Disease control emergency plan instruction disassembling method based on large model

Abstract

The technical scheme of the invention discloses a disease control emergency plan instruction disassembling method based on a large model. Compared with the existing technical scheme, the method disclosed by the invention has the following advantages: the method has semantic comprehension and generalization ability; the method has the capabilities of few sample learning and rapid adaptation; the method has the characteristics of non-formatted text input and structured output, key information can be extracted from a lengthy non-structured text, and a corresponding disposal method is generated. The method has knowledge integration and reasoning capabilities. Through medical knowledge in the large model, simple medical logical reasoning can be carried out, the method can be used for complementing missing parts and correcting obvious errors, meanwhile, contradictory conditions in various kinds of information can be found and corrected, and the result accuracy is improved.

Description

TECHNICAL FIELD

[0001] The application relates to an automatic method capable of disassembling emergency plans into specific execution actions, and belongs to the technical fields of artificial intelligence, natural language processing (NLP), large models and medical informatization. BACKGROUND

[0002] At present, artificial intelligence technology is being popularized in disease prevention and control systems, which can help disease control agencies quickly summarize historical plans, associate knowledge base documents, and provide assistance for disease control in handling emergencies. However, in real-world applications, large models in the disease control field still face problems such as insufficient data and low intelligence. Using a large model to disassemble emergency plans can directly convert emergency plans into a series of specific instructions for execution, improving work efficiency and providing technical support for infectious disease prevention and control. SUMMARY

[0003] The purpose of the present application is to automatically identify and structure key information related to disease control emergency plan disposal from a large amount of unstructured medical text (such as historical cases, rule documents, disposal guidelines, etc.) related to infectious diseases, and to learn from it in order to disassemble subsequent input emergency plans into specific execution instructions and improve the efficiency of handling emergencies.

[0004] In order to achieve the above purpose, the technical scheme of the present application discloses a disease control emergency plan instruction disassembly method based on a large model, characterized in that it comprises the following steps: Step 1, preliminary processing of the obtained data, converting documents of different formats into pure text; Step 2, classifying and processing the cleaned pure text to identify the document type; Step 3, creating different prompt words and processing rules for different document types and extraction targets, and constructing the corresponding dictionary; Step 4, large-scale input of data into the model for learning, and checking the results during the learning process to check whether the results meet the expectations; Step 5, testing according to the medical knowledge base and disposal guidelines to verify the rationality and usability of the disposal results given by the model; Step 6, deploying the verified model in a server, designing the format of the call and return to ensure that the formatted results output can be directly used by other systems.

[0005] Preferably, in step 1, the preliminary processing further comprises: removing information in the pure text that is irrelevant to model training; correcting errors and spelling mistakes in the pure text; For non-critical missing parts in the plain text, they are supplemented based on reasoning; for plain text with missing critical parts, they are directly discarded to ensure the accuracy of the text.

[0006] Preferably, in step 2, when classifying and processing, the focus is on the emergency response guidelines and response records.

[0007] Preferably, in step 3, a small number of emergency event samples are used for testing to ensure that the accuracy and precision of the results meet the requirements.

[0008] Preferably, in step 5, for systematic errors, the newly labeled cases are added to the training set, the model is retrained, and the data is processed.

[0009] Compared with existing technical solutions, the present invention has the following beneficial effects: 1) The method disclosed in this invention has semantic understanding and generalization capabilities. The large model can accurately capture complex expressions, synonyms, abbreviations and other information in disease control-related texts, such as "HFD" and "hand-foot-mouth disease", and "H1N1" refers to H1N1 influenza. It can also infer the severity of the outbreak from the description without relying on a large number of preset rules.

[0010] 2) The method disclosed in this invention has the ability to learn with few samples and adapt quickly. Traditional methods require the construction of a large number of rules or labeled data for each event type. However, the large model, through carefully designed prompts, can perform new information extraction tasks with very few or even no task-related training examples, greatly reducing the dependence on labeled data and reducing the amount of preliminary work. The extracted information can also be applied to model improvement and broaden the application scenarios.

[0011] 3) The method disclosed in this invention has the characteristics of unformatted text input and structured output, which can extract key information from lengthy unstructured text and generate corresponding processing methods.

[0012] 4) The method disclosed in this invention possesses knowledge integration and reasoning capabilities. Through the medical knowledge within the large model, it can perform simple medical logical reasoning, which can be used to fill in missing parts and correct obvious errors. At the same time, it can also discover and correct contradictory situations among various types of information, thereby improving the accuracy of the results. Attached Figure Description

[0013] Figure 1 This is a flowchart of the present invention. Detailed Implementation

[0014] The present invention will be further illustrated below with reference to specific embodiments. It should be understood that these embodiments are for illustrative purposes only and are not intended to limit the scope of the invention. Furthermore, it should be understood that after reading the teachings of this invention, those skilled in the art can make various alterations or modifications to the invention, and these equivalent forms also fall within the scope defined by the appended claims.

[0015] like Figure 1 As shown in the figure, an embodiment of the present invention discloses a method for generating disease control plans based on a large model, which specifically includes the following steps: Step 1: Perform preliminary processing on the acquired data. Convert documents of different formats, including Word and PDF, into plain text, remove information irrelevant to model training, and correct typos and spelling errors. For non-critical missing parts, supplement them based on inference. For missing parts with critical components, discard them directly to ensure the accuracy of the text.

[0016] Step 2: Classify the cleaned text and identify document types (such as emergency situations, handling guidelines, handling records, related literature, etc.) so that different extraction rules can be used for subsequent processing. This invention mainly focuses on the handling of disease control emergency plans, therefore, the handling guidelines and handling records of emergencies should be given special attention.

[0017] Step 3: Create different prompts and processing rules for different text types and extraction targets, build corresponding dictionaries, and test them using a small number of emergency event samples to ensure that the accuracy and refinement of the results meet the requirements.

[0018] Step 4: Input large-scale data for learning, and conduct spot checks during the learning process to see if the results meet expectations.

[0019] Step 5: Test the model against the medical knowledge base and treatment guidelines to verify the rationality and usability of the treatment results provided. For systematic errors, newly labeled cases can be added to the training set to retrain the model and process the data.

[0020] Step 6: Deploy the model on the server and design its call and return formats to ensure that the formatted output can be directly used by other systems.

Claims

1. A method for decomposing disease control emergency response plan instructions based on a large model, characterized in that, Includes the following steps: Step 1: Perform preliminary processing on the acquired data, converting documents of different formats into plain text. Step 2: Classify the cleaned plain text and identify document types. Step 3: Create different prompts and processing rules for different document types and extraction targets, and construct corresponding dictionaries. Step 4: The model learns from large-scale input data, and performs spot checks during the learning process to verify whether the results meet expectations. Step 5: Test the model based on medical knowledge bases and treatment guidelines to verify the rationality and usability of the treatment results provided by the model. Step 6: Deploy the validated model on a server, designing its call and return formats to ensure that the formatted output results can be directly used by other systems.

2. The method for decomposing disease control emergency response plan instructions based on a large model as described in claim 1, characterized in that, In step 1, the preliminary processing also includes: removing information from the plain text that is irrelevant to model training; correcting typos and spelling errors in the plain text; supplementing non-critical missing parts of the plain text based on reasoning; and directly discarding plain text with missing critical parts to ensure the accuracy of the text.

3. The method for decomposing disease control emergency response plan instructions based on a large model as described in claim 1, characterized in that, In step 2, when classifying and processing incidents, the focus is on the emergency response guidelines and records.

4. The method for decomposing disease control emergency response plan instructions based on a large model as described in claim 1, characterized in that, In step 3, a small number of emergency event samples are used for testing to ensure that the accuracy and precision of the results meet the requirements.

5. The method for decomposing disease control emergency response plan instructions based on a large model as described in claim 1, characterized in that, In step 5, for systematic errors, newly labeled cases are added to the training set, the model is retrained, and the data is processed.

Images