Adaptive complex scene information analysis method, electronic equipment, storage medium and program product

Abstract

The invention provides a self-adaptive complex scene information analysis method, electronic equipment, a storage medium and a program product. The method comprises the following steps: respectively acquiring scene data from a plurality of data sources through a data source module; converting the acquired scene data into a preset data format through a data conversion module, and storing the scene data into a data storage module; an analysis module responds to the received analysis request, a historical analysis request with the similarity with the analysis request larger than a first similarity threshold value is inquired from a historical storage module, and the analysis request is analyzed into a query task by referring to an analysis record of the inquired historical analysis request; calling a query tool in a query tool library through an adaptive query module according to the query task, and performing data query on a data storage module to obtain query result data; and analyzing and processing the query result data through an analysis module to obtain analysis result data.

Description

Technical Field

[0001] The present disclosure generally relates to the field of data analysis technology, and more particularly, to an adaptive complex scenario information analysis method, an electronic device, a storage medium, and a program product. Background Art

[0002] In complex scenarios such as emergency management, smart cities, and enterprise risk control, data analysis faces multiple challenges such as large data volume, high heterogeneity, strong real-time requirements, and complex event correlations. Traditional analysis methods often use fixed rules to implement data query and analysis, making it difficult to meet the requirements for efficient, intelligent, and adaptive analysis in these scenarios, which restricts the intelligent development of fields such as emergency management and smart cities.

[0003] Therefore, there is an urgent need to develop an intelligent analysis system that can integrate multi-source data, support real-time analysis, and have an adaptive learning ability. Summary of the Invention

[0004] The present disclosure provides an adaptive complex scenario information analysis method, an electronic device, a storage medium, and a program product for solving at least one of the above problems.

[0005] According to a first aspect of an embodiment of the present disclosure, there is provided an adaptive complex scenario information analysis method for an information analysis system including a data source module, a data conversion module, a data storage module, an analysis module, a query tool library, an adaptive query module, and a historical storage module for storing historical analysis records of analysis requests. The adaptive complex scenario information analysis method includes: obtaining scenario data from multiple data sources respectively through the data source module; converting the obtained scenario data to a preset data format through the data conversion module and storing it in the data storage module; in response to receiving an analysis request through the analysis module, querying the historical storage module for historical analysis requests with a similarity greater than a first similarity threshold to the analysis request, and parsing the analysis request into a query task with reference to the analysis records of the queried historical analysis requests; according to the query task, calling a query tool in the query tool library through the adaptive query module to perform a data query on the data storage module to obtain query result data; and performing analysis processing on the query result data through the analysis module to obtain analysis result data.

[0006] Optionally, converting the obtained scenario data to a preset data format by the data conversion module and storing it in the data storage module includes: for the scenario data of each object, generating an object identifier of the object as a key by the data conversion module, and using the scenario data of the object as a value, and storing it in the data storage module in the form of key-value pairs.

[0007] Optionally, in response to receiving an analysis request, the analysis module queries historical analysis requests in the historical storage module whose similarity to the analysis request is greater than a first similarity threshold, and refers to the analysis records of the queried historical analysis requests to parse the analysis request into a query task, including: in response to receiving the analysis request, the analysis module evaluates whether the analysis request is a complex request, where the complex request is an analysis request that needs to be parsed into at least N query tasks, and N is an integer greater than 1; in the case where the analysis request is not the complex request, parsing the analysis request into a query task; in the case where the analysis request is the complex request, querying historical analysis requests in the historical storage module whose similarity to the analysis request is greater than the first similarity threshold, and referring to the analysis records of the queried historical analysis requests to parse the analysis request into the query task.

[0008] Optionally, in the case where the analysis request is the complex request, querying historical analysis requests in the historical storage module whose similarity to the analysis request is greater than the first similarity threshold, and referring to the analysis records of the queried historical analysis requests to parse the analysis request into the query task includes: in the case where the analysis request is the complex request, determining whether there is a historical analysis request in the historical storage module that meets a preset condition, where the preset condition includes: the difference between the request time of the corresponding historical analysis request and the request time of the analysis request is less than a time difference threshold, and the similarity between the corresponding historical analysis request and the analysis request is greater than a second similarity threshold, where the second similarity threshold is greater than the first similarity threshold; if there is no historical analysis request that meets the preset condition, querying historical analysis requests in the historical storage module whose similarity to the analysis request is greater than the first similarity threshold, and referring to the analysis records of the queried historical analysis requests to parse the analysis request into the query task; if there is a historical analysis request that meets the preset condition, abandoning the parsing of the query task, and obtaining the analysis result data of the historical analysis request that meets the preset condition as the analysis result data of the analysis request.

[0009] Optionally, parsing the analysis request into a query task includes: using a large language model to parse the analysis request into the query task; and / or using a predefined template to parse the analysis request into the query task; and / or based on a knowledge graph, parsing the analysis request into the query task.

[0010] Optionally, the adaptive query module calling the query tools in the query tool library according to the query task includes: the adaptive query module determining query tool requirements according to the query task, where the query tool requirements represent the functions of the query tools to be called; retrieving query tools that meet the query tool requirements from the query tool library; and dynamically generating query tools that meet the query tool requirements in case of retrieval failure.

[0011] Optionally, the call information of the dynamically generated query tools is also recorded in the query tool library, and the call information includes the call times and call times. The adaptive complex scenario information analysis method further includes: for the dynamically generated query tools in the query tools, determining the call frequency of the corresponding query tools according to the call information; deleting the dynamically generated query tools whose call frequencies meet the preset deletion conditions, where the preset deletion conditions are used to represent low-frequency calls.

[0012] Optionally, the information analysis system further includes a domain knowledge base. Among them, the adaptive query module calling the query tools in the query tool library according to the query task to perform data query on the data storage module to obtain query result data includes: the adaptive query module calling the query tools in the query tool library according to the query task to perform data query on the data storage module as preliminary query data; analyzing whether the preliminary query data meets the analysis requirements of the analysis request; if it meets the analysis requirements of the analysis request, using the preliminary query data as the query result data; if it does not meet the analysis requirements of the analysis request, the analysis module combines the data in the data storage module and the domain knowledge base to analyze and reason about the preliminary query data to obtain an updated query task, and the adaptive query module re-calls the query tools in the query tool library according to the updated query task to perform data query on the data storage module to update the preliminary query data, and repeating the step of analyzing whether the preliminary query data meets the analysis requirements of the analysis request until the query result data is obtained or the maximum update times are reached.

[0013] According to a second aspect of the embodiments of the present disclosure, an adaptive complex scenario information analysis device is provided. The adaptive complex scenario information analysis device is used for an information analysis system, and the information analysis system includes a data source module, a data conversion module, a data storage module, an analysis module, a query tool library, an adaptive query module, and a historical storage module for storing analysis records of historical analysis requests. The adaptive complex scenario information analysis device includes: an acquisition unit configured to respectively acquire scenario data from multiple data sources through the data source module; a conversion unit configured to convert the acquired scenario data into a preset data format through the data conversion module and store it in the data storage module; an analysis unit configured to, in response to receiving an analysis request through the analysis module, query historical analysis requests with a similarity greater than a first similarity threshold from the historical storage module, and refer to the analysis records of the queried historical analysis requests to parse the analysis request into a query task; a query unit configured to, according to the query task, call a query tool in the query tool library through the adaptive query module to perform data query on the data storage module to obtain query result data; an analysis unit configured to perform analysis processing on the query result data through the analysis module to obtain analysis result data.

[0014] According to a third aspect of the embodiments of the present disclosure, an electronic device is provided, including: at least one processor; at least one memory storing computer-executable instructions, wherein when the computer-executable instructions are run by the at least one processor, the at least one processor is caused to execute the adaptive complex scenario information analysis method according to the exemplary embodiments of the present disclosure.

[0015] According to a fourth aspect of the embodiments of the present disclosure, a computer-readable storage medium is provided. When the instructions in the computer-readable storage medium are run by at least one processor, the at least one processor is caused to execute the adaptive complex scenario information analysis method according to the exemplary embodiments of the present disclosure.

[0016] According to a fifth aspect of the embodiments of the present disclosure, a computer program product is provided, including computer instructions, which when run by at least one processor, cause the at least one processor to execute the adaptive complex scenario information analysis method according to the exemplary embodiments of the present disclosure.

[0017] The technical solutions provided by the embodiments of the present disclosure at least bring the following beneficial effects: According to the adaptive complex scenario information analysis method, electronic device, storage medium, and program product of the present disclosure, by acquiring the scenario data of multiple data sources and uniformly performing format conversion and storage, multi-source data can be effectively integrated, which helps to improve the data query efficiency. By storing the analysis records of different analysis requests processed previously in the historical storage module, finding similar historical analysis requests when receiving an analysis request, and parsing the currently received analysis request with reference to the analysis records of the historical analysis requests, it can adaptively learn how to optimize task parsing, obtain a query task that better meets the requirements, helps to reduce the uncertainty in the reasoning process, and query the required scenario data to meet the real-time analysis requirements of complex scenarios.

[0018] It should be understood that the above general description and the following detailed description are only exemplary and explanatory, and cannot limit the present disclosure. Brief Description of the Drawings

[0019] The drawings herein are incorporated into the specification and constitute a part of this specification, showing embodiments consistent with the present disclosure, and are used together with the specification to explain the principles of the present disclosure, and do not constitute an improper limitation to the present disclosure.

[0020] Figure 1 is a flowchart of an adaptive complex scenario information analysis method according to an exemplary embodiment of the present disclosure.

[0021] Figure 2 is an overall architecture diagram of an information analysis system according to an exemplary embodiment of the present disclosure.

[0022] Figure 3 is a flowchart of adaptive query processing according to a specific embodiment of the present disclosure.

[0023] Figure 4 is a block diagram of an adaptive complex scenario information analysis device according to an exemplary embodiment of the present disclosure.

[0024] Figure 5 is a block diagram of an electronic device according to an exemplary embodiment of the present disclosure. Detailed Embodiments

[0025] In order to enable those of ordinary skill in the art to better understand the technical solutions of the present disclosure, the technical solutions in the embodiments of the present disclosure will be clearly and completely described below with reference to the drawings.

[0026] It should be noted that the terms "first", "second", etc. in the description, claims and the above-mentioned drawings of the present disclosure are used to distinguish similar objects and do not necessarily describe a specific order or sequence. It should be understood that the data used in this way can be interchanged under appropriate circumstances so that the embodiments of the present disclosure described herein can be implemented in an order other than those illustrated or described herein. The embodiments described in the following examples do not represent all embodiments consistent with the present disclosure. On the contrary, they are merely examples of devices and methods consistent with some aspects of the present disclosure as detailed in the appended claims.

[0027] It should be noted here that "at least one of several items" in the present disclosure all represents three parallel situations, including "any one of the several items", "a combination of any multiple of the several items", and "all of the several items". For example, "including at least one of A and B" includes the following three parallel situations: (1) including A; (2) including B; (3) including A and B. Another example is "performing at least one of step one and step two", which means the following three parallel situations: (1) performing step one; (2) performing step two; (3) performing step one and step two.

[0028] In the fields of emergency management, smart cities, and enterprise risk control, etc., data analysis systems need to process multi-source heterogeneous data to achieve real-time analysis and dynamic reasoning. However, there are still many deficiencies in the existing technologies in aspects such as data integration, real-time response, interactive query, and complex event correlation reasoning, making it difficult to meet the requirements of high-dynamic and high-uncertainty scenarios.

[0029] Next, various existing analysis techniques will be introduced one by one. It should be noted that the following is only an introduction to the existing technologies and does not imply that the analysis method of the present disclosure can solve all the problems mentioned next.

[0030] Traditional model-driven methods analyze data through predefined rules or mathematical models, such as optimizing evacuation routes based on traffic flow or predicting the spread of fire using a fire spread model. Such methods have clear logic and interpretable results, but their applicability is limited to specific fields and it is difficult to flexibly adapt to emergencies and complex environments. In addition, they have insufficient ability to integrate multi-modal data and are particularly difficult to process unstructured data, such as social media information or video surveillance data. Moreover, this method usually relies on manual parameter setting and it is difficult to automatically adjust the analysis strategy according to real-time data.

[0031] Traditional machine learning methods rely on historical data to train models for pattern recognition, classification, and prediction, such as traffic congestion prediction or optimization of emergency resource allocation. Although this method can discover potential patterns in data and improve prediction accuracy, it highly depends on high-quality datasets and is difficult to handle data-scarce scenarios or emergencies. At the same time, existing machine learning methods lack the ability to deeply model complex relationships between events and cannot adaptively adjust when facing new situations. In addition, traditional machine learning models mainly rely on batch data training and are difficult to meet the needs of online learning and real-time analysis.

[0032] Digital twin technology realizes real-time monitoring, analysis, and simulation by constructing digital models of cities or facilities. For example, a disaster monitoring system uses sensor data to construct disaster scenarios. However, the deployment cost of this method is high, relying on high-precision data collection and fixed visualization models, and it is difficult to quickly build corresponding models when dealing with emergencies. In addition, digital twin systems usually rely on fixed data patterns and cognitive models, are difficult to learn new environmental changes in real time, and have limited ability to predict the evolution trend of complex events.

[0033] With the development of large language models (LLMs), technology based on large model intelligent analysis has been widely used in complex data analysis, intelligent query, and situation awareness scenarios. This method can combine multi-modal data to achieve comprehensive perception and analysis of text, images, and sensor information. However, existing large model-based methods mainly rely on fixed prompt templates or specially trained domain models, lacking an adaptive fine-grained task decomposition and optimization mechanism, and are difficult to dynamically adjust analysis strategies according to different scenario requirements.

[0034] Inference large model technology, represented by OpenAI o1 and other inference large models, demonstrates strong capabilities in intelligent analysis, dynamic query, and decision-making inference, but there are still limitations. First, when dealing with complex task chains in long text inference, such models may experience loss of context information and a decline in inference consistency. Second, dynamic task decomposition still relies on static prompts and rule templates, lacking the ability to adaptively optimize different analysis scenarios. Third, the accuracy of multi-modal data fusion is limited, especially in cross-modal understanding of images, text, and sensor data, and additional data preprocessing or dedicated model support is still required. Fourth, the intelligence of tool invocation is insufficient, mainly relying on static matching and single-step invocation, and it is difficult to perform multi-round task optimization, resulting in limited usability and efficiency of complex queries. In high-dynamic and high-uncertainty application scenarios, stronger adaptive task planning, intelligent tool management, and associated inference optimization technologies still need to be combined to improve the practicality of the model and the accuracy of decision-making.

[0035] Using large language models to call tools based on large model tool call technology makes them more flexible in data query, analysis, and complex task execution. Existing methods mainly rely on natural language matching functions for tool calls, but semantic deviations may lead to incorrect calls and affect query accuracy. In addition, most methods still rely on static tool libraries or simple registration mechanisms, lacking dynamic optimization capabilities, making it difficult to effectively control the tool call frequency and affecting the utilization efficiency of computing resources. Even systems such as AutoGPT provide certain dynamic tool creation capabilities, but such methods often manage tools rather crudely, failing to effectively control tool call frequency, performance evaluation, and update strategies, resulting in redundant tools occupying resources while critical tools are not given priority, making it difficult to be truly practical. In addition, existing methods mainly trigger calls based on single queries, making it difficult to automatically disassemble complex tasks or optimize query order, leading to inefficient analysis processes and wasted computing resources. In multi-round interactive query scenarios, the lack of state retention capabilities results in poor coherence between consecutive analysis tasks, affecting the comprehensiveness and accuracy of decision-making.

[0036] In summary, there are still many deficiencies in the existing technologies for complex scenario analysis. Traditional model-driven methods have poor adaptability and are difficult to handle dynamic environments; machine learning methods rely on high-quality data and lack the ability of associative reasoning and adaptive adjustment; digital twin technology is limited in real-time updates during emergencies; although large models have improved the ability to integrate multi-modal data and perform intelligent analysis, existing methods rely on fixed prompts or static models and are difficult to perform fine-grained task disassembly and dynamic optimization. In addition, the current large model tool call technology still lacks precise tool matching, dynamic management, and multi-round interaction support, making it difficult to efficiently execute complex queries and associative reasoning. Overall, existing methods are difficult to achieve real-time, accurate, and adaptive analysis and decision support in highly heterogeneous and dynamically changing scenarios, restricting the intelligent development of fields such as emergency management and smart cities.

[0037] Next, an adaptive complex scenario information analysis method, an electronic device, a storage medium, and a program product according to an exemplary embodiment of the present disclosure will be described in detail with reference to the accompanying drawings.

[0038] Figure 1 is a flowchart of an adaptive complex scenario information analysis method according to an exemplary embodiment of the present disclosure. This method can be executed on an electronic device with sufficient computing power, and an information analysis system is configured on this electronic device. Figure 2 is an overall architecture diagram of an information analysis system according to an exemplary embodiment of the present disclosure.

[0039] Referring to Figure 2 , the information analysis system includes a data source module, a data conversion module, a data storage module (which can be in the form of a cache, corresponding to Figure 2data cache), analysis module, query tool library, adaptive query module, and historical storage module (which can be in the form of a cache, corresponding to Figure 2 historical cache in). The data source module is mainly used to obtain scenario data from multiple data sources respectively; the data conversion module is mainly used to convert the data format; the data storage module is mainly used to store the data after format conversion; the analysis module (which can also be called the analysis agent) is mainly used to obtain and parse the analysis request, and hand the parsed query task to the adaptive query module (which can also be called the adaptive query agent). The adaptive query module calls the query tools in the query tool library to query the required data from the scenario data stored in the data storage module, and then hands it to the analysis module for analysis and processing to obtain the analysis result data (which can be, for example, a query result or a summary of the query result, etc.); the historical storage module is mainly used to store the analysis records of historical analysis requests, which can provide reference for the task parsing of the analysis module.

[0040] Such as Figure 2 shown, the analysis module can also be connected to the user interaction module and the downstream business module to interact with the user and the downstream business module, so as to obtain the analysis request and deliver the analysis result data.

[0041] Specifically, the user can initiate an analysis request through the user interaction module to obtain an intelligent summary or query data. The user can also adjust the analysis direction of the analysis request based on the feedback given by the system. The interaction result between the user and the analysis module can be further used as the input of the analysis to optimize the behavior of the agent, so as to achieve the effect of the system's active learning and improvement, making it more in line with the actual business needs.

[0042] The downstream business module represents the scalability of the system, allowing the analysis result data to be used in other business systems (such as scheduling systems, decision support systems, etc.).

[0043] It should be understood that the user interaction module and the downstream business module can exist simultaneously or be used alternatively, and the present disclosure does not limit this. As an example, if the system is mainly used for expert-assisted decision-making, the external interface can be omitted and only the user visualization interaction function can be provided. This method can simplify the system architecture and reduce the complexity of external system integration. In a highly automated system (such as an intelligent scheduling system), the user often does not directly operate the information analysis system, but the downstream business module receives the analysis result data and automatically executes relevant operations, such as adjusting the deployment of rescue resources and optimizing the logistics route, etc. This method can improve the automation degree of the system, but reduces the flexibility of the user to adjust the analysis strategy.

[0044] As will be introduced later, the information analysis system can further include Figure 2 the domain knowledge base in to provide more information references, which will not be elaborated here for the time being.

[0045] Next, an adaptive complex scenario information analysis method according to an exemplary embodiment of the present disclosure will be introduced.

[0046] Refer to Figure 1 , in step S101, scenario data is obtained from multiple data sources through a data source module respectively.

[0047] In a complex information analysis scenario, data sources are often highly heterogeneous, including both structured data such as sensors and scheduling systems, unstructured data such as social media and news reports, and may also contain multimodal information such as images and videos. These data have different formats and inconsistent update frequencies. Traditional analysis methods are difficult to integrate efficiently, resulting in information fragmentation, which in turn affects the accuracy and timeliness of decision-making. The data source module of the present disclosure can continuously collect situation data from multiple data sources (such as including but not limited to sensors, monitoring systems, scheduling systems, social media, etc.), so as to provide support for decision-making.

[0048] In step S102, the obtained scenario data is converted to a preset data format through a data conversion module and stored in a data storage module.

[0049] By performing standardization processing (i.e., converting to a preset data format) on the obtained scenario data through the data conversion module, it can conform to a unified data format specification, effectively integrate multi-source data, and help improve data query efficiency. This process includes the parsing of structured data, the natural language processing (NLP) extraction of unstructured text, and the conversion of multimodal data (such as videos and images). For multimodal data, a multimodal large model (such as GPT4o, QwenVL, etc.) can be combined to generate corresponding natural language descriptions for subsequent query and analysis.

[0050] As an example, a multimodal large model can be used to process scenario data, including structured data, unstructured data, and visual data (images, videos), etc. In some scenarios, other smaller-scale models can also be used to replace the multimodal large model to reduce computational resource overhead. For example, if the input scenario data is mainly text (such as social media, reports, logs, etc.), the multimodal large model can be not used, but a pure large language model can be used to process all text data, which can reduce the multimodal training and inference costs, but will sacrifice the ability to understand visual data; for some stable visual scenarios (such as disaster detection, target recognition, etc.), a dedicated visual model can be used to replace the multimodal large model, thereby reducing computational complexity.

[0051] The scenario data after conversion is stored in the data storage module. The data storage module can support efficient indexing and retrieval mechanisms (such as high-speed storage based on Redis) to reduce direct queries to the original data source, improve query efficiency. In addition, through the TTL (Time-to-Live) mechanism of Redis, it supports automatic expiration cleaning of data, ensures the management efficiency of real-time data, and prevents redundant or outdated data from occupying storage resources. As an example, the data storage module specifically adopts a cache form, corresponding to Figure 2 the data cache in it, which helps to better meet the real-time requirements of information analysis. It can specifically include scenario data cache and scenario description cache. The former is used to store scenario data obtained from different data sources and serves as the main object of data query. The latter is used to save the key information and / or descriptive information extracted from scenario data at different times. These extracted information can be applied to scenarios such as trend analysis and event backtracking, enabling the system to not only handle the current situation but also use historical data for reasoning and prediction, so that the system can conduct deeper analysis in combination with historical data. The data storage module can also combine with a vector database (Facebook AI Similarity Search, abbreviated as FAISS) for data storage, or other solutions can be selected according to storage requirements and performance requirements. For example, if the analysis task does not require high-frequency queries but mainly relies on historical data, it can be combined with persistent storage (such as MySQL), use a database such as MySQL as the main storage, and combine Redis or other high-speed cache mechanisms as the hot data cache. This method is suitable for long-term data storage, reducing memory occupancy, but the query speed may be slower. Another example is that for applications relying on similarity search (such as situation analysis, case matching), only a vector database (such as Milvus) can be used without using Redis. This method is suitable for scenarios that require efficient retrieval of similar events but is not suitable for data storage with high-frequency updates.

[0052] In addition, in related technologies, the relevance of different data sources is weak, making it difficult to unify modeling and real-time analysis, reducing the utilization efficiency of data. In some embodiments of the present disclosure, optionally, step S102 includes: for the scenario data of each object, generate an object identifier of the object as the key through the data conversion module, and use the scenario data of the object as the value, and store it in the data storage module in the format of a key-value pair. By taking the object as the clue, generating a unique object identifier for each object as the key (Key), and using the scenario data of the same object as the value (Value), and storing it in the structured form of a key-value pair (KV pair), this storage method can adapt to heterogeneous data types, while maintaining good scalability, it also provides efficient query and aggregation capabilities, which helps to improve the utilization efficiency of data.

[0053] The following is an example of a key-value pair:

[0054] In this example, the key is the object identifier of the tent, rescue_tent:001, and the value is the scene data related to the tent, including name, type, location, inventory, and demand.

[0055] In step S103, in response to receiving an analysis request, the analysis module queries the historical storage module for historical analysis requests whose similarity to the analysis request is greater than the first similarity threshold, and refers to the analysis records of the queried historical analysis requests to parse the analysis request into a query task.

[0056] The user interaction module or downstream business modules can initiate analysis requests at any time, and these requests may involve requirements such as data query, trend prediction, event backtracking, risk assessment, etc. The system can parse the analysis requests, refer to similar historical analysis requests and their analysis records in the historical storage module, and combine the tool functions in the query tool library to determine the optimal query method and obtain query tasks. For complex requests involving multiple query targets or different data sources, they can be decomposed into multiple query tasks and executed sequentially. By storing the analysis records of different previously processed analysis requests in the historical storage module, finding similar historical analysis requests when receiving an analysis request, and referring to the analysis records of the historical analysis requests to parse the currently received analysis request, it can adaptively learn how to optimize task parsing, obtain query tasks that better meet the requirements, help reduce the uncertainty in the reasoning process, query the required scene data, and meet the real-time analysis requirements of complex scenarios.

[0057] As an example, when calculating the similarity between two analysis requests, the similarity between the embedding features of the corresponding request texts can be calculated, such as cosine similarity. In addition, the first similarity threshold can be reasonably set to ensure that historical analysis requests with a similarity greater than the first similarity threshold can be queried.

[0058] In step S104, the adaptive query module calls the query tools in the query tool library according to the query task to query the data storage module for query result data.

[0059] The adaptive query module is an important part of the system and the core of the dynamic query mechanism. It can, according to the query tasks parsed by the analysis module (when actually executed, the analysis module can send a query request to the adaptive query module), call appropriate query tools from the query tool library by registering / deregistering specific query modes (query tools include, for example, but are not limited to the form of query functions), and perform intelligent queries on the scenario data cached in the data storage module, so as to efficiently respond to the interaction requirements or analysis requirements of users or downstream business modules. The query task or query request can be in the form of natural language, for example. The adaptive query module supports the dynamic registration of query tools to ensure that the query logic can flexibly adapt to the constantly changing business requirements. As an example, the adaptive query module can adjust query parameters according to historical analysis records and current scenario data to improve the accuracy and timeliness of query results. In this way, it can maintain high efficiency in complex analysis tasks, while enhancing the decision support ability and making the query process more intelligent and adaptive.

[0060] The query tool library contains various query tools that can be called by the adaptive query module to support various query needs. Allowing the registration or deregistration of query tools on demand enables the system to automatically adjust its query capabilities as the analysis requirements change.

[0061] In step S105, the analysis module analyzes and processes the query result data to obtain the analysis result data.

[0062] The query result data is passed to the analysis module for the analysis module to perform summary extraction based on information such as user requirements, historical scenario descriptions, and domain knowledge, ensuring that the output information is focused, logically clear, and highly readable.

[0063] As an example, for different types of analysis requests, summaries with different granularities can be provided, including, for example, but not limited to, an overview summary and a detailed analysis. The former can extract key information, give the main conclusions and suggestions, and the latter can list relevant data, historical cases, reasoning processes, and attach decision-making reference suggestions.

[0064] As an example, a large language model can be used to generate an intelligent summary as the analysis result data, or an automatic summary based on keyword extraction can be used. For example, algorithms such as TF-IDF (Term Frequency-Inverse Document Frequency) and TextRank (a graph-based ranking model) can be used to extract key information without relying on a large language model, which can reduce the computational cost. A pre-trained summary model can also be used to generate a summary. For example, models such as T5 (Text-to-Text Transfer Transformer) and BERT (Bidirectional Encoder Representations from Transformers) can be used to generate a summary without relying on a general large language model. This method can improve the summary speed in a specific field. In actual execution, the generation method can be selected as needed, and the present disclosure does not limit this.

[0065] The final analysis summary and query result data will be returned to the submitter of the analysis request (including users and / or downstream business modules). The submitter can further adjust the query requirements based on the results. For example, the query conditions can be refined (such as focusing on the risk changes in a specific area) or the causal relationship of related events can be explored (such as analyzing the potential impact of a disaster on the supply chain).

[0066] If the system cannot fully meet the proposed analysis request, it can also provide the specific reasons for non-compliance and recommended solutions. For example, if the data is insufficient, the system can suggest adjusting the query scope or providing more relevant inputs; if the query tool is limited, the system can recommend using more advanced analysis methods or dynamically invoking new tools. In addition, this feedback information can also be synchronously transmitted to relevant users or downstream business modules to further optimize the business decision-making process.

[0067] At the same time, the system will record the analysis request, its corresponding analysis result data, and the feedback from the submitter of the analysis request through the historical storage module for optimizing the subsequent query process. This adaptive mechanism ensures that the system can continuously optimize the query logic during long-term operation, making it more in line with the actual business needs and improving the intelligent analysis ability and overall operation efficiency.

[0068] Next, step S103 will be further introduced, that is, how to parse the analysis request to obtain the query task.

[0069] In some embodiments, optionally, step S103 includes: in response to receiving an analysis request, the analysis module evaluates whether the analysis request is a complex request, where a complex request is an analysis request that needs to be resolved into at least N query tasks, and N is an integer greater than 1; in the case where the analysis request is not a complex request, the analysis request is resolved into query tasks; in the case where the analysis request is a complex request, a historical analysis request with a similarity greater than a first similarity threshold to the analysis request is queried from the historical storage module, and with reference to the analysis record of the queried historical analysis request, the analysis request is resolved into query tasks. By directly performing task resolution on simple requests without querying and using historical analysis requests, and querying and referring to historical analysis requests for complex requests, the computing resources consumed by querying historical analysis requests can be reduced. Of course, in other embodiments, historical analysis requests can also be queried and referred to for simple requests, and according to the actual situation of simple requests, historical analysis requests may not be queried and referred to when the query tasks can be directly resolved, and historical analysis requests may be queried and referred to when the query tasks cannot be directly resolved. This is also an implementation manner of the present disclosure.

[0070] Further optionally, the operation of, in the case where the analysis request is a complex request, querying from the historical storage module a historical analysis request with a similarity greater than a first similarity threshold to the analysis request and, with reference to the analysis record of the queried historical analysis request, resolving the analysis request into query tasks in step S103 includes: in the case where the analysis request is a complex request, determining whether there is a historical analysis request in the historical storage module that meets a preset condition, where the preset condition includes: the difference between the request time of the corresponding historical analysis request and the request time of the analysis request is less than a time difference threshold, and the similarity between the corresponding historical analysis request and the analysis request is greater than a second similarity threshold, where the second similarity threshold is greater than the first similarity threshold; if there is no historical analysis request that meets the preset condition, then a historical analysis request with a similarity greater than a first similarity threshold to the analysis request is queried from the historical storage module, and with reference to the analysis record of the queried historical analysis request, the analysis request is resolved into query tasks; if there is a historical analysis request that meets the preset condition, then the resolution of the query task is abandoned, and the analysis result data of the historical analysis request that meets the preset condition is obtained as the analysis result data of the analysis request. By further configuring a preset condition for complex requests to represent recently processed identical or highly similar historical analysis requests and preferentially searching for historical analysis requests that meet the preset condition, the analysis result data thereof can be preferentially reused, and there is no need to perform task resolution and subsequent data querying and analysis, thereby greatly reducing the consumption of computing resources and greatly improving the response rate of information analysis.

[0071] As an example, regarding the first condition in the preset conditions, that is, the time condition, for different types of analysis requests, different time difference thresholds can be set according to the sensitivity to time, so as to obtain more reference-worthy historical analysis requests and ensure the reliability of the reused analysis result data.

[0072] As an example, if a historical analysis request that meets the preset conditions with a similarity greater than the second similarity threshold but does not meet the time condition is queried, that is, an earlier identical or highly similar historical analysis request is queried, then on the one hand, this historical analysis request can be used as a reference for parsing the query task, and on the other hand, after obtaining the final analysis result data in the subsequent step S105, the content that is different from the analysis result data of this historical analysis request can be marked in the analysis result data to highlight the changes in the situation.

[0073] As an example, other conditions can also be added to the preset conditions as needed to meet more requirements and improve the flexibility of the solution. For example, for different types of analysis requests, specified background information can be configured, and the preset conditions can be made to include that the similarity between the specified background information of the corresponding historical analysis request and the specified background information of the analysis request is greater than the third similarity threshold, so as to improve the reliability of the reused analysis results.

[0074] Generally speaking, the historical storage module is the core module in the system dedicated to storing, reusing, and optimizing historical analysis requests. It can not only improve the query efficiency and reduce repeated calculations, but also provide the context information for multi-round conversations, enhancing the intelligence and coherence of the system in interactive analysis.

[0075] During the operation of the system, the historical storage module will store past analysis requests, analysis result data, and task decomposition records, and combine the parsed query tasks, query results, and data sources to support intelligent reuse. At the same time, the historical storage module will also record the query tools called and the user's feedback on the output analysis result data to optimize future task decomposition and query strategies.

[0076] As an example, the historical storage module specifically adopts a caching form, corresponding to Figure 2 the historical cache in it, which helps to better meet the real-time requirements of information analysis. To improve the storage and query efficiency, the historical storage module can be implemented, for example, in a way that combines a vector database (such as FAISS) with a memory queue. Among them, the vector database is responsible for storing the embedding vectors of analysis requests and supports efficient similarity queries, enabling the system to quickly retrieve similar historical analysis records and improving the accuracy of analysis strategies. The memory queue is mainly used to store the most recent analysis requests and results. For example, it can be a recent set period of time, or it can be a certain number of the most recent analysis requests, and at the same time provide the context information for multi-round interactions to support continuous conversations, which helps to ensure the consistency of continuous queries.

[0077] In addition to the combination of the vector database and the memory queue, the historical storage module can also be implemented in other ways as needed. For example, for systems with less historical analysis requests, analysis reuse based on log files can be adopted, directly recording request logs and parsing the logs through a rule engine (such as the ELK Stack) to replace the vector database storage. Another example is that since some large language models (such as GPT-4o, etc.) support long-term memory, the historical analysis requests can be directly stored based on the memory mechanism of the large language model through an API (Application Programming Interface), without using local caches. This method is suitable for SaaS (Software as a Service) deployment, but it depends on external large language model services and is difficult to flexibly customize the memory strategy.

[0078] When executing a new analysis request, the system will first query the historical storage module to determine whether the existing analysis results can be directly reused or incrementally updated based on historical data. For example, when querying the rescue situation in a certain area, if the system finds that information such as the current weather and infrastructure status has not changed significantly recently, it can reuse the historical description to optimize computing resources and improve the analysis speed.

[0079] In addition, the historical storage module also has the ability to detect changes. By comparing the current query data with the historical cache records, it can automatically identify key change points. For example, if the system detects a significant fluctuation in the rescue supply inventory or the addition of a new high-risk area, it will automatically mark these contents so that users can quickly focus on the latest developments when obtaining the analysis results. The similarity calculation of the vector database combined with the time window mechanism can effectively extract the most influential changes in the short term, ensuring the real-time and accuracy of the analysis.

[0080] To improve the storage management efficiency, the historical storage module will regularly clean up redundant or infrequently used data. For example, the TTL mechanism is used to control the data life cycle, and the LRU (Least Recently Used) algorithm is combined to dynamically optimize the records of historical analysis requests, ensuring the efficient use of storage space while ensuring the timeliness of the analysis result data of historical analysis requests.

[0081] When performing step S103 in combination with the historical storage module, when the system receives an analysis request submitted by a user or a downstream business module, it can first evaluate the complexity of the analysis request to determine the optimal processing method. If the request is relatively simple, such as "query the current location of rescue vehicles" or "obtain the current inventory of a certain warehouse", it can be directly parsed into a query task, then access the data storage module or call the query tool library for query without further decomposition, so as to improve the response speed and reduce the consumption of computing resources.

[0082] However, for more complex analysis requests such as "obtain the overall situation of the current rescue scenario", the system needs to perform deeper parsing. For this purpose, the historical storage module can be queried first to check whether there is a recent analysis record, so as to judge whether the existing analysis results can be reused. If it is found that the same or highly similar analysis request has been processed recently, the historical analysis results can be directly returned to avoid repeated calculations. If the recent analysis record cannot be directly reused, a similarity query can be further performed in the vector database to find historical analysis records close to the content of the current analysis request. Through cosine similarity calculation, the system can quickly screen out the most relevant analysis cases and optimize the current analysis strategy by combining these historical data. For example, if the current analysis request is "analyze the flood impact in a certain area", the system may find analysis records for adjacent areas or similar disasters in the past and use them as a reference to reduce the uncertainty in the reasoning process.

[0083] For complex scenarios, due to the huge volume of data, especially in the event of emergencies or urgent situations, relevant information accumulates rapidly. Traditional analysis tools are difficult to screen and process in a timely manner, forcing decision-makers to spend a great deal of time manually extracting key information. Due to the lack of effective automatic summarization and priority screening mechanisms, redundant information easily interferes with the analysis process, making it difficult to quickly obtain core data, thus affecting the efficiency of emergency response and the timeliness of decision-making. In terms of event correlation analysis, existing systems mainly rely on static rules or simple statistical associations, lacking the ability to deeply model and reason about the relationships between complex events. For example, traffic congestion in a certain area may affect the dispatching of rescue vehicles, and negative public opinion on social media may reflect the lag in rescue work. However, these potential correlation information is often difficult to be automatically discovered and utilized by traditional analysis tools, resulting in the formulation of emergency response strategies lacking a global perspective and potentially missing key influencing factors. In addition, online analysis has strong dynamics and interactivity. In the face of a dynamically changing environment, existing systems mostly rely on preset fixed analysis models and adopt fixed query patterns, usually lacking the ability of adaptive learning and the prediction ability of event evolution trends. They cannot automatically optimize the analysis process according to the changes in actual needs and are difficult to adapt to complex and changing analysis requirements, limiting their application in long-term emergency management and risk assessment. For example, commanders may need to query in real time information such as "the nearest available rescue vehicle to a certain rescue point", "the rescue personnel with the longest current working hours", or "the latest distress information and its key elements". Traditional methods usually require predefined query logic and are difficult to flexibly respond to the temporary needs of users. Another example is that as the event develops, the focus of attention may shift from the affected area to resource allocation and then to post-disaster recovery. Existing systems usually require manual adjustment of the analysis logic and cannot adaptively adjust the data processing strategy. This limitation makes traditional analysis tools difficult to adapt to rapidly changing business scenarios in practical applications, leading to a decline in analysis efficiency, difficulty in meeting the adaptability requirements for long-term operation, and reducing the system's adaptability and user experience.

[0084] When the present disclosure analyzes an analysis request by referring to the analysis record of a historical analysis request in a complex scenario, as an example, the task disassembling method of the queried historical analysis request can be directly adopted; or the content of the multiple query tasks obtained by disassembling can be further adjusted according to the changes in time, content, background information, etc. of the current analysis request relative to the queried historical analysis request (regarding time, for example, the focus of the query task can be adjusted according to the length of the interval time to estimate whether the event has developed to different stages; regarding content, for example, the relevant content in the query task can be adjusted according to the content difference; regarding background information, for example, for an analysis request for formulating a rescue plan, the focus of the query task can be adjusted by referring to the changes in background information such as the weather and infrastructure status on the same day, and the background information can be stored in the scenario description cache in the data storage module introduced above); the reasoning ability of the large language model can also be utilized, and prompt words can be used to enable the large language model to automatically disassemble tasks with the analysis record of the queried historical analysis request as a reference.

[0085] Regarding the change of the above background information, as an example, for the embodiment where the preset conditions described above include that the similarity of the specified background information is greater than the third similarity threshold, if a historical analysis request that does not meet this preset condition but meets the preset condition that the similarity is greater than the second similarity threshold is queried, then although the analysis result data of this historical analysis request cannot be directly reused, this historical analysis request can be directly used as a reference for parsing the query task (regardless of whether it meets the time condition in the preset condition), because this historical analysis request must meet the condition that the similarity is greater than the first similarity threshold, thereby reducing the computational resource consumption for querying historical analysis requests. At the same time, the parsed query task can be adjusted according to the change of the specified background information to improve the task parsing quality.

[0086] For task parsing, in some embodiments, optionally, the operation of parsing the analysis request into query tasks in step S103 includes: using a large language model to parse the analysis request into query tasks. When processing complex requests, the system can combine the reasoning ability of the large language model and automatically disassemble the analysis request into finer-grained query tasks through a predefined prompt word template. This task disassembling method can improve the accuracy of querying, and can utilize historical data and similar cases to optimize the analysis strategy, making the disassembled tasks more in line with business logic, which helps to improve the flexibility and reliability of task disassembling.

[0087] Taking "obtaining the overall situation of the current rescue scenario" as an example, the system can use prompt words similar to the following:

[0088] Based on these prompt words, the large language model can break down high-level analysis requests into multiple specific query tasks, and the analysis module then sends the decomposed query tasks to the adaptive query module. Still taking "obtain the overall situation of the current rescue scenario" as an example, it can be broken down into the following specific query tasks: 1. Obtain the total number of available rescue supplies: including the quantity statistics of food, water, medicine, tents, etc. 2. Obtain the total number of rescue personnel: including on-site rescue personnel, logistics support personnel, their locations, statuses, and distribution of professional skills. 3. Obtain weather and environmental conditions: including the current weather, future forecasts, and the likelihood of potential secondary disasters. 4. Obtain the scale of the affected population: including the number of affected people, missing people, casualties, etc. 5. Obtain key rescue targets: such as information on key facilities that need to be protected first, like hospitals, schools, community centers, etc.

[0089] It should be understood that the large language models used here include both pure large language models and multimodal large models. With the development of technology, the capabilities of multimodal large models and machine performance will gradually improve. In queries, the capabilities of multimodal large models can also be combined, and conditions or results in multiple modalities can be used as elements of the query. From the perspective of underlying implementation, the parameters of the query tool can also be represented by multimodal data.

[0090] For task parsing, in some other embodiments, optionally, the operation of parsing the analysis request into query tasks in step S103 includes: using a predefined template to parse the analysis request into query tasks. For fixed query requirements (such as inventory queries, personnel scheduling), by using predefined query templates, such as SQL queries or API endpoints, instead of using the large language model to dynamically parse queries, the query efficiency can be improved.

[0091] In still some other embodiments, optionally, the operation of parsing the analysis request into query tasks in step S103 includes: based on a knowledge graph, parsing the analysis request into query tasks. By adopting query optimization based on a knowledge graph, in scenarios of complex relational data (such as the disaster impact propagation path), "knowledge graph + SQL" queries can be used, or a knowledge graph that supports the large language model (such as GraphRAG) can be used to replace the free text parsing of the large language model, thereby enhancing the causal reasoning ability.

[0092] Generally speaking, step S103 can achieve the following aspects.

[0093] Cache content management: Store the embedding vectors of the analysis request based on a vector database, and combine an in-memory queue to store the most recent query tasks to improve query efficiency.

[0094] Intelligent reuse and incremental update: When parsing new tasks, first query the cache to determine whether existing results can be directly reused or incremental optimization can be performed based on existing data, thereby reducing the computational overhead.

[0095] Change detection: By combining the similarity calculation of the vector database and the time window mechanism, identify key change points, such as fluctuations in rescue resources and changes in high-risk areas, and mark key information to highlight the dynamic evolution trend.

[0096] Task decomposition optimization: Based on the historical task completion situation, automatically iteratively optimize the prompt words to improve the accuracy of task decomposition and the adaptability of query tasks.

[0097] Multi-round interaction support: When the user makes consecutive queries, store the dialogue context information to ensure the coherence of the queries, so that the user does not need to repeatedly describe the query background, improving the intelligent interaction experience.

[0098] The above mechanisms provide an adaptive scenario analysis method for the system.

[0099] Next, step S104 will be further introduced, that is, how to call the query tool through the adaptive query module for data query.

[0100] In some embodiments, optionally, the operation of calling the query tool in the query tool library by the adaptive query module according to the query task in step S104 includes: determining the query tool requirements by the adaptive query module according to the query task, where the query tool requirements represent the functions of the query tool to be called; retrieving the query tool that meets the query tool requirements from the query tool library; and dynamically generating the query tool that meets the query tool requirements in case of retrieval failure. By clarifying the query tool requirements, a basis can be provided for determining the query tool. By combining the query tool library, checking whether there is a suitable query tool that can directly execute these queries, and dynamically generating a new query tool when it is found that there is no registered matching query tool in the query tool library, real-time query and adaptive expansion functions in complex scenarios can be realized while making full use of the existing query tools. As an example, after dynamically generating a new query tool, its logical correctness and computational efficiency can be further tested to ensure its usability. As an example, the adaptive query module can receive natural language query requests, retrieve or dynamically generate query tools using large language models (such as GPT4o, Qwen2.5, etc.), and then call the appropriate tool to extract the required data from the real-time situation data cache and historical data storage in the data storage module.

[0101] It should be understood that newly dynamically generated query tools can also be saved in the query tool library to expand the query tool library. In the initial state, the query tool library can first store multiple pre-defined query tools that are commonly and easily used. This means that the query tool library includes two types of tools: pre-defined query tools and dynamically generated query tools. Since the dynamically generated query tools are generated for the query tasks of specific analysis requests, the possibilities of reusing them vary. For tools that are difficult to reuse, if they are stored in the query tool library for a long time, it will not only cause waste of storage space, but also increase the computational resource consumption and retrieval time for tool retrieval.

[0102] To solve this problem, optionally, the query tool library also records the call information of the dynamically generated query tools. The call information includes the call times and call time. The adaptive complex scenario information analysis method according to an exemplary embodiment of the present disclosure further includes: for the dynamically generated query tools in the query tools, determining the call frequency of the corresponding query tools according to the call information; deleting the dynamically generated query tools whose call frequency meets the preset deletion condition, where the preset deletion condition is used to represent low-frequency calls. By counting the call frequency of the dynamically generated query tools and deleting the low-frequency call tools, that is, clearing the infrequently used query tools, storage space can be freed up, thereby improving the usability and call efficiency of the tools in the library. It should be understood that this operation can be performed by the query tool library or by the adaptive query module. In actual implementation, as an example, a dynamic generation mark can be configured for the dynamically generated query tools, or the dynamically generated query tools and the pre-defined query tools can be stored separately for easy determination of the call frequency and tool deletion only for the dynamically generated query tools. Additionally, a default call frequency can be configured for the pre-defined query tools and the default value can be set to a sufficiently large value. The present disclosure does not limit this. Regarding the preset deletion condition, it can be set as needed. For example, it can include a call frequency lower than the frequency threshold, or it can include that the call frequency ranks among the last several positions among all the dynamically generated query tools (the specific number of positions can be determined according to the number of tools to be actually deleted). The present disclosure does not limit this either. Regarding the execution timing of this operation, it can also be set as needed, such as including but not limited to regular cleaning, cleaning when generating new tools. The present disclosure does not limit this either.

[0103] In addition to dynamically generating query tools using a large language model, in some other embodiments, if the query requirements are relatively fixed, a fixed tool set can be adopted, a group of query tools can be pre-defined and updated manually regularly, without the need to dynamically generate tools. This method is applicable to systems with high stability requirements.

[0104] In some other embodiments, plug-in tools can be used for management, that is, the tools can be deployed as independent plug-ins (such as Python API), supporting dynamic loading, rather than relying entirely on the large language model for generation. This approach improves flexibility but requires additional tool management mechanisms.

[0105] In addition, in some embodiments, optionally, the information analysis system further includes a domain knowledge base. Step S104 includes: calling the query tools in the query tool library according to the query task through the adaptive query module to perform data query on the data storage module as preliminary query data; analyzing whether the preliminary query data meets the analysis requirements of the analysis request; if it meets the analysis requirements of the analysis request, taking the preliminary query data as the query result data; if it does not meet the analysis requirements of the analysis request, analyzing and reasoning the preliminary query data by combining the data in the data storage module and the domain knowledge base through the analysis module to obtain an updated query task, and calling the query tools in the query tool library again according to the updated query task through the adaptive query module to perform data query on the data storage module to update the preliminary query data, and repeating the step of analyzing whether the preliminary query data meets the analysis requirements of the analysis request until the query result data is obtained or the maximum number of updates is reached. By not directly applying the queried data but first analyzing it to determine whether it can meet the analysis requirements, it is possible to pre-check the queried data in advance, and when it does not meet the requirements, combine multiple pieces of information to re-infer a new query task, and so on in a loop, which can effectively improve the quality of the analysis results.

[0106] The domain knowledge base is used to store industry-specific knowledge to help the analysis module enhance its reasoning ability. For example, standard procedures for rescue operations, equipment scheduling rules, public opinion analysis methods, etc. The domain knowledge base can usually be implemented based on RAG (Retrieval-Augmented Generation) or knowledge graph technology (such as GraphRAG for graph retrieval and augmentation).

[0107] After the preliminary query is completed, the system first checks whether the current query result data meets the analysis requirements of the analysis request. If the result already meets the analysis requirements, the system directly proceeds to the next step; otherwise, the system will decide whether further reasoning is needed to dig deeper information or optimize the analysis logic.

[0108] When further reasoning is required, the system can utilize the scenario data cache, scenario description cache in the data storage module, and the domain knowledge base to comprehensively analyze and deeply reason about the current preliminary query data to identify potential associated factors. For example, when analyzing whether traffic congestion in a certain area affects rescue operations, the system may call a large language model to perform multi-step reasoning, automatically generate new queries, or adjust the existing analysis logic to ensure that the reasoning process is more in line with the actual situation. During the reasoning process, the system can dynamically evaluate whether the current information is sufficient to support the final conclusion. If it is found that key data is insufficient, the system can automatically backtrack to the adaptive query process and trigger a new data query task to supplement the missing information. This backtracking mechanism ensures the continuity of queries and reasoning, while avoiding analysis biases caused by insufficient information. The system can also control the reasonable utilization of computing resources according to the preset maximum query / reasoning depth, on the premise of ensuring the integrity of the analysis, to avoid meaningless reasoning loops. This reflective reasoning mechanism not only improves the system's intelligent adaptability in complex analysis tasks, but also enhances the accuracy and interpretability of decisions.

[0109] In addition to using large language model reasoning for complex event analysis, other methods can also be adopted as needed. For example, decision support based on a rule engine. Specifically, for some domains (such as disaster warning, logistics scheduling, etc.), an expert rule engine (such as Drools) can be used to replace large language model reasoning, which can improve interpretability to a certain extent. Another example is that in event correlation analysis, a causal reasoning model (such as Bayesian network, Granger causality analysis, etc.) can be used to replace large language model reasoning to improve the transparency and interpretability of causal reasoning.

[0110] Figure 3 It is a flowchart of adaptive query processing according to a specific embodiment of the present disclosure.

[0111] In this specific embodiment, the query tool library stores all predefined and dynamically generated query tools, and records the description and invocation method of each tool to support the matching process. The main contents of the query tool library include the tool description (Description) and the tool invocation method (Invocation). The former briefly describes the function and applicable scenario of the tool, and the latter provides the specific tool invocation format. The query tool library manages query tools through metadata. The metadata includes, for example, the tool type (predefined or dynamically generated), invocation count, and last invocation time. The system can use the metadata to dynamically optimize the tool functions in the query tool library, such as clearing infrequently used tools (mainly dynamically generated query tools), thereby improving the usability of the tools and the efficiency of tool invocation.

[0112] After receiving a query task, the adaptive query module decides which query tool to use for the query. The system supports dynamic registration / deregistration of the query tool library, and can adjust the use of query tools according to task requirements, making the query method more flexible. For example, when receiving "query idle vehicles in disaster area A" or "view the recent inventory consumption trend", it will trigger the adaptive query processing flow as shown in Figure 3 This process first converts the natural language query request into a query tool call requirement, and retrieves available tools in the query tool library; if there is no matching tool, it will call the large language model to dynamically generate a new query tool, and test and verify its logical correctness, performance, etc. After completing tool matching or generation, the system will call the corresponding tool with necessary input parameters (such as "vehicle model = truck") to obtain the query result data from the data storage module in real time. At the same time, the system will update metadata such as the call count and timestamp of the tool, and use the LRU algorithm to clean up infrequently used tools to ensure the efficient operation of the tool library.

[0113] Next, the detailed content of the adaptive query processing flow of this specific embodiment will be introduced.

[0114] Refer to Figure 3 , in step S301, receive a natural language query request.

[0115] The user can initiate an analysis request by describing it in natural language. For simple requests, they can be directly used as natural language query requests. For complex requests, they need to be parsed into multiple query tasks through steps such as Figure 1 shown in step S103. These query tasks can also be query requests in natural language form. The natural language query request interface converts the request into a query tool call requirement. For example, for the query request "query idle vehicles in disaster area A", the function call generated by the large language model is as follows:

[0116] In step S302, construct a tool matching prompt word.

[0117] Extract the instructions and call methods of all tools from the query tool library, splice them into a Chinese prompt word and append the query request. For example, when the input is "query idle vehicles in disaster area A", the spliced prompt word may be as follows:

[0118] The large model selects the best query tool according to the spliced Chinese prompt word, or returns the "<:NO_MATCH>" mark.

[0119] In step S303, match the tool.

[0120] The large language model returns a matching result according to the prompt. If a tool is successfully matched, the tool is called and the process proceeds to step S307. If the match fails, the process proceeds to step S304 to start the dynamic tool generation process.

[0121] Query tools include predefined query tools ( ), and dynamically generated query tools ( ), in two types. Predefined query tools ( ) include commonly used query tool functions built into the system to handle common requirements and remain resident long-term; dynamically generated query tools ( ) include tool functions dynamically generated by the large language model according to specific query requirements and support an automatic cleaning mechanism (for details, see step S309).

[0122] In step S304, a tool is dynamically generated.

[0123] Call the large language model to dynamically generate a new query tool , and the tool logic incorporates the input parameters of the query request. Examples of the newly generated query tool functions are:

[0124] In step S305, the newly generated tool is tested and verified.

[0125] For the newly generated query tool, the large language model generates a verification program that covers various input scenarios and boundary conditions. For example:

[0126] The test is executed in a separate process, and mechanisms such as setting an execution timeout check are established to ensure the robustness of the system.

[0127] In step S306, failure feedback and reflection for improvement are performed.

[0128] When the newly generated tool fails the test, the system records the reason for failure and the test feedback information. The reasons for failure usually include syntax errors, logical errors, performance issues, and errors exceeding the range. Among them, a logical error means that the generated tool does not meet the query requirements, a syntax error means that there is a syntax error in the generated tool function and it cannot be compiled and run normally, a performance issue means that the tool execution time or resource consumption exceeds expectations, and an error issue means that the error between the tool output result and the expected result exceeds the allowable range , which can be expressed by the formula .

[0129] Based on the error feedback, the system modifies the tool to generate prompt words for improvement. For example, the system constructs new prompt words by combining the failure reasons and feedback information, and the following content can be supplemented in the prompt words:

[0130] Based on the improved prompt words, the failure reasons and test feedback can be passed to the large language model for reflection and improvement, starting the iterative process of "recreate -> test", that is, enabling the large language model to regenerate an improved version of the tool. If the newly generated tool still fails to pass the test program verification, this process is repeated until the newly generated tool passes the test or reaches the preset maximum number of retries.

[0131] In step S307, the tool function is executed.

[0132] This step calls the matching tool or the newly generated tool, passes in the required parameters, and extracts the results from the data storage module. This process can be expressed as the following formula:

[0133] In the above formula, represents the query result, represents the query tool logic function, represents the additional input parameters provided by the user, such as "vehicle model = truck", represents the situation data cache in the data storage module (such as Redis data).

[0134] During the execution process, the query tool combines and to extract the qualified results.

[0135] In step S308, the tool metadata is updated.

[0136] After the query tool finishes execution, the tool metadata is updated, such as the call count and the last call time.

[0137] In step S309, the tool is cleaned up.

[0138] Over time, there may be too many generated tools accumulated in the query tool library, which may lead to a decline in the performance of the large language model. For example, overly long prompt words may affect the model's parsing ability; at the same time, the understanding accuracy of the large language model when processing long texts will also decrease. To avoid the excessive number of generated tools affecting the accuracy of tool calls, the system adopts a preset upper limit of the number of tools combined with the LRU algorithm. When generating a new tool each time, it automatically checks and cleans up the infrequently used dynamically generated query tools, thereby ensuring the efficient operation of the query tool library.

[0139] The conditional formula for tool cleaning is as follows:

[0140] In this formula, represents the total number of tools in the current query tool library, represents the upper limit of the number of tools. When the total number of tools exceeds , the cleaning mechanism is triggered to remove the least recently used tool. It should be understood that when the total number of tools reaches the upper limit of the tool library, each time a new tool is generated, this condition will be triggered.

[0141] The LRU cleaning priority formula is defined as:

[0142] In this formula, represents the set of tools to be cleaned, represents the set of dynamically generated tools, represents the tool 's call frequency, which is defined as:

[0143] In this formula, represents the number of calls of the tool , represents the last call time of the tool , represents the registration time of the tool , represents the number of times the tool is called per unit time during its survival period.

[0144] The set of tools to be cleaned is determined by the call frequency of the tool , and the registration time of the tool is comprehensively considered. The tool with the lowest call frequency is removed first.

[0145] The cleaning operation can be expressed by the formula:

[0146] Through the above formula, while ensuring the efficient operation of the query tool library, the system limits the number of dynamically generated tools, effectively reducing the risk of large language model parsing tool errors caused by overly long tool call prompts.

[0147] Generally speaking, considering that in complex scenario analysis, the data sources are diverse and the analysis requirements are highly dynamic, the traditional fixed query mode is difficult to meet the needs of real-time adjustment. According to the adaptive complex scenario information analysis method of the exemplary embodiments of the present disclosure, an approach combining adaptive query and intelligent analysis is adopted to dynamically parse the analysis requests of users or downstream business modules, and optimize the query strategy by combining historical data, domain knowledge, and intelligent reasoning.

[0148] Specifically, the present disclosure adopts adaptive analysis of complex scenario data. By integrating multi-modal large models, comprehensive analysis and understanding of structured, unstructured text, and visual data are achieved. Compared with traditional methods, this adaptive analysis mechanism significantly improves data processing efficiency and real-time performance. The system can automatically adjust the analysis process according to real-time data to ensure the accuracy and timeliness of decisions. For example, when processing multi-source heterogeneous data, the system can automatically identify data changes and dynamically adjust the analysis strategy, avoiding decision-making delays caused by data fragmentation in traditional methods.

[0149] The present disclosure also adopts a dynamic query tool management mechanism. The system uses large language models to dynamically generate query tools, supporting real-time generation or adjustment of the tool set according to different analysis requirements. Compared with traditional fixed tool sets, this dynamic generation mechanism reduces manual intervention, improves response speed, and enhances the scalability of the system. The system can automatically generate new query tools according to user needs and automatically clean up infrequently used tools after the task is completed to ensure the efficient operation of the query tool library. This mechanism enables the system to quickly adapt to changing analysis requirements and improves overall flexibility.

[0150] The present disclosure also adopts adaptive scenario summary generation. By combining data caching and vector databases, the system achieves efficient caching, retrieval, and dynamic management, and generates analysis summaries based on large language models, which have significant advantages in relevance and accuracy. The system can adaptively optimize summary generation according to historical analysis requests and context information, enabling users to quickly obtain key information and enhancing the intelligence and coherence of analysis.

[0151] The present disclosure also adopts flexible configuration of alternative solutions. The system provides multiple alternative solutions, such as only retaining the downstream business module or the user interaction module, and selecting to use a pure large language model or a multi-modal large model according to data characteristics. This flexibility enables the system to better adapt to different application scenarios. For example, in scenarios with a small amount of data, using a pure large language model can reduce computational costs, while in scenarios with rich visual data, the application of a multi-modal large model can significantly improve the accuracy of analysis.

[0152] The present disclosure also innovates in the overall architecture of the system. The system adopts a modular design and has high scalability. This design enables the system to easily integrate new data sources and query tools, adapting to the development of future technologies and changes in business requirements. For example, the scalability of the system enables it to easily integrate new data sources or query analysis algorithms, enhancing the system's data processing ability and adaptability.

[0153] In summary, the exemplary embodiments of the present disclosure use large language models combined with adaptive tool invocation, dynamic task decomposition, dynamic tool registration, and multimodal data fusion technologies to construct an efficient intelligent analysis system. This system can not only automatically extract key information, optimize the query process, realize the reasoning and correlation analysis of complex events, solve the deficiencies of the prior art in data integration, real-time analysis, interactive query, and complex event reasoning, further improve the overall performance and application scope of the system, but also has an adaptive learning ability. It can dynamically adjust the analysis strategy according to changes in scenarios and actual needs, provide more intelligent decision-making support, thereby improving the information utilization efficiency and decision-making support ability in complex environments, and meeting the requirements of high-dynamic and high-uncertainty application scenarios such as emergency management and smart cities. It should be understood that different implementation methods are introduced for each step in this article. In actual applications, different solutions can be selected according to specific requirements and resource conditions.

[0154] Figure 4 is a block diagram of an adaptive complex scenario information analysis device according to an exemplary embodiment of the present disclosure. This device is used for an information analysis system, and the information analysis system includes a data source module, a data conversion module, a data storage module, an analysis module, a query tool library, an adaptive query module, and a historical storage module for storing analysis records of historical analysis requests. Refer to Figure 4 , the adaptive complex scenario information analysis device 400 includes an acquisition unit 401, a conversion unit 402, an analysis unit 403, a query unit 404, and an analysis unit 405.

[0155] The acquisition unit 401 is configured to obtain scenario data from multiple data sources respectively through the data source module.

[0156] The conversion unit 402 is configured to convert the obtained scenario data to a preset data format through the data conversion module and store it in the data storage module.

[0157] The analysis unit 403 is configured to, in response to receiving an analysis request through the analysis module, query historical analysis requests in the historical storage module whose similarity to the analysis request is greater than a first similarity threshold, and refer to the analysis records of the queried historical analysis requests to parse the analysis request into a query task.

[0158] The query unit 404 is configured to call the query tools in the query tool library according to the query task through the adaptive query module, query the data storage module for data, and obtain query result data.

[0159] The analysis unit 405 is configured to analyze and process the query result data through the analysis module to obtain analysis result data.

[0160] Optionally, the conversion unit 402 is further configured to, for the scenario data of each object, generate the object identifier of the object as the key through the data conversion module, and use the scenario data of the object as the value, and store it in the data storage module in the format of key-value pairs.

[0161] Optionally, the parsing unit 403 is further configured to: in response to receiving an analysis request through the analysis module, evaluate whether the analysis request is a complex request, where a complex request is an analysis request that needs to be parsed into at least N query tasks, and N is an integer greater than 1; in the case where the analysis request is not a complex request, parse the analysis request into a query task; in the case where the analysis request is a complex request, query the historical analysis requests in the historical storage module whose similarity to the analysis request is greater than the first similarity threshold, and refer to the analysis records of the queried historical analysis requests to parse the analysis request into a query task.

[0162] Optionally, the parsing unit 403 is further configured to: in the case where the analysis request is a complex request, determine whether there is a historical analysis request in the historical storage module that meets the preset conditions, where the preset conditions include: the difference between the request time of the corresponding historical analysis request and the request time of the analysis request is less than the time difference threshold, and the similarity between the corresponding historical analysis request and the analysis request is greater than the second similarity threshold, where the second similarity threshold is greater than the first similarity threshold; if there is no historical analysis request that meets the preset conditions, query the historical analysis requests in the historical storage module whose similarity to the analysis request is greater than the first similarity threshold, and refer to the analysis records of the queried historical analysis requests to parse the analysis request into a query task; if there is a historical analysis request that meets the preset conditions, abandon parsing the query task, and obtain the analysis result data of the historical analysis request that meets the preset conditions as the analysis result data of the analysis request.

[0163] Optionally, the parsing unit 403 is further configured to: use a large language model to parse the analysis request into a query task; and / or use a predefined template to parse the analysis request into a query task; and / or based on a knowledge graph, parse the analysis request into a query task.

[0164] Optionally, the query unit 404 is further configured to: determine query tool requirements according to a query task through an adaptive query module, where the query tool requirements represent the functions of the query tools to be called; retrieve query tools that meet the query tool requirements from a query tool library; and dynamically generate query tools that meet the query tool requirements in case of retrieval failure.

[0165] Optionally, the call information of the dynamically generated query tools is also recorded in the query tool library, and the call information includes the call times and call time. The adaptive complex scenario information analysis device 400 further includes a tool management unit (not shown in the figure), which is configured to: for the dynamically generated query tools in the query tools, determine the call frequency of the corresponding query tools according to the call information; and delete the dynamically generated query tools whose call frequency meets a preset deletion condition, where the preset deletion condition is used to represent low-frequency calls.

[0166] Optionally, the information analysis system further includes a domain knowledge base, and the query unit 404 is further configured to: call the query tools in the query tool library according to a query task through an adaptive query module to perform data query on the data storage module as preliminary query data; analyze whether the preliminary query data meets the analysis requirements of the analysis request; if the preliminary query data meets the analysis requirements of the analysis request, use the preliminary query data as query result data; if the preliminary query data does not meet the analysis requirements of the analysis request, analyze and reason about the preliminary query data by combining the data in the data storage module and the domain knowledge base through an analysis module to obtain an updated query task, and call the query tools in the query tool library again according to the updated query task through the adaptive query module to perform data query on the data storage module to update the preliminary query data, and repeat the step of analyzing whether the preliminary query data meets the analysis requirements of the analysis request until query result data is obtained or the maximum update times is reached.

[0167] Regarding the device in the above embodiments, the specific manners in which each unit performs operations have been described in detail in the embodiments related to the method, and will not be elaborated herein.

[0168] Figure 5 The block diagram of an electronic device 500 according to an exemplary embodiment of the present disclosure is shown.

[0169] Refer to Figure 5 , the electronic device 500 includes: at least one memory 501 and at least one processor 502, and computer-executable instructions are stored in the at least one memory 501. When the computer-executable instructions are run by the at least one processor 502, the at least one processor is caused to execute the adaptive complex scenario information analysis method as described in the above exemplary embodiment.

[0170] As an example, the electronic device 500 can be a PC computer, a tablet device, a personal digital assistant, a smart phone, or other devices capable of executing the above instruction set. Here, the electronic device 500 does not have to be a single electronic device 500, but can also be any collection of devices or circuits that can execute the above instructions (or instruction sets) individually or jointly. The electronic device 500 can also be a part of an integrated control system or system manager, or can be configured as a portable electronic device 500 that interfaces with a local or remote device (e.g., via wireless transmission).

[0171] In the electronic device 500, the processor 502 can include a central processing unit (CPU), a graphics processing unit (GPU), a programmable logic device, a dedicated processor system, a microcontroller, or a microprocessor. By way of example and not limitation, the processor 502 can also include an analog processor, a digital processor, a microprocessor, a multi-core processor, a processor array, a network processor, and the like.

[0172] The processor 502 can run instructions or code stored in the memory 501, where the memory 501 can also store data. The instructions and data can also be sent and received over a network via a network interface device, where the network interface device can employ any known transmission protocol.

[0173] The memory 501 can be integrated with the processor 502, for example, by arranging RAM or flash memory within an integrated circuit microprocessor or the like. Additionally, the memory 501 can include a stand-alone device, such as an external disk drive, a storage array, or other storage devices usable by any database system. The memory 501 and the processor 502 can be operatively coupled or can communicate with each other, for example, via an I / O port, a network connection, etc., such that the processor 502 can read files stored in the memory.

[0174] In addition, the electronic device 500 can also include a video display (such as a liquid crystal display) and a user interaction interface (such as a keyboard, a mouse, a touch input device, etc.). All components of the electronic device 500 can be connected to each other via a bus and / or a network.

[0175] According to an exemplary embodiment of the present disclosure, a computer-readable storage medium storing instructions may also be provided, wherein the instructions, when run by at least one processor, cause the at least one processor to execute the adaptive complex scenario information analysis method as described in the above exemplary embodiment. Examples of the computer-readable storage medium here include: read-only memory (ROM), programmable read-only memory (PROM), electrically erasable programmable read-only memory (EEPROM), random access memory (RAM), dynamic random access memory (DRAM), static random access memory (SRAM), flash memory, non-volatile memory, CD-ROM, CD-R, CD+R, CD-RW, CD+RW, DVD-ROM, DVD-R, DVD+R, DVD-RW, DVD+RW, DVD-RAM, BD-ROM, BD-R, BD-R LTH, BD-RE, Blu-ray or optical disc memory, hard disk drive (HDD), solid state drive (SSD), cartridge memory (such as, multimedia card, secure digital (SD) card or extreme digital (XD) card), magnetic tape, floppy disk, magneto-optical data storage device, optical data storage device, hard disk, solid state disk and any other device configured to store a computer program and any associated data, data files and data structures in a non-transitory manner and provide the computer program and any associated data, data files and data structures to a processor or computer such that the processor or computer can execute the computer program. The computer program in the above computer-readable storage medium may run in an environment deployed in computer devices such as clients, hosts, proxy devices, servers, etc. In addition, in one example, the computer program and any associated data, data files and data structures are distributed on a networked computer system such that the computer program and any associated data, data files and data structures are stored, accessed and executed in a distributed manner by one or more processors or computers.

[0176] According to an exemplary embodiment of the present disclosure, a computer program product including computer instructions may also be provided, and the computer instructions, when run by at least one processor, execute the adaptive complex scenario information analysis method as described in the above exemplary embodiment.

[0177] Those skilled in the art will readily conceive of other embodiments of the present disclosure after considering the specification and practicing the invention disclosed herein. The present disclosure is intended to cover any variations, uses, or adaptations of the present disclosure that follow the general principles of the present disclosure and include known common knowledge or conventional technical means in the technical field not disclosed in the present disclosure. The specification and embodiments are only to be considered as exemplary, and the true scope and spirit of the present disclosure are pointed out by the appended claims.

[0178] It should be understood that the present disclosure is not limited to the exact structures described above and shown in the drawings, and various modifications and changes can be made without departing from its scope. The scope of the present disclosure is limited only by the appended claims.

Claims

1. An adaptive complex scene information analysis method, characterized in that: The adaptive complex scene information analysis method is used in an information analysis system, which includes a data source module, a data conversion module, a data storage module, an analysis module, a query tool library, an adaptive query module, and a history storage module for storing analysis records of historical analysis requests. The adaptive complex scene information analysis method includes: Acquire scene data from multiple data sources respectively through the data source module; The acquired scene data is converted into a preset data format by the data conversion module, and stored in the data storage module; In response to receiving the analysis request, the analysis module searches for historical analysis requests whose similarity to the analysis request is greater than a first similarity threshold from the history storage module, and resolves the analysis request into a query task with reference to the analysis records of the searched historical analysis requests; The adaptive query module calls the query tool in the query tool library according to the query task, performs data query on the data storage module, and obtains query result data; The query result data is analyzed and processed by the analysis module to obtain analysis result data.

2. The adaptive complex scene information analysis method according to claim 1, characterized in that: The step of converting the acquired scene data into a preset data format through the data conversion module and storing the data in the data storage module includes: For the scene data of each object, the object identifier of the object is generated by the data conversion module as a key, and the scene data of the object is used as a value and stored in the data storage module in a key-value pair format.

3. The adaptive complex scene information analysis method according to claim 1, characterized in that: The step of querying, by the analysis module in response to receiving the analysis request, historical analysis requests having a similarity with the analysis request greater than a first similarity threshold from the history storage module, and parsing the analysis request into a query task with reference to the analysis records of the queried historical analysis requests includes: In response to receiving the analysis request, evaluating, by the analysis module, whether the analysis request is a complex request, wherein the complex request is an analysis request that needs to be parsed into at least N query tasks, where N is an integer greater than 1; If the analysis request is not the complex request, parsing the analysis request into a query task; In the case where the analysis request is the complex request, determining whether there is a historical analysis request that meets a preset condition in the history storage module, wherein the preset condition includes: a difference between a request time of a corresponding historical analysis request and a request time of the analysis request is less than a time difference threshold, and a similarity between the corresponding historical analysis request and the analysis request is greater than a second similarity threshold, wherein the second similarity threshold is greater than the first similarity threshold; If there is no historical analysis request that meets the preset condition, querying the historical analysis request whose similarity with the analysis request is greater than the first similarity threshold from the historical storage module, and referring to the analysis records of the queried historical analysis requests to parse the analysis request into the query task; If there is a historical analysis request that meets the preset condition, the parsing query task is abandoned, and the analysis result data of the historical analysis request that meets the preset condition is obtained as the analysis result data of the analysis request.

4. The adaptive complex scene information analysis method according to claim 1, characterized in that: The step of parsing the analysis request into a query task includes: Using a large language model, parsing the analysis request into the query task; and / or Using a predefined template, parsing the analysis request into the query task; and / or Based on the knowledge graph, the analysis request is parsed into the query task.

5. The adaptive complex scene information analysis method according to claim 1, characterized in that: The step of calling the query tool in the query tool library according to the query task by the adaptive query module includes: Determining query tool requirements according to the query task through the adaptive query module, wherein the query tool requirements represent the functions of the query tool to be called; Retrieving a query tool that meets the query tool requirement from the query tool library; In case of retrieval failure, a query tool that meets the requirements of the query tool is dynamically generated.

6. The adaptive complex scene information analysis method according to claim 5, characterized in that: The query tool library also records the call information of the dynamically generated query tool, wherein the call information includes the number of calls and the call time. The adaptive complex scene information analysis method further includes: For the dynamically generated query tools in the query tools, determining the calling frequency of the corresponding query tools according to the calling information; A dynamically generated query tool is provided for deleting calls whose frequency satisfies a preset deletion condition, wherein the preset deletion condition is used to represent low-frequency calls.

7. The adaptive complex scene information analysis method according to any one of claims 1 to 6, characterized in that: The information analysis system further includes a domain knowledge base, wherein the adaptive query module calls the query tool in the query tool library according to the query task, performs data query on the data storage module, and obtains query result data, including: The adaptive query module calls the query tool in the query tool library according to the query task to query the data in the data storage module as preliminary query data; Analyzing whether the preliminary query data meets the analysis requirements of the analysis request; If the analysis requirement of the analysis request is met, the preliminary query data is used as the query result data; If the analysis requirements of the analysis request are not met, the preliminary query data is analyzed and inferred by the analysis module in combination with the data in the data storage module and the domain knowledge base to obtain an updated query task, and the adaptive query module re-calls the query tool in the query tool library according to the updated query task to perform data query on the data storage module to update the preliminary query data, and repeats the step of analyzing whether the preliminary query data meets the analysis requirements of the analysis request until the query result data is obtained or the maximum number of updates is reached.

8. An electronic device, characterized in that: include: at least one processor; at least one memory storing computer executable instructions, Wherein, when the computer executable instructions are executed by the at least one processor, the at least one processor is prompted to execute the adaptive complex scene information analysis method as described in any one of claims 1 to 7.

9. A computer-readable storage medium, characterized in that: When the instructions in the computer-readable storage medium are executed by at least one processor, the at least one processor is prompted to execute the adaptive complex scene information analysis method as described in any one of claims 1 to 7.

10. A computer program product comprising computer instructions, characterized in that: When the computer instructions are executed by at least one processor, the at least one processor is prompted to execute the adaptive complex scene information analysis method as described in any one of claims 1 to 7.

Images