An AI model-based analysis method, system, device, and storage medium

Abstract

The application relates to an AI model-based analysis method, system, device and storage medium, and belongs to the technical field of business decision making by using AI. The method comprises the following steps: acquiring an analysis instruction, inputting initial content contained in the analysis instruction into a pre-constructed analysis model; based on a pre-constructed decision tree, a decision node contained in the decision tree, and a plurality of preset basic questions corresponding to the decision node, analyzing and determining answer content associated with the basic questions by analyzing the initial content after disassembling the initial content by using the analysis model; integrating the answer content to generate an analysis result, and outputting the analysis result. The application has the effect of improving the analysis efficiency and accuracy of the AI model for the general and complex initial content.

Description

Technical Field

[0001] This application relates to the technical field of using AI for business decision-making, and in particular to an analysis method, system, device and storage medium based on an AI model. Background Technology

[0002] With the continuous development of deep learning technology, AI models are increasingly replacing manual analysis in addressing problems, thereby improving efficiency. For example, the problem can be text-based materials such as reports, which may include patient medical examination reports or construction project reports. By inputting a patient's medical examination report into an AI model, the model can extract and analyze specific types of data to determine the patient's condition. Similarly, by inputting a construction project report into an AI model, the model can analyze specific types of data to determine the construction quality.

[0003] However, when an AI model is needed to analyze a complex problem, such as when a report containing a variety of complex data from multiple sources is input into the AI ​​model all at once, the AI ​​model may be unable to extract the required data for analysis efficiently and accurately. Therefore, this needs to be improved. Summary of the Invention

[0004] To improve the efficiency and accuracy of AI models in analyzing general and complex content, this application provides an AI model-based analysis method, system, device, and storage medium.

[0005] Firstly, this application provides an analysis method based on an AI model, employing the following technical solution:

[0006] Obtain analysis instructions and input the initial content contained in the analysis instructions into the pre-built analysis model;

[0007] Based on the pre-constructed decision tree, the decision nodes contained in the decision tree, and the pre-set basic questions corresponding to the decision nodes, the initial content is decomposed and analyzed through the analysis model to determine the answer content associated with the basic questions.

[0008] The solutions are integrated to generate analysis results, which are then output.

[0009] By adopting the above technical solution, when analyzing complex and general initial content, this application proposes to input it into a pre-constructed analysis model, and then, based on predefined basic questions and a decision tree used to define the questioning order and questioning logic for the basic questions, specifically decompose the initial content into several detailed contents (where each detailed content is related to any basic question). Then, each of the decomposed detailed contents is analyzed separately. The analysis method is to extract the answer to the corresponding basic question (i.e., the answer content). Finally, the integrated result obtained by integrating all the answer contents is the analysis result of the initial content. Ultimately, this achieves the analysis and judgment of the aforementioned complex and general initial content, improving the accuracy, efficiency, and relevance of the analysis of complex and general real-time content.

[0010] Optionally, the step of decomposing and analyzing the initial content using the analysis model to determine the solution content associated with the basic question includes:

[0011] A preliminary set of questions is determined. Through analysis of the model, solutions related to the basic questions contained in the question set are extracted from the initial content. The question set includes several basic questions.

[0012] Whenever a solution related to the target fundamental question is determined, it is determined whether the first solution meets a preset stop analysis condition; wherein, the target fundamental question refers to any fundamental question in the question set, and the first solution refers to the solution related to the target fundamental question; the stop analysis condition includes at least: the decision node corresponding to the first solution does not contain a lower-level node in the decision tree;

[0013] If the first solution does not meet the preset stop analysis condition, then based on the first solution, the basic questions contained in the question set are adjusted so that the decision nodes corresponding to the basic questions contained in the adjusted question set are all decision nodes related to the first solution, until the stop analysis condition is met.

[0014] The process of integrating the solutions to generate analysis results includes:

[0015] When the first solution content meets the preset stop analysis condition, the solution content is integrated to generate an analysis result.

[0016] By adopting the above technical solution, the completion of the analysis operation is determined based on the answer content corresponding to each basic question. The basis for the determination is whether the answer content obtained from each analysis meets the stop analysis condition, that is, the decision node corresponding to the first answer content does not contain lower-level nodes in the decision tree. In other words, the first answer content is clear and unique, and there are no other questions that need to be asked later. In summary, the answer content enables dynamic adjustment of the basic questions to be asked, achieving targeted analysis and improving analysis efficiency.

[0017] Optionally, adjusting the basic questions included in the question set based on the content of the first solution includes:

[0018] Based on the first solution content, the decision nodes of all lower-level nodes in the decision tree that are the decision nodes corresponding to the first solution content are taken as target decision nodes, and the basic questions corresponding to the target decision nodes are added to the question set.

[0019] If there exists a first target decision node among the target decision nodes, and the first target decision node satisfies the following condition: the first target decision node corresponds to a sub-decision tree, then the guiding question corresponding to the sub-decision tree is added to the question set;

[0020] Whenever a second solution is determined that is associated with the target guidance question, it is determined whether the second solution meets the preset guidance conditions. If it does, the basic questions of all decision nodes contained in the sub-decision tree corresponding to the target guidance question are added to the question set.

[0021] By adopting the above technical solution, this application limits a single decision node to further include independent sub-decision trees. When querying the guiding question corresponding to the sub-decision tree and determining the second answer, if the second answer meets the preset guiding conditions, the basic questions of all decision nodes included in the sub-decision tree are added to the question set to achieve further detailed analysis of the initial content.

[0022] Optionally, the step of extracting the answer content related to the basic question pointed to by the preset query engine from the initial content through the analysis model further includes:

[0023] In the initial content, a label is added to each answer; wherein the label is used to characterize the relationship between the answer and the corresponding basic question.

[0024] By adopting the above technical solution, after extracting the answer content related to the basic question from the initial content, the position of the answer content corresponding to the basic question in the initial content is marked by adding tags (i.e., marking the source of the answer to the basic question), so that experts can know the source of the answer content when reviewing it.

[0025] Optionally, inputting the initial content contained in the analysis instruction into the pre-built analysis model includes:

[0026] Based on the report type of the initial content in the analysis instruction, a report template is matched. Based on the matched report template, the basic data in the initial content of the analysis instruction is classified and integrated, and the initial content is input into the pre-built classification model.

[0027] By adopting the above technical solution, before using the analysis model to analyze the initial content, a report template can be matched for the initial content, and the basic data in the initial content can be classified and integrated according to the preset layout of the report template. Through the pre-performed classification and integration operation, the subsequent analysis model can more efficiently realize the association between basic questions and answer content.

[0028] Optionally, the method further includes:

[0029] Receive manual editing instructions, add decision nodes to the decision tree based on the manual editing instructions, and set trigger conditions for the currently added decision nodes according to the node attributes contained in the manual editing instructions;

[0030] Whenever an analysis command is received, a preliminary analysis is performed on the initial content through the analysis model to determine all decision nodes that meet the trigger conditions, and a decision tree is generated using the decision nodes that meet the trigger conditions.

[0031] By adopting the above technical solution, experts can trigger editing instructions during review, such as adding decision nodes and defining the applicable scenarios (i.e., triggering conditions) of decision nodes. Whenever an analysis instruction is received, it can be determined whether all decision nodes meet the triggering conditions based on the specific situation of the analysis report in the analysis instruction, and only all decision nodes that meet the triggering conditions will be used to build a decision tree, thereby achieving adaptability and dynamic adjustment of the decision tree.

[0032] Optionally, the initial content may also include basic information about the object being analyzed and the report generation time;

[0033] The method further includes:

[0034] Upon receiving an analysis command, the basic attributes of the object being analyzed are determined based on the analysis results corresponding to the initial content. These basic attributes include at least time-transitional attributes.

[0035] For initial content with time-transitional basic attributes, the decision node corresponding to the end of the analysis operation of the initial content by the analysis model is taken as the end node. Based on the answer content corresponding to the end node, the change trend of basic data in the initial content corresponding to the analyzed object over time is predicted, and a prediction index is generated. The prediction index contains several basic questions, and the basic questions include at least the basic question corresponding to the end node.

[0036] The analysis results, analysis results and corresponding estimated indexes corresponding to the initial content are used to generate analysis records, and the analysis records are stored.

[0037] The step of extracting the answer content related to the basic question pointed to by the preset query engine from the initial content through the analysis model includes:

[0038] Determine whether the object being analyzed corresponding to the initial content has an analysis record in the historical period. If it does, retrieve the most recent analysis record before the current time, and based on the analysis record, use a preset query engine to point to the basic questions contained in the estimated index in the analysis record one by one.

[0039] The analysis model extracts answers from the initial content that are related to the basic questions pointed to by the preset query engine.

[0040] By adopting the above technical solution, if the object being analyzed corresponding to the initial content changes over time, its corresponding initial content (i.e., the basic data within the initial content) will change. In order to avoid repeating the basic questions that have been asked before, this application proposes to predict the changing trend of the initial content of the object being analyzed, and reorganize several basic questions to generate a prediction index based on the prediction results. The basic questions included in the prediction index can be understood as the basic questions that need to be asked when analyzing the initial content generated next time for the object being analyzed. These basic questions can cover or partially cover the basic questions included in the current analysis process. That is, a set of basic questions is derived by combining the results of the current analysis with the prediction of future changing trends. By making full use of historical analysis results, unnecessary questions in the next analysis process can be reduced as much as possible, thereby improving analysis efficiency.

[0041] Secondly, this application provides an analysis system based on an AI model, including:

[0042] An analysis instruction acquisition module is used to acquire analysis instructions and input the initial content contained in the analysis instructions into a pre-built analysis model;

[0043] The initial content analysis module is used to analyze and determine the answer content associated with the basic questions by decomposing the initial content based on the pre-built decision tree, the decision nodes contained in the decision tree, and the preset basic questions corresponding to the decision nodes through the analysis model.

[0044] The analysis result verification module is used to integrate the answer content to generate analysis results and output the analysis results.

[0045] Thirdly, this application provides an AI model-based analysis device, including a memory and a processor, wherein the memory stores a computer program that can be loaded by the processor and executed as any of the methods described in the first aspect.

[0046] Fourthly, this application provides a computer-readable storage medium, characterized in that it stores a computer program capable of being loaded by a processor and executing any of the methods described in the first aspect.

[0047] In summary, this application includes the following beneficial technical effects:

[0048] In this application, when analyzing complex and general initial content, this application proposes to decompose a general and complicated initial content into several basic and simple details. By analyzing each detail, the analysis and judgment of the aforementioned complex and general initial content can be finally realized, thereby improving the accuracy and efficiency of the analysis of complex and general initial content. Attached Figure Description

[0049] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying 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.

[0050] Figure 1 This is a flowchart illustrating an AI model-based analysis method disclosed in an embodiment of this application.

[0051] Figure 2 This is a schematic diagram illustrating a decision tree structure in a medical scenario, as exemplified in the embodiments of this application.

[0052] Figure 3 This is a structural block diagram of an AI model-based analysis system disclosed in an embodiment of this application.

[0053] Figure labeling: 201, Analysis instruction acquisition module; 202, Initial content analysis module; 203, Analysis result verification module. Detailed Implementation

[0054] The following is in conjunction with the appendix Figure 1-3 This application will be described in further detail.

[0055] This application discloses an analysis method based on an AI model. (Refer to...) Figure 1 The AI-based analysis method is used to analyze general and complex initial content. It breaks down the content, analyzes each sub-content, and then integrates the results to form the analysis outcome of the initial content. This achieves efficient and accurate analysis of initial content, especially general and complex content. The execution entity of this AI-based analysis method is an AI-based analysis system (hereinafter referred to as the analysis system). The following will describe this in conjunction with the attached... Figure 1-2 The process steps of the AI ​​model-based analysis method are explained in detail.

[0056] S101, Obtain the analysis instructions and input the initial content contained in the analysis instructions into the pre-built analysis model.

[0057] In implementation, users can access the analysis system via a webpage or app and trigger analysis commands. The initial content can be text content, such as an electronic data report, and must include at least the report generation time and basic information of the analyzed object. The analyzed object can be considered the descriptive subject of all basic data in the initial content. For example, in a medical scenario, the analyzed object can be a patient, and its basic information can be the patient's identity information. The corresponding initial content can be the patient's medical examination data report, which includes basic data such as the patient's various examination data and surgical details. In a construction engineering scenario, the analyzed object can be the building under construction, and the corresponding initial content can be a construction monitoring report, which includes basic data such as various measurement data during the construction process.

[0058] After receiving the initial content, the analysis system inputs the initial content into a pre-built analysis model, specifically an AI large model, to extract relevant features from the initial content for analysis. The specific analysis steps are as follows:

[0059] S102, based on the pre-built decision tree, the decision nodes contained in the decision tree, and the pre-set basic questions corresponding to the decision nodes, the initial content is decomposed and analyzed through the analysis model to determine the answer content related to the basic questions.

[0060] Specifically, S102, "By analyzing the model, breaking down the initial content and determining the solution content related to the basic question," includes the following steps:

[0061] S1021, Initially determine the problem set, and through analysis model, extract the solution content related to the basic questions contained in the problem set from the initial content; wherein, the problem set includes several basic questions;

[0062] S1022, whenever a solution related to the target basic question is determined, it is determined whether the first solution meets the preset stop analysis condition; wherein, the target basic question refers to any basic question in the question set, and the first solution refers to the solution related to the target basic question; the stop analysis condition includes at least: the decision node corresponding to the first solution does not contain a lower-level node in the decision tree;

[0063] S1023, if the content of the first solution does not meet the preset stop analysis condition, then based on the content of the first solution, adjust the basic questions contained in the question set so that the decision nodes corresponding to the basic questions contained in the adjusted question set are all decision nodes related to the content of the first solution, until the stop analysis condition is met.

[0064] S1023's "adjusting the basic questions included in the question set based on the content of the first solution" further includes:

[0065] Based on the content of the first solution, the decision nodes of all lower-level nodes in the decision tree that correspond to the decision node of the first solution are taken as the target decision nodes, and the basic questions corresponding to the target decision nodes are added to the question set.

[0066] If there exists a first target decision node among the target decision nodes, and the first target decision node satisfies the following condition: the first target decision node corresponds to a sub-decision tree, then add the guiding question corresponding to the sub-decision tree to the problem set;

[0067] Whenever a second solution is determined that is associated with the target guidance question, it is determined whether the second solution meets the preset guidance conditions. If it does, the basic questions of all decision nodes contained in the sub-decision tree corresponding to the target guidance question are added to the question set.

[0068] In implementation, the analysis system is pre-configured with a decision tree, which is a tree-structured classification model, referring to... Figure 2 The diagram illustrates a decision tree structure in a medical scenario. It includes several decision nodes connected by wires, forming hierarchical relationships. Each decision node's lowest-level node is connected to a stop node, which is a node without any directly connected lower-level nodes. Figure 2 The node indicated by the middle arrow F.

[0069] The analysis system pre-stores basic questions corresponding to each decision node, as well as baseline answers for each basic question. The basic questions and baseline answers can be predefined by humans. For example, the basic question corresponding to the decision node mentioned by arrow A can be "whether to cough" and its corresponding baseline answers can be a set, which can include "cough" and "no cough".

[0070] Furthermore, the analysis system also pre-establishes and stores the association between the baseline answer corresponding to each decision node and its lower-level nodes; for example, in the baseline answer corresponding to arrow A, the baseline answer with the content "cough" is associated with the nodes shown by arrows B and C, and the baseline answer with the content "no cough" is associated with the decision node shown by arrow D.

[0071] Accordingly, the analysis model is a large AI model, used to find the solutions to the basic questions from the initial content. This operation involves decomposing the initial content and then analyzing the decomposed content. The solutions can be considered as any set of content consistent with the benchmark answers corresponding to the basic questions. It should be noted that, by default, the basic questions and their benchmark answers for all decision nodes in the decision tree cover the initial content, meaning that the basic questions can be used to decompose the initial content. Furthermore, the benchmark answers to the basic questions mentioned in this application have unique solutions within the initial content, and all solutions are included within the initial content.

[0072] Regarding the selection of basic questions and the order of inquiry, this application proposes the following: First, a question set is constructed, and a preliminary selection of some basic questions is added to the question set. The selected basic questions can be considered as the basic questions corresponding to the topmost decision nodes in the decision tree (i.e., the root node of the decision tree, such as the decision node directly connected to the starting node). Figure 2 The decision node indicated by the middle arrow E.

[0073] Furthermore, the analysis system has a pre-set query engine. The query engine is used to point to the basic questions in the question set in order of priority of the upper level and the lower level according to the hierarchical relationship of the decision nodes corresponding to the basic content in the decision tree. The analysis model is used to analyze the basic questions pointed to by the query engine, and whenever the analysis result is obtained (that is, the answer content related to the corresponding basic question is obtained), the corresponding basic question is deleted from the question set.

[0074] Furthermore, whenever the analysis model arrives at a solution, the analysis system will determine whether to terminate the analysis operation based on the solution content. That is, whether the solution content meets the preset stop analysis condition. The stop analysis condition refers to the decision node in the decision tree that is the lower-level node of the decision node corresponding to the current solution content (hereinafter referred to as the first node). Here, the lower-level node refers to the decision node that is directly connected to the first node.

[0075] If the stop analysis condition is not met, the basic questions included in the question set will be adjusted based on the current solution content (i.e., the first solution content mentioned above). The specific adjustment rules are as follows:

[0076] Whenever the analysis model yields a solution, it identifies the set of baseline answers corresponding to the decision node, as well as the set of content consistent with that solution. Combining this with the relationships described earlier, it adds the basic question corresponding to the decision node associated with that set (hereinafter referred to as the second node) to the question set. Then, based on the added second node, it determines whether the second node corresponds to a sub-decision tree. If so, then this second node is the first target decision node mentioned above (e.g., ...). Figure 2 (The decision node indicated by the middle arrow A) At this point, the preset guiding question corresponding to the second node is added to the question set. The guiding question can be considered as the basis for determining whether further queries are needed on the basic questions in the sub-decision tree. If the analysis model determines that the second answer content associated with the guiding question meets the preset guiding conditions, the analysis system pre-stores the baseline answer for each guiding question. If the second answer content is consistent with the baseline answer, it is considered that the preset guiding conditions are met. If the guiding conditions are met, the basic questions corresponding to all decision nodes in the sub-decision tree corresponding to the second node are added to the question set. Finally, the basic content in the query set is sequentially pointed to according to the order of the query engine described above until the stop analysis condition is met.

[0077] In another embodiment, all decision nodes that are directly or indirectly connected to the first node can be used as the second node. The basic questions corresponding to the second node are added to the question set and cover all the basic questions in the original question set (in other embodiments, all the basic questions in the original question set can be retained).

[0078] S103: When the first solution meets the preset stop analysis condition, the solution is integrated to generate the analysis result and the analysis result is output.

[0079] In implementation, when the analysis model completes its analysis operation, the analysis system merges all the solutions obtained by the analysis model to generate the analysis result. In other embodiments, the analysis system also sends the analysis result to a review terminal, so that experts can review the analysis result through the review terminal and then send the analysis result to the review terminal for manual access (such as experts) to view the analysis result. During the viewing, the manual can modify the analysis result, such as triggering a modification command to the analysis system through the review terminal and editing the analysis result. The final modified analysis result is the analysis result, and the analysis system user output displays the final analysis result.

[0080] Optionally, after "extracting the answer content related to the basic question pointed to by the preset query engine from the initial content through the analysis model" in S1021, it also includes:

[0081] In the initial content, add a label to each answer; the label is used to represent the relationship between the answer and the corresponding basic question.

[0082] In practice, whenever the analysis model outputs a solution, the analysis system marks the solution in the initial content (i.e. adds a label), such as underlining the solution and adding a label in the form of a comment, and adding the corresponding basic question in the comment so that the source of the solution can be known during manual review.

[0083] Optionally, AI model-based analysis methods may also include the following steps:

[0084] The phrase "inputting the initial content contained in the analysis instructions into the pre-built analysis model" in S101 specifically includes the following:

[0085] Based on the report type of the initial content in the analysis instruction, a report template is matched. Based on the matched report template, the basic data in the initial content of the analysis instruction is classified and integrated, and the initial content is input into the pre-built classification model.

[0086] In implementation, the initial content can be pre-classified to form corresponding report types (such as patient visit reports) according to different application scenarios. In medical scenarios, by default, each basic data in the initial content corresponds to a data generation time. The order of these data generation times can be used to reflect the order of patient visits and treatment (such as the order of registration, examination, and prescription / surgery). When the analysis system receives the analysis command, it can adjust the distribution of the basic data in the initial content according to the order of their data generation times, ensuring that the basic data with earlier generation times are placed earlier in the initial content. This order of data generation is used as the basis for generating the report template.

[0087] In other embodiments, to facilitate the analysis model in quickly finding answers related to the basic questions from the initial content, a report template can be generated in advance. When an analysis command is received, the report template is sent to the user who triggered the analysis command. For example, the report template for the initial content can be in the form of a table, and input instructions can be displayed in the cells corresponding to rows and columns E at a specified location (e.g., E row E column) to prompt the user to enter the basic data in the corresponding cells. This limits the position of the basic data in the initial content, realizes the classification and integration of basic data, and improves the efficiency of associating the answer content.

[0088] Optionally, AI model-based analysis methods may also include the following steps:

[0089] S104, Receive manual editing instructions, add decision nodes in the decision tree based on the manual editing instructions, and set trigger conditions for the currently added decision nodes according to the node attributes contained in the manual editing instructions;

[0090] S105: Whenever an analysis instruction is received, a preliminary analysis is performed on the initial content through the analysis model to determine all decision nodes that meet the trigger conditions, and a decision tree is generated using the decision nodes that meet the trigger conditions.

[0091] In implementation, experts can trigger manual editing commands through the review terminal. The timing of these commands is not limited to the review period; experts can trigger them at any time to edit nodes in the decision tree. Editing can involve adding decision or answer nodes, or defining trigger conditions for any decision or answer node. Trigger conditions refer to the criteria used to determine whether a corresponding decision node should appear in the decision tree each time it is used by the preset query engine and analyzed by the analysis model. In other words, whenever the analysis system receives an analysis command and calls the decision tree, it needs to determine whether all decision nodes meet the trigger conditions and select only those nodes to form the decision tree. For example, in a medical scenario, a complete decision tree can include decision nodes corresponding to all medical subjects (such as internal medicine, surgery, otolaryngology, gynecology, etc.). The purpose of this solution is to screen the decision tree structure through the determination of trigger conditions. For instance, if the analyzed object or initial content only involves otolaryngology, then only decision nodes related to otolaryngology are used to form the decision tree, helping to reduce unnecessary decision nodes and improve query analysis efficiency. It should also be noted that the aforementioned "related to otolaryngology" refers to other non-otolaryngological medical subjects that are directly related to otolaryngology or that may cause complications leading to otolaryngology-related diseases.

[0092] The triggering conditions can be customized. For example, in a medical scenario, the triggering condition can be the applicable department of the node, such as otolaryngology. It can also be further refined to include thyroid diseases, etc. The aforementioned content can be used as the specific content of the triggering condition. In a medical scenario, the default initial content includes content related to the triggering condition. In other embodiments, whenever an analysis command is received, the analysis model can interact with the user who triggered the analysis command to ask questions. That is, AI is used to output a preset question to the user. The preset question corresponds to the triggering condition and is stored in the analysis system in advance. The analysis model receives the feedback content related to the preset question from the user and compares the feedback content with the triggering condition to determine whether the corresponding triggering condition is met. In other embodiments, when a user triggers the analysis model, the user can choose the analysis scope of the current initial content or an expert can help the user set the analysis scope in advance before review. The analysis scope here can be considered as the content corresponding to the triggering condition and used to determine whether the triggering condition is met. For example, in a medical scenario, the analysis scope can be the patient's historical disease type and / or the department visited (such as otolaryngology, internal medicine), and the triggering condition can be considered as a subset of the analysis scope.

[0093] Optionally, AI model-based analysis methods may also include the following steps:

[0094] Upon receiving an analysis command, the basic attributes of the object being analyzed are determined based on the analysis results corresponding to the initial content. The basic attributes include at least time-transitional attributes.

[0095] For initial content with time-transitional basic attributes, the decision node corresponding to the end of the analysis operation of the initial content by the analysis model is taken as the end node. Based on the answer content corresponding to the end node, the change trend of basic data in the initial content corresponding to the analyzed object over time is predicted, and a prediction index is generated. The prediction index contains several basic questions, and the basic questions include at least the basic questions corresponding to the end node.

[0096] The analysis results corresponding to the initial content, the analysis results, and the corresponding estimated index are used to generate analysis records, and the analysis records are stored.

[0097] The step S105, "Extracting the answer content related to the basic question pointed to by the preset query engine from the initial content through the analysis model," includes the following steps:

[0098] Determine if the object being analyzed corresponding to the initial content has an analysis record in the historical period. If it does, retrieve the most recent analysis record before the current time, and based on the analysis record, use the preset query engine to point to the basic questions contained in the estimated index in the analysis record one by one.

[0099] The analysis model extracts answers from the initial content that are related to the basic questions pointed to by the preset query engine.

[0100] In practice, time-transitional initial content refers to the fact that the specific content of the basic data within the initial content changes over time. That is, after a certain period of time, when the initial content of the object being analyzed is obtained again, the types and / or specific content of the basic data within the initial content have changed compared to the previous initial content.

[0101] For time-transitional initial content, after each analysis operation of the initial content is completed, the system predicts the trend of the initial content corresponding to the analyzed object over time based on the end node and its corresponding solution content, and generates a predictive index. For example, the analysis system can pre-store the predictive index for each decision node as an end node. The generation method of the predictive index can vary depending on the actual application scenario of the corresponding initial content. For instance, in a medical scenario, the basic data (such as vital signs data) in the patient's initial content may change over time, such as improvement or relapse. Therefore, the predictive index will be a set of basic questions generated under different changing conditions. This set can be determined by experts during review or automatically generated by the analysis system. For example, the analysis system can pre-store the predictive index for each decision node as an end node. All estimated indexes are retrieved. Then, when the initial content of the same object is received again, the analysis system can retrieve the analysis records stored for that object in historical periods. Based on the estimated indexes contained in the most recent analysis record, during analysis using the analysis model, the system prioritizes pointing to the basic questions contained in the estimated indexes using a preset query engine. After pointing to all basic questions in the estimated index in the order of their corresponding decision nodes in the decision tree (from top to bottom), if the corresponding answer does not meet the preset stopping condition, the system uses the basic question corresponding to the lower-level node (decision node) of the decision node corresponding to the last basic question pointed to in the estimated index as the basic question pointed to by the preset query engine, as described above, until the preset stopping condition is met, thus completing the analysis operation.

[0102] Optionally, AI model-based analysis methods may also include the following steps:

[0103] Whenever an analysis command is received, the basic question pointed to by the query engine is determined in real time, and a query index is generated when the analysis model completes the analysis operation. The query index is a collection of all basic questions that the query engine points to in chronological order.

[0104] The step S105, "Extracting the answer content related to the basic question pointed to by the preset query engine from the initial content through the analysis model," includes the following steps:

[0105] S1051, whenever the query engine updates the underlying question it points to, for the underlying data corresponding to the answer content related to the query content pointed to by the current query engine, analyze whether there is underlying data that is related to the underlying data and is not included in any answer content. The relationship includes at least temporal relationship and causal relationship.

[0106] S1052, if it exists, generate a logical index and match it with the query index stored in the historical period to determine the first query index containing the logical index. Select the target basic question from the basic questions contained in the first query index through the preset jump query engine and point to the target basic question so that the analysis model can extract the answer content related to the target basic question from the initial content. The target basic question is a basic question that has not been pointed to by the query engine before the current time. The pointing priority of the jump query engine is higher than the pointing priority of the query engine. The analysis model prioritizes extracting the answer content related to the basic question with higher pointing priority from the initial content.

[0107] The step S1052, "selecting the target basic question from the basic questions contained in the first query index and pointing to the target basic question," specifically includes the following steps:

[0108] When there is a merging node between the first query index and other query indices, or when the first query index contains a merging node and the underlying question corresponding to the merging node was not pointed to by the query engine before the current time, then any underlying question is selected as the target underlying question from the query content corresponding to all decision nodes in the first query index that are located at and after the merging node, and is pointed to the target underlying question; where, a merging node refers to a decision node that takes several upper-level decision nodes as input.

[0109] In implementation, the analysis system uses the basic question pointed to by the current preset query engine and its corresponding decision node (hereinafter referred to as the third node) to determine the decision node (hereinafter referred to as the fourth node, and by default the fourth node is a lower-level node of the third node) that is related to the third node. The third node, the fourth node and their connection are used as logical indexes. For example, if the basic question corresponding to the decision node pointed to by arrow J is the basic question pointed to by the current preset query engine, that is, the decision node pointed to by arrow J is the third node, and its corresponding lower-level decision node is two (that is, there are two fourth nodes), namely the decision node pointed to by arrow K. If the decision node pointed to by arrow L is a node, then the decision node pointed to by arrow J and the decision node pointed to by arrow K, along with their connecting lines, constitute a logical index (hereinafter referred to as logical index one). The decision node pointed to by arrow J and the decision node pointed to by arrow L, along with their connecting lines, constitute a logical index (hereinafter referred to as logical index two). The analysis system is used to determine the corresponding first query index for each logical index. The query index is composed of decision nodes and the connecting lines between decision nodes. The first query index is a query index that takes the decision node corresponding to the logical index as the top-level node and includes all decision nodes in the corresponding logical index.

[0110] After determining the query index, target decision nodes are selected from the decision nodes contained in the query index. The underlying question corresponding to the target decision node is the target underlying question. The selection method for target decision nodes is as follows: determine whether there is a merging node between the first query index and other first query indices, or whether there is a merging node among the decision nodes contained in the first query index. If so, the merging node is used as the target decision node. That is, the number of target decision nodes depends on the number of merging nodes, and the merging node satisfies the following conditions: 1. The merging node is a decision node; 2. The number of upper-level decision nodes directly connected to the merging node is at least 2.

[0111] After identifying the merging node, a decision node will be selected from the first query index to which the merging node belongs, and from the merging node and its lower-level nodes as the target decision node. The jump query engine will then be used to point to the target basic question corresponding to the target decision node. At this point, there will be a situation where the jump query engine and the preset query engine each point to a basic question. Based on the pointing priority, the analysis model will prioritize the basic question pointed to by the jump query engine and extract the corresponding answer content from the initial content.

[0112] Furthermore, for any target decision node corresponding to a target basic question, once the solution to the target basic question is found in the initial content through the analysis model, the first query index of the target decision node corresponding to the target basic question is used as the second query index, and the basic question pointed to by the preset query engine is changed to the basic question corresponding to the lower-level decision node connected to the target decision node.

[0113] This application also discloses an analysis system based on an AI model. (Refer to...) Figure 3 ,include:

[0114] The analysis instruction acquisition module 201 is used to acquire analysis instructions and input the initial content contained in the analysis instructions into the pre-built analysis model;

[0115] The initial content analysis module 202 is used to analyze and determine the answer content related to the basic questions by decomposing the initial content based on the pre-built decision tree, the decision nodes contained in the decision tree, and the pre-set basic questions corresponding to the decision nodes through the analysis model.

[0116] The analysis result verification module 203 is used to integrate the answer content to generate analysis results and output the analysis results.

[0117] Optionally, the initial content analysis module 202 is also used to initially determine the question set, and through the analysis model, extract the answer content related to the basic questions contained in the question set from the initial content; wherein, the question set includes several basic questions;

[0118] Whenever a solution related to the target fundamental question is determined, it is determined whether the first solution meets the preset stopping analysis condition. Here, the target fundamental question refers to any fundamental question in the question set, and the first solution refers to the solution related to the target fundamental question. The stopping analysis condition includes at least the following: the decision node corresponding to the first solution does not contain any lower-level nodes in the decision tree. If the first solution does not meet the preset stopping analysis condition, the fundamental questions included in the question set are adjusted based on the first solution so that the decision nodes corresponding to the fundamental questions included in the adjusted question set are all decision nodes related to the first solution, until the stopping analysis condition is met.

[0119] The analysis result verification module 203 is used to integrate the solution content and generate analysis results when the first solution content meets the preset stop analysis conditions.

[0120] Optionally, the analysis result verification module 203 is further configured to, based on the first solution content, take all the lower-level nodes of the decision nodes in the decision tree that are the decision nodes corresponding to the first solution content as target decision nodes, and add the basic questions corresponding to the target decision nodes to the question set; if there is a first target decision node among the target decision nodes, and the first target decision node satisfies that the first target decision node corresponds to a sub-decision tree, then add the guiding questions corresponding to the sub-decision tree to the question set; and is further configured to, whenever a second solution content associated with the target guiding question is determined, determine whether the second solution content satisfies the preset guiding conditions, and if so, add the basic questions of all decision nodes contained in the sub-decision tree corresponding to the target guiding question to the question set.

[0121] Optionally, the initial content analysis module 202 is also used to add tags to each answer in the initial content, so that humans can know the relationship between the answer and the basic question based on the tags.

[0122] Optionally, a report template creation module is also included, which is used to determine whether there is a relationship between the decision nodes corresponding to the decision tree. If there is, a report template is generated based on the relationship.

[0123] The analysis instruction acquisition module 201 is also used to classify and integrate the basic data in the initial content of the analysis instruction when a report template exists, and then input the initial content into the pre-built classification model.

[0124] Optionally, the decision tree adjustment module is used to receive manual editing instructions, add decision nodes to the decision tree based on the manual editing instructions, and set the trigger conditions for the currently added decision nodes according to the node attributes contained in the manual editing instructions; it is also used to update the decision tree with all decision nodes that meet the trigger conditions before extracting the answer content related to the basic question pointed to by the preset query engine from the initial content through the analysis model based on the pre-built decision tree and the basic question corresponding to each decision node in the decision tree, so that the updated decision tree only contains decision nodes that meet the corresponding trigger conditions.

[0125] Optionally, it also includes an analysis record storage module, used to determine the basic attributes of the analyzed object based on the analysis results corresponding to the initial content whenever an analysis instruction is received. The basic attributes include at least time-transitional attributes. It is also used to, for initial content whose basic attributes are time-transitional, take the decision node corresponding to the end of the analysis operation of the analysis model as the end node, predict the change trend of the initial content corresponding to the analyzed object over time based on the solution content corresponding to the end node, and generate a prediction index. The prediction index contains several basic questions, and the basic questions include at least the basic question corresponding to the end node. It is also used to generate analysis records from the analysis results corresponding to the initial content, the analysis results, and the corresponding prediction index, and store the analysis records.

[0126] The decision tree adjustment module is also used to determine whether the object being analyzed corresponding to the initial content has an analysis record in the historical period. If it does, it retrieves the most recent analysis record before the current time and, based on the analysis record, uses a preset query engine to successively point to the basic questions contained in the estimated index of the analysis record; and extracts the answer content related to the basic questions pointed to by the preset query engine from the initial content through the analysis model.

[0127] This application also discloses an AI model-based analysis device, which includes a memory and a processor. The memory stores a computer program that can be loaded by the processor and executed as described above for the AI ​​model-based analysis method.

[0128] This application also discloses a computer-readable storage medium that stores a computer program that can be loaded by a processor and executed as described above regarding the AI ​​model-based analysis method. The computer-readable storage medium includes, for example, various media capable of storing program code, such as a USB flash drive, a portable hard drive, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk.

[0129] It should be noted that in this paper, relational terms such as first and second are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations.

[0130] The above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit the scope of protection of the application. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on these embodiments, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

Claims

1. An analysis method based on an AI model, characterized in that, include: Obtain analysis instructions and input the initial content contained in the analysis instructions into the pre-built analysis model; Based on the pre-constructed decision tree, the decision nodes contained in the decision tree, and the pre-set basic questions corresponding to the decision nodes, the initial content is decomposed and analyzed through the analysis model to determine the answer content associated with the basic questions. The solutions are integrated to generate analysis results, which are then output. The step of using the analysis model to decompose and analyze the initial content to determine the answer content associated with the basic question includes: A preliminary question set is determined. Using a pre-set query engine, the decision nodes corresponding to the basic content in the question set are sequentially directed to the basic questions in the question set according to their hierarchical relationship in the decision tree. The basic questions directed by the query engine are analyzed by an analysis model to obtain the corresponding answer content. The question set includes several basic questions. Whenever an analysis command is received, the basic question pointed to by the query engine is determined in real time, and a query index is generated when the analysis model completes the analysis operation. The query index is a collection of all basic questions that the query engine points to in chronological order. Whenever the query engine updates the underlying question it points to, for the underlying data corresponding to the answer content related to the underlying question pointed to by the current query engine, analyze whether there is underlying data that is related to the underlying data and is not included in any answer content. The relationship includes at least temporal relationship and causal relationship. If it exists, a logical index is generated and matched with the query indexes stored in the historical period to determine the first query index containing the logical index. The preset jump query engine selects the target basic question from the basic questions contained in the first query index and points to the target basic question so that the analysis model can extract the answer content related to the target basic question from the initial content. The target basic question is a basic question that has not been pointed to by the query engine before the current time. The jump query engine has a higher pointing priority than the query engine, and the analysis model prioritizes extracting the answer content related to the basic question with higher pointing priority from the initial content. For any target decision node corresponding to a target basic question, once the solution to the target basic question is found in the initial content through the analysis model, the first query index of the target decision node corresponding to the target basic question is used as the second query index, and the basic question pointed to by the preset query engine is changed to the basic question corresponding to the lower-level decision node connected to the target decision node.

2. The analysis method based on an AI model according to claim 1, characterized in that, The step of using the analysis model to decompose and analyze the initial content to determine the answer content associated with the basic question also includes: Whenever a solution related to the target fundamental question is determined, it is determined whether the first solution meets a preset stop analysis condition; wherein, the target fundamental question refers to any fundamental question in the question set, and the first solution refers to the solution related to the target fundamental question; the stop analysis condition includes at least: the decision node corresponding to the first solution does not contain a lower-level node in the decision tree; If the first solution does not meet the preset stop analysis condition, then based on the first solution, the basic questions contained in the question set are adjusted so that the decision nodes corresponding to the basic questions contained in the adjusted question set are all decision nodes related to the first solution, until the stop analysis condition is met. The process of integrating the solutions to generate analysis results includes: When the first solution content meets the preset stop analysis condition, the first solution content is integrated to generate the analysis result.

3. The analysis method based on an AI model according to claim 2, characterized in that, The adjustment of the basic questions included in the question set based on the content of the first solution includes: Based on the first solution content, the decision nodes of all lower-level nodes in the decision tree that are the decision nodes corresponding to the first solution content are taken as target decision nodes, and the basic questions corresponding to the target decision nodes are added to the question set. If there exists a first target decision node among the target decision nodes, and the first target decision node satisfies the following condition: the first target decision node corresponds to a sub-decision tree, then the guiding question corresponding to the sub-decision tree is added to the question set; Whenever a second solution is determined that is associated with the target guidance question, it is determined whether the second solution meets the preset guidance conditions. If it does, the basic questions of all decision nodes contained in the sub-decision tree corresponding to the target guidance question are added to the question set.

4. The analysis method based on an AI model according to claim 2, characterized in that, The process of extracting answers related to the basic questions pointed to by the preset query engine from the initial content using the analysis model further includes: In the initial content, a label is added to each answer; wherein the label is used to characterize the relationship between the answer and the corresponding basic question.

5. The analysis method based on an AI model according to claim 4, characterized in that, The step of inputting the initial content contained in the analysis instructions into the pre-built analysis model includes: Based on the report type of the initial content in the analysis instruction, a report template is matched. Based on the matched report template, the basic data in the initial content of the analysis instruction is classified and integrated, and the initial content is input into the pre-built classification model.

6. The analysis method based on an AI model according to claim 2, characterized in that, The method further includes: Receive manual editing instructions, add decision nodes to the decision tree based on the manual editing instructions, and set trigger conditions for the currently added decision nodes according to the node attributes contained in the manual editing instructions; Whenever an analysis command is received, a preliminary analysis is performed on the initial content through the analysis model to determine all decision nodes that meet the trigger conditions, and a decision tree is generated using the decision nodes that meet the trigger conditions.

7. The analysis method based on an AI model according to claim 2, characterized in that, The initial content also includes basic information about the object being analyzed, as well as the report generation time; The method further includes: Upon receiving an analysis command, the basic attributes of the object being analyzed are determined based on the analysis results corresponding to the initial content. These basic attributes include at least time-transitional attributes. For initial content with time-transitional basic attributes, the decision node corresponding to the end of the analysis operation of the initial content by the analysis model is taken as the end node. Based on the answer content corresponding to the end node, the change trend of basic data in the initial content corresponding to the analyzed object over time is predicted, and a prediction index is generated. The prediction index contains several basic questions, and the basic questions include at least the basic question corresponding to the end node. The analysis results corresponding to the initial content and the corresponding estimated index are used to generate analysis records, and the analysis records are stored. The step of extracting the solution content associated with the basic questions contained in the question set from the initial content using the analysis model includes: Determine whether the object being analyzed corresponding to the initial content has an analysis record in the historical period. If it does, retrieve the most recent analysis record before the current time, and based on the analysis record, use a preset query engine to point to the basic questions contained in the estimated index in the analysis record one by one. The analysis model extracts answers from the initial content that are related to the basic questions pointed to by the preset query engine.

8. An AI model-based analysis system, applied to the AI ​​model-based analysis method as described in claim 1, characterized in that, include, The analysis instruction acquisition module (201) is used to acquire analysis instructions and input the initial content contained in the analysis instructions into the pre-built analysis model; The initial content analysis module (202) is used to analyze and determine the answer content associated with the basic questions by decomposing the initial content based on the pre-built decision tree, the decision nodes contained in the decision tree, and the preset basic questions corresponding to the decision nodes through the analysis model. The analysis result verification module (203) is used to integrate the answer content to generate analysis results and output the analysis results.

9. An analysis device based on an AI model, characterized in that, It includes a memory and a processor, wherein the memory stores a computer program that can be loaded by the processor and executed by any one of the methods described in claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer program is stored that can be loaded by a processor and executed by any one of the methods described in claims 1 to 7.

Images