A general artificial intelligence based text content analysis and inference system

Abstract

A general artificial intelligence-based text content analysis and reasoning system comprises an examination engine unit, a text analysis unit, a rule analysis unit, a reasoning unit and a thread management unit, the examination engine unit establishes a preprocessing process and a rule acquisition process, and when performing an examination task, the examination engine unit quantifies the task quantity of the last moment or the last period, obtains a quantization mark, and then allocates the resource occupation of the preprocessing process and the rule acquisition process according to the quantization mark; the text analysis unit pre-processes a to-be-processed text to obtain a data signal; the rule analysis unit crawls basic rules, and performs semantic analysis on the basic rules to obtain reasoning rules; the reasoning unit inputs the data signal and the vector corresponding to the reasoning rules into a reasoning neural network to perform logical calculation and obtain a reasoning result; and the thread management unit manages the threads in the preprocessing process and the rule acquisition process. The present application reasonably allocates resources and improves the task execution speed.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence technology, and more specifically, to a general artificial intelligence-based text content analysis and reasoning system. Background Art

[0002] With the continuous updating of laws, regulations and industry standards, enterprises need to ensure that the laws and regulatory documents involved in their operations comply with the requirements of relevant laws, regulations and / or internal policies. For example, whether the contract complies with the requirements of relevant laws, regulations and internal policies; whether the established processes comply with the requirements of internal policies; and whether the formulated rules and regulations comply with relevant laws and regulations.

[0003] Traditional compliance reviews of legal and regulatory documents rely mainly on manual work. Reviewers need to review relevant laws, regulations, or internal policy documents and compare them with them. This is inefficient, costly, and prone to errors, making it difficult to cope with complex legal environments and rapidly changing compliance requirements.

[0004] With the development of information technology, although there are some simple text retrieval tools to assist in review, they are unable to achieve in-depth semantic analysis and intelligent judgment, and it is difficult to meet the growing needs of institutional compliance review.

[0005] Therefore, there is an urgent need for an intelligent text content analysis and reasoning assistance system. Summary of the Invention

[0006] (1) Purpose of the Invention: To address the problems existing in the above-mentioned prior art, the purpose of the present invention is to provide a general artificial intelligence-based text content analysis and reasoning system that can automatically identify, analyze and evaluate the compliance of institutional texts and provide improvement suggestions.

[0007] (II) Technical Solution: To solve the above technical problems, this technical solution provides a general artificial intelligence-based text content analysis and reasoning system, including a review engine unit, a text analysis unit, a rule analysis unit, a reasoning unit, and a thread management unit.

[0008] The review engine unit allocates the resource usage of each unit when performing the review task, and establishes a preprocessing process and a rule acquisition process; the text analysis unit is used to preprocess the text to be processed to obtain a data signal; the rule analysis unit is used to crawl basic rules and perform semantic analysis on the basic rules to obtain inference rules; the inference unit inputs the vectors corresponding to the data signal and the inference rule into the inference neural network, performs logical calculations, and obtains the inference results; the thread management unit manages the threads in the preprocessing process and the rule acquisition process respectively;

[0009] When the review engine unit performs the review task, it quantifies the task amount at the previous moment or the previous cycle to obtain a quantification mark, and then allocates the resource occupancy of the preprocessing process and the rule acquisition process according to the quantification mark.

[0010] The general artificial intelligence-based text content analysis and reasoning system, wherein preprocessing of the text to be processed includes semantic analysis of the text to be processed, extracting key data of the text; and arranging and formatting the key data to obtain the data signal.

[0011] The general artificial intelligence-based text content analysis and reasoning system, wherein the rule analysis unit converts each reasoning rule into a vector according to a domain-specific language, each reasoning rule corresponds to a vector, and all vectors constitute a rule vector space.

[0012] The general artificial intelligence-based text content analysis and reasoning system, wherein the reasoning unit constructs a knowledge graph based on the association relationship between entities in the reasoning results of the same text to be processed.

[0013] The general artificial intelligence-based text content analysis and reasoning system, wherein the display unit converts the knowledge graph into a target report through a visualization engine according to the input unit, and displays the target report.

[0014] The general text content analysis and reasoning system based on artificial intelligence, wherein the review engine unit allocates the resource occupancy of the preprocessing process and the rule acquisition process, specifically including:

[0015] The review engine unit obtains the amount of tasks that need to be processed at the last moment or the last cycle;

[0016] The review engine unit quantifies the amount of tasks that need to be processed at the last moment or in the last cycle to obtain a quantitative mark;

[0017] The review engine unit calculates the number of quantization marks at the previous moment or the previous cycle to obtain a quantization value;

[0018] The review engine unit calculates the resource occupancy ratio of the rule analysis unit at the last moment or the last cycle according to the quantified value at the last moment or the last cycle;

[0019] The review engine unit allocates resource occupancy of the rule analysis unit and the rule analysis unit at a current moment or a current period according to the rule analysis unit and the resource occupancy ratio of the rule analysis unit at a previous moment or a previous period.

[0020] The general artificial intelligence-based text content analysis and reasoning system, wherein the review engine unit quantifies the amount of tasks that need to be processed at the previous moment or the previous cycle to obtain a quantitative mark. Specifically, the review engine unit quantifies the number of texts to be processed and the memory usage at the previous moment or the previous cycle to obtain a quantitative mark of the texts to be processed; the review engine unit quantifies the number of basic rules to be processed and the memory usage at the previous moment or the previous cycle to obtain a quantitative mark of the basic rules.

[0021] A general artificial intelligence-based text content analysis and reasoning system, wherein the review engine unit calculates the number of quantitative marks at the previous moment or the previous period to obtain a quantitative value. Specifically, the review engine unit calculates the number of quantitative marks of the text expected to be processed at the previous moment or the previous week to obtain a first quantitative value; the review engine unit calculates the number of quantitative marks of the basic rules expected to be processed at the previous moment or the previous week to obtain a second quantitative value.

[0022] The general artificial intelligence-based text content analysis and reasoning system, wherein the review engine unit quantifies the amount of tasks that need to be processed at the previous moment or the previous cycle to obtain a quantitative mark, specifically includes: the review engine unit generates the same number of first quantitative marks as the number of tasks that need to be processed at the previous moment or the previous cycle; the review engine unit divides the size of each task that needs to be processed at the previous moment or the previous cycle by a first quantitative threshold to obtain a quantitative quotient value; the review engine unit generates a target second quantitative mark based on the quantitative quotient value; the first quantitative mark and the second quantitative mark constitute the quantitative mark.

[0023] The general artificial intelligence-based text content analysis and reasoning system, wherein when the quantization quotient value of each task is less than 1, the number of the second quantization mark is zero, and in this case, the second quantization mark is zero;

[0024] When the quantization quotient values ​​of all tasks have a quantization quotient value greater than 1, the number of the second quantization marks is greater than zero: when the quantization quotient value of a task is an integer greater than 1, the review engine unit generates a second quantization mark equal to the quantization quotient value of the task minus 1, which is the second quantization mark of the task; when the quantization quotient value of a task is a non-integer greater than 1, the review engine unit generates a second quantization mark equal to the quantization quotient value of the task, which is the second quantization mark of the task;

[0025] The second quantized labels of all tasks constitute the target second quantized label.

[0026] (3) Beneficial effects: The present invention provides a general text content analysis and reasoning system based on artificial intelligence, which executes the preprocessing task and rule acquisition task of the text to be processed through two relatively independent processes, and allocates the resource occupancy at the current moment or the current cycle according to the task volume of the preprocessing task and the rule acquisition task at the previous moment or the previous cycle, so that the resource allocation of the preprocessing process and the rule acquisition process is more reasonable, and the task execution speed is improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] Figure 1 This is a schematic diagram of the steps of allocating resources between the pre-processing process and the rule acquisition process by the review engine unit of a general artificial intelligence-based text content analysis and reasoning system of the present invention;

[0028] Figure 2 This is a schematic diagram of partitions of coordinates to be assigned in a general artificial intelligence-based text content analysis and reasoning system of the present invention. DETAILED DESCRIPTION

[0029] The present invention is further described in detail below in conjunction with preferred embodiments. More details are set forth in the following description to facilitate a full understanding of the present invention. However, the present invention can obviously be implemented in a variety of other ways different from the description. Those skilled in the art can make similar generalizations and deductions based on actual application situations without violating the connotation of the present invention. Therefore, the scope of protection of the present invention should not be limited by the content of this specific embodiment.

[0030] The accompanying drawings are schematic diagrams of embodiments of the present invention. It should be noted that the drawings are merely examples and are not drawn to scale, and should not be used to limit the actual scope of protection claimed in the present invention.

[0031] A general AI-based text content analysis and reasoning system is used to analyze and reason on the text to be processed, review the compliance of the processed text, and complete the review task. A general AI-based text content analysis and reasoning system includes a review engine unit, a text analysis unit, a rule analysis unit, an inference unit, a display unit, and a thread management unit.

[0032] The review engine unit is deployed on the underlying operating platform and performs core scheduling for the system. Specifically, it allocates resources to various units when performing review tasks. These units include a text analysis unit, a rule analysis unit, an inference unit, and a display unit. The underlying operating platform includes desktop operating systems and mobile operating systems.

[0033] The text analysis unit performs a preprocessing task on the to-be-processed text, that is, the text analysis unit is used to preprocess the to-be-processed text to obtain a data signal. Preprocessing the to-be-processed text includes performing semantic analysis on the to-be-processed text to extract key data of the text; and after collating the key data, format conversion is performed to obtain the data signal. The key data refers to variable data in the to-be-processed text, for example, when the to-be-processed text is a contract, the key data refers to specific clauses or specific numerical values in the clauses or clauses; when the to-be-processed text is business data, the key data refers to core data in specific business data. The key data can be obtained by performing semantic analysis on the to-be-processed text, or can be directly obtained from the to-be-processed text, which is not specifically shown here. The data signal is the variable information required in the reasoning unit, and its format is a format that can be recognized by the reasoning unit when performing logical calculation. Semantic analysis on the to-be-processed text can be performed by artificial intelligence, such as DeepSeek. The collation of the key data includes preliminary processing such as calculation and sorting of the key data.

[0034] The rule analysis unit performs a rule acquisition task, that is, the rule analysis unit is used to crawl basic rules and perform semantic analysis on the basic rules to obtain reasoning rules. The basic rules include text logical rules, calculation rules, business analysis rules, and other legal regulations or industry rules. The rule analysis unit performs semantic analysis on the basic rules to obtain reasoning rules, which can be specifically performed by artificial intelligence, such as DeepSeek.

[0035] The reasoning rules can also be input by the input unit.

[0036] The rule analysis unit stores the reasoning rules in a rule library, and the rule library includes reasoning rules of different industries and different classifications. When the rule analysis unit performs semantic analysis on the basic rules, the obtained reasoning rules are compared with the reasoning in the rule library, and the reasoning rules in the rule library that conflict or are not complete are replaced, so as to ensure that the reasoning rules in the rule library comply with the latest legal regulations.

[0037] The rule analysis unit can also convert each reasoning rule into a vector according to DSL (Domain Specific Language), and each reasoning rule corresponds to a vector, and all vectors form a vector space of rules.

[0038] The reasoning unit inputs the data signal and the vector corresponding to the reasoning rule into a reasoning neural network to perform logical calculation and obtain a reasoning result. The reasoning neural network takes the vector corresponding to the reasoning rule as a calculation function and takes the data signal as an independent variable to perform calculation, and the calculation result is the reasoning result.

[0039] The reasoning unit can also construct a knowledge graph based on the reasoning results of the same text to be processed. The reasoning unit constructs the knowledge graph based on the association relationship between entities in the reasoning results of the same text to be processed.

[0040] The display unit converts the knowledge graph into a target report through a visualization engine according to the output report format selected by the input unit, and displays the target report. The target report includes multiple formats, such as text format, presentation format, chart format, etc.

[0041] The review engine unit supports calling system APIs and allocates system resources for the preprocessing tasks of the text analysis unit and the rule acquisition tasks of the rule analysis unit. It establishes a process corresponding to the preprocessing tasks of the text analysis unit, the preprocessing process, and a process corresponding to the rule analysis unit's rule tasks, the rule acquisition process. The preprocessing process and the rule acquisition process are two independent processes in a general AI-based text content analysis and reasoning system.

[0042] The preprocessing process executes the text analysis unit to preprocess multiple to-be-processed texts at the same time. The preprocessing of each to-be-processed text is performed by a different thread in the preprocessing process, thereby achieving the preprocessing of each to-be-processed text separately.

[0043] The rule acquisition process executes the rule analysis unit to perform semantic analysis on multiple basic rules at the same time. The semantic analysis of each basic rule is performed by a different thread in the rule acquisition process, thereby achieving semantic analysis of each basic rule separately.

[0044] The thread management unit manages the threads in the preprocessing process and the rule acquisition process respectively, including: occupying and allocating current process resources for each thread in the preprocessing process and the rule acquisition process.

[0045] The thread management unit allocates resource occupancy of each thread in the process according to the number of tasks that need to be processed in the process. Specifically, the thread management unit manages each thread in the preprocessing process according to the number of pending texts that need to be processed by the text analysis unit, and allocates each thread's use of current process resources; the thread management unit manages each thread in the rule acquisition process according to the number of basic rules that need to be processed by the rule analysis unit, and allocates each thread's use of current process resources.

[0046] The thread management unit allocates resource occupancy to each thread in the process, specifically, allocation of CPU or kernel resource occupancy.

[0047] When the review engine unit performs the review task, it allocates the rule analysis unit and the resource occupancy of the rule analysis unit, that is, it allocates the resources of the preprocessing process and the rule acquisition process. When allocating the CPU or kernel resource occupancy of the preprocessing process and the rule acquisition process, the review engine unit quantifies the amount of tasks that need to be processed in different time periods, and then allocates the resource occupancy according to the quantification of the amount of tasks. Specifically, the amount of tasks at the previous moment or the previous cycle is quantified, and after obtaining the quantification mark, the resource occupancy is allocated according to the quantification mark. The same moment of the previous cycle refers to the previous cycle adjacent to the cycle of the current moment.

[0048] like Figure 1 As shown, the specific implementation process is as follows:

[0049] The review engine unit obtains the amount of tasks to be processed at the previous moment or cycle. Specifically, the review engine unit obtains the number of documents to be processed and the size of each document to be processed at the previous moment or cycle. The review engine unit also obtains the number of basic rules to be processed and the size of each basic rule to be processed at the previous moment or cycle. The size of the documents to be processed refers to the amount of storage space occupied by the documents to be processed, and the size of the basic rules to be processed refers to the amount of storage space occupied by the basic rules to be processed.

[0050] The review engine unit quantifies the amount of tasks that need to be processed at the previous moment or cycle to obtain a quantitative mark. Specifically, the review engine unit quantifies the amount of text to be processed and the memory usage at the previous moment or cycle to obtain a quantitative mark of the text to be processed; the review engine unit quantifies the amount of basic rules to be processed and the memory usage at the previous moment or cycle to obtain a quantitative mark of the basic rules.

[0051] The review engine unit calculates the number of quantization marks at the previous moment or the previous period to obtain a quantization value. Specifically, the review engine unit calculates the number of quantization marks of the text expected to be processed at the previous moment or the previous week to obtain a first quantization value; the review engine unit calculates the number of quantization marks of the basic rule expected to be processed at the previous moment or the previous week to obtain a second quantization value.

[0052] The review engine unit calculates the resource usage ratio of the rule analysis unit and the rule analysis unit at the previous moment or the previous cycle based on the quantized value at the previous moment or the previous cycle. Specifically, the review engine unit calculates the ratio of the first quantized value to the second quantized value, which is the resource usage ratio of the rule analysis unit and the rule analysis unit.

[0053] The review engine unit allocates resource occupancy of the rule analysis unit and the rule analysis unit at a current moment or a current period according to the rule analysis unit and the resource occupancy ratio of the rule analysis unit at a previous moment or a previous period.

[0054] The review engine unit includes a first quantization threshold.

[0055] The review engine unit quantifies the amount of tasks that need to be processed at the previous moment or the previous cycle to obtain a quantization mark, specifically including: the review engine unit generates the same number of first quantization marks as the number of tasks that need to be processed at the previous moment or the previous cycle; the review engine unit divides the size of each task that needs to be processed at the previous moment or the previous cycle by a first quantization threshold to obtain a quantization quotient value; the review engine unit generates a target second quantization mark based on the quantization quotient value; the first quantization mark and the second quantization mark constitute the quantization mark.

[0056] When the quantization quotient value of each task is less than 1, the number of second quantization marks is zero. In this case, the number of second quantization marks is zero. When the quantization quotient value of all tasks is greater than 1, the number of second quantization marks is greater than zero. When the quantization quotient value of a task is an integer greater than 1, the review engine unit generates a second quantization mark equal to the quantization quotient value of the task minus 1, which is the second quantization mark of the task. When the quantization quotient value of a task is a non-integer greater than 1, the review engine unit generates a second quantization mark equal to the quantization quotient value of the task, which is the second quantization mark of the task. The second quantization marks of all tasks constitute the target second quantization mark.

[0057] The quantization mark can be a bar graph or other image representing quantity, which is not specifically limited here. The number of the first quantization mark is the same as the number of tasks that need to be processed at the previous moment or the previous cycle. Taking the size of the task that needs to be processed at the previous moment or the previous cycle as the quotient of the first quantization threshold means taking the size of each task that needs to be processed at the previous moment or the previous cycle as the dividend and the first quantization threshold as the divisor to perform a division operation. The size of each task refers to the memory usage of each task.

[0058] Specifically, the review engine unit quantifies the number of texts to be processed and the memory usage at the previous moment or the previous cycle to obtain a quantitative mark of the texts to be processed; the review engine unit quantifies the number of basic rules to be processed and the memory usage at the previous moment or the previous cycle to obtain a quantitative mark of the basic rules.

[0059] The review engine unit generates the same number of first quantization marks of texts as the number of texts to be processed that need to be processed at the previous moment or the previous cycle; the review engine unit divides the size of each text to be processed that needs to be processed at the previous moment or the previous cycle by the first quantization threshold to obtain a text quantization quotient value; the review engine unit generates a second quantization mark of the target text based on the text quantization quotient value; the first quantization mark of the text and the second quantization mark of the target text constitute the quantization mark of the text to be processed.

[0060] When the text quantization quotient of each to-be-processed text is less than 1, the number of the second quantization marks of the text is zero, and the second quantization marks of the text are zero. When the text quantization quotient of all the to-be-processed texts has a text quantization quotient greater than 1, the number of the second quantization marks of the text is greater than zero: when the text quantization quotient of the to-be-processed text is an integer greater than 1, the review engine unit generates the text quantization quotient of the to-be-processed text minus 1, which is the second quantization mark of the text to be processed; when the text quantization quotient of the to-be-processed text is a non-integer greater than 1, the review engine unit generates the text quantization quotient of the to-be-processed text as many as 1, which is the second quantization mark of the text to be processed. The second quantization marks of all the to-be-processed texts constitute the second quantization mark of the target text.

[0061] The review engine unit generates the same number of first quantization marks for rules according to the number of pending basic rules that need to be processed at the previous moment or the previous cycle; the review engine unit divides the size of each pending basic rule that needs to be processed at the previous moment or the previous cycle by the first quantization threshold to obtain a rule quantization quotient value; the review engine unit generates a second quantization mark for the target rule based on the rule quantization quotient value; the first quantization mark for the rule and the second quantization mark for the target rule constitute the quantization mark of the basic rule.

[0062] When the rule quantization quotient value of each basic rule to be processed is less than 1, the number of rule second quantization marks is zero, and the rule second quantization mark is zero. When the rule quantization quotient value of all basic rules to be processed has a rule quantization quotient value greater than 1, the number of rule second quantization marks is greater than zero: when the rule quantization quotient value of the basic rule to be processed is an integer greater than 1, the review engine unit generates a rule second quantization mark of the rule quantization quotient value of the basic rule to be processed minus 1, which is the rule second quantization mark of the basic rule to be processed; when the rule quantization quotient value of the basic rule to be processed is a non-integer greater than 1, the review engine unit generates a rule second quantization mark of the rule quantization quotient value of the basic rule to be processed, which is the rule second quantization mark of the basic rule to be processed. The rule second quantization marks of all basic rules to be processed constitute the target rule second quantization mark.

[0063] When the thread management unit allocates resource occupancy for each thread in a process, it allocates resource occupancy for each thread in the preprocessing process and the rule acquisition process. Resource occupancy for each thread refers to allocation of thread running resources, including but not limited to registers, stack space, etc.

[0064] Specifically include:

[0065] The thread management unit divides a buffer pool for each process, and the buffer pool is used to store threads in the corresponding process, and each thread corresponds to a preprocessing task or a rule acquisition task.

[0066] The thread management unit creates a thread for each task in the process cache pool and allocates running resources to the created thread. After obtaining the running resources, each thread executes the thread task.

[0067] After the thread completes its task, the thread management unit destroys the thread and reclaims the running resources.

[0068] Each thread in the cache pool includes corresponding task data, which refers to the time when the thread's corresponding task enters the cache pool, the memory usage of the task file, etc.

[0069] When the thread management unit allocates running resources to the created threads, it allocates running resources to the created threads by coordinate partitioning. Specifically, the following steps are included:

[0070] Establish a thread resource allocation coordinate system. The thread resource allocation coordinate system is a two-dimensional coordinate system, with the axes representing the time when the thread's corresponding task enters the cache pool and the memory usage of the task file, respectively. This example uses the x-axis representing the time when the thread's corresponding task enters the cache pool, and the y-axis representing the memory usage of the task file, as an example.

[0071] Based on the task data corresponding to each thread, threads created in the cache pool are marked in the thread resource coordinate system to obtain a coordinate system to be allocated. Each thread corresponds to a coordinate point in the thread resource coordinate system. The allocation coordinate system includes the thread resource coordinate system and the coordinate points corresponding to all threads of the process.

[0072] The coordinate system to be allocated is hierarchically partitioned, where all partitions include all coordinate points, or in other words, the area formed by all partitions covers all coordinate points in the coordinate system to be allocated. The hierarchical partitions include at least three, and here, three hierarchical partitions are used as an example, specifically including a first partition, a second partition, and a third partition. The area formed by the first partition, the second partition, and the third partition covers all coordinate points in the coordinate system to be allocated.

[0073] According to the partition where the coordinate point is located, running resource allocation is performed for the thread corresponding to the coordinate point.

[0074] As shown in Figure 2 the first partition is a "「」" shaped partition composed of a one-one partition and a one-two partition. The one-one partition has a range on the x-axis that is greater than or equal to the minimum entering time of all task entering cache pool times and less than or equal to the maximum entering time of all task entering cache pool times; the one-one partition has a range on the y-axis that is less than or equal to the maximum memory occupation of all task file memory occupations and greater than or equal to the first memory threshold. The one-two partition has a range on the x-axis that is greater than or equal to the minimum entering time of all task entering cache pool times and less than or equal to the first time threshold; the one-two partition has a range on the y-axis that is less than or equal to the first memory threshold and greater than or equal to the minimum memory occupation of all task file memory occupations.

[0075] The second partition is a "「」" shaped partition composed of a two-one partition and a two-two partition. The two-one partition has a range on the x-axis that is greater than the first time threshold and less than or equal to the maximum entering time of all task entering cache pool times; the two-one partition has a range on the y-axis that is less than the first memory threshold and greater than or equal to the second memory threshold. The two-two partition has a range on the x-axis that is greater than the first time threshold and less than or equal to the second time threshold; the two-two partition has a range on the y-axis that is less than the first memory threshold and greater than or equal to the minimum memory occupation of all task file memory occupations.

[0076] The third partition is a "|" shaped partition. The third partition has a range on the x-axis that is greater than the second time threshold and less than or equal to the maximum entering time of all task entering cache pool times; the third partition has a range on the y-axis that is less than the second memory threshold and greater than or equal to the maximum memory occupation of all task file memory occupations.

[0077] The difference between the maximum memory occupation and the first memory threshold, the difference between the first memory threshold and the second memory threshold, and the difference between the second memory threshold and the minimum memory occupation are equal.

[0078] The difference between the minimum entering time and the first time threshold, the difference between the first time threshold and the second time threshold, and the difference between the second time threshold and the maximum entering time are equal.

[0079] When running resource allocation is performed for the thread corresponding to the coordinate point according to the partition where the coordinate point is located, the running resource allocation is performed in the order of the first partition, the second partition, and the third partition according to the running resource required by the thread corresponding to the coordinate point.

[0080] Specifically,

[0081] The thread management unit calculates the runtime resources required by the threads corresponding to the coordinate points of the first partition, the second partition, and the third partition, respectively. When the sum of the runtime resources required by the threads corresponding to the coordinate points of the first partition, the second partition, and the third partition is less than or equal to the runtime resources of the current process, the thread management unit simultaneously allocates runtime resources to the threads corresponding to the coordinate points of the first partition, the second partition, and the third partition.

[0082] When the sum of the running resources required by the threads corresponding to the coordinate points of the first partition, the second partition, and the third partition is greater than the running resources of the current process:

[0083] The sum of the running resources required by the threads corresponding to the coordinate points of the first partition and the second partition is calculated. When the sum of the running resources required by the threads corresponding to the coordinate points of the first partition and the second partition is less than or equal to the running resources of the current process, the thread management unit simultaneously allocates running resources to the threads corresponding to the coordinate points of the first partition and the second partition; after the thread management unit completes the allocation of running resources to the threads corresponding to the coordinate points of the first partition and the second partition, it allocates running resources to the threads corresponding to the coordinate points of the third partition in descending order of the y coordinates of the coordinate points until the remaining running resources of the process are less than the running resources required by the threads corresponding to all the coordinate points.

[0084] When the sum of the running resources required by the threads corresponding to the coordinate points of the first and second partitions is greater than the running resources of the current process:

[0085] The sum of the running resources required by the threads corresponding to the coordinate points of the first partition is calculated. When the sum of the running resources required by the threads corresponding to the coordinate points of the first partition is less than or equal to the running resources of the current process, the thread management unit simultaneously allocates running resources to the threads corresponding to the coordinate points of the first partition. After the thread management unit completes the allocation of running resources to the threads corresponding to the coordinate points of the first partition, the thread management unit allocates running resources to the threads corresponding to the coordinate points of the second partition and the third partition in the order of the y coordinates of the coordinate points from large to small and from the second partition to the third partition, until the remaining running resources of the process are less than the running resources required by the threads corresponding to all the coordinate points.

[0086] When the total running resources required by the threads corresponding to the first partition coordinate point are greater than the running resources of the current process:

[0087] The thread management unit allocates running resources to the threads corresponding to the coordinate points of the first partition, the second partition, and the third partition in the order of the y coordinates of the coordinate points from large to small, the x coordinates from small to large, and from the first partition to the second partition, and from the second partition to the third partition, until the remaining running resources of the process are less than the running resources required by the threads corresponding to all the coordinate points.

[0088] By allocating resources to threads within each process in this manner, we ensure that sufficient resources are available to ensure thread execution. This planning improves thread execution speed and significantly increases system speed. Furthermore, by allocating resources to threads based on their availability, we ensure the progress of all threads.

[0089] In a general artificial intelligence-based text content analysis and reasoning system, a storage unit is also included, which is used to store data pre-existing in the system, as well as task data such as process data, result data, etc. in the review task.

[0090] The review engine unit establishes the preprocessing process and the rule acquisition process, and sets a shared storage area in the storage unit. The shared storage area stores the data signal of the text to be processed, the rule base, and / or the vector space. The preprocessing process and the rule acquisition process include access permissions to the shared storage area. The text analysis unit stores the data signal in the shared storage area, and the rule analysis unit stores the rule base and / or the vector space in the shared storage area.

[0091] A general AI-based text content analysis and reasoning system performs preprocessing and rule acquisition tasks for the text to be processed through two relatively independent processes, improving the system's task execution speed. Furthermore, when allocating resources between the two processes, the current resource usage is allocated based on the workload of the preprocessing and rule acquisition tasks at the previous moment or cycle. This makes resource allocation more rational, improves the execution speed of these two processes, and further ensures the system's task execution speed.

[0092] The above content is an explanation of the preferred embodiments of the present invention, which can help those skilled in the art to more fully understand the technical solutions of the present invention. However, these embodiments are merely illustrative, and it cannot be determined that the specific implementation methods of the present invention are limited to the description of these embodiments. For those skilled in the art of the present invention, without departing from the concept of the present invention, several simple deductions and transformations can be made, which should be deemed to fall within the scope of protection of the present invention.

Claims

1. A general artificial intelligence-based text content analysis and reasoning system, characterized by: It includes review engine unit, text analysis unit, rule analysis unit, reasoning unit and thread management unit. The review engine unit allocates the resource usage of each unit when performing the review task, and establishes a preprocessing process and a rule acquisition process; the text analysis unit is used to preprocess the text to be processed to obtain a data signal; the rule analysis unit is used to crawl basic rules and perform semantic analysis on the basic rules to obtain inference rules; the inference unit inputs the vectors corresponding to the data signal and the inference rule into the inference neural network, performs logical calculations, and obtains the inference results; the thread management unit manages the threads in the preprocessing process and the rule acquisition process respectively; When the review engine unit performs the review task, it quantifies the task amount at the previous moment or the previous cycle to obtain a quantification mark, and then allocates the resource occupancy of the preprocessing process and the rule acquisition process according to the quantification mark.

2. A general artificial intelligence-based text content analysis and reasoning system according to claim 1, characterized in that: The preprocessing of the text to be processed includes performing semantic analysis on the text to be processed and extracting key data of the text; arranging the key data and converting the format to obtain the data signal.

3. A general artificial intelligence-based text content analysis and reasoning system according to claim 1, characterized in that: The rule analysis unit converts each inference rule into a vector according to a domain specific language. Each inference rule corresponds to a vector, and all vectors constitute a vector space of the rule.

4. A general artificial intelligence-based text content analysis and reasoning system according to claim 1, characterized in that: The reasoning unit constructs a knowledge graph based on the association relationship between entities in the reasoning results of the same text to be processed.

5. A general artificial intelligence-based text content analysis and reasoning system according to claim 1, characterized in that: It also includes a display unit, which converts the knowledge graph into a target report through a visualization engine according to the input unit and displays the target report.

6. A general artificial intelligence-based text content analysis and reasoning system according to claim 1, characterized in that: The review engine unit allocates resource occupancy of the pre-processing process and the rule acquisition process, specifically including: The review engine unit obtains the amount of tasks that need to be processed at the last moment or the last cycle; The review engine unit quantifies the amount of tasks that need to be processed at the last moment or in the last cycle to obtain a quantitative mark; The review engine unit calculates the number of quantization marks at the previous moment or the previous cycle to obtain a quantization value; The review engine unit calculates the resource occupancy ratio of the rule analysis unit at the last moment or the last cycle according to the quantified value at the last moment or the last cycle; The review engine unit allocates resource occupancy of the rule analysis unit and the rule analysis unit at a current moment or a current period according to the rule analysis unit and the resource occupancy ratio of the rule analysis unit at a previous moment or a previous period.

7. A general artificial intelligence-based text content analysis and reasoning system according to claim 6, characterized in that: The review engine unit quantifies the amount of tasks to be processed at the previous moment or the previous cycle to obtain a quantitative mark. Specifically, the review engine unit quantifies the amount of text to be processed and the memory usage at the previous moment or the previous cycle to obtain a quantitative mark of the text to be processed; The review engine unit quantifies the number of basic rules to be processed at the last moment or the last cycle and the memory usage to obtain a quantified mark of the basic rules.

8. A general artificial intelligence-based text content analysis and reasoning system according to claim 7, characterized in that: The review engine unit calculates the number of quantization marks at the previous moment or the previous period to obtain a quantization value. Specifically, the review engine unit calculates the number of quantization marks of the text expected to be processed at the previous moment or the previous week to obtain a first quantization value; the review engine unit calculates the number of quantization marks of the basic rules expected to be processed at the previous moment or the previous week to obtain a second quantization value.

9. A general artificial intelligence-based text content analysis and reasoning system according to claim 6, characterized in that: The review engine unit quantifies the amount of tasks that need to be processed at the previous moment or the previous cycle to obtain a quantization mark, specifically including: the review engine unit generating the same number of first quantization marks according to the number of tasks that need to be processed at the previous moment or the previous cycle; The review engine unit calculates the quotient of the size of each task that needs to be processed at the last moment or the last cycle by the first quantization threshold to obtain a quantization quotient value; The review engine unit generates a target second quantization mark according to the quantization quotient value; The first quantization flag and the second quantization flag constitute the quantization flag.

10. A general artificial intelligence-based text content analysis and reasoning system according to claim 9, characterized in that: When the quantization quotient value of each task is less than 1, the number of the second quantization mark is zero; in this case, the second quantization mark is zero; When the quantization quotient values ​​of all tasks have a quantization quotient value greater than 1, the number of the second quantization marks is greater than zero: when the quantization quotient value of a task is an integer greater than 1, the review engine unit generates a second quantization mark equal to the quantization quotient value of the task minus 1, which is the second quantization mark of the task; when the quantization quotient value of a task is a non-integer greater than 1, the review engine unit generates a second quantization mark equal to the quantization quotient value of the task, which is the second quantization mark of the task; The second quantized labels of all tasks constitute the target second quantized label.

Images