Thinking framework-based reasoning method realized through computing power of intelligent computing center

Abstract

The invention provides a thinking framework-based reasoning method realized through computing power of an intelligent computing center. The method comprises the following steps: S1, obtaining a to-be-analyzed problem input by a user and related data; s2, analyzing and reasoning the to-be-analyzed problem based on the own thinking framework and the memory bank to obtain and output the insight, the step S2 comprising: obtaining the preprocessed to-be-analyzed problem; processing the data to obtain a corresponding cause and effect graph; and generating a corresponding hypothesis based on the causal graph, the preprocessed to-be-analyzed problem, the data and the memory bank, constructing a corresponding analysis entity, carrying out analysis and reasoning on the hypothesis and the preprocessed to-be-analyzed problem based on the analysis entity, and obtaining and outputting insight. Therefore, through a structured thinking framework and a systematic analysis process, complex problems can be effectively analyzed, high-quality insight is provided, and the accuracy and efficiency of problem solving are remarkably improved.

Description

Technical Field

[0001] The present invention relates to the technical fields of intelligent computing centers, intelligent computing centers and computing power infrastructure, and specifically relates to a reasoning method based on a thinking framework through the computing power of an intelligent computing center. Background Art

[0002] With the rapid development of artificial intelligence technology, "intelligent computing centers" and "intelligent computing centers" have emerged as the times require.

[0003] An "intelligent computing center" refers to a facility that uses large-scale heterogeneous computing power resources, including general computing power and intelligent computing power, to mainly provide the required computing power, data, and algorithms for artificial intelligence applications (such as scenarios like artificial intelligence deep learning model development, model training, and model inference). An intelligent computing center covers facilities, hardware, and software, and can provide full-stack capabilities from underlying computing power to top-level application enabling.

[0004] An "intelligent computing center" includes but is not limited to an "intelligent computing center".

[0005] An "intelligent computing center", that is, an artificial intelligence computing center, is a type of computing power infrastructure based on artificial intelligence theory, adopting an artificial intelligence computing architecture, and providing computing power services, data services, and algorithm services required for artificial intelligence applications.

[0006] "Computing power" is the core of "intelligent computing centers" and "intelligent computing centers". It is the ability of a computer device or a computing / data center to process information, the ability of computer hardware and software to cooperate to jointly execute a certain computing requirement, the computing ability to achieve the output of a target result by processing information data, a new type of productive force integrating information computing power, network carrying capacity, and data storage capacity, and mainly provides services to society through computing power infrastructure.

[0007] In today's era of rapid digital and intelligent development, intelligent agents, as an important part of the field of artificial intelligence, are playing an increasingly important role.

[0008] An "intelligent agent" refers to an agent that can perceive the environment and take actions to achieve specific goals. It can be software, hardware, or a system, and has autonomy, adaptability, and interaction capabilities. An intelligent agent perceives changes in the environment (such as through sensors or data input), makes judgments and decisions based on the knowledge and algorithms learned by itself, and then executes actions to affect the environment or achieve a predetermined goal. Intelligent agents are widely used in the field of artificial intelligence, commonly found in automation systems, robots, virtual assistants, and game characters, etc. The core lies in the ability to learn independently and evolve continuously to better complete tasks and adapt to complex environments.

[0009] Existing agents, when analyzing problems, mainly rely on pre-set algorithms, rules, and large amounts of training data. First, they will parse the input complex tasks or problems, extract key information and features, and then search for matching patterns or examples in their own knowledge systems or training data. In this way, they attempt to understand the essence and requirements of the problems and generate corresponding answers or action plans based on existing logics and algorithms.

[0010] However, when dealing with complex tasks or problems, agents face numerous technical challenges. Complex problems often have characteristics such as ambiguity, uncertainty, and multi-domain intersections. This makes it difficult for agents to deeply understand the complete context and true intentions of the problems without a thinking framework. Existing agents usually can only parse and answer problems from the surface or literal meaning, with a superficial and unstructured understanding of the problems, unable to think systematically. This superficial and unstructured processing method easily leads to the "hallucination" phenomenon, that is, agents generate seemingly reasonable but actually inaccurate or irrelevant answers, and it is difficult to generate profound insights or provide accurate and structured answers.

[0011] In summary, due to the lack of a thinking framework, existing agents are unable to obtain profound insights, accurate, and structured answers when dealing with problems (especially complex problems). Summary of the Invention

[0012] The present invention provides a reasoning method based on a thinking framework through the computing power of an intelligent computing center to solve the technical problem that existing agents, due to the lack of a thinking framework, are unable to obtain profound insights, accurate, and structured answers when dealing with problems (especially complex problems).

[0013] To solve the above technical problems, the present invention is implemented as follows:

[0014] In a first aspect, the present invention provides a reasoning method based on a thinking framework through the computing power of an intelligent computing center. The method is applied to an agent and includes:

[0015] Step S1: Obtain the problem to be analyzed input by the user and the data related to the problem to be analyzed;

[0016] Step S2: Analyze and reason about the problem to be analyzed based on its own thinking framework and memory bank, and obtain and output insights. Among them, the thinking framework is a structured way of thinking of the agent.

[0017] Among them, the step S2 includes:

[0018] Step S21: After preprocessing the problem to be analyzed, obtain the preprocessed problem to be analyzed;

[0019] Step S22: Process the data based on the causal inference model and the memory bank to obtain a corresponding causal graph;

[0020] Step S23: Generate a corresponding hypothesis based on the causal graph, the preprocessed problem to be analyzed, the data, and the memory bank. Construct a corresponding analysis entity based on the hypothesis, the preprocessed problem to be analyzed, and the memory bank. Analyze and reason about the hypothesis and the preprocessed problem to be analyzed based on the analysis entity, and obtain and output the insight.

[0021] Optionally, the analysis entity includes: subspace, breakdown, and metric. The subspace is used to define the data object to be analyzed. The breakdown is used to define the relevant data objects of the dimensions related to the data object to be analyzed. The metric is used to define the analysis metrics of the data object to be analyzed. The object to be analyzed and the relevant data objects are determined according to the hypothesis and the preprocessed problem to be analyzed;

[0022] The insight includes: the subspace, the breakdown, the metric, type, and score. The type is used to define the type of the insight, and the score is used to define the confidence level of the insight.

[0023] Optionally, step S21 includes:

[0024] Step S211: Standardize the problem to be analyzed to obtain a standard problem to be analyzed;

[0025] Step S212: Judge the complexity of the standard problem to be analyzed. If the complexity of the standard problem to be analyzed reaches a preset complexity threshold, determine the standard problem to be analyzed as a complex problem to be analyzed; if the complexity of the standard problem to be analyzed does not reach the complexity threshold, determine the standard problem to be analyzed as a simple problem to be analyzed;

[0026] Step S213: When the standard problem to be analyzed is the complex problem to be analyzed, perform intent recognition on the complex problem to be analyzed. According to the result of the intent recognition, convert the format of the complex problem to be analyzed, disassemble the complex problem to be analyzed after format conversion into multiple sequentially arranged sub-problems to be analyzed, and determine the multiple sequentially arranged sub-problems to be analyzed as the preprocessed problem to be analyzed;

[0027] When the standard problem to be analyzed is the simple problem to be analyzed, perform intent recognition on the simple problem to be analyzed. According to the result of the intent recognition, convert the format of the simple problem to be analyzed, and determine the simple problem to be analyzed after format conversion as the preprocessed problem to be analyzed.

[0028] Optionally, step S211 includes at least one of the following two items:

[0029] The first item includes:

[0030] Step S2111: Perform a first analysis on the current problem to be analyzed based on natural language processing technology to determine whether the clarity of the problem to be analyzed reaches a preset clarity threshold; if not, then execute step S2112;

[0031] Step S2112: Output a first prompt message, where the first prompt message is used to prompt the user to re-enter the problem to be analyzed;

[0032] Step S2113: Perform a first analysis on the problem to be analyzed re-entered by the user based on the natural language processing technology to determine whether the clarity of the problem to be analyzed re-entered by the user reaches the preset clarity threshold. If not, repeatedly execute step S2112 and step S2113 until the clarity of the current problem to be analyzed reaches the preset clarity threshold;

[0033] Step S2114: Determine the current problem to be analyzed whose clarity reaches the preset clarity threshold as the standard problem to be analyzed;

[0034] Or,

[0035] Step S2111’: Perform a first analysis on the current problem to be analyzed based on natural language processing technology to determine whether the clarity of the problem to be analyzed reaches a preset clarity threshold; if not, then execute step S2112’;

[0036] Step S2112’: Interpret the problem to be analyzed based on the thinking framework and the memory library, output the interpreted problem to be analyzed, and receive the feedback information input by the user; if the feedback information indicates that the interpreted problem to be analyzed is incorrect, repeatedly execute step S2112’ until the feedback information indicates that the current interpreted problem to be analyzed is correct;

[0037] Step S2113’: Determine the interpreted problem to be analyzed corresponding to the feedback information indicating that the current interpreted problem to be analyzed is correct as the standard problem to be analyzed;

[0038] The second item includes:

[0039] Step S211a: Perform a second analysis on the problem to be analyzed based on natural language processing technology to determine whether the problem to be analyzed is meaningful. If not, send a second prompt message for prompting the user that the problem to be analyzed is meaningless. Wherein, the meaningfulness of the problem to be analyzed includes: the clarity of the intention of the problem to be analyzed reaches a preset intention clarity threshold, the operability of the problem to be analyzed reaches an operability threshold, and the problem to be analyzed conforms to common sense and logic, where the common sense and the logic are stored in the memory bank.

[0040] Step S211b: Determine the meaningful problem to be analyzed as the standard problem to be analyzed.

[0041] Optionally, the step S1 includes: Step S11: Obtain the problem to be analyzed input by the user, the data related to the problem to be analyzed, and the background knowledge related to the problem to be analyzed.

[0042] The step S23 includes: Step S231: Generate corresponding hypotheses based on the causal graph, the preprocessed problem to be analyzed, the data, the background knowledge, and the memory bank. Construct corresponding analysis entities based on the hypotheses, the preprocessed problem to be analyzed, and the memory bank. Analyze and reason about the hypotheses and the preprocessed problem to be analyzed based on the analysis entities, and obtain and output the insight.

[0043] Optionally, the preprocessed problem to be analyzed includes: M sequentially arranged sub-problems to be analyzed, where M is a positive integer greater than or equal to 2.

[0044] The step S231 includes:

[0045] Step S2311: Generate corresponding hypotheses based on the causal graph, the first sub-problem to be analyzed among the M sequentially arranged sub-problems to be analyzed, the background knowledge, the data, and the memory bank. Construct corresponding analysis entities based on the hypotheses, the first sub-problem to be analyzed, and the memory bank. Analyze and reason about the hypotheses and the first sub-problem to be analyzed based on the analysis entities to obtain an insight.

[0046] Step S2312: Generate corresponding new hypotheses based on the insight, the causal graph, the Nth sub-problem to be analyzed among the M sequentially arranged sub-problems to be analyzed, the background knowledge, the data, and the memory bank. Construct corresponding new analysis entities based on the new hypotheses, the Nth sub-problem to be analyzed, and the memory bank. Analyze and reason about the new hypotheses and the Nth sub-problem to be analyzed again based on the new analysis entities to obtain a new insight.

[0047] Step S2313: Repeat the said step S2312 until the value of N reaches M. When the step S2312 is executed for the first time, the corresponding value of N is 2. Each time the step S2312 is repeated, the value of N is incremented by 1, where N ∈ [2, M] and N is a positive integer;

[0048] Step S2314: Output all the obtained insights.

[0049] Optionally, the number of preprocessed problems to be analyzed is one, and the step S231 includes:

[0050] Step S231a: Generate corresponding hypotheses based on the causal graph, the preprocessed problem to be analyzed, the background knowledge, the data, and the memory bank; construct corresponding analysis entities based on the hypotheses, the preprocessed problem to be analyzed, and the memory bank; perform analysis and reasoning on the hypotheses and the preprocessed problem to be analyzed based on the analysis entities to obtain insights;

[0051] Step S231b: Generate corresponding new hypotheses based on the insights, the causal graph, the preprocessed problem to be analyzed, the background knowledge, the data, and the memory bank; construct corresponding new analysis entities based on the new hypotheses, the preprocessed problem to be analyzed, and the memory bank; perform analysis and reasoning on the new hypotheses and the preprocessed problem to be analyzed again based on the new analysis entities to obtain new insights;

[0052] Step S231c: Repeat the execution of step S231b until the number of executions of step S231b reaches a preset number of times, or until the total execution time of step S231b reaches a preset duration;

[0053] Step S231d: Output all the obtained insights.

[0054] Optionally, all the obtained insights include:

[0055] Summarize all the insights and generate a structured insight report;

[0056] Output the insight report, where the types of the insight report include at least one of the following: basic insight, related insight, summary insight, and nested insight.

[0057] Optionally, the thinking framework can continuously perform self-iteration based on the autonomous learning process of the intelligent agent, the memory bank, the analysis and reasoning process, and the user's feedback on the insights to obtain a new thinking framework.

[0058] In a second aspect, the present invention provides an inference device based on a thinking framework implemented through the computing power of an intelligent computing center. The device is applied to an intelligent agent and includes:

[0059] An acquisition module for performing step S1: acquiring the problem to be analyzed input by the user and the data related to the problem to be analyzed;

[0060] An execution module for performing step S2: analyzing and inferring the problem to be analyzed based on its own thinking framework and memory bank, and obtaining and outputting insights. Among them, the thinking framework is a structured way of thinking of the intelligent agent;

[0061] Among them, step S2 includes:

[0062] Step S21: After preprocessing the problem to be analyzed, obtaining the preprocessed problem to be analyzed;

[0063] Step S22: Processing the data based on the causal inference model and the memory bank to obtain a corresponding causal graph;

[0064] Step S23: Generating corresponding hypotheses based on the causal graph, the preprocessed problem to be analyzed, the data, and the memory bank, constructing corresponding analysis entities based on the hypotheses, the preprocessed problem to be analyzed, and the memory bank, and analyzing and inferring the hypotheses and the preprocessed problem to be analyzed based on the analysis entities to obtain and output the insights.

[0065] In a third aspect, the present invention provides a server, including: a processor, a memory, and a program stored on the memory and executable on the processor. When the program is executed by the processor, it implements the steps of an inference method based on a thinking framework implemented through the computing power of an intelligent computing center as described in the first aspect above.

[0066] In a fourth aspect, the present invention provides a computer-readable storage medium on which a computer program is stored. When the computer program is executed by a processor, it implements the steps of an inference method based on a thinking framework implemented through the computing power of an intelligent computing center as described in the first aspect above.

[0067] In a fifth aspect, the present invention provides a computer program product including computer instructions. When the computer instructions are executed by a processor, they implement the steps of an inference method based on a thinking framework implemented through the computing power of an intelligent computing center as described in the first aspect above.

[0068] In the present invention, with the powerful computing power of the intelligent computing center, the intelligent agent can achieve the following: quickly process and analyze massive data, significantly improving the efficiency of problem analysis; based on the causal inference model and the memory bank, quickly generate high-precision causal diagrams to reveal the complex relationships in the data; conduct complex multi-step reasoning and hypothesis verification to ensure the logical rigor of the analysis process and the credibility of the results; generate high-quality and interpretable insights to provide valuable conclusions and suggestions for users.

[0069] In summary, with the computing power of the intelligent computing center and combined with a structured thinking framework and a systematic analysis process, the intelligent agent can more effectively handle complex problems and provide higher-quality insights. This improves the accuracy and efficiency of analyzing and solving complex problems. BRIEF DESCRIPTION OF THE DRAWINGS

[0070] By reading the following detailed description of the preferred embodiments, various other advantages and benefits will become clear to those of ordinary skill in the art. The drawings are only for the purpose of showing the preferred embodiments and are not considered to be a limitation of the present invention. Moreover, throughout the drawings, the same reference numerals are used to represent the same components. In the drawings:

[0071] Figure 1 is a flowchart of a reasoning method based on a thinking framework realized by the computing power of the intelligent computing center provided by the present invention;

[0072] Figure 2 is a flowchart of a reasoning method based on a thinking framework realized by the computing power of the intelligent computing center provided by the present invention;

[0073] Figure 3 is a schematic diagram of an intelligent agent memory bank provided by the present invention;

[0074] Figure 4 is a flowchart of a reasoning method based on a thinking framework realized by the computing power of the intelligent computing center provided by the present invention;

[0075] Figure 5 is a schematic diagram of twelve insight mode types provided by the present invention;

[0076] Figure 6 is a schematic diagram showing that the thinking framework of the intelligent agent can be self-iterative provided by the present invention;

[0077] Figure 7 is a structural block diagram of a reasoning device based on a thinking framework realized by the computing power of the intelligent computing center provided by the present invention;

[0078] Figure 8 is a structural schematic diagram of an electronic device of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0079] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, rather than all of the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0080] First, the technical terms related to the present invention will be briefly explained below.

[0081] The "computing power" referred to in the present invention means: the ability of a computer device or a computing / data center to process information, the ability of computer hardware and software to cooperate to jointly execute a certain computing requirement, the computing ability to process information data and output a target result, a new type of productive force integrating information computing power, network carrying capacity, and data storage capacity, and mainly providing services to society through computing power infrastructure.

[0082] The "computational power" (Computational Power, CP) referred to in the present invention means: the ability of a data center server to process data and output results, a comprehensive index to measure the computing ability of a data center, including general computing ability, supercomputing ability, and intelligent computing ability. The commonly used measurement unit is the number of floating-point operations per second (FLOPS, 1 EFLOPS = 10^18 FLOPS), and the larger the value, the stronger the comprehensive computing ability. It is estimated that 1 EFLOPS is approximately the computing power output of 5 Tianhe-2A or 500,000 mainstream server CPUs or 2 million mainstream laptops. The calculation formula is: CP = CP_general + CP_intelligent + CP_super

[0083] The "network power" (Network Power, NP) referred to in the present invention means: the manifestation of the data transmission ability of computing power facilities, a comprehensive ability including network architecture, network bandwidth, transmission delay, intelligent management and scheduling, etc., involving network transmission inside and between data centers, and a comprehensive index to measure the network transmission scheduling ability.

[0084] The "Storage Power" (SP) described in the present invention refers to the comprehensive ability of a data center in four aspects: data storage capacity, performance, security and reliability, and green and low-carbon. It is a comprehensive indicator for measuring the data storage capacity of a data center, including external storage devices such as storage arrays and built-in storage devices of servers. The commonly used measurement unit for storage capacity is exabyte (EB, 1EB = 2^60 bytes), the commonly used measurement unit for performance is the number of read / write operations per second per unit capacity (IOPS / TB, Input / Output Operations Per Second / TB), and the disaster recovery ratio is an important manifestation of security and reliability.

[0085] The "computing power infrastructure" described in the present invention refers to a new type of information infrastructure that integrates information computing power, network carrying capacity, and data storage power, and can realize the centralized computing, storage, transmission, and application of information.

[0086] The "new type of information infrastructure" described in the present invention mainly includes network infrastructures such as 5G networks, fiber broadband networks, backbone networks, international communication networks, and satellite Internet, computing power infrastructures such as data centers, general computing power centers, intelligent computing centers, and supercomputing centers, and new technology infrastructures such as artificial intelligence, blockchain, and quantum computing.

[0087] The "computing power" described in the present invention includes general computing power, intelligent computing power, and super computing power.

[0088] The "general computing power" described in the present invention refers to the computing power provided by servers based on CPU (Central Processing Unit) chips, which is used to support basic general computing such as cloud computing and edge computing.

[0089] The "intelligent computing power" described in the present invention refers to a computing platform that is scaled for various artificial intelligence innovation applications based on dedicated chips such as GPU (Graphics Processing Unit), FPGA (Field Programmable Gate Array), and ASIC (Application Specific Integrated Circuit), such as natural language processing and machine vision.

[0090] The "super computing power" described in the present invention mainly refers to the computing power provided by high-performance computing clusters such as supercomputers. It utilizes the centralized computing resources of multiple computer systems working in parallel and processes extremely complex or data-intensive problems through a dedicated operating system. It is mainly used for computing in cutting-edge scientific fields, such as planetary simulation, drug molecule design, and gene analysis.

[0091] The "Intelligent Computing Center" as described in the present invention refers to a facility that, by using large-scale heterogeneous computing power resources, including general computing power (CPU) and intelligent computing power (GPU, FPGA, ASIC, etc.), mainly provides the required computing power, data, and algorithms for artificial intelligence applications (such as scenarios like artificial intelligence deep learning model development, model training, and model inference). The intelligent computing center covers facilities, hardware, and software, and can provide full-stack capabilities from underlying computing power to top-level application enablement.

[0092] The "Intelligent Computing Center" as described in the present invention includes, but is not limited to, intelligent computing centers.

[0093] The "Intelligent Computing Center" as described in the present invention, namely the artificial intelligence computing center, is a type of computing power infrastructure that, based on artificial intelligence theory and adopting an artificial intelligence computing architecture, provides computing power services, data services, and algorithm services required for artificial intelligence applications.

[0094] The "Computing Power Center" as described in the present invention refers to a facility mainly composed of infrastructure such as wind, fire, water, and electricity and IT software and hardware devices, and having computing power, carrying capacity, and storage capacity, including general data centers, intelligent computing centers, supercomputing centers, etc.

[0095] The "Supercomputing Center" as described in the present invention refers to, namely the supercomputing data center, a data center based on supercomputers or large-scale computing clusters, which can provide functions such as large-scale computing, storage, and network services, and is widely used in application scenarios such as aerospace, national defense, oil exploration, climate modeling, and genome sequencing.

[0096] The "Computing Power Resources" as described in the present invention refers to technologies and facilities with information computing, transmission, storage, and application capabilities required for the development of the digital society, including but not limited to computing resources such as CPU and GPU, network resources such as switches and routers, storage resources such as storage arrays and distributed storage, security resources such as firewalls and intrusion detection systems, and support and guarantee resources such as wind, fire, water, and electricity.

[0097] The "Large Language Model" as described in the present invention refers to the Large Language Model (LLM), which is a language model with a relatively large number of parameters, aiming to understand and generate human language, trained through a large amount of text data, and can perform a wide range of tasks including text summarization, translation, sentiment analysis, etc.

[0098] The "Agent" described in the present invention refers to an entity that can perceive the environment and take actions to achieve specific goals. It can be software, hardware, or a system, and has autonomy, adaptability, and interaction capabilities. The agent perceives changes in the environment (such as through sensors or data input), makes judgments and decisions based on the knowledge and algorithms it has learned, and then executes actions to affect the environment or achieve a predetermined goal. Agents are widely used in the field of artificial intelligence and are commonly found in automated systems, robots, virtual assistants, and game characters. The core lies in the ability to learn autonomously and evolve continuously to better complete tasks and adapt to complex environments.

[0099] Figure 1 Figure 4 shows a reasoning method based on a thinking framework implemented through the computing power of an intelligent computing center according to the present invention, as Figure 1 and with reference to Figure 2 shown, the method is applied to an agent, and the method includes:

[0100] Step S1: Obtain the problem to be analyzed input by the user and the data related to the problem to be analyzed;

[0101] Step S2: Analyze and reason about the problem to be analyzed based on its own thinking framework and memory bank, and obtain and output insights;

[0102] Among them, the thinking framework is a structured way of thinking of the agent;

[0103] It should be noted that in step S1, the agent first needs to obtain the problem to be analyzed from the user. The problem to be analyzed can be a specific query, decision request, or complex problem to be solved. It also needs to collect data related to this problem from the user, aiming to provide the necessary context for the analysis of the problem.

[0104] In addition, in a specific application scenario, the agent can also obtain background knowledge related to the problem to be analyzed. As Figure 2 shown, users can be divided into common users and expert users, and expert users can provide the agent with background knowledge related to the problem to be analyzed. The background knowledge includes types of knowledge such as domain, social, fact, personal, and tool. Thus, by comprehensively collecting information, the agent can ensure that it has sufficient context and data support when analyzing complex problems, thereby improving the accuracy and relevance of the analysis.

[0105] In step S2, the Thinking Framework is a structured way of thinking for an agent, which helps the agent systematically analyze problems, organize information, and solve problems. It provides a clear and logical thinking path, making complex problems easier to understand and handle. A typical thinking framework usually includes the following parts: Problem Definition: Clearly define the problem or goal to be solved; Information Gathering: Obtain data and information related to the problem; Analysis and Reasoning: Draw conclusions through logical reasoning or data analysis; Verification and Optimization: Verify the rationality of the conclusions and optimize the solutions; Result Output: Generate the final answer or action suggestions.

[0106] Insight refers to the meaningful and valuable information or knowledge discovered through the analysis and interpretation of data. It usually reveals patterns, trends, anomalies, or relationships in the data, and can help users better understand the data and support decision-making. For example: Discovering a significant decline in sales during a certain period; Identifying that the customer satisfaction in a certain region is significantly lower than that in other regions.

[0107] Specifically, insight can also refer to the following: A deep understanding of a situation (or person or thing), discovering patterns in data, or discovering relationships between variables that were previously unknown, there may be an exciting "aha moment" (which refers to the moment of suddenly understanding or discovering a key insight or solution), it is the connection between knowledge units, can distinguish domain knowledge from analytical knowledge, and is an interpretation of visualization rather than visualization itself. Generally speaking, insight can reflect the interesting aspects of a specific data theme from a certain perspective.

[0108] The memory bank refers to a structured database or knowledge base used by the agent to store and manage historical data, knowledge, experience, and context information. It is one of the core components for the agent to achieve autonomous learning, reasoning, and decision-making, and can support the agent to quickly retrieve and utilize relevant information when analyzing problems.

[0109] The core components of the memory bank are (as Figure 3 shown): Working Memory: Stores active and readily available information: perceptual input, real-time knowledge, goals, and intermediate reasoning results; Semantic Memory: Stores facts and concepts about the world; Episodic Memory: Stores past behavioral experiences: input-output pairs, historical event streams, action trajectories, etc.; Procedural Memory: Stores implicit knowledge in the weights of the large language model and explicit knowledge or skills written in the agent's code. The agent needs to analyze and reason about the problem to be analyzed based on its own thinking framework and memory bank, and obtain and output insights.

[0110] Among them, as Figure 4 shown, step S2 includes:

[0111] Step S21: After preprocessing the problem to be analyzed, obtain the preprocessed problem to be analyzed;

[0112] Step S22: Process the data based on the causal inference model and the memory bank to obtain the corresponding causal graph;

[0113] Step S23: Generate the corresponding hypothesis based on the causal graph, the preprocessed problem to be analyzed, the data, and the memory bank, construct the corresponding analysis entity based on the hypothesis, the preprocessed problem to be analyzed, and the memory bank, and perform analysis and reasoning on the hypothesis and the preprocessed problem to be analyzed based on the analysis entity to obtain and output insights.

[0114] In step S21, the problem to be analyzed will go through preprocessing processes such as standardization, clarity improvement, and complexity judgment. The intelligent agent will identify the key information of the problem, remove redundant information, and transform the problem into a standard format for analysis. Through preprocessing, the intelligent agent can ensure that the problem to be analyzed has a clear expression and structure, thus providing a good basis for subsequent analysis.

[0115] In step S22, the intelligent agent uses the causal inference model to process the collected data. By analyzing the relationships between the data, the intelligent agent constructs a causal graph to represent the causal relationships between different variables. The causal graph can help the intelligent agent understand the potential relationships and patterns in the data and provide important support for subsequent hypothesis generation and analysis.

[0116] Among them, the causal inference model is a statistical model used to analyze and infer the causal relationships between variables. Different from traditional correlation analysis, the causal inference model aims to identify and quantify the direct impact of one variable on another variable, thereby revealing the causal mechanism.

[0117] In step S23, the intelligent agent generates the corresponding hypothesis based on the causal graph, the preprocessed problem to be analyzed, the data, and the memory bank. The hypothesis can be understood as a potential answer or explanation to the problem to be analyzed; then, the intelligent agent constructs the corresponding analysis entity based on these hypotheses, the preprocessed problem to be analyzed, and the memory bank. The analysis entity is a structured representation of the problem and the hypothesis, containing relevant variables and relationships; finally, the intelligent agent performs analysis and reasoning on the hypothesis and the preprocessed problem to be analyzed based on the analysis entity to obtain and output insights. Thus, by generating hypotheses and constructing analysis entities, the intelligent agent can obtain insights and systematically conduct in-depth analysis of the problem to be analyzed, thereby improving the comprehensiveness and depth of reasoning.

[0118] In a possible implementation, the analysis entity includes: subspace, breakdown, and measure, that is: AnalysisEntity := {subspace, breakdown, measure}, where the subspace is used to define the data object to be analyzed, the breakdown is used to define the relevant data objects of the dimensions related to the data object to be analyzed, and the measure is used to define the analysis metrics of the data object to be analyzed. And the object to be analyzed and the relevant data objects are determined according to the problem to be analyzed after hypothesis and preprocessing.

[0119] Insight includes: subspace, breakdown, measure, type, and score, that is: Insight := {subspace, breakdown, measure, type, score}, where the first three are the same as the definitions in the analysis entity. The type is used to define the category of the insight (such as: trend, anomaly, or correlation), and the score is used to define the confidence level of the insight. It can be seen that the analysis entity is the basis of the insight, defining the scope, dimensions, and metrics of the analysis, while the insight is valuable information further interpreted and discovered based on the analysis entity.

[0120] In a possible implementation, there are 12 basic types of insights: Trend: The pattern of data changing over time, such as the annual increase or decrease in sales; Outlier: The points in the data that significantly deviate from other values, such as an unusually high or low sales volume in a certain month; Change: The significant change in the data at a certain time point or under certain conditions, such as the change in user behavior after the implementation of a policy; Distribution: The distribution characteristics of the data, such as whether the distribution of user ages conforms to a normal distribution; Ranking: The ranking of the data in a certain dimension, such as the product category with the highest sales volume; Correlation: The relationship between two or more variables, such as the positive correlation between advertising investment and sales volume; Difference: The data difference between different groups or categories, such as the difference in user activity levels in different regions; Proportion: The relationship between the part and the whole, such as the proportion of a certain product category in the total sales volume; Extremum: The maximum or minimum value of the data, such as the highest or lowest temperature on a certain day; Association: The patterns that appear simultaneously in the data, such as users who purchase product A also tend to purchase product B; Anomaly: The points in the data that do not conform to the expected pattern, such as a sudden drop in sales volume in a certain region; Clustering: The natural groupings formed in the data, such as the clustering of user behavior patterns, as Figure 5 shown.

[0121] The relationships among hypotheses, analytical entities, and insights will be described below.

[0122] A hypothesis is a preliminary conjecture or speculation about data or phenomena, usually proposed based on domain knowledge, experience, or preliminary observations. It is the starting point of data analysis and is used to guide subsequent analysis and verification. A hypothesis can be qualitative (e.g., "The decline in sales in a certain region may be related to the economic environment") or quantitative (e.g., "A 10% increase in advertising investment will lead to a 5% increase in sales"). The hypothesis provides the direction and goal for the analysis. For example, if the hypothesis is "The decline in sales in a certain region may be related to the economic environment", then the analytical entity may focus on the sales data (subspace) in that region and further analyze it by time or economic indicators (breakdown), and sales can be used as the main metric (measure). The analytical entity is the concretization of the hypothesis, which transforms the abstract conjecture into an operable analysis framework. Through the scope, dimensions, and indicators defined by the analytical entity, users can systematically verify the rationality of the hypothesis and obtain insights. Insights are the results of hypothesis verification. Insights are valuable information extracted from the data of the analytical entity, which may be the verification, modification, or negation of the hypothesis. For example, the analysis may reveal that "The decline in sales in a certain region is indeed related to the deterioration of the economic environment", which is an insight. Insights can also trigger new hypotheses, forming an iterative analysis process. For example, after discovering the decline in sales, it may be further hypothesized that "The economic environment has a greater impact on high-priced products", thus initiating a new analysis (to be described later).

[0123] In the present invention, with the powerful computing power of the intelligent computing center, the intelligent agent can achieve the following: quickly process and analyze massive data, significantly improving the efficiency of problem analysis; quickly generate high-precision causal graphs based on the causal inference model and memory bank to reveal complex relationships in the data; perform complex multi-step reasoning and hypothesis verification to ensure the logical rigor of the analysis process and the credibility of the results; generate high-quality and interpretable insights to provide valuable conclusions and suggestions for users.

[0124] In summary, with the computing power of the intelligent computing center, combined with a structured thinking framework and a systematic analysis process, it is possible to more effectively handle complex problems, provide higher-quality insights, and improve the accuracy and efficiency of the intelligent agent in analyzing and solving complex problems.

[0125] In one possible implementation, step S21 includes:

[0126] Step S211: Standardize the problem to be analyzed to obtain a standard problem to be analyzed;

[0127] Step S212: Determine the complexity of the standard problem to be analyzed. If the complexity of the standard problem to be analyzed reaches the preset complexity threshold, then determine the standard problem to be analyzed as a complex problem to be analyzed; if the complexity of the standard problem to be analyzed does not reach the complexity threshold, then determine the standard problem to be analyzed as a simple problem to be analyzed.

[0128] Step S213: When the standard problem to be analyzed is a complex problem to be analyzed, perform intention recognition on the complex problem to be analyzed. According to the result of the intention recognition, convert the format of the complex problem to be analyzed, decompose the format-converted complex problem to be analyzed into multiple sequentially arranged sub-problems to be analyzed, and determine the multiple sequentially arranged sub-problems to be analyzed as the preprocessed problems to be analyzed.

[0129] When the standard problem to be analyzed is a simple problem to be analyzed, perform intention recognition on the simple problem to be analyzed. According to the result of the intention recognition, convert the format of the simple problem to be analyzed, and determine the format-converted simple problem to be analyzed as the preprocessed problem to be analyzed.

[0130] For reference Figure 2 , the intelligent agent first standardizes the problem to be analyzed to obtain the standard problem to be analyzed. Then, the intelligent agent judges the complexity of the standard problem to be analyzed according to the preset complexity threshold, and then classifies it as a complex problem to be analyzed or a simple problem to be analyzed. If it is a complex problem to be analyzed, after intention recognition and format conversion, it will be decomposed into multiple sequentially arranged sub-problems to be analyzed for step-by-step analysis; if it is a simple problem to be analyzed, after intention recognition and format conversion, it will be determined as the preprocessed problem to be analyzed. Through this series of processes, the intelligent agent can effectively identify the nature of the problem, ensure the pertinence and systematicness of the analysis process, and thus improve the accuracy and efficiency of subsequent reasoning.

[0131] It should be noted that intention recognition refers to using natural language processing (NLP) technology to understand the purpose or intention of the user's input problem, so as to determine the information the user wants to obtain or the task to be completed. And simple problems usually refer to problems with clear expressions and clear intentions, which usually only require one-step operations to answer, so the intention can be directly recognized and the answer can be returned without decomposition. Complex problems refer to problems with vague expressions or containing multiple sub-tasks that need to be further decomposed and refined, so the problems can be decomposed and solved step by step.

[0132] In a possible implementation, step S211 includes at least one of the following two items:

[0133] The first item includes:

[0134] Step S2111: Perform a first analysis on the current problem to be analyzed based on natural language processing technology to determine whether the clarity of the problem to be analyzed reaches a preset clarity threshold; if not, then execute Step S2112;

[0135] Step S2112: Output a first prompt message, where the first prompt message is used to prompt the user to re-enter the problem to be analyzed;

[0136] Step S2113: Perform a first analysis on the problem to be analyzed re-entered by the user based on natural language processing technology to determine whether the clarity of the problem to be analyzed re-entered by the user reaches the preset clarity threshold. If not, repeat Step S2112 and Step S2113 until the clarity of the current problem to be analyzed reaches the preset clarity threshold;

[0137] Step S2114: Determine the current problem to be analyzed whose clarity reaches the preset clarity threshold as the standard problem to be analyzed;

[0138] Or,

[0139] Step S2111’: Perform a first analysis on the current problem to be analyzed based on natural language processing technology to determine whether the clarity of the problem to be analyzed reaches a preset clarity threshold; if not, then execute Step S2112’;

[0140] Step S2112’: Interpret the problem to be analyzed based on the thinking framework and the memory bank, output the interpreted problem to be analyzed, and receive the feedback information input by the user; if the feedback information indicates that the interpreted problem to be analyzed is incorrect, then repeat Step S2112’ until the feedback information indicates that the current interpreted problem to be analyzed is correct;

[0141] Step S2113’: Determine the interpreted problem to be analyzed corresponding to the feedback information indicating that the current interpreted problem to be analyzed is correct as the standard problem to be analyzed;

[0142] The second item includes:

[0143] Step S211a: Perform a second analysis on the problem to be analyzed based on natural language processing technology to determine whether the problem to be analyzed is meaningful. If not, then issue a second prompt message, where the second prompt message is used to prompt the user that the problem to be analyzed is meaningless; where a meaningful problem to be analyzed includes: the intention clarity of the problem to be analyzed reaches a preset intention clarity threshold, the operability of the problem to be analyzed reaches an operability threshold, and the problem to be analyzed conforms to common sense and logic, where the common sense and logic are stored in the memory bank;

[0144] Step S211b: Determine the meaningful problem to be analyzed as the standard problem to be analyzed.

[0145] It should be noted that the purpose of this possible implementation is to obtain the problem to be analyzed with clarity reaching a preset threshold and / or a meaningful problem to be analyzed.

[0146] Specifically, natural language processing technology can be used to preliminarily analyze the problem to be analyzed, judge whether its clarity meets the preset standard. If it does not meet the standard, a prompt message is output, asking the user to re-enter the problem to be analyzed, analyze the problem re-entered by the user again, and judge its clarity. If it still does not meet the standard, the user is repeatedly prompted until the problem is clear. Once the clarity of the problem reaches the standard, it is determined as the standard problem to be analyzed. Or, similarly analyze the clarity of the problem. If it does not meet the standard, the problem is explained based on the thinking framework and the memory bank, and the user feedback is received. If the user feedback indicates that the explanation is incorrect, continue to adjust until the user confirms that the explanation is correct, and determine the problem after the explanation confirmed by the user as the standard problem to be analyzed.

[0147] Natural language processing technology can also be used to judge whether the problem to be analyzed is meaningful, including the judgment of intention clarity, operability, and logical common sense. If the problem is meaningless, a prompt message is issued. If the problem is meaningful, the meaningful problem is determined as the standard problem to be analyzed.

[0148] Through this series of steps, the intelligent agent can ensure that the problem to be analyzed meets the preset standards in terms of clarity and meaning, thereby improving the effectiveness of subsequent analysis. When the user inputs a problem, timely feedback and guidance can be obtained, ensuring that the finally formed standard problem to be analyzed is both clear and has practical operation value. Thus, the analysis errors caused by fuzzy or meaningless problems are effectively reduced, and the quality and efficiency of overall problem analysis are improved.

[0149] In a possible implementation, step S1 includes: step S11: obtaining the problem to be analyzed input by the user, the data related to the problem to be analyzed, and the background knowledge related to the problem to be analyzed;

[0150] Correspondingly, step S23 includes: step S231: generating corresponding hypotheses based on the causal graph, the preprocessed problem to be analyzed, the data, the background knowledge, and the memory bank, constructing corresponding analysis entities based on the hypotheses, the preprocessed problem to be analyzed, and the memory bank, and performing analysis and reasoning on the hypotheses and the preprocessed problem to be analyzed based on the analysis entities, to obtain and output insights.

[0151] That is to say, the intelligent agent can also obtain the background knowledge related to the problem to be analyzed. Such as Figure 2As shown, users can be divided into common users and expert users, and expert users can provide background knowledge related to the problem to be analyzed for the intelligent agent. The background knowledge includes knowledge of types such as domain, social, fact, personal, and tool. Thus, by comprehensively collecting information, the intelligent agent can ensure that it has sufficient context and data support when analyzing complex problems, thereby improving the accuracy and relevance of the analysis.

[0152] Now, the inference analysis logic for simple and complex problems to be analyzed will be specifically described. Refer to Figure 2 。

[0153] In a possible implementation, the preprocessed problem to be analyzed includes: M sequentially arranged sub-problems to be analyzed (corresponding to complex problems to be analyzed), where M is a positive integer greater than or equal to 2;

[0154] Step S231 includes:

[0155] Step S2311: Based on the causal graph, the first sub-problem to be analyzed among the M sequentially arranged sub-problems to be analyzed, background knowledge, data, and the memory bank, generate corresponding hypotheses; based on the hypotheses, the first sub-problem to be analyzed, and the memory bank, construct corresponding analysis entities; based on the analysis entities, analyze and reason about the hypotheses and the first sub-problem to be analyzed to obtain insights;

[0156] Step S2312: Based on the insights, the causal graph, the Nth sub-problem to be analyzed among the M sequentially arranged sub-problems to be analyzed, background knowledge, data, and the memory bank, generate corresponding new hypotheses; based on the new hypotheses, the Nth sub-problem to be analyzed, and the memory bank, construct corresponding new analysis entities; based on the new analysis entities, re-analyze and reason about the new hypotheses and the Nth sub-problem to be analyzed to obtain new insights;

[0157] Step S2313: Repeat Step S2312 until the value of N reaches M. Among them, when Step S2312 is executed for the first time, the corresponding value of N is 2. Each time Step S2312 is repeated, the value of N + 1, N ∈ [2, M], and N is a positive integer;

[0158] Step S2314: Output all the obtained insights.

[0159] Thus, by sequentially processing sub-problems, complex problems can be analyzed step by step in depth, and the analysis of each sub-problem is based on the previous insight, which can ensure the coherence of the analysis and reasoning process. Finally, all sub-tasks of the problem (M sequentially arranged sub-problems to be analyzed) can be covered, and comprehensive insights can be generated.

[0160] In a possible implementation, the number of problems to be analyzed after preprocessing is one (corresponding to a simple problem to be analyzed), and step S231 includes:

[0161] Step S231a: Generate corresponding hypotheses based on the causal graph, the problem to be analyzed after preprocessing, background knowledge, data, and the memory bank; construct corresponding analysis entities based on the hypotheses, the problem to be analyzed after preprocessing, and the memory bank; perform analysis and reasoning on the hypotheses and the problem to be analyzed after preprocessing based on the analysis entities to obtain insights;

[0162] Step S231b: Generate corresponding new hypotheses based on the insights, the causal graph, the problem to be analyzed after preprocessing, background knowledge, data, and the memory bank; construct corresponding new analysis entities based on the new hypotheses, the problem to be analyzed after preprocessing, and the memory bank; perform analysis and reasoning on the new hypotheses and the problem to be analyzed after preprocessing again based on the new analysis entities to obtain new insights;

[0163] Step S231c: Repeat step S231b until the number of executions of step S231b reaches a preset number, or until the total execution time of step S231b reaches a preset duration;

[0164] Step S231d: Output all the obtained insights.

[0165] Thus, through multiple iterations, the hypotheses and insights can be gradually optimized, the analysis accuracy can be gradually improved, and it supports dynamically adjusting the reasoning depth according to preset conditions (such as the number of times or time), and can avoid infinite iterations. Finally, high-quality insights can be quickly generated for simple problems.

[0166] It should be noted that in the execution of steps S2311 to S2314, or steps S231a to S231d, background knowledge may not be considered. It can be selected according to the actual application scenario.

[0167] In a possible implementation, all the obtained insights output include: summarizing all the insights and generating a structured insight report; outputting the insight report, where the types of the insight report include at least one of the following: basic insight, related insight, summary insight, and nested insight.

[0168] It should be noted that all the generated insights can be classified and sorted according to types, topics, or logical relationships, and the summarized and sorted insights can be organized into a structured report according to the logical levels. The report can be output in formats such as PDF, PPT, etc. (for reference, see Figure 2 ).

[0169] Moreover, the types of insight reports can be: Base Insight, which is the most fundamental and core insight directly extracted from data, such as "The total number of housing sales in California, USA in 2019 was X units."; Related Insights, which are other insights related to the Base Insight and are usually supplements or expansions of the Base Insight; Summary Insight is a comprehensive summary of the Base Insight and Related Insights, providing higher-level insights; Nested Insight is an insight that is further refined based on a certain insight and is usually an in-depth analysis of the insight.

[0170] In summary, by generating structured insight reports, the agent can transform complex data analysis results into knowledge that is understandable and actionable by users. The report types include Base Insight, Related Insights, Summary Insight, and Nested Insight, which provide core facts, supplementary information, a global perspective, and in-depth analysis respectively. This structured output method significantly improves the clarity, comprehensiveness, operability, and aesthetics of the report, providing users with more reliable data support and decision-making basis.

[0171] In a possible implementation, as Figure 6 shown, the thinking framework can continuously self-iterate based on the agent's autonomous learning process, memory bank, analysis and reasoning process, and user feedback on insights to obtain a new thinking framework.

[0172] It should be noted that the thinking framework is self-iterative, that is, self-growing. The thinking framework of the agent can achieve continuous growth through the closed-loop iteration of autonomous learning, memory bank, reasoning process, and user feedback. Autonomous learning can extract patterns from new data and rely on the memory bank (which stores historical knowledge and experience) to provide context support; the reasoning process can generate insights in combination with the knowledge in the memory bank and output them to the user; user feedback (such as satisfaction scores) can directly calibrate the confidence of the insight and trigger the dynamic update of the memory bank (for example: high-confidence results are precipitated as empirical knowledge, and low-scoring conclusions drive model retraining); the updated memory bank and optimized model can feed back to the new round of autonomous learning. Thus, in the process of continuously solving new problems, the agent can gradually improve the accuracy, adaptability, and decision-making efficiency of problem-solving, and through the continuous evolution of the thinking framework, the agent's ability to solve complex problems can be improved.

[0173] In the present invention, with the powerful computing power of the intelligent computing center, the intelligent agent can achieve the following: quickly process and analyze massive amounts of data, significantly improving the efficiency of problem analysis; quickly generate high-precision causal diagrams based on causal inference models and memory banks to reveal complex relationships in the data; perform complex multi-step reasoning and hypothesis verification to ensure the logical rigor of the analysis process and the credibility of the results; generate high-quality, interpretable insights to provide valuable conclusions and suggestions for users.

[0174] In summary, with the computing power of the intelligent computing center and combined with a structured thinking framework and a systematic analysis process, it is possible to more effectively handle complex problems and provide higher-quality insights. This improves the accuracy and efficiency of the intelligent agent in analyzing and solving complex problems.

[0175] Figure 7 There is shown an inference device based on a thinking framework realized through the computing power of an intelligent computing center. The device is applied in an intelligent agent and includes:

[0176] An acquisition module 701, configured to execute step S1: acquire the problem to be analyzed input by the user and the data related to the problem to be analyzed;

[0177] An execution module 702, configured to execute step S2: perform analysis and reasoning on the problem to be analyzed based on its own thinking framework and memory bank, and obtain and output insights, where the thinking framework is a structured way of thinking of the intelligent agent;

[0178] Among them, step S2 includes:

[0179] Step S21: After preprocessing the problem to be analyzed, obtain the preprocessed problem to be analyzed;

[0180] Step S22: Process the data based on the causal inference model and the memory bank to obtain the corresponding causal diagram;

[0181] Step S23: Generate corresponding hypotheses based on the causal diagram, the preprocessed problem to be analyzed, the data, and the memory bank, construct corresponding analysis entities based on the hypotheses, the preprocessed problem to be analyzed, and the memory bank, and perform analysis and reasoning on the hypotheses and the preprocessed problem to be analyzed based on the analysis entities to obtain and output insights.

[0182] In a possible implementation manner, the analysis entity includes: subspace, subdivision, and metric. The subspace is used to define the data object to be analyzed, the subdivision is used to define the related data objects of the dimensions related to the data object to be analyzed, and the metric is used to define the analysis metrics of the data object to be analyzed; the object to be analyzed and the related data objects are determined according to the hypotheses and the preprocessed problem to be analyzed;

[0183] Insights include: subspace, subdivision, metric, type, and score. The type is used to define the kind of insight, and the score is used to define the confidence level of the insight.

[0184] In a possible implementation, step S21 includes:

[0185] Step S211: Standardize the problem to be analyzed to obtain a standard problem to be analyzed;

[0186] Step S212: Judge the complexity of the standard problem to be analyzed. If the complexity of the standard problem to be analyzed reaches a preset complexity threshold, then determine the standard problem to be analyzed as a complex problem to be analyzed; if the complexity of the standard problem to be analyzed does not reach the complexity threshold, then determine the standard problem to be analyzed as a simple problem to be analyzed;

[0187] Step S213: When the standard problem to be analyzed is a complex problem to be analyzed, perform intention recognition on the complex problem to be analyzed. According to the result of the intention recognition, convert the format of the complex problem to be analyzed, disassemble the complex problem to be analyzed after format conversion into multiple sequentially arranged sub-problems to be analyzed, and determine the multiple sequentially arranged sub-problems to be analyzed as the preprocessed problems to be analyzed;

[0188] When the standard problem to be analyzed is a simple problem to be analyzed, perform intention recognition on the simple problem to be analyzed. According to the result of the intention recognition, convert the format of the simple problem to be analyzed, and determine the simple problem to be analyzed after format conversion as the preprocessed problem to be analyzed.

[0189] In a possible implementation, step S211 includes at least one of the following two items:

[0190] The first item includes:

[0191] Step S2111: Perform a first analysis on the current problem to be analyzed based on natural language processing technology to judge whether the clarity of the problem to be analyzed reaches a preset clarity threshold; if not, then execute step S2112;

[0192] Step S2112: Output a first prompt message, where the first prompt message is used to prompt the user to re-enter the problem to be analyzed;

[0193] Step S2113: Perform a first analysis on the problem to be analyzed re-entered by the user based on natural language processing technology to judge whether the clarity of the problem to be analyzed re-entered by the user reaches a preset clarity threshold. If not, repeat steps S2112 and S2113 until the clarity of the current problem to be analyzed reaches the preset clarity threshold;

[0194] Step S2114: Determine the current problem to be analyzed with clarity reaching the preset clarity threshold as the standard problem to be analyzed;

[0195] Or,

[0196] Step S2111’: Conduct a first analysis on the current problem to be analyzed based on natural language processing technology to determine whether the clarity of the problem to be analyzed reaches the preset clarity threshold; if not, execute Step S2112’;

[0197] Step S2112’: Based on the thinking framework and memory bank, interpret the problem to be analyzed, output the interpreted problem to be analyzed, and receive the feedback information input by the user; if the feedback information indicates that the interpreted problem to be analyzed is incorrect, repeat Step S2112’ until the feedback information indicates that the current interpreted problem to be analyzed is correct;

[0198] Step S2113’: Determine the interpreted problem to be analyzed corresponding to the feedback information indicating that the current interpreted problem to be analyzed is correct as the standard problem to be analyzed;

[0199] The second item includes:

[0200] Step S211a: Conduct a second analysis on the problem to be analyzed based on natural language processing technology to determine whether the problem to be analyzed is meaningful; if not, send a second prompt message for prompting the user that the problem to be analyzed is meaningless; where the problem to be analyzed being meaningful includes: the intention clarity of the problem to be analyzed reaches the preset intention clarity threshold, the operability of the problem to be analyzed reaches the operability threshold, and the problem to be analyzed conforms to common sense and logic, where the common sense and logic are stored in the memory bank;

[0201] Step S211b: Determine the meaningful problem to be analyzed as the standard problem to be analyzed.

[0202] In a possible implementation, Step S1 includes: Step S11: Obtain the problem to be analyzed input by the user, the data related to the problem to be analyzed, and the background knowledge related to the problem to be analyzed;

[0203] Step S23 includes: Step S231: Generate corresponding hypotheses based on the causal graph, the preprocessed problem to be analyzed, the data, the background knowledge, and the memory bank, construct corresponding analysis entities based on the hypotheses, the preprocessed problem to be analyzed, and the memory bank, and conduct analysis and reasoning on the hypotheses and the preprocessed problem to be analyzed based on the analysis entities to obtain and output insights.

[0204] In a possible implementation, the preprocessed problem to be analyzed includes: M serially arranged sub-problems to be analyzed, where M is a positive integer greater than or equal to 2;

[0205] Step S231 includes:

[0206] Step S2311: Generate a corresponding hypothesis based on the causal graph, the first sub-problem to be analyzed among the M sequentially arranged sub-problems to be analyzed, background knowledge, data, and the memory bank; construct a corresponding analysis entity based on the hypothesis, the first sub-problem to be analyzed, and the memory bank; perform analysis and reasoning on the hypothesis and the first sub-problem to be analyzed based on the analysis entity to obtain insights;

[0207] Step S2312: Generate a corresponding new hypothesis based on the insights, the causal graph, the Nth sub-problem to be analyzed among the M sequentially arranged sub-problems to be analyzed, background knowledge, data, and the memory bank; construct a corresponding new analysis entity based on the new hypothesis, the Nth sub-problem to be analyzed, and the memory bank; perform re-analysis and reasoning on the new hypothesis and the Nth sub-problem to be analyzed based on the new analysis entity to obtain new insights;

[0208] Step S2313: Repeat Step S2312 until the value of N reaches M, where when Step S2312 is executed for the first time, the corresponding value of N is 2, and each time Step S2312 is repeated, the value of N is incremented by 1, N ∈ [2, M], and N is a positive integer;

[0209] Step S2314: Output all the obtained insights.

[0210] In a possible implementation, the number of preprocessed sub-problems to be analyzed is one, and Step S23 includes:

[0211] Step S231a: Generate a corresponding hypothesis based on the causal graph, the preprocessed sub-problem to be analyzed, background knowledge, data, and the memory bank; construct a corresponding analysis entity based on the hypothesis, the preprocessed sub-problem to be analyzed, and the memory bank; perform analysis and reasoning on the hypothesis and the preprocessed sub-problem to be analyzed based on the analysis entity to obtain insights;

[0212] Step S231b: Generate a corresponding new hypothesis based on the insights, the causal graph, the preprocessed sub-problem to be analyzed, background knowledge, data, and the memory bank; construct a corresponding new analysis entity based on the new hypothesis, the preprocessed sub-problem to be analyzed, and the memory bank; perform re-analysis and reasoning on the new hypothesis and the preprocessed sub-problem to be analyzed based on the new analysis entity to obtain new insights;

[0213] Step S231c: Repeat Step S231b until the number of executions of Step S231b reaches a preset number of times, or until the total execution time of Step S231b reaches a preset duration;

[0214] Step S231d: Output all the obtained insights.

[0215] In a possible implementation, all the obtained insights include:

[0216] Summarize all the insights and generate a structured insight report;

[0217] Output the insight report, where the types of the insight report include at least one of the following: basic insight, related insight, summary insight, and nested insight.

[0218] In a possible implementation, the thinking framework can continuously self-iterate based on the autonomous learning process of the intelligent agent, the memory bank, the process of analytical reasoning, and the user's feedback on the insights to obtain a new thinking framework.

[0219] In the present invention, with the powerful computing power of the intelligent computing center, the intelligent agent can achieve the following: quickly process and analyze massive data, significantly improve the efficiency of problem analysis; quickly generate high-precision causal graphs based on the causal inference model and the memory bank to reveal the complex relationships in the data; perform complex multi-step reasoning and hypothesis verification to ensure the logical rigor of the analysis process and the credibility of the results; generate high-quality and interpretable insights to provide valuable conclusions and suggestions for users.

[0220] In summary, with the computing power of the intelligent computing center, combined with the structured thinking framework and the systematic analysis process, it is possible to more effectively handle complex problems and provide higher-quality insights. This improves the accuracy and efficiency of the intelligent agent in analyzing and solving complex problems.

[0221] Please refer to Figure 8 , the present invention also provides an electronic device 80, including a processor 801, a memory 802, and a computer program stored on the memory 802 and executable on the processor 801. When the computer program is executed by the processor 801, it implements the steps of the above-mentioned reasoning method based on the thinking framework realized by the computing power of the intelligent computing center and can achieve the same technical effects. To avoid repetition, it will not be elaborated here.

[0222] The present invention also provides a computer-readable storage medium. A computer program is stored on the computer-readable storage medium. When the computer program is executed by a processor, it implements the steps of the above-mentioned reasoning method based on the thinking framework realized by the computing power of the intelligent computing center and can achieve the same technical effects. To avoid repetition, it will not be elaborated here. Among them, the computer-readable storage medium is, for example, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disc, etc.

[0223] The present invention also provides a computer program product, including computer instructions which, when executed by a processor, implement the steps of the above-mentioned inference method based on a thinking framework by means of the computing power of an intelligent computing center, and can achieve the same technical effects. To avoid repetition, it will not be elaborated here.

[0224] It should be noted that in this document, the terms "include", "comprise" or any other variant thereof are intended to cover non-exclusive inclusion, such that a process, method, article or device including a series of elements not only includes those elements but also includes other elements not expressly listed, or further includes elements inherent to such process, method, article or device. Without further limitation, an element defined by the phrase "including a..." does not exclude the presence of additional identical elements in the process, method, article or device including such element.

[0225] Through the description of the above embodiments, those skilled in the art can clearly understand that the above method can be implemented by means of software plus a necessary general hardware platform. Of course, it can also be implemented by hardware, but in many cases the former is a better implementation. Based on such an understanding, the technical solution of the present invention, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disc) and includes several instructions for causing a terminal (which can be a mobile phone, computer, server, air conditioner, or network device, etc.) to execute the method described in the present invention.

[0226] The present invention has been described above in conjunction with the accompanying drawings, but the present invention is not limited to the above specific embodiments. The above specific embodiments are merely illustrative and not restrictive. Under the inspiration of the present invention, those of ordinary skill in the art can also make many forms without departing from the spirit and scope protected by the present invention and the claims, and all of them belong to the protection scope of the present invention.

Claims

1. A reasoning method based on a thinking framework is realized by using the computing power of an intelligent computing center, characterized in that: The method is applied to an intelligent agent, and the method comprises: Step S1: obtaining the problem to be analyzed input by the user and data related to the problem to be analyzed; Step S2: Analyze and reason the problem to be analyzed based on its own thinking framework and memory library to obtain and output insights, wherein the thinking framework is a structured way of thinking of the intelligent agent; Wherein, the step S2 comprises: Step S21: preprocessing the problem to be analyzed to obtain the preprocessed problem to be analyzed; Step S22: Processing the data based on the causal inference model and the memory library to obtain a corresponding causal graph; Step S23: Generate corresponding hypotheses based on the causal diagram, the preprocessed problem to be analyzed, the data and the memory bank, construct corresponding analysis entities based on the hypotheses, the preprocessed problem to be analyzed and the memory bank, analyze and reason on the hypotheses and the preprocessed problem to be analyzed based on the analysis entities, and obtain and output the insights.

2. The method according to claim 1, wherein the analysis entity comprises: Subspace, segmentation and metric, the subspace is used to define the data object to be analyzed, the segmentation is used to define the data objects related to the dimension related to the data object to be analyzed, and the metric is used to define the analysis index of the data object to be analyzed; The object to be analyzed and the related data object are determined according to the assumption and the preprocessed problem to be analyzed; The insight includes: the subspace, the segmentation, the metric, a type, and a score, wherein the type is used to define the type of the insight, and the score is used to define the confidence of the insight.

3. The method according to claim 1, characterized in that The step S21 comprises: Step S211: performing standardization processing on the problem to be analyzed to obtain a standard problem to be analyzed; Step S212: determining the complexity of the standard problem to be analyzed. If the complexity of the standard problem to be analyzed reaches a preset complexity threshold, the standard problem to be analyzed is determined as a complex problem to be analyzed; if the complexity of the standard problem to be analyzed does not reach the complexity threshold, the standard problem to be analyzed is determined as a simple problem to be analyzed. Step S213: when the standard problem to be analyzed is the complex problem to be analyzed, perform intent recognition on the complex problem to be analyzed, perform format conversion on the complex problem to be analyzed according to the result of intent recognition, decompose the complex problem to be analyzed after format conversion into a plurality of sequentially arranged sub-problems to be analyzed, and determine the plurality of sequentially arranged sub-problems to be analyzed as the preprocessed problem to be analyzed; When the standard problem to be analyzed is the simple problem to be analyzed, intent recognition is performed on the simple problem to be analyzed, and according to the result of intent recognition, the simple problem to be analyzed is format converted, and the simple problem to be analyzed after format conversion is determined as the preprocessed problem to be analyzed.

4. The method according to claim 3, characterized in that The step S211 includes at least one of the following two items: The first includes: Step S2111: performing a first analysis on the current question to be analyzed based on natural language processing technology to determine whether the clarity of the question to be analyzed reaches a preset clarity threshold; if not, executing step S2112; Step S2112: outputting first prompt information, wherein the first prompt information is used to prompt the user to re-enter the question to be analyzed; Step S2113: performing a first analysis on the question to be analyzed re-entered by the user based on the natural language processing technology to determine whether the clarity of the question to be analyzed re-entered by the user reaches the preset clarity threshold; if not, repeatedly executing steps S2112 and S2113 until the clarity of the current question to be analyzed reaches the preset clarity threshold; Step S2114: determining the current problem to be analyzed whose clarity reaches the preset clarity threshold as the standard problem to be analyzed; or, Step S2111': performing a first analysis on the current question to be analyzed based on natural language processing technology to determine whether the clarity of the question to be analyzed reaches a preset clarity threshold; if not, executing step S2112'; Step S2112': based on the thinking framework and the memory bank, interpret the problem to be analyzed, output the interpreted problem to be analyzed, and receive feedback information input by the user; if the feedback information is used to indicate that the interpreted problem to be analyzed is incorrect, then repeat the step S2112' until the feedback information is used to indicate that the currently interpreted problem to be analyzed is correct; Step S2113': determining the interpreted problem to be analyzed corresponding to the feedback information indicating that the currently interpreted problem to be analyzed is correct as the standard problem to be analyzed; The second item includes: Step S211a: Perform a second analysis on the problem to be analyzed based on natural language processing technology to determine whether the problem to be analyzed is meaningful. If not, issue a second prompt message, wherein the second prompt message is used to prompt the user that the problem to be analyzed is meaningless; wherein the problem to be analyzed is meaningful includes: the intention clarity of the problem to be analyzed reaches a preset intention clarity threshold, the operability of the problem to be analyzed reaches an operability threshold, and the problem to be analyzed is consistent with common sense and logic, wherein the common sense and the logic are stored in the memory bank; Step S211b: Determine the meaningful problem to be analyzed as the standard problem to be analyzed.

5. The method according to claim 1, characterized in that The step S1 comprises: step S11: obtaining the problem to be analyzed input by the user, data related to the problem to be analyzed, and background knowledge related to the problem to be analyzed; The step S23 includes: step S231: generating corresponding hypotheses based on the causal diagram, the preprocessed problem to be analyzed, the data, the background knowledge and the memory base, constructing corresponding analysis entities based on the hypotheses, the preprocessed problem to be analyzed and the memory base, analyzing and reasoning the hypotheses and the preprocessed problem to be analyzed based on the analysis entities, and obtaining and outputting the insights.

6. The method according to claim 5, characterized in that The preprocessed problem to be analyzed includes: M sequentially arranged sub-problems to be analyzed, where M is a positive integer greater than or equal to 2; The step S231 includes: Step S2311: Generate a corresponding hypothesis based on the causal graph, the first sub-problem to be analyzed among the M sequentially arranged sub-problems to be analyzed, the background knowledge, the data and the memory bank; construct a corresponding analysis entity based on the hypothesis, the first sub-problem to be analyzed and the memory bank; perform analysis and reasoning on the hypothesis and the first sub-problem to be analyzed based on the analysis entity to obtain insights; Step S2312: Generate a corresponding new hypothesis based on the insight, the causal diagram, the Nth sub-problem to be analyzed among the M sequentially arranged sub-problems to be analyzed, the background knowledge, the data and the memory bank; construct a corresponding new analysis entity based on the new hypothesis, the Nth sub-problem to be analyzed and the memory bank; re-analyze and reason the new hypothesis and the Nth sub-problem to be analyzed based on the new analysis entity to obtain a new insight; Step S2313: Repeat step S2312 until the value of N reaches M, wherein when step S2312 is executed for the first time, the corresponding value of N is 2, and each time step S2312 is repeated, the value of N increases by 1, N∈[2, M], where N is a positive integer; Step S2314: Output all insights obtained.

7. The method according to claim 5, characterized in that The number of the pre-processed problem to be analyzed is one, and the step S231 includes: Step S231a: Generate a corresponding hypothesis based on the causal diagram, the preprocessed problem to be analyzed, the background knowledge, the data and the memory bank; construct a corresponding analysis entity based on the hypothesis, the preprocessed problem to be analyzed and the memory bank; perform analysis and reasoning on the hypothesis and the preprocessed problem to be analyzed based on the analysis entity to obtain insights; Step S231b: Generate a corresponding new hypothesis based on the insight, the causal diagram, the preprocessed problem to be analyzed, the background knowledge, the data and the memory bank, and construct a corresponding new analysis entity based on the new hypothesis, the preprocessed problem to be analyzed and the memory bank; re-analyze and reason the new hypothesis and the preprocessed problem to be analyzed based on the new analysis entity to obtain a new insight; Step S231c: Repeat step S231b until the number of executions of step S231b reaches a preset number of times, or until the total execution time of step S231b reaches a preset duration; Step S231d: Output all insights obtained.

8. The method according to claim 6 or 7, characterized in that: The output provides all the insights including: Summarize all insights and generate structured insight reports; The insight report is output, wherein the type of the insight report includes at least one of the following: basic insight, related insight, summary insight, and nested insight.

9. The method according to claim 1, characterized in that: The thinking framework can continuously iterate itself based on the autonomous learning process of the intelligent agent, the memory bank, the analytical reasoning process and the user's feedback on the insight to obtain a new thinking framework.

10. A reasoning device based on a thinking framework is realized by using the computing power of an intelligent computing center, characterized in that: The device is applied to an intelligent body, and the device comprises: The acquisition module is used to execute step S1: acquire the problem to be analyzed input by the user and data related to the problem to be analyzed; An execution module, configured to execute step S2: analyzing and reasoning the problem to be analyzed based on its own thinking framework and memory library to obtain and output insights, wherein the thinking framework is a structured way of thinking of the intelligent agent; Wherein, the step S2 comprises: Step S21: preprocessing the problem to be analyzed to obtain the preprocessed problem to be analyzed; Step S22: Processing the data based on the causal inference model and the memory library to obtain a corresponding causal graph; Step S23: Generate corresponding hypotheses based on the causal diagram, the preprocessed problem to be analyzed, the data and the memory bank, construct corresponding analysis entities based on the hypotheses, the preprocessed problem to be analyzed and the memory bank, analyze and reason on the hypotheses and the preprocessed problem to be analyzed based on the analysis entities, and obtain and output the insights.

11. A server, characterized in that: include: A processor, a memory, and a program stored in the memory and executable on the processor, wherein the program, when executed by the processor, implements the steps of a reasoning method based on a thinking framework through the computing power of an intelligent computing center as described in any one of claims 1 to 9.

12. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps of a reasoning method based on a thinking framework through the computing power of an intelligent computing center as described in any one of claims 1 to 9.

13. A computer program product, characterized in that It includes computer instructions, which, when executed by a processor, implement the steps of a reasoning method based on a thinking framework through the computing power of an intelligent computing center as described in any one of claims 1-9.

Images