Emergency data analysis method based on digital plan

Abstract

The invention discloses an emergency data analysis method based on a digital plan, and relates to the technical field of digital emergency plans, and the method comprises the steps: employing a plurality of forms of emergency plan data entry modes, and achieving the standardization of original emergency plan data according to a fixed format; importance scoring is carried out on the emergency plan keywords based on semantic recognition, and an important reference is provided for subsequent emergency plan scoring; an emergency response module is adopted to be in butt joint with an emergency platform to collect emergency event information, and retrieval results are graded and sorted on the basis of the number proportion and importance scores of successfully matched keywords; and finally, inputting the emergency event information and the optimal emergency plan into a large language model, and outputting the emergency event emergency plan by the large language model. The implementation form of the emergency response plan generation module can be a large model arranged on a computer in a networked state. And finally, a reliable rapid and accurate matching solution of the digital plan under the emergency event is formed, and the response capability of the emergency event is improved.

Description

Technical Field

[0001] The invention relates to the technical field of digital emergency plans, and in particular to an emergency event data analysis method based on a digital plan. Background Art

[0002] Digital emergency plans play a vital role in handling emergency incidents. Through information technology, they digitize the contents of the plans, thereby improving the efficiency and accuracy of their execution.

[0003] At present, many regions and departments have established information platforms for emergency response, which can realize the rapid retrieval, call and update of plans.

[0004] Through mobile applications, on-site personnel can quickly access the system, execute plans, and report situations, thereby improving on-site response speed.

[0005] The defects are: the emergency plans provided are only a simple display of relevance, and for specific emergency events, it is impossible to search the emergency plan database to provide an emergency plan that is more suitable for the current emergency event, and it cannot meet the needs of digital emergency plans in emergency event handling. Summary of the invention

[0006] In order to solve the technical problem of digital management of emergency plans, the present invention provides an emergency event data analysis method based on digital plans. The following technical solutions are adopted:

[0007] A method for analyzing emergency event data based on a digital emergency plan comprises the following steps:

[0008] Step 1, input various emergency plan data to form an emergency plan original database;

[0009] Step 2, extracting multiple emergency plan keywords of all emergency plans in the original emergency plan database, scoring the importance of the emergency plan keywords based on semantic recognition, and classifying the multiple emergency plan keywords and the corresponding importance scores in the emergency plan data to form an emergency plan structured database;

[0010] Step 3: Use the emergency response module to connect to the emergency platform to collect emergency event information;

[0011] Step 4: The emergency response module extracts emergency event keywords from the emergency event information, and performs matching retrieval in the emergency plan structured database based on the emergency event keywords. The retrieval results are ranked based on the ratio of the number of successfully matched keywords and the importance score, and a recommended list of emergency response plans is output based on the ranking results.

[0012] Step 5: The staff selects an emergency plan from the emergency response plan recommendation list, and the selected emergency plan is the optimal emergency plan;

[0013] Step 6: The emergency response plan generation module inputs the emergency event information and the optimal emergency plan into the language model, and the language model outputs the emergency plan for the emergency event.

[0014] Optionally, in step 1, the method of entering emergency plan data includes electronic file entry and paper file entry. The paper file entry uses a high-resolution scanner to scan the paper plan into a digital image, and then applies OCR technology to convert the image into an editable text format. The original emergency plan data in the editable text format of the electronic file entry and the paper file entry are standardized and stored in the emergency plan original database.

[0015] By adopting the above technical solution, various forms of emergency plan data entry methods can be adopted. The emergency plan can be entered according to a template-based document, which can be in text form, picture form, etc. It can also be a paper emergency plan document. The paper plan is scanned into a digital image by a high-resolution scanner to achieve electronic entry. The editable text format can be in a fixed format to achieve standardization of the original emergency plan data. The standardization can be in various forms, and the purpose is to increase the success rate and accuracy of subsequent emergency plan keywords and subsequent keyword matching;

[0016] By scoring the importance of emergency plan keywords based on semantic recognition, it is possible to evaluate the importance of emergency event keywords and provide an important reference for subsequent emergency plan scoring. Semantic recognition can identify the meaning of keywords and score them according to their importance in the emergency plan. For example, the keyword "fire" is the most important keyword in the emergency plan for fire emergency response. In the same emergency plan, the keyword "rescue" is less important because "rescue" exists in many emergency events, while "fire" is likely to exist only in the emergency plan for fire emergency response. After categorizing multiple emergency plan keywords and corresponding importance scores in the emergency plan data, a structured database of emergency plans can be formed.

[0017] The emergency response module is used to connect to the emergency platform to collect emergency event information. The emergency response module can be in the form of a networked computer, which can connect to the relevant emergency event information of the emergency platform. The emergency event information generally includes a brief description of the emergency event. The emergency event keywords are extracted from the emergency event information, which can be obtained based on semantic recognition and other forms. Since it is an emergency event, the relevant description is generally brief and not necessarily very accurate. Therefore, matching retrieval is required in the matching process of the structured database of emergency plans. Generally, it is based on semantic similarity retrieval. The retrieval results are scored and sorted based on the proportion of the number of successfully matched keywords and the importance score. The set number of emergency plans with the highest matching degree can be quickly and accurately screened out. These emergency plans form a recommended list of emergency response plans. The recommended list of emergency response plans usually includes the name of the emergency plan, a brief description of the content, etc. At this point, the staff needs to select the emergency plan from the recommended list of emergency response plans after a brief analysis to avoid uncontrollable risks caused by relying entirely on artificial intelligence in the determination of emergency plans.

[0018] Finally, the emergency event information and the optimal emergency plan are input into the language big model, and the language big model outputs the emergency plan. The implementation form of the emergency response plan generation module can be a big model arranged on a networked computer.

[0019] Ultimately, a reliable digital emergency plan can be quickly and accurately matched to improve the response capabilities of emergency events.

[0020] Optionally, standardization of original emergency plan data includes standardization of plan naming, plan purpose and plan content. Standardization of plan naming includes plan type, plan version number and plan date; standardization of plan purpose includes the scope of applicable emergency events and an overview of emergency response goals; standardization of plan content includes emergency response procedures and supporting emergency disposal measures.

[0021] Optionally, in step 2, multiple emergency plan keywords of the emergency plan are extracted using TF-IDF and TextRank; and importance scoring of the emergency plan keywords based on semantic recognition includes the following sub-steps:

[0022] Step 21, extracting emergency plan keywords using TF-IDF or TextRank method;

[0023] Step 22, constructing a knowledge graph based on the content and associated knowledge of the emergency plan;

[0024] The entities and relations in the knowledge graph have been extracted, and the relation weight w(r) is determined according to the type and frequency of the relation. The formula is as follows:

[0025]

[0026] Where N is the number of all relations in the knowledge graph, n r is the number of occurrences of a particular relation r;

[0027] Step 23: Use the word embedding model to calculate the semantic similarity Sim(w 1 , w 2 ):

[0028]

[0029] where w 1 Represents the emergency plan keyword, w 2 is the keyword of the knowledge graph;

[0030] Step 24, combining keyword extraction, semantic similarity and the relationship in the knowledge graph, define the keyword importance scoring formula:

[0031] Imp(t,d)=α×TF﹣IDE(t,d)+β×STScore(t)+γ×KGScore(t);

[0032] Where Imp(t, d) is the keyword importance score of emergency plan keyword t in emergency plan d; α, β and γ are weight coefficients respectively; TF-IDE(t, d) is the TF-IDE score of emergency plan keyword t in the corresponding emergency plan d; STScore(t) is the score of emergency plan keyword t based on speech similarity; KGScore(t) is the score of emergency plan keyword t based on knowledge graph.

[0033] By adopting the above technical solutions, in order to combine semantic recognition and knowledge graph to score the importance of keywords in the emergency plan, the emergency plan text needs to be preprocessed first: including word segmentation, removal of stop words, part-of-speech tagging, etc. Keyword extraction: Use TF-IDF, TextRank and other methods to extract keywords.

[0034] Build a knowledge graph: Build a knowledge graph based on the content of the emergency plan and related domain knowledge. Build a knowledge graph: Use a graph database (such as Neo4j) and knowledge graph building tools (such as OWL, RDF, etc.). Word embedding model: Use pre-trained word embedding models such as Word2Vec, GloVe, or BERT. Use libraries such as scikit-learn to calculate TF-IDF.

[0035] The knowledge graph score can be calculated based on the following formula:

[0036] KGScore(t)=∑ (t，r，e)∈KG w(r)×sim(t,e);

[0037] Where (t, r, e) represents a triple in the knowledge graph, t is the keyword, r is the relationship, e is the entity, w(r) is the weight of the relationship, and sim(t, e) is the semantic similarity between keyword t and entity e.

[0038] For example, in the knowledge graph, there is a relationship "lead to" between "fire" and "evacuation", and there is a relationship "ensure" between "evacuation" and "safety".

[0039] The importance score for "Evacuate" is calculated as:

[0040] Imp(evacuation, d) = α × TF-IDE(evacuation, d) + β × STScore(evacuation) + γ × KGScore(evacuation);

[0041] We can combine semantic information and relationships in the knowledge graph to assign importance scores to keywords in the emergency plan. This score is closer to reality than the mechanically assigned score and can more accurately achieve subsequent emergency plan matching.

[0042] Optionally, in step 23, a semantic similarity threshold is set, Sim(w 1 , w 2 ) is greater than the set semantic similarity threshold, it is determined that the emergency plan keywords and the keywords in the knowledge graph are successfully matched, and STScore(t) and KGScore(t) are calculated in the keyword importance scoring formula.

[0043] Optionally, in step 4, the emergency response module uses the TF-IDF or TextRank method to extract emergency event keywords from the emergency event information, and performs matching retrieval in the emergency plan structured database based on a similarity matching algorithm. Results with a similarity greater than a set threshold are judged as successful keyword matching.

[0044] By adopting the above technical solution, the semantic similarity threshold is selected between 0.8-0.9, and the keyword match below the set semantic similarity threshold is considered unsuccessful, and the keyword match greater than the semantic similarity threshold is considered successful.

[0045] Optionally, in step 4, the search result score is calculated using the following formula:

[0046]

[0047] Where EeScore(d) is the search result score of emergency event d, a 1 , a 2 , …, a n They represent the importance scores of the n successfully matched keywords, and X is the total number of emergency plan keywords.

[0048] By adopting the above technical solution, the search result scoring needs to consider the proportion of successfully matched keywords to the emergency plan keywords, which can more accurately reflect the degree of correlation between emergency events and emergency plans.

[0049] Optionally, in step 4, for the current emergency event, the search result scores of multiple emergency plans in the emergency plan structured database are calculated and recorded as EeScore(d) 1 ,EeScore(d) 2 ,…,EeScore(d) y , where y is the total number of emergency plans.

[0050] Optionally, EeScore(d) 1 ,EeScore(d) 2 ,…,EeScore(d) y Sort by size, and select the top five emergency plans with the highest search result scores to form a recommended list of emergency response plans.

[0051] By adopting the above technical solution, the top five emergency plans with the highest search result scores are selected to form a recommended list of emergency response plans, which usually includes the five closest emergency plans to provide streamlined options for staff.

[0052] Optionally, the specific method of step 6 is: use OpenAI or BERT to generate emergency response plans.

[0053] By adopting the above technical solutions, in order to combine emergency event information with the optimal emergency plan and use the language model to output the corresponding emergency plan for emergency events, first collect and organize a large amount of emergency event information and the corresponding emergency plan text. Clean the data, ensure the quality of the text, and perform necessary text standardization. Fine-tune the pre-trained language model to learn how to generate the corresponding emergency plan based on the input emergency event information. Use the validation set to evaluate the performance of the model to ensure the accuracy and practicality of the generated emergency plan. Deploy the trained model to practical applications to generate emergency plans for emergency events.

[0054] You can choose from the following pre-trained language models: GPT-3 (OpenAI); BERT (Google);

[0055] It can efficiently and accurately correspond the current emergency events to the selected emergency plans, making it easier for staff to implement the emergency plans more efficiently.

[0056] In summary, the present invention includes at least one of the following beneficial technical effects:

[0057] The present invention can provide an emergency event data analysis method based on a digital plan, adopting various forms of emergency plan data entry methods to realize standardization of original emergency plan data according to a fixed format, so as to increase the success rate and accuracy of subsequent emergency plan keywords and subsequent keyword matching;

[0058] By scoring the importance of emergency plan keywords based on semantic recognition, it is possible to evaluate the importance of emergency event keywords and provide an important reference for subsequent emergency plan scoring. Semantic recognition can identify the meaning of keywords and score them according to their importance in the emergency plan. After multiple emergency plan keywords and corresponding importance scores are classified in the emergency plan data, a structured database of emergency plans can be formed.

[0059] The emergency response module is used to connect to the emergency platform to collect emergency event information. The search results are scored and sorted based on the proportion of the number of successfully matched keywords and the importance score. This can quickly and accurately screen out a set number of emergency plans with the highest matching degree. These emergency plans form a recommended list of emergency response plans. The recommended list of emergency response plans usually includes the name of the emergency plan, a brief description of the content, etc. The staff is required to select the emergency plan from the recommended list of emergency response plans after a brief analysis to avoid uncontrollable risks caused by relying entirely on artificial intelligence in the determination of emergency plans;

[0060] Finally, the emergency event information and the optimal emergency plan are input into the language big model, and the language big model outputs the emergency plan. The implementation form of the emergency response plan generation module can be a big model arranged on a networked computer.

[0061] Ultimately, a reliable solution for fast and accurate matching of digital emergency plans will be formed to improve the response capabilities of emergency events. BRIEF DESCRIPTION OF THE DRAWINGS

[0062] Figure 1 It is a flow chart of an emergency event data analysis method based on a digital plan of the present invention; DETAILED DESCRIPTION

[0063] The present invention is further described in detail below in conjunction with the accompanying drawings.

[0064] The embodiment of the present invention discloses an emergency event data analysis method based on a digital plan.

[0065] Reference Figure 1 Embodiment 1: A method for analyzing emergency event data based on a digital emergency plan, comprising the following steps:

[0066] Step 1, input various emergency plan data to form an emergency plan original database;

[0067] Step 2, extracting multiple emergency plan keywords of all emergency plans in the original emergency plan database, scoring the importance of the emergency plan keywords based on semantic recognition, and classifying the multiple emergency plan keywords and the corresponding importance scores in the emergency plan data to form an emergency plan structured database;

[0068] Step 3: Use the emergency response module to connect to the emergency platform to collect emergency event information;

[0069] Step 4: The emergency response module extracts emergency event keywords from the emergency event information, and performs matching retrieval in the emergency plan structured database based on the emergency event keywords. The retrieval results are ranked based on the ratio of the number of successfully matched keywords and the importance score, and a recommended list of emergency response plans is output based on the ranking results.

[0070] Step 5: The staff selects an emergency plan from the emergency response plan recommendation list, and the selected emergency plan is the optimal emergency plan;

[0071] Step 6: The emergency response plan generation module inputs the emergency event information and the optimal emergency plan into the language model, and the language model outputs the emergency plan for the emergency event.

[0072] In Example 2, in step 1, the method of entering emergency plan data includes electronic file entry and paper file entry. The paper file entry uses a high-resolution scanner to scan the paper plan into a digital image, and then applies OCR technology to convert the image into an editable text format. The original emergency plan data in the editable text format of the electronic file entry and the paper file entry are standardized and stored in the emergency plan original database.

[0073] Various forms of emergency plan data entry methods can be used. Emergency plans can be entered according to template documents, which can be in text form, picture form, etc., or paper emergency plan documents. Paper plans can be scanned into digital images by high-resolution scanners to achieve electronic entry. Editable text formats can be standardized according to fixed formats to achieve standardization of original emergency plan data. Standardization can be in various forms, with the purpose of increasing the success rate and accuracy of subsequent emergency plan keywords and subsequent keyword matching;

[0074] By scoring the importance of emergency plan keywords based on semantic recognition, it is possible to evaluate the importance of emergency event keywords and provide an important reference for subsequent emergency plan scoring. Semantic recognition can identify the meaning of keywords and score them according to their importance in the emergency plan. For example, the keyword "fire" is the most important keyword in the emergency plan for fire emergency response. In the same emergency plan, the keyword "rescue" is less important because "rescue" exists in many emergency events, while "fire" is likely to exist only in the emergency plan for fire emergency response. After categorizing multiple emergency plan keywords and corresponding importance scores in the emergency plan data, a structured database of emergency plans can be formed.

[0075] The emergency response module is used to connect to the emergency platform to collect emergency event information. The emergency response module can be in the form of a networked computer, which can connect to the relevant emergency event information of the emergency platform. The emergency event information generally includes a brief description of the emergency event. The emergency event keywords are extracted from the emergency event information, which can be obtained based on semantic recognition and other forms. Since it is an emergency event, the relevant description is generally brief and not necessarily very accurate. Therefore, matching retrieval is required in the matching process of the structured database of emergency plans. Generally, it is based on semantic similarity retrieval. The retrieval results are scored and sorted based on the proportion of the number of successfully matched keywords and the importance score. The set number of emergency plans with the highest matching degree can be quickly and accurately screened out. These emergency plans form a recommended list of emergency response plans. The recommended list of emergency response plans usually includes the name of the emergency plan, a brief description of the content, etc. At this point, the staff needs to select the emergency plan from the recommended list of emergency response plans after a brief analysis to avoid uncontrollable risks caused by relying entirely on artificial intelligence in the determination of emergency plans.

[0076] Finally, the emergency event information and the optimal emergency plan are input into the language big model, and the language big model outputs the emergency plan. The implementation form of the emergency response plan generation module can be a big model arranged on a networked computer.

[0077] Ultimately, a reliable digital emergency plan can be quickly and accurately matched to improve the response capabilities of emergency events.

[0078] Example 3, the standardization of original emergency plan data includes standardization of plan naming, standardization of plan purposes and standardization of plan contents. The standardization of plan naming includes plan type, plan version number and plan date; the standardization of plan purposes includes the scope of applicable emergency events and an overview of emergency response goals; the standardization of plan contents includes emergency response processes and supporting emergency disposal measures.

[0079] In Example 4, in step 2, TF-IDF and TextRank are used to extract multiple emergency plan keywords of the emergency plan; and importance scoring of the emergency plan keywords based on semantic recognition includes the following sub-steps:

[0080] Step 21, extracting emergency plan keywords using TF-IDF or TextRank method;

[0081] Step 22, constructing a knowledge graph based on the content and associated knowledge of the emergency plan;

[0082] The entities and relations in the knowledge graph have been extracted, and the relation weight w(r) is determined according to the type and frequency of the relation. The formula is as follows:

[0083]

[0084] Where N is the number of all relations in the knowledge graph, n r is the number of occurrences of a particular relation r;

[0085] Step 23: Use the word embedding model to calculate the semantic similarity Sim(w 1 , w 2 ):

[0086]

[0087] where w 1 Represents the emergency plan keyword, w 2 is the keyword of the knowledge graph;

[0088] Step 24, combining keyword extraction, semantic similarity and the relationship in the knowledge graph, define the keyword importance scoring formula:

[0089] Imp(t,d)=α×TF﹣IDE(t,d)+β×STScore(t)+γ×KGScore(t);

[0090] Where Imp(t, d) is the keyword importance score of emergency plan keyword t in emergency plan d; α, β and γ are weight coefficients respectively; TF-IDE(t, d) is the TF-IDE score of emergency plan keyword t in the corresponding emergency plan d; STScore(t) is the score of emergency plan keyword t based on speech similarity; KGScore(t) is the score of emergency plan keyword t based on knowledge graph.

[0091] To combine semantic recognition and knowledge graph to score the importance of keywords in the emergency plan, the emergency plan text needs to be preprocessed first: including word segmentation, removal of stop words, part-of-speech tagging, etc. Keyword extraction: Use TF-IDF, TextRank and other methods to extract keywords.

[0092] Build a knowledge graph: Build a knowledge graph based on the content of the emergency plan and related domain knowledge. Build a knowledge graph: Use a graph database (such as Neo4j) and knowledge graph building tools (such as OWL, RDF, etc.). Word embedding model: Use pre-trained word embedding models such as Word2Vec, GloVe, or BERT. Use libraries such as scikit-learn to calculate TF-IDF.

[0093] The knowledge graph score can be calculated based on the following formula:

[0094] KGScore(t)=∑ (t，r，e)∈KG w(r)×sim(t,e);

[0095] Where (t, r, e) represents a triple in the knowledge graph, t is the keyword, r is the relationship, e is the entity, w(r) is the weight of the relationship, and sim(t, e) is the semantic similarity between keyword t and entity e.

[0096] For example, in the knowledge graph, there is a relationship "lead to" between "fire" and "evacuation", and there is a relationship "ensure" between "evacuation" and "safety".

[0097] The importance score for "Evacuate" is calculated as:

[0098] Imp(evacuation, d) = α × TF-IDE(evacuation, d) + β × STScore(evacuation) + γ × KGScore(evacuation);

[0099] We can combine semantic information and relationships in the knowledge graph to assign importance scores to keywords in the emergency plan. This score is closer to reality than the mechanically assigned score and can more accurately achieve subsequent emergency plan matching.

[0100] In Example 5, a semantic similarity threshold is set in step 23. If Sim(w1, w2) is greater than the set semantic similarity threshold, it is determined that the emergency plan keywords and the keywords in the knowledge graph are successfully matched in similarity, and STScore(t) and KGScore(t) are calculated in the keyword importance scoring formula.

[0101] In step 4, the emergency response module uses the TF-IDF or TextRank method to extract emergency event keywords from the emergency event information, and performs matching retrieval in the emergency plan structured database based on the similarity matching algorithm. The results with similarity greater than the set threshold are judged as successful keyword matching.

[0102] The semantic similarity threshold is selected between 0.8 and 0.9. A keyword match below the set semantic similarity threshold is considered unsuccessful, and a keyword match greater than the semantic similarity threshold is considered successful.

[0103] In Example 6, in step 4, the search result score is calculated using the following formula:

[0104]

[0105] Where EeScore(d) is the search result score of emergency event d, a 1 , a 2 , …, a n They represent the importance scores of the n successfully matched keywords, and X is the total number of emergency plan keywords.

[0106] The search result scoring needs to consider the proportion of successfully matched keywords to the emergency plan keywords, which can more accurately reflect the degree of correlation between emergency events and emergency plans.

[0107] In Example 7, in step 4, for the current emergency event, the search result scores of multiple emergency plans in the emergency plan structured database are calculated and recorded as EeScore(d) respectively. 1 ,EeScore(d) 2 ,…,EeScore(d) y , where y is the total number of emergency plans.

[0108] Example 8, respectively EeScore (d) 1 ,EeScore(d) 2 ,…,EeScore(d) y Sort by size, and select the top five emergency plans with the highest search result scores to form a recommended list of emergency response plans.

[0109] The top five emergency plans with the lowest search result scores are selected to form a recommended list of emergency response plans, which usually includes the five closest emergency plans to provide streamlined options for staff.

[0110] In Example 9, the specific method of step 6 is: using OpenAI or BERT to generate an emergency response plan for emergencies.

[0111] To combine emergency event information with the optimal emergency plan and use a large language model to output the corresponding emergency plan, first collect and organize a large amount of emergency event information and the corresponding emergency plan text. Clean the data, ensure the text quality, and perform necessary text standardization. Fine-tune the pre-trained language model to learn how to generate the corresponding emergency plan based on the input emergency event information. Use the validation set to evaluate the performance of the model to ensure the accuracy and practicality of the generated emergency plan. Deploy the trained model to practical applications to generate emergency plans for emergencies.

[0112] You can choose from the following pre-trained language models: GPT-3 (OpenAI); BERT (Google); Specifically, the following steps are included

[0113] 1. Data preparation: collect emergency incident cases and corresponding emergency response plans. Label the data and match emergency incident information with emergency response plans.

[0114] 2. Data preprocessing: Use natural language processing tools to clean the text, such as removing useless characters, unifying terminology, etc. Use word segmentation tools to segment Chinese words.

[0115] 3. Model training, using the following formula as the training target:

[0116] Input:EmergencyEvnet(e)→Model→EmergencyPlan(p)

[0117] The emergency event information e is taken as input and the corresponding emergency plan p is taken as output.

[0118] Training is done using the cross entropy loss function:

[0119] L(θ)=-∑ (e，p)∈D logP(p|e;θ);

[0120] Where θ is the model parameter and D is the training dataset.

[0121] 4. Model evaluation, use cross-validation or leave-out method to evaluate model performance.

[0122] The evaluation metric can be BLEU, ROUGE, or a custom practicality score.

[0123] 5. Deploy the application and deploy the model to be called in actual emergency response.

[0124] Design a user interface that allows users to input emergency event information and receive the generated emergency response plan.

[0125] It can efficiently and accurately correspond the current emergency events to the selected emergency plans, making it easier for staff to implement the emergency plans more efficiently.

[0126] The above are all preferred embodiments of the present invention, and are not intended to limit the protection scope of the present invention. Therefore, any equivalent changes made based on the structure, shape, and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for analyzing emergency event data based on digital emergency plans, characterized in that: The following steps are involved: Step 1, input various emergency plan data to form an emergency plan original database; Step 2, extracting multiple emergency plan keywords of all emergency plans in the original emergency plan database, scoring the importance of the emergency plan keywords based on semantic recognition, and classifying the multiple emergency plan keywords and the corresponding importance scores in the emergency plan data to form an emergency plan structured database; Step 3: Use the emergency response module to connect to the emergency platform to collect emergency event information; Step 4: The emergency response module extracts emergency event keywords from the emergency event information, and performs matching retrieval in the emergency plan structured database based on the emergency event keywords. The retrieval results are ranked based on the ratio of the number of successfully matched keywords and the importance score, and a recommended list of emergency response plans is output based on the ranking results. Step 5: The staff selects an emergency plan from the emergency response plan recommendation list, and the selected emergency plan is the optimal emergency plan; Step 6: The emergency response plan generation module inputs the emergency event information and the optimal emergency plan into the language model, and the language model outputs the emergency plan for the emergency event.

2. The method for analyzing emergency event data based on a digital plan according to claim 1, characterized in that: In step 1, the methods of entering emergency plan data include electronic file entry and paper file entry. The paper file entry uses a high-resolution scanner to scan the paper plan into a digital image, and then uses OCR technology to convert the image into an editable text format. The original emergency plan data in the editable text format of the electronic file entry and the paper file entry are standardized and stored in the emergency plan original database.

3. The method for analyzing emergency event data based on a digital plan according to claim 2, characterized in that: The standardization of original emergency plan data includes the standardization of plan naming, plan purpose and plan content. The standardization of plan naming includes the plan type, plan version number and plan date; the standardization of plan purpose includes the scope of applicable emergency events and an overview of emergency response goals; the standardization of plan content includes the emergency response process and supporting emergency disposal measures.

4. The method for analyzing emergency event data based on a digital plan according to claim 3, characterized in that: In step 2, TF-IDF and TextRank are used to extract multiple emergency plan keywords. The importance scoring of emergency plan keywords based on semantic recognition includes the following sub-steps: Step 21, extracting emergency plan keywords using TF-IDF or TextRank method; Step 22, constructing a knowledge graph based on the content and associated knowledge of the emergency plan; The entities and relations in the knowledge graph have been extracted, and the relation weights w ( r ) It is determined based on the type and frequency of the relationship, and the formula is as follows: Where N is the number of all relations in the knowledge graph, n r is the number of occurrences of a particular relation r; Step 23: Use the word embedding model to calculate the semantic similarity Sim between the emergency plan keywords and the keywords in the knowledge graph ( w1,w 2) : Among them, w1 represents the emergency plan keyword, and w2 is the keyword of the knowledge graph; Step 24, combining keyword extraction, semantic similarity and the relationship in the knowledge graph, define the keyword importance scoring formula: Imp ( t,d ) =α×TF﹣IDE ( t,d ) +β×STScore ( t ) +γ×KGScore ( t ) ; Imp ( t,d ) is the importance score of the emergency plan keyword t in the emergency plan d; α, β and γ are weight coefficients respectively; TF﹣IDE ( t, d0 is the TF-IDE score of the emergency plan keyword t in the corresponding emergency plan d; STScore ( t0 is the score of the emergency plan keyword t based on speech similarity; KGScore ( t ) It is the score of the emergency plan keyword t based on the knowledge graph.

5. The method for analyzing emergency event data based on a digital plan according to claim 4, characterized in that: In step 23, a semantic similarity threshold is set. If Sim(w1, w2) is greater than the set semantic similarity threshold, it is determined that the emergency plan keywords and the keywords in the knowledge graph are successfully matched in similarity, and STScore(t) and KGScore(t) are calculated in the keyword importance scoring formula.

6. The method for analyzing emergency event data based on a digital plan according to claim 5, characterized in that: In step 4, the emergency response module uses the TF-IDF or TextRank method to extract emergency event keywords from the emergency event information, and performs matching retrieval in the emergency plan structured database based on the similarity matching algorithm. The results with similarity greater than the set threshold are judged as successful keyword matching.

7. The method for analyzing emergency event data based on a digital plan according to claim 6, characterized in that: In step 4, the search results are scored using the following formula: EeScore ( d ) is the search result score of emergency event d, a1, a2, …, a n They represent the importance scores of the n successfully matched keywords, and X is the total number of emergency plan keywords.

8. The method for analyzing emergency event data based on a digital plan according to claim 7, characterized in that: In step 4, for the current emergency event, the search result scores of multiple emergency plans in the emergency plan structured database are calculated and recorded as EeScore ( d )1 , EeScore ( d )2 ,…,EeScore ( d ) y , where y is the total number of emergency plans.

9. The method for analyzing emergency event data based on a digital plan according to claim 8, characterized in that: EeScore ( d ) 1. EeScore ( d ) 2, …, EeScore ( d ) y Sort by size, and select the top five emergency plans with the highest search result scores to form a recommended list of emergency response plans.

10. The method for analyzing emergency event data based on a digital plan according to claim 9, characterized in that: The specific method of step 6 is: use OpenAI or BERT to generate emergency response plans.

Images