Dynamic rule base construction method and device based on semantic contact calculation and medium

Abstract

The invention discloses a dynamic rule base construction method and device based on semantic contact computing and a medium, and relates to the technical field of artificial intelligence, the method comprises the following steps: carrying out manifold subspace construction on an initial concept manifold to obtain a concept manifold; on the basis of the concept manifold, determining a continuous thinking chain through micronization of a concept neighborhood; dividing the language field into a token vector field and a semantic contact field according to the continuous thinking chain; according to the semantic contact field, a semantic field arrangement statement is obtained through semantic relation optimization; and based on the token vector field, performing prediction probability analysis of the next token on the semantic field arrangement statement to determine the generation probability of the next token. According to the method, the technical problem that the universality is poor is solved through large-model reasoning.

Description

Technical Field

[0001] The present application relates to the field of artificial intelligence technology, and in particular to a method, device, and medium for constructing a dynamic rule base based on semantic connection computing. Background Art

[0002] The "reasoning" capabilities of large language models are based on statistical pattern recognition, possessing implicit reasoning and emergence capabilities. However, they lack true causal reasoning and are subject to hallucinations and difficulties with multi-step reasoning. Strictly speaking, large language models themselves lack "reasoning" capabilities. For the time being, they can be considered a form of "inference," still relying on cognitive approximations based on correlations. Therefore, building true reasoning capabilities in large models is a top priority.

[0003] Currently, integrating symbolic logic into neural network training presents significant technical challenges, and the complex and diverse constraints corresponding to rule-based judgments in various scenarios present challenges that must be addressed. The industry is still exploring the implementation of weak / strong coupling for neural symbolic logic. Technical solutions such as neural symbolic programming, neural rule engines, and differentiable transformers are all areas of ongoing research. The difficulties in widespread adoption of large-scale model inference and judgment in existing technologies can be attributed to the following: large models do not understand the abstract process of a concept, probabilistic models cannot effectively adapt to all conceptual reasoning, and precision is difficult to embed. These issues are caused by the fact that each concept has its own sub-attributes, each concept has its own context and scope of application, and precision in thought chains is difficult to embed. Summary of the Invention

[0004] The embodiments of the present application provide a method, device and medium for constructing a dynamic rule base based on semantic connection calculation, which solves the technical problem of poor universality of large model inference judgment.

[0005] In a first aspect, an embodiment of the present application provides a method for constructing a dynamic rule base based on semantic connection calculation, characterized in that the method includes: constructing a manifold subspace of an initial concept manifold to obtain a concept manifold; based on the concept manifold, determining a continuous thinking chain through concept neighborhood differentiability; according to the continuous thinking chain, dividing the language field into a token vector field and a semantic connection field; according to the semantic connection field, obtaining a semantic field arrangement sentence through semantic relationship optimization; based on the token vector field, performing a predicted probability analysis of the next token on the semantic field arrangement sentence to determine the probability of generating the next token.

[0006] In an implementation manner of the present application, a manifold subspace is constructed for the initial concept manifold to obtain a concept manifold, which specifically includes: obtaining the local active coordinate frame of the initial concept manifold, and selecting the sub-attributes of the concept as the unit vectors of the local active coordinate frame to determine the local concept space; semantically mapping the local concept space into an N-dimensional space to obtain the to-be-defined concept manifold; where the N-dimensional space includes: N-dimensional Euclidean space, geometric space; according to the preset application scenario, performing a manifold definition on the to-be-defined concept manifold to obtain a concept manifold; where the types of the concept manifold include: parametric manifold, data manifold, concept manifold.

[0007] In an implementation manner of the present application, based on the concept manifold, a continuous thought chain is determined through concept neighborhood differentiability, which specifically includes: based on the concept manifold, determining a token chain, and performing text coherence screening on the token chain to obtain a discontinuous token chain; determining the disconnection region of the discontinuous token chain, and based on the disconnection region, invoking the latent space; according to the latent space, performing differentiable supplementation of the disconnection region in the cross-neighborhood latent space to determine a continuous thought chain.

[0008] In an implementation manner of the present application, according to the continuous thought chain, the language field is divided into a token vector field and a semantic connection field, which specifically includes: setting a lexeme function, and based on the continuous thought chain and the lexeme function, constructing a semantic connection field through lexeme semantic organization; according to the semantic connection field, constructing a token vector field through state vector analysis of tokens.

[0009] In an implementation manner of the present application, according to the semantic connection field, through semantic relationship optimization, a semantic field arrangement statement is obtained, which specifically includes: obtaining a lexeme, and based on the lexeme, defining the geometric structure of the semantic manifold; constructing the connection coefficient of the local coordinate system for the semantically manifold after geometric structure definition to determine the semantic connection characteristics; arranging and organizing the semantic connection characteristics of the lexeme to obtain a semantic field arrangement statement.

[0010] In an implementation manner of the present application, based on the token vector field, a prediction probability analysis of the next token is performed on the semantic field arrangement statement to determine the generation probability of the next token, which specifically includes: based on the token vector field, performing a prediction probability analysis of the next token on the semantic field arrangement statement to obtain the generation state of the next token; obtaining a lexeme function and the corresponding lexeme complex function of the lexeme function, and according to the lexeme function and the lexeme complex function, determining the generation probability of the next token through token collapse analysis.

[0011] In an implementation manner of the present application, according to the lemma function and the lemma complex function, the generation probability of the next token is determined through token collapse analysis, specifically including: when the token prediction state is that the next token has not been generated, the token prediction state is uncertain; when the token prediction state is that the next token has been generated, the token prediction state is collapsed; in the case where the token prediction state is collapsed, according to the lemma function and the lemma complex function, the field function corresponding to the semantic field arrangement statement is determined, and the square of the modulus of the field function of the next token and the field function corresponding to the semantic field arrangement statement is set as the generation probability of the next token.

[0012] In an implementation manner of the present application, based on the token vector field, the prediction probability analysis of the next token for the semantic field arrangement statement is performed to determine the generation probability of the next token. The method further includes: constructing a semantic dynamic rule library based on the token vector field and the semantic connection field.

[0013] In a second aspect, an embodiment of the present application further provides a device for constructing a dynamic rule library based on semantic connection calculation, characterized in that the device includes: at least one processor; and a memory communicatively connected to the at least one processor; wherein, the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can: construct a manifold subspace for the initial concept manifold to obtain a concept manifold; determine a continuous thinking chain based on the concept manifold through concept neighborhood differentiability; divide the language field into a token vector field and a semantic connection field according to the continuous thinking chain; obtain a semantic field arrangement statement through semantic relationship optimization according to the semantic connection field; perform prediction probability analysis of the next token for the semantic field arrangement statement based on the token vector field to determine the generation probability of the next token.

[0014] In a third aspect, an embodiment of the present application further provides a non-volatile computer storage medium for a method of constructing a dynamic rule library based on semantic connection calculation, storing computer-executable instructions, characterized in that the computer-executable instructions are set to: construct a manifold subspace for the initial concept manifold to obtain a concept manifold; determine a continuous thinking chain based on the concept manifold through concept neighborhood differentiability; divide the language field into a token vector field and a semantic connection field according to the continuous thinking chain; obtain a semantic field arrangement statement through semantic relationship optimization according to the semantic connection field; perform prediction probability analysis of the next token for the semantic field arrangement statement based on the token vector field to determine the generation probability of the next token.

[0015] The embodiments of the present application provide a method, device and medium for constructing a dynamic rule base based on semantic connection calculation. By using a neuro-symbolic unified modeling method based on intrinsic geometry to improve the data accuracy of the spatial framework, a continuous reasoning method in the cross-neighborhood latent space to dynamically fill in missing tokens, and a unified language field construction method based on semantic connection to perform token prediction and lemmatization organization analysis, the technical problem of poor universality of large model reasoning and judgment is solved, the construction of the concept tangent space, the differentiability of the neighborhood, and the dynamic formation of a continuous reasoning chain across knowledge domains are realized. It can be applied to scenarios such as compliance review and sourcing control, and calculating semantic connection can realize the self-assembly and self-construction of the knowledge structure, mine strong associations of information, and be applied to scenarios such as compliance analysis and early warning. Description of the Drawings

[0016] The drawings described herein are used to provide a further understanding of the present application and constitute a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation to the present application. In the drawings: Figure 1 It is a flowchart of a method for constructing a dynamic rule base based on semantic connection calculation provided by an embodiment of the present application; Figure 2 It is a schematic diagram of the construction of a concept manifold provided by an embodiment of the present application; Figure 3 It is a schematic diagram of the differentiability of a concept neighborhood provided by an embodiment of the present application; Figure 4 It is a schematic diagram of the differentiated state of a language field provided by an embodiment of the present application; Figure 5 It is a schematic diagram of the internal structure of a device for constructing a dynamic rule base based on semantic connection calculation provided by an embodiment of the present application. Detailed Embodiments

[0017] To make the objectives, technical solutions and advantages of the present application clearer, the technical solutions of the present application will be clearly and completely described below in conjunction with the specific embodiments of the present application and the corresponding drawings. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present application.

[0018] The embodiments of the present application provide a method, device, and medium for constructing a dynamic rule base based on semantic liaison calculation. By using a neuro-symbolic unified modeling method based on intrinsic geometry to improve the data accuracy of the spatial framework, a continuous reasoning method in the cross-neighborhood latent space to dynamically fill in missing tokens, and a unified language field construction method based on semantic liaison to perform token prediction and lemmatization organization analysis, the technical problem of poor universality of large model reasoning and judgment is solved, the construction of the concept cut space, the differentiability of the neighborhood, and the dynamic formation of a continuous reasoning chain across knowledge domains are realized. It can be applied to scenarios such as compliance auditing and sourcing control, and calculating semantic liaison can achieve self-assembly and self-construction of the knowledge structure, mine strong associations of information, and be applied to scenarios such as compliance analysis and early warning.

[0019] The technical solutions proposed in the embodiments of the present application will be described in detail below with reference to the accompanying drawings.

[0020] Figure 1 It is a flowchart of a method for constructing a dynamic rule base based on semantic liaison calculation provided by the embodiments of the present application. As Figure 1 shown, a method for constructing a dynamic rule base based on semantic liaison calculation provided by the embodiments of the present application specifically includes the following steps: Step 101, construct a manifold subspace for the initial concept manifold to obtain a concept manifold.

[0021] Exemplarily, in order to no longer study the properties of the concept manifold by constructing an external coordinate system and only establish a local active coordinate frame on the manifold surface, the present application constructs a manifold subspace for the initial concept manifold to obtain a concept manifold, getting rid of the dependence on the coordinate system, directly starting from the geometric structure of the model space to describe the local geometric properties of the concept manifold, and realizing data dimensionality reduction through intrinsic geometry, so that the data can be better analyzed, visualized, and understood while maintaining its geometric features.

[0022] Specifically, constructing a manifold subspace for the initial concept manifold to obtain a concept manifold includes: obtaining the local active coordinate frame of the initial concept manifold, and selecting the sub-attributes of the concept as the unit vectors of the local active coordinate frame to determine the local concept space; semantically mapping the local concept space into an N-dimensional space to obtain the to-be-defined concept manifold; where the N-dimensional space includes: N-dimensional Euclidean space, geometric space; according to the preset application scenario, perform manifold definition on the to-be-defined concept manifold to obtain a concept manifold; where the types of the concept manifold include: parametric manifold, data manifold, concept manifold.

[0023] In one embodiment, to incorporate the ability of continuous reasoning, a conceptual manifold needs to be established. The conceptual manifold has no external coordinates, and there is a moving coordinate frame on the manifold. The sub-properties of the concept are used as the unit vectors of the local moving coordinate frame. The concept points in the concept region can all be represented in the local moving coordinate frame. For the local space of an n-dimensional manifold, it is homeomorphic to an n-dimensional Euclidean space. Figure 2 FIG. is a schematic diagram for constructing a conceptual manifold provided by an embodiment of the present application, as Figure 2 shown, in the subspace of the conceptual manifold, "penguin" and "swallow" are both "birds", and the concept of "bird" includes "beak", "wings", and "feathers"; similarly, the subspace relationship of insects can be deduced by analogy, which will not be elaborated here.

[0024] After the construction of the original conceptual manifold is completed, many concepts are distributed in different subspaces. At this time, the construction of the subspaces is not yet perfect. Therefore, it is necessary to select the sub-properties of the concept as the unit vectors of the local moving coordinate frame to determine the local concept space, which can not only get rid of the coordinate limitation but also meet the construction of the concept hierarchy in the local manifold (such as the first level, the second level, etc.).

[0025] It should be noted that the construction of the conceptual manifold is a manifestation of intrinsic geometry. High-dimensional data distributions usually exhibit a certain geometric structure. In the specific process of analyzing the structure, algorithms such as manifold learning can be used to reduce its dimension and explain its intrinsic geometric structure.

[0026] Furthermore, As the representation of the conceptual manifold, is the dimension of the manifold, and essentially represents a real number (determined by the mathematical concept of the manifold); at this time, the construction of the conceptual manifold is completed. For the concepts "insect" and "bird", there is a commonality (intersection) of being able to fly.

[0027] Through , , the concept is homeomorphically mapped into the Euclidean geometric space to obtain three ; the three can also be mutually mapped to each other, that is, and . and are , 's inverse. In the mutual mapping process, the , of the homeomorphic mapping can be cancelled out; finally, the middle circular is after the mutual mapping is completed, representing that the intersection of the concepts "insect" and "bird" is "fly".

[0028] Note that in a smooth manifold, for the coordinate mapping in a minimal space, since the curvature tends to 0, it can be mapped through mapping methods such as three-dimensional geometry.

[0029] Furthermore, the local conceptual space can be studied in an n-dimensional Euclidean space, and the definition of a manifold can be diversified according to the application scenario.

[0030] Parameterized manifold: For example, the family of probability distributions in a statistical model (such as the parameter space of a Gaussian distribution), and its tangent space describes the impact of tiny parameter perturbations.

[0031] Data manifold: After a high-dimensional data set is embedded in a low-dimensional space through dimensionality reduction (such as t-SNE, UMAP in manifold learning), its local linear structure is characterized by the tangent space.

[0032] Conceptual manifold: Abstract semantics (such as lexical semantics, cognitive structure) can be geometrically modeled as a manifold.

[0033] This application uses a neuro-symbolic unified modeling method based on intrinsic geometry. Instead of studying the properties of a conceptual manifold by constructing an external coordinate system, it only establishes a local active coordinate frame on the manifold surface. This method gets rid of the dependence on the coordinate system and directly starts from the geometric structure of the model space to describe the local geometric properties of the conceptual manifold. It focuses on processing high-dimensional data and revealing its underlying low-dimensional manifold structure. Its core idea is to assume that high-dimensional data samples are distributed on a potential low-dimensional manifold. Even though the data is in a high-dimensional space, it can be described by a lower-dimensional structure. This method drives manifold learning algorithms, aiming to discover the intrinsic low-dimensional representation of data, enabling the data to be better analyzed, visualized, and understood while maintaining its geometric features. This method is beneficial for studying the self-driven mechanism of changes in the language space structure and the field source of changes in the semantic field.

[0034] Step 102: Based on the conceptual manifold, determine a continuous thinking chain through the differentiability of the conceptual neighborhood.

[0035] Exemplarily, this application determines a continuous thinking chain based on the conceptual manifold through the differentiability of the conceptual neighborhood, realizes the supplementation of discontinuous spaces in the conceptual manifold, the model conducts reasoning in the latent space without generating intermediate language expressions, skips the discontinuous regions in the token space, and improves the overall stability of the conceptual manifold.

[0036] Specifically, determining a continuous thinking chain based on the conceptual manifold through the differentiability of the conceptual neighborhood includes: based on the conceptual manifold, determining a token chain, and performing text coherence screening on the token chain to obtain a discontinuous token chain; determining the disconnection region of the discontinuous token chain, and based on the disconnection region, invoking the latent space; according to the latent space, performing differentiable supplementation of the disconnection region across the neighborhood latent space to determine a continuous thinking chain.

[0037] In one embodiment, when a large language model (LLM) reasons using language, a major problem arises: the amount of reasoning required for each specific reasoning token varies greatly, but the current LLM architectures allocate almost the same computational budget for predicting each token. These problems are addressed by prompting the LLM to generate concise reasoning chains or perform additional reasoning before generating some key tokens, but these solutions are still limited to the language space and do not solve the fundamental problem. Instead, ideally, the LLM should be able to reason freely without any language restrictions and then translate its findings into language only when necessary. LLMs are limited to reasoning in the "language space", and they typically use the chain of thought (CoT) to express the reasoning process to solve complex reasoning problems. However, the language space may not always be the best space for reasoning. Most tokens in the reasoning chain are generated only for fluency and contribute little to the actual reasoning process.

[0038] Figure 3 A schematic diagram of concept neighborhood differentiability provided by an embodiment of this application.

[0039] Most word tokens are mainly used for text coherence and are not essential for reasoning, while some key tokens require complex planning and pose a great challenge to LLMs. To explore the potential of LLMs to reason in an unrestricted latent space rather than using natural language, a new paradigm, Coconut (Continuous Chain of Thought), is introduced at the moment when the "hidden state" generated by the Transformer layer is converted into a token. This conversion process means that information is compressed from an infinite mathematical space composed of continuous numerical values into a finite set composed of only about tens of thousands of discrete vocabulary words. This "dimensionality reduction" not only causes an efficiency loss but also may compress or distort the internal representation of the model. To this end, the model structure is modified: bypass the step of generating tokens, directly loop the hidden state back to the input embedding, and process it through the Transformer layer again. In this way, the model can always process information within the continuous mathematical space without being restricted by the discrete boundaries set by human language.

[0040] The construction of the concept manifold realizes the construction without an external coordinate framework, the concept neighborhood differentiability realizes the supplementary processing of problems such as discontinuous language data in the framework, and the latent space is used to supplement the discontinuous part of the token space.

[0041] This application constructs a language model that mainly conducts reasoning in the latent space through a continuous reasoning method in the cross-neighborhood latent space. When training the model, it not only learns to reason in the latent space but also can autonomously determine when to end the reasoning and return to language generation. The output of the intermediate module of the model is not immediately converted into tokens but is sent back into the module internally in the form of embedding vectors for further processing. The model needs to autonomously learn how many times of cyclic reasoning should be performed in different tasks. When the change of the embedding vectors in the cyclic output tends to be stable, the reasoning ends. In this process, the model gradually establishes the mutual connections between each token. The final layer synthesizes all information to generate a new sequence of embedding vectors. The model outputs multiple tokens to gradually display its solution path. The last hidden state of the model is regarded as the representation of "continuous thinking" and is directly used as the embedding of the next input instead of being decoded into tokens. In this way, the model can reason in the latent space without generating intermediate language expressions and skip the discontinuous regions in the token space. This not only helps to reveal the reasoning mechanism of the model but also significantly improves the accuracy of the answers.

[0042] Step 103: Divide the language field into a token vector field and a semantic connection field according to the continuous thought chain.

[0043] Exemplarily, this application proposes the concept of a language field and divides the language field into a token vector field and a semantic connection field. Tokens are the basic building blocks of the language edifice, and the semantic connection field is responsible for organizing the tokens to form sentences with fixed semantics. The large model predicts the probability of generating the next token in the token vector field.

[0044] Specifically, dividing the language field into a token vector field and a semantic connection field according to the continuous thought chain includes: setting a token function, and based on the continuous thought chain and the token function, constructing the semantic connection field through token semantic organization; constructing the token vector field according to the semantic connection field through the state vector analysis of tokens.

[0045] Step 104: Obtain the semantic field arrangement statement through semantic relationship optimization according to the semantic connection field.

[0046] Specifically, obtaining the semantic field arrangement statement through semantic relationship optimization according to the semantic connection field includes: obtaining tokens, and based on the tokens, defining the geometric structure of the semantic manifold; constructing the connection coefficients of the local coordinate system for the geometric structure-defined semantic manifold to determine the semantic connection characteristics; arranging and organizing the semantic connection characteristics of the tokens to obtain the semantic field arrangement statement.

[0047] In one embodiment, the semantic connection is a specific implementation of the affine connection on the manifold in the semantic space, used to describe the covariant derivative and parallel transport rules of vector fields in the semantic space. The core lies in defining the geometric structure of the semantic manifold (such as the word embedding space, concept space) and constructing the connection coefficients in the local coordinate system. The semantic space is modeled as a smooth manifold, and its local coordinates correspond to semantic features (such as the dimensions of word vectors, abstract attributes of concepts). For example, in the Word2Vec model, the word embedding space can be regarded as a submanifold of the semantic space, reflecting the semantic relationships between words. The calculation of the semantic connection is the core task of geometric semantics, which requires integrating differential geometry, statistical learning, and domain knowledge. In a certain semantic field, word elements need to act through the semantic field as a medium, and are arranged and organized by the semantic field to form a sentence with a certain semantics.

[0048] The semantic field arranging the statement means that under the configuration of the semantic field, the word elements should be arranged in a certain state and the output result.

[0049] Step 105: Based on the token vector field, perform a prediction probability analysis on the semantic field arranging statement to determine the generation probability of the next token.

[0050] Exemplarily, the word element function is a vector in the complex Hilbert space. The probability that the model generates the next token based on the previous statement is the square of the modulus of the inner product of the next token vector and the previous statement vector. Before the next token is generated, its state is uncertain. When the token is generated, the state collapses. The word element complex function will rotate under the action of the semantic field and be in different semantic states. The modulus of the word element complex function is its basic meaning.

[0051] Specifically, based on the token vector field, performing a prediction probability analysis on the semantic field arranging statement to determine the generation probability of the next token specifically includes: based on the token vector field, performing a prediction probability analysis on the semantic field arranging statement to obtain the generation state of the next token; obtaining the word element function and the corresponding word element complex function, and determining the generation probability of the next token through token collapse analysis according to the word element function and the word element complex function.

[0052] Further, according to the token function and the token complex function, the generation probability of the next token is determined through token collapse analysis, specifically including: when the token prediction state is that the next token has not been generated, the token prediction state is uncertain; when the token prediction state is that the next token has been generated, the token prediction state undergoes collapse; in the case where the token prediction state undergoes collapse, according to the token function and the token complex function, the field function corresponding to the semantic field arrangement statement is determined, and the square of the modulus of the field function of the next token and the field function corresponding to the semantic field arrangement statement is set as the generation probability of the next token.

[0053] Further, based on the token vector field, the prediction probability analysis of the next token for the semantic field arrangement statement is performed to determine the generation probability of the next token. The method further includes: constructing a semantic dynamic rule library based on the token vector field and the semantic connection field.

[0054] Figure 4 This is a schematic diagram of the discrimination state of a language field provided by an embodiment of the present application.

[0055] In one embodiment, the token function is a vector in the complex Hilbert space. The probability that the model generates the next token based on the previous statement is the square of the modulus of the inner product of the vector of the next token and the vector of the previous statement. Before the next token is generated, its state is uncertain, and when the token is generated, the state undergoes collapse. The token complex function will rotate under the action of the semantic field and be in different semantic states.

[0056] Therefore, the probability that the large model predicts the generation of the next token is the square of the modulus of the inner product of the field function of the next token and the field function of the previous statement (for the next token that conforms to the quantum superposition state, observation will cause collapse, and its state cannot be determined without observation).

[0057] The above is the method embodiment proposed by the present application. Based on the same inventive concept, the embodiment of the present application also provides a device for constructing a dynamic rule library based on semantic connection calculation, and its structure is as Figure 5 shown.

[0058] Figure 5 This is a schematic diagram of the internal structure of a device for constructing a dynamic rule library based on semantic connection calculation provided by an embodiment of the present application. As Figure 5 shown, the device includes: At least one processor 501; And a memory 502 communicatively connected to at least one processor; Among them, the memory 502 stores instructions executable by at least one processor. The instructions are executed by at least one processor 501 so that at least one processor 501 can: Construct a manifold subspace for the initial concept manifold to obtain a concept manifold; based on the concept manifold, determine a continuous thought chain through concept neighborhood differentiability; according to the continuous thought chain, divide the language field into a token vector field and a semantic connection field; according to the semantic connection field, obtain a semantic field arrangement statement through semantic relationship optimization; based on the token vector field, perform a prediction probability analysis of the next token on the semantic field arrangement statement to determine the generation probability of the next token.

[0059] Some embodiments of the present application provide a non-volatile computer storage medium corresponding to Figure 1 a dynamic rule base construction based on semantic connection calculation, storing computer-executable instructions, and the computer-executable instructions are set to: Construct a manifold subspace for the initial concept manifold to obtain a concept manifold; based on the concept manifold, determine a continuous thought chain through concept neighborhood differentiability; according to the continuous thought chain, divide the language field into a token vector field and a semantic connection field; according to the semantic connection field, obtain a semantic field arrangement statement through semantic relationship optimization; based on the token vector field, perform a prediction probability analysis of the next token on the semantic field arrangement statement to determine the generation probability of the next token.

[0060] The various embodiments in the present application are described in a progressive manner. The same or similar parts among the various embodiments can be referred to each other, and the key points of each embodiment are the differences from other embodiments. In particular, for the embodiments of the Internet of Things devices and media, since they are basically similar to the method embodiments, the description is relatively simple, and the relevant parts can refer to the partial description of the method embodiments.

[0061] The systems and media provided by the embodiments of the present application correspond one-to-one with the methods. Therefore, the systems and media also have beneficial technical effects similar to the corresponding methods. Since the beneficial technical effects of the methods have been described in detail above, the beneficial technical effects of the systems and media will not be repeated here.

[0062] Those skilled in the art should understand that the embodiments of the present application can be provided as methods, systems, or computer program products. Therefore, the present application can take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present application can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk memories, CD-ROMs, optical memories, etc.) containing computer-usable program codes.

[0063] This application is described with reference to the flowcharts and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the present application. It should be understood that each flow and / or block in the flowchart and / or block diagram can be implemented by computer program instructions, as well as the combination of flows and / or blocks in the flowchart and / or block diagram. These computer program instructions can be provided to the processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing device to generate a machine, such that the instructions executed by the processor of the computer or other programmable data processing device produce means for implementing the functions specified in one Figure 1 flow or multiple flows and / or blocks Figure 1 or multiple blocks.

[0064] These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing device to work in a specific manner, such that the instructions stored in the computer-readable memory produce a manufactured article including instruction means that implement the functions specified in one Figure 1 flow or multiple flows and / or blocks Figure 1 or multiple blocks.

[0065] These computer program instructions can also be loaded onto a computer or other programmable data processing device, such that a series of operation steps are executed on the computer or other programmable device to generate a computer-implemented process, and thus the instructions executed on the computer or other programmable device provide steps for implementing the functions specified in one Figure 1 flow or multiple flows and / or blocks Figure 1 or multiple blocks.

[0066] In a typical configuration, a computing device includes one or more processors (CPUs), an input / output interface, a network interface, and memory.

[0067] The memory may include non-permanent memory in the form of computer-readable media, random access memory (RAM), and / or non-volatile memory, such as read-only memory (ROM) or flash RAM. The memory is an example of computer-readable media.

[0068] Computer-readable media includes both permanent and non-permanent, removable and non-removable media that can be implemented using any method or technology to store information. Information can be computer-readable instructions, data structures, program modules, or other data. Examples of computer storage media include, but are not limited to, phase-change RAM (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, compact disc read-only memory (CD-ROM), digital versatile disc (DVD) or other optical storage, magnetic cassettes, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer-readable media does not include transitory computer-readable media such as modulated data signals and carrier waves.

[0069] It should also be noted that the terms "comprises," "includes," or any other variations thereof are intended to encompass non-exclusive inclusion, such that a process, method, commodity, or apparatus that includes a series of elements includes not only those elements but also other elements not explicitly listed, or includes elements inherent to such process, method, commodity, or apparatus. In the absence of further limitations, an element defined by the phrase "comprises a ..." does not exclude the presence of other identical elements in the process, method, commodity, or apparatus that includes the element.

[0070] The above are merely embodiments of the present application and are not intended to limit the present application. For those skilled in the art, the present application may have various changes and variations. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present application should all be included within the scope of the claims of the present application.

Claims

1. A method for constructing a dynamic rule base based on semantic connection calculation, characterized in that The method includes: Constructing a manifold subspace for the initial concept manifold to obtain a concept manifold; Determining a continuous thought chain based on the concept manifold through concept neighborhood differentiabilization; Dividing the language field into a token vector field and a semantic connection field according to the continuous thought chain; Obtaining a semantic field arrangement statement through semantic relationship optimization according to the semantic connection field; Based on the token vector field, performing a prediction probability analysis of the next token on the semantic field arrangement statement to determine the generation probability of the next token.

2. The method for constructing a dynamic rule base based on semantic liaison calculation according to claim 1, wherein Constructing a manifold subspace for the initial concept manifold to obtain a concept manifold, specifically including: Obtaining the local active coordinate frame of the initial concept manifold, and selecting sub-attributes of the concept as the unit vectors of the local active coordinate frame to determine a local concept space; Semantically mapping the local concept space into an N-dimensional space to obtain a to-be-defined concept manifold; where the N-dimensional space includes: an N-dimensional Euclidean space, a geometric space; Performing a manifold definition on the to-be-defined concept manifold according to a preset application scenario to obtain the concept manifold; where the types of the concept manifold include: a parametric manifold, a data manifold, a concept manifold.

3. A method for constructing a dynamic rule base based on semantic connection calculation according to claim 1, characterized in that Determining a continuous thought chain based on the concept manifold through concept neighborhood differentiabilization, specifically including: Based on the concept manifold, determining a token chain, and performing text coherence screening on the token chain to obtain a discontinuous token chain; Determining the disconnection region of the discontinuous token chain, and based on the disconnection region, invoking a latent space; According to the latent space, performing differentiable supplementation of the disconnection region in the cross-neighborhood latent space to determine the continuous thought chain.

4. A method for constructing a dynamic rule base based on semantic liaison calculation according to claim 1, characterized in that Dividing the language field into a token vector field and a semantic connection field according to the continuous thought chain, specifically including: Setting a token element function, and based on the continuous thought chain and the token element function, constructing the semantic connection field through token element semantic organization; Constructing the token vector field through state vector analysis of the token according to the semantic connection field.

5. A method for constructing a dynamic rule base based on semantic liaison calculation according to claim 1, characterized in that Obtaining a semantic field arrangement statement through semantic relationship optimization according to the semantic connection field, specifically including: Obtaining a token element, and based on the token element, defining the geometric structure of the semantic manifold; Constructing the connection coefficient of the local coordinate system for the semantic manifold after the geometric structure definition to determine the semantic connection feature; Performing arrangement and organization of the semantic connection feature on the token element to obtain the semantic field arrangement statement.

6. A method for constructing a dynamic rule base based on semantic connection calculation according to claim 1, characterized in that Based on the token vector field, performing a prediction probability analysis of the next token on the semantic field arrangement statement to determine the generation probability of the next token, specifically including: Based on the token vector field, performing a prediction probability analysis of the next token on the semantic field arrangement statement to obtain the generation state of the next token; Obtaining a token element function and the corresponding token element complex function of the token element function, and according to the token element function and the token element complex function, determining the generation probability of the next token through token collapse analysis.

7. A method for constructing a dynamic rule base based on semantic liaison calculation according to claim 6, characterized in that According to the token function and the token complex function, the generation probability of the next token is determined through token collapse analysis, specifically including: When the token prediction state is that the next token has not been generated, the token prediction state is uncertain; When the token prediction state is that the next token is to be generated, the token prediction state undergoes collapse; In the case where the token prediction state undergoes collapse, according to the token function and the token complex function, the field function corresponding to the semantic field arrangement statement is determined, and the square of the modulus of the field function of the next token and the field function corresponding to the semantic field arrangement statement is set as the generation probability of the next token.

8. A method for constructing a dynamic rule base based on semantic connection calculation according to claim 1, characterized in that Based on the token vector field, a prediction probability analysis of the next token for the semantic field arrangement statement is performed to determine the generation probability of the next token. The method further includes: Based on the token vector field and the semantic connection field, a semantic dynamic rule library is constructed.

9. A device for constructing a dynamic rule base based on semantic connection calculation, characterized in that The device includes: At least one processor; And a memory communicatively connected to the at least one processor; Wherein, the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can: Construct a manifold subspace for the initial concept manifold to obtain a concept manifold; Based on the concept manifold, determine a continuous thinking chain through concept neighborhood differentiability; According to the continuous thinking chain, divide the language field into a token vector field and a semantic connection field; According to the semantic connection field, obtain a semantic field arrangement statement through semantic relationship optimization; Based on the token vector field, perform a prediction probability analysis of the next token for the semantic field arrangement statement to determine the generation probability of the next token.

10. A non-volatile computer storage medium for a method of constructing a dynamic rule base based on semantic liaison calculation, storing computer-executable instructions, characterized in that, The computer-executable instructions are set to: Construct a manifold subspace for the initial concept manifold to obtain a concept manifold; Based on the concept manifold, determine a continuous thinking chain through concept neighborhood differentiability; According to the continuous thinking chain, divide the language field into a token vector field and a semantic connection field; According to the semantic connection field, obtain a semantic field arrangement statement through semantic relationship optimization; Based on the token vector field, perform a prediction probability analysis of the next token for the semantic field arrangement statement to determine the generation probability of the next token.

Images