge pre-trained language model to obtain corresponding action events and deployment events;

[0017] Finally, using computers to construct a keyword library for status and plan based on domain expert knowledge, the current status and future plans of the target are detected to obtain signs of crisis events.

[0018] The spatiotemporal data analysis specifically includes the following sub-steps:

[0019] Guided by the crisis event indicators of the crisis event model and based on the characteristics of crisis events, the system uses computers to perform distance calculation, area calculation, time calculation, and combined analysis of the three. Distance calculation refers to the distance of the target from a specified area, area calculation refers to whether it is in a specified area, and time calculation refers to the target's activity events.

[0020] Single-target events are mainly calculated based on distance and region. If the target is located in a certain region, group-target events are mainly calculated based on a combination of distance, region, and time. If group targets meet, and the final calculation result meets the requirements of crisis event indicators, the corresponding crisis event indicators will be used as signs of a crisis event.

[0021] Furthermore, in step S2, tracking the development trend of the crisis event specifically includes the following sub-steps:

[0022] First, an expert knowledge base is built using computers, specifying rules for crisis occurrence, crisis escalation, crisis mitigation, and crisis resolution. Then, semantic matching calculations are performed between a pre-trained language model and text data according to the specified rules.

[0023] Then, the computer is used to determine whether a crisis event has occurred. If no crisis event has occurred, semantic matching is performed on the text data. If a crisis event has occurred, the computer is used to calculate whether the crisis event has intensified, weakened, or resolved. If the crisis event has resolved, the analysis results of occurrence, intensification, weakening, and resolution are saved. At the same time, guided by the rules in the expert knowledge base, the computer is used to calculate the development status of the crisis event through spatiotemporal data.

[0024] Finally, the crisis event stages obtained from text data matching and spatiotemporal data calculation are sorted and deduplicated according to chronological order to obtain the final description of the crisis event development stages: occurrence, aggravation, weakening, and resolution.

[0025] Furthermore, in step S3, the use of a computer to perform temporal, synergistic, and causal relationship analysis based on the basic crisis clues, important crisis clues, and development trends strengthens the connections between crisis events. The computer also uses crisis event classification to generate crisis event evolution trajectories based on different dimensions such as event type and target type. This specifically includes the following sub-steps:

[0026] First, computers are used to define the synergistic and causal relationships of crisis events within a domain based on expert knowledge and experience;

[0027] Then, based on the crisis event set formed by the description of crisis event signs and crisis event development stages, the crisis event set is sorted by time and the crisis event labels and relationships are classified based on the BERT pre-trained model, forming three types of crisis event labels: background, behavior, and state. The synergistic and causal relationships between crisis events are given. Finally, the crisis events are constructed according to six dimensions: background, behavior, state, goal, relationship, and signs, to obtain a complete and rich evolutionary path of crisis events.

[0028] Furthermore, the calculation of the development status of a crisis event using spatiotemporal data specifically includes the following sub-steps:

[0029] First, the computer determines the type of crisis event. If it is a single-target event, the location of the crisis event is further determined. Then, the crisis event stage calculation is performed to check whether the distance and area meet the rule requirements. If the calculation result does not meet the requirements for the occurrence of a crisis event, the calculation of the occurrence of a crisis event continues. If the calculation result meets the requirements for the occurrence of a crisis event, the calculation of whether the crisis event is aggravated, weakened, or resolved continues. If the crisis event is not resolved, the calculation of the crisis event continues. If the crisis event is resolved, the above analysis results of occurrence, aggravation, weakening, and resolution are saved, and the results are generated into a text description according to the format.

[0030] If it is a group target event, the location of the crisis event is further determined. Then, the spatiotemporal data of different targets are aligned at the time granularity. The spatiotemporal data of different targets are classified according to intervals. Then, the median value of each interval of different targets is used as the data value after data granularity alignment, so as to achieve consistency of the time scale of different targets. Then, the crisis event stage calculation is performed to calculate whether the distance, region and time between group targets meet the rule requirements. If the calculation result does not meet the requirements for the occurrence of a crisis event, the crisis event occurrence calculation continues. If the calculation result meets the requirements for the occurrence of a crisis event, the calculation continues to determine whether the crisis event is aggravated, weakened or resolved. If the crisis event is not resolved, the crisis event calculation continues. If the crisis event is resolved, the occurrence, aggravation, weakening and resolution analysis results are saved and the results are generated into a text description according to the format.

[0031] A domain expert knowledge-driven crisis event evolution context generation device includes a processor and a memory, wherein the memory stores a computer program that, when loaded by the processor, executes the method described in any of the preceding claims.

[0032] The beneficial effects of this invention include:

[0033] (1) Domain-specificity: The device and method of the present invention take into account the importance and indispensability of domain knowledge in the crisis event analysis process. Domain knowledge is both the foundation and the core. Compared with data-driven or data-based analysis methods with knowledge as a supplement, the device and method are more domain-specific, and the analysis results in the fields of military, politics, economy, diplomacy, security, science and technology will be more reliable.

[0034] (2) Completeness: The device and method of the present invention take the analysis of the background and current state of the crisis event as the center, that is, it considers the continuous evolution process of the current state, and also deeply explores the historical background of the crisis event, comprehensively analyzes the various logical relationships between events, and forms a complete and rich evolutionary context of the crisis event.

[0035] (3) Practicality: The device and method of the present invention are based on the actual needs of crisis analysis in different fields, take the knowledge of domain experts as the starting point, and set up an overall analysis approach in the program flow, which is more focused and practical than general methods. Attached Figure Description

[0036] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0037] Figure 1 This is a flowchart illustrating the overall steps of the method in an embodiment of the present invention;

[0038] Figure 2 Build a framework for crisis event models;

[0039] Figure 3 A schematic diagram illustrating the principles of crisis event analysis;

[0040] Figure 4 A flowchart for crisis event tracking;

[0041] Figure 5 A schematic diagram illustrating the principles of crisis event evolution analysis;

[0042] Figure 6 This is a diagram illustrating the evolution of a crisis event. Detailed Implementation

[0043] All features disclosed in all embodiments of this specification, or steps in all methods or processes implied in the disclosure, may be combined and / or extended or replaced in any way, except for mutually exclusive features and / or steps.

[0044] In view of the current situation, the specific implementation process of the present invention is as follows:

[0045] like Figure 1 As shown, according to the present invention, a computer device is provided and the running program performs the following steps: First, based on the knowledge and experience of domain experts, a basic data model containing the characteristics and indicators of crisis events within the domain is constructed to support the analysis of the entire event evolution; then, based on Internet data, event detection and analysis are performed to form event sentences, and basic crisis clues such as relevant major meetings, political statements, and holidays, as well as important crisis clues such as behavior and status, are mined; at the same time, the development trend of crisis events is continuously tracked, including the occurrence, aggravation, weakening, and resolution of crisis events; finally, based on the preceding basic crisis clues, important crisis clues, and development trends, the relationship analysis of time sequence, synergy, and causality is performed to strengthen the connection between crisis events, and through crisis event classification, the evolutionary context of crisis events is generated in different dimensions such as event type and target type.

[0046] like Figure 2As shown, a crisis event model is first constructed as the basic data model to support subsequent crisis event mining and evolution analysis. This invention constructs the data model from two dimensions: crisis event characteristics and crisis event indicators, establishing a unified crisis event model template. Then, crisis event models from different fields are built upon this template, driven by expert knowledge, to concretize and instantiate the crisis event models. Specifically, crisis event characteristics are divided into four categories: participating countries, participating targets, occurrence region, and crisis type. The crisis is categorized into participating countries into groups of countries and individual countries, with groups of countries constructed based on national interests. Participating targets are categorized into individuals, ships, and aircraft; the geographical areas are categorized into urban areas and maritime areas. Crisis types are categorized into group-target events and single-target events. Group-target events focus on individuals, ships, and aircraft, constrained by groups of countries. If the geographical area is urban, the focus is on analyzing the activities of individuals; if the geographical area is maritime, the focus is on analyzing the activities of ships and aircraft. Single-target events, while also focusing on individuals, ships, and aircraft, are constrained by a single country. Crisis event indicators are divided into three main categories: background indicators, behavioral indicators, and status indicators. The background indicators mainly include three aspects: important political figures' statements, major social activities, and important holidays. Important political figures' statements refer to major statements made by key political figures from different countries on potential crisis risks such as the environment, resources, economy, society, and territory. Different fields can build databases of important political figures and crisis risk points based on expert knowledge according to their needs and priorities, and maintain databases of political figures and crisis risk points. Major social activities include visits, meetings, and other activities. Important holidays include major traditional holidays in different countries, such as the Spring Festival and Thanksgiving. The behavioral indicators are divided into two categories: actions and deployments. Actions mainly refer to a specific activity, while deployments mainly refer to prior arrangements. Deployments can be a specific activity or a combination of multiple activities. By subdividing actions and deployments, a clearer evolutionary path can be obtained. Different fields can also be unified into behavioral indicators based on practical needs, without making specific distinctions between the two. The status indicators are divided into two categories: current status and future plans. Current status refers to the current activity of the target, reflecting a static characteristic, such as the target being in a certain region. Future plans refer to the future activity of the target, reflecting a dynamic characteristic, such as visiting a certain country in the next few days.

[0047] like Figure 3As shown, based on the aforementioned crisis event model, symptom analysis is performed on textual and spatiotemporal data obtained from the internet to uncover the preceding background of crisis events, i.e., crisis symptoms. Specifically, textual data analysis is guided by the crisis event indicators of the crisis event model. First, background-related symptoms are mined. Based on pre-trained language models such as BERT, event detection is performed on the textual data to obtain event sentences. Then, events are classified to obtain major meeting events and speech events, and the emotional polarity of the speech is analyzed, such as positive or negative. At the same time, an important holiday database is built based on expert knowledge to obtain current important holiday information in real time. Next, behavioral symptoms are mined. Based on pre-trained language models such as Bge, semantic calculations are performed on the text and behavioral indicators to obtain corresponding action events and deployment events. Finally, based on domain expert knowledge, a database of state and plan keywords is built to detect the target's current state and future plans, thereby obtaining crisis event symptoms. Spatiotemporal data analysis is guided by crisis event indicators from crisis event models and based on the characteristics of crisis events. It performs distance calculation, regional calculation, and time calculation, as well as combined analysis of the three. Distance calculation refers to the distance of the target from a specified area, regional calculation refers to whether it is in a specified area, and time calculation refers to the target's activity events. Single target events are mainly calculated by distance and regional calculation. If the target is in a certain area, group target events are mainly calculated by combining the three dimensions of distance, region, and time. If group targets meet, and the final calculation result meets the requirements of crisis event indicators, then the corresponding crisis event indicators are taken as signs of a crisis event.

[0048] like Figure 4As shown, based on the aforementioned crisis event model, textual and spatiotemporal data obtained from the internet are tracked and analyzed to determine the development status of crisis events. Specifically, an expert knowledge base is first constructed, specifying rules for crisis occurrence, crisis escalation, crisis mitigation, and crisis resolution. According to the specified rules, a pre-trained language model is used to perform semantic matching calculations with the text data to determine whether a crisis event has occurred. If the crisis event has not occurred, semantic matching continues from the text data. If the crisis event has occurred, the same calculations are performed to determine whether the crisis event has escalated, weakened, or resolved. If the crisis event has resolved, the aforementioned analysis results for occurrence, escalation, weakening, and resolution are saved. Simultaneously, guided by rules in the expert knowledge base, the development status of crisis events is calculated using spatiotemporal data. First, the type of crisis event is determined. If it's a single-target event, the location is further identified, followed by crisis event stage calculations. Distance and area are checked to see if they meet rule requirements. If the calculation results don't meet the requirements for crisis event occurrence, the crisis event occurrence calculation continues. If the results do meet the requirements, the crisis event is further calculated to determine whether it intensifies, weakens, or resolves. If the crisis event is not resolved, the calculation continues. If the crisis event resolves, the aforementioned analysis results (occurrence, intensification, weakening, resolution) are saved, and a text description is generated in the format "Time + Target + Crisis Event Name + Occurrence / Intensification / Weakening / Resolution" (e.g., "On a certain date, a certain target went on an outing"). Similarly, if it's a group-target event, the location is further determined, and then the spatiotemporal data of different targets are aligned with the time granularity, specifically using the time specified by expert knowledge and experience (e.g., 30...). Using minutes as intervals, a 24-hour day is divided into intervals. For example, 0:00 to 0:30 is the first interval, 0:31 to 1:00 is the second interval, and so on. The spatiotemporal data of different targets are categorized according to intervals. Then, the median value of each interval for different targets is used as the data value after data granularity alignment, so as to achieve consistency of time scale for different targets. Then, crisis event stage calculation is performed to calculate whether the distance, region, time, etc. between group targets meet the rule requirements. If the calculation result does not meet the requirements for the occurrence of a crisis event, the calculation of the occurrence of a crisis event continues. If the calculation result meets the requirements for the occurrence of a crisis event, the calculation of whether the crisis event is aggravated, weakened, or resolved continues. If the crisis event is not resolved, the calculation of the crisis event continues. If the crisis event is resolved, the above analysis results of occurrence, aggravation, weakening, and resolution are saved and the results are generated into a text description in the format of "time + target + crisis event name + occurrence / aggravation / weakening / resolution" (e.g., on a certain day of a certain month of a certain year, a certain target met with another target). Finally, the crisis event stages obtained by matching the above text data and calculating the spatiotemporal data are sorted and deduplicated according to chronological order to obtain the final description of the crisis event development stages, such as occurrence, aggravation, weakening, and resolution.

[0049] like Figure 5 As shown, firstly, based on expert knowledge and experience, the collaborative relationships (such as a coordinated relationship between two objectives) and causal relationships (such as the occurrence of type A event leading to the occurrence of type B event) of crisis events within the domain are defined. Then, based on the crisis event set formed by the above descriptions of crisis event symptoms and development stages, the time sequence of crisis events is performed, and crisis event label classification and crisis event relationship classification are performed based on the BERT pre-trained model, forming three categories of crisis event labels: background, behavior, and state. The collaborative and causal relationships between crisis events are given. Finally, the crisis events are constructed according to six dimensions: background, behavior, state, objective, relationship, and symptoms, resulting in a complete and rich evolutionary path of crisis events. For a detailed result illustration, please refer to the diagram. Figure 6 .

[0050] The units described in the embodiments of the present invention can be implemented in software or hardware, and the described units can also be located in a processor. The names of these units do not necessarily limit the specific unit itself.

[0051] According to one aspect of the present invention, a computer program product or computer program is provided, the computer program product or computer program including computer instructions stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium, and executes the computer instructions, causing the computer device to perform the methods provided in the various optional implementations described above.

[0052] In another aspect, embodiments of the present invention also provide a computer-readable medium, which may be included in the electronic device described in the above embodiments; or it may exist independently and not assembled into the electronic device. The computer-readable medium carries one or more programs, which, when executed by the electronic device, cause the electronic device to perform the methods described in the above embodiments.

Claims

1. A method for generating a crisis event evolution context driven by domain expert knowledge, characterized in that, The method comprises the following steps: S1, based on the experience of field experts, a basic data model containing the characteristics and indicators of crisis events in the field is constructed using a program, and the model is run based on a computer processor to support the entire event evolution context analysis; S2, using a computer to analyze events based on Internet data to form event sentences, and using a computer to run the model to mine basic crisis clues and important crisis clues; meanwhile, the development trend of the crisis event is tracked; S3, using a computer to analyze the time sequence, coordination and causality based on the basic crisis clues, important crisis clues and development trend, to strengthen the connection between crisis events, and to realize the generation of crisis event evolution context in different dimensions of event type and target type through crisis event classification using a computer; The basic crisis clues include relevant meetings, speeches and holidays; the important crisis clues include behaviors and states; and the development trend of the crisis event includes the occurrence, intensification, weakening and resolution state of the crisis event; In step S2, the use of a computer to analyze events based on Internet data to form event sentences, and the use of a computer to run the model to mine basic crisis clues and important crisis clues, specifically includes text data analysis and spatiotemporal data analysis; The text data analysis specifically includes the following sub-steps: First, using a computer to mine background signs based on the crisis event indicators of the crisis event model, detecting events based on the Bert pre-trained language model to obtain event sentences, and then classifying events to obtain important meeting events and speech events, and analyzing the sentiment polarity of speeches, while constructing an important holiday database based on field expert knowledge to obtain real-time important holiday information; Then, using a computer to mine behavior signs, and based on the Bge pre-trained language model, performing semantic calculation on text and behavior indicators to obtain corresponding action events and deployment events; Finally, using a computer to construct a state and plan keyword database based on field expert knowledge to detect the current state and future plan of the target to obtain crisis event signs; The spatiotemporal data analysis specifically includes the following sub-steps: Based on the crisis event indicators of the crisis event model and the characteristics of the crisis event, using a computer to perform distance calculation, region calculation, time calculation and combined analysis of the three, distance calculation refers to the distance of the target from a specified region, region calculation refers to whether the target is in a specified region, and time calculation refers to the activity time of the target; Single-target events mainly use distance calculation and region calculation, such as a target being in a certain region, and group-target events mainly use comprehensive calculation of distance, region and time, such as group targets meeting, and finally the calculation result meets the requirements of the crisis event indicators, and the corresponding crisis event indicators are used as crisis event signs; In step S2, the development trend of the crisis event is tracked, specifically including the following sub-steps: Firstly, an expert knowledge base is built by using a computer to specify crisis occurrence rules, crisis event intensification rules, crisis event weakening rules and crisis event resolution rules, and semantic matching calculation is performed on pre-trained language models and text data according to the specified rules; Then, it is judged by using a computer whether a crisis event occurs, if the crisis event does not occur, semantic matching is continuously performed on the text data, if the crisis event occurs, it is further calculated by using a computer whether the crisis event is intensified, weakened and resolved, if the crisis event is resolved, the occurrence, intensification, weakening and resolution analysis results are saved; meanwhile, the development state of the crisis event is calculated by using a computer through space-time data under the traction of the rules in the expert knowledge base; Finally, the crisis event stages obtained through text data matching and space-time data calculation are sorted and de-duplicated according to time sequence to obtain the final crisis event occurrence, intensification, weakening and resolution crisis event development stage description.

2. The field expert knowledge driven crisis event evolution thread generation method according to claim 1, characterized in that, In step S1, the basic data model containing crisis event characteristics and crisis event indicators in the field is built by using programs based on the experience of field experts, specifically including the following sub-steps: The data model is built from two dimensions of crisis event characteristics and crisis event indicators by using computer programming, and a unified crisis event model template is established; then, different field crisis event models are based on this, and the crisis event models are concretized and instantiated by using field expert knowledge.

3. The field expert knowledge driven crisis event evolution thread generation method according to claim 1, characterized in that, In step S3, the time sequence, synergy and causal relationship analysis are performed by using a computer based on the basic crisis clues, important crisis clues and development trend, the connection between crisis events is strengthened, and the crisis event evolution context is generated in different dimensions of event type and target type by classifying the crisis events through a computer, specifically including the following sub-steps: Firstly, the synergistic relationship and causal relationship existing in the field are defined by using a computer based on expert knowledge and experience; Then, the crisis event set formed based on the crisis event signs and crisis event development stage description is time-sequenced, and the crisis event label classification and crisis event relationship classification are performed based on the Bert pre-training model, forming three types of crisis event labels of background, behavior and state, giving the synergistic relationship and causal relationship between crisis events, and finally constructing the context in 6 dimensions of background, behavior, state, target, relationship and sign to obtain a complete and rich crisis event evolution context.

4. The field expert knowledge driven crisis event evolution thread generation method according to claim 1, characterized in that, The development state of the crisis event is calculated through space-time data, specifically including the following sub-steps: Firstly, the type of the crisis event is judged by using a computer, if it is a single-target event, the location of the crisis event is further determined, then the crisis event stage is calculated, whether the distance and area meet the rule requirements, if the calculation result does not meet the crisis event occurrence requirement, the crisis event occurrence calculation is continuously performed, if the calculation result meets the crisis event occurrence requirement, the crisis event whether is intensified, weakened and resolved is calculated, if the crisis event is not resolved, the crisis event calculation is continuously performed, if the crisis event is resolved, the occurrence, intensification, weakening and resolution analysis results are saved and the results are generated in a text description format; If it is a group target event, then the crisis event location is further determined, then the time granularity of different target space-time data is aligned, the space-time data of different targets is classified according to the interval, then the median value of each interval of different targets is taken as the data value after the data granularity alignment of each target, the time scale of different targets is unified, then the crisis event stage calculation is carried out, whether the distance, area and time between the group targets meet the rule requirements is calculated, if the calculation result does not meet the crisis event occurrence requirement, the crisis event occurrence calculation is continued, if the calculation result meets the crisis event occurrence requirement, whether the crisis event is intensified, weakened and eliminated is calculated, if the crisis event is not eliminated, the crisis event calculation is continued, if the crisis event is eliminated, the occurrence, intensification, weakening and elimination analysis result is saved and the result is generated in a text description according to the format.

5. A domain expert knowledge driven crisis event evolution thread generation apparatus, characterized in that, The system comprises a processor and a memory, wherein the memory stores a computer program which, when loaded by the processor, executes the method according to any one of claims 1-4.

Images