Document proofreading system and method based on artificial intelligence

Abstract

The invention discloses a document proofreading system and method based on artificial intelligence, and relates to the technical field of natural language processing, and the method comprises the steps: analyzing a document, dividing semantic units, and obtaining a document with a context label based on context type library labeling; performing initial error detection by using the corresponding rule set, and comparing the format of the document and the text quality characteristics with a standard sample vector space to obtain an initial error set; constructing a defect conduction chain network, and locating root cause error nodes causing a plurality of secondary errors by tracing the defect conduction chain network; generating an intelligent proofreading report; receiving a new-version document revised by the user, comparing the new-version document with the original-version document, positioning a change area, and analyzing a quality difference of the change area on a related quality dimension; based on the quality difference, an incremental proofreading report is output, and the incremental proofreading report comprises newly introduced errors and unsolved root cause errors.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of natural language processing, and particularly relates to a document proofreading system and method based on artificial intelligence. BACKGROUND

[0002] With the popularization of digital office, the writing and auditing work of various documents is increasingly heavy. The traditional document proofreading mainly relies on manual work, which is not only low in efficiency, but also prone to problems such as missed detection and false detection due to proofreading fatigue. Especially for technical documents, academic papers and other professional content, the proofreading work not only involves basic grammar and spelling checking, but also needs to comprehensively evaluate the logical consistency, data accuracy, format specification and other dimensions of the document.

[0003] Some automatic proofreading tools existing in the market at present are mostly based on a pre-defined rule library or a basic statistical model, and can only detect superficial grammar errors and spelling errors, and cannot deeply understand the semantic content and logical structure of the document. Such tools are not capable of dealing with complex logical contradictions and data inconsistencies. In addition, the existing tools usually treat various errors in isolation, lack analysis of the correlation between errors, cannot identify and locate the root cause of a series of problems, and require users to spend a lot of time on repetitive correction work. Moreover, the existing intelligent proofreading system lacks the ability of continuous learning. They cannot learn from the user's correction behavior, cannot adapt to the document proofreading requirements of different fields and different styles, and the accuracy and adaptability of the system are greatly limited. SUMMARY

[0004] The present application aims to provide a document proofreading system and method based on artificial intelligence to solve the problems in the prior art.

[0005] To achieve the above-mentioned purpose, the present application provides the following technical scheme: a document proofreading method based on artificial intelligence, the method comprising: S100, analyzing the document, dividing semantic units, and labeling semantic unit types based on a context type library to obtain document data with context labels; the context type library is obtained by training based on a standard document sample library; the training process is to correlate the text features and semantic functions of the semantic units in the standard document sample library to establish a mapping relationship from the text features to the context types; S200, based on the context labels, using the corresponding rule set to perform initial error detection, and comparing the format and text quality features of the document with a standard sample vector space to calculate the quality deviation of each dimension and obtain an initial error set; S300, taking the initial error set as nodes, analyzing the causal relationship between the error nodes according to the defect propagation rule library, constructing a defect propagation chain network, and locating the root error node causing multiple secondary errors by tracing the network; S400, generating an intelligent proofreading report based on the root error node and the defect propagation chain, the intelligent proofreading report including error positioning, a priority correction strategy for the root error, and a global correction suggestion; S500, receiving a new version of the document revised by the user based on the intelligent proofreading report, comparing the new version of the document with the original version of the document, locating the changed area, and calculating the quality difference of the changed area in the related quality dimension; S600, based on the quality difference, outputting an incremental proofreading report, the incremental proofreading report including newly introduced errors and unsolved root errors; using the quality difference data to continuously optimize the context type labeling, the defect propagation rule library and the root error determination logic.

[0006] According to the above content, step S100 includes: S110, performing syntax structure and semantic dependency analysis on the input document, identifying and dividing independent semantic units constituting the logical structure of the document; S120, extracting text features of each semantic unit, the text features including syntax structure mode, keyword distribution, semantic role and logical association relationship with adjacent units; the syntax structure mode refers to the sentence structure and component collocation form of the semantic unit; the keyword distribution refers to the position and frequency characteristics of the core term in the semantic unit; the semantic role refers to the functions of each component in the semantic unit, such as statement, modification and limitation; the logical association relationship refers to the association form of cause and effect, progression and parallelism between adjacent semantic units; S130, inputting the text features into a pre-trained context classification model, the context classification model learning a large number of standard document samples through a machine learning algorithm, and establishing a mapping relationship from the text features to the context type; the pre-training process of the context classification model adopts a supervised learning method, and the semantic unit type manually labeled in the standard document sample library is used as a label to optimize the model parameters to improve the classification accuracy; S140, based on the output of the context classification model, assigning a specific context type label to each semantic unit to generate document data with context labels; the context types include technical definition statement, data argumentation statement, reasoning argumentation statement, instance explanation statement and background introduction statement.

[0007] According to the above content, step S200 includes: S210, according to the context type label of the semantic unit, calling the special error checking rule set pre-bound with the context type label, performing syntax rule compliance check and semantic logic consistency check on the semantic unit; each context type label corresponds to a unique special error checking rule set, and the checking dimension and judgment standard of the special error checking rule set are customized according to the semantic function and quality requirement of the corresponding context type, to ensure the pertinence and accuracy of the checking; S220, extracting global format specification features and text quality statistical features from the document, the format specification features including title hierarchy structure, numbering sequence consistency and chart label specification, and the text quality statistical features including term usage consistency, paragraph length distribution and logical connection word density; S230, mapping the extracted format specification features and text quality statistical features to a multi-dimensional vector space constructed based on a large number of standard document samples, calculating the Euclidean distance between the feature vector of the document to be corrected and the clustering center of the standard sample feature vector in each dimension as the quality deviation degree of each dimension; the multi-dimensional standard sample vector space is constructed by clustering analysis on the feature data of all samples in the standard document sample library, and the clustering center is the standard reference value of each feature dimension, to ensure the objectivity and uniformity of quality evaluation; S240, integrating the rule violation errors identified by the special error checking rule set and the statistical abnormal errors identified by the quality deviation degree calculation to generate an initial error set; during the integration process, the error location and error type are used for deduplication processing, and the feature dimension data and context label corresponding to each error are recorded.

[0008] According to the above content, the context types include technical definition statement, data argument statement, reasoning argument statement, instance explanation statement and background introduction statement; the special error checking rule set is configured based on different context type labels; The rule set configured for the technical definition statement type focuses on the uniqueness, completeness and unambiguity of the core term definition; the rule set configured for the data argument statement type focuses on the unit uniformity of numerical data, the specification of data source citation, and the rationality of the support relationship between data and conclusion; the rule set configured for the reasoning argument statement type focuses on the logical completeness of the cause-effect relationship in the argument chain and the effectiveness of the reasoning premise; the rule set configured for the instance explanation statement type focuses on the relevance of the instance and the abstract concept or rule explained; the rule set configured for the background introduction statement type focuses on the objectivity of historical fact statement and the basis grammar compliance.

[0009] According to the above content, step S300 includes: S310, abstract each error instance in the initial error set as a network node, and record the error type, semantic unit to which the network node belongs, and context label attribute of the network node; each network node has a unique identifier, and the attribute information thereof corresponds to the error details in the initial error set one by one; S320, call a defect propagation rule library, the defect propagation rule library defines the cause-effect relationship logic between different error type nodes in a specific context; the defect propagation rule library includes error type association rules, context adaptation rules, and cause-effect strength determination rules, the error type association rules explicitly define the propagation possibility between different error types, the context adaptation rules limit the semantic scene in which the propagation relationship is established, and the cause-effect strength determination rules quantify the propagation association degree between error nodes; S330, based on the defect propagation rule library, traverse all error node pairs, if the node pair satisfies the cause-effect relationship condition defined in the defect propagation rule library, establish a directed propagation edge between the two nodes, and construct a directed defect propagation chain network; the direction of the propagation edge is from the cause error node to the effect error node, and the rule basis corresponding to the propagation relationship and the cause-effect strength value are recorded at the same time; S340, calculate the out-degree of each node in the defect propagation chain network, and identify the circular dependency path in the network; the out-degree of the node is as follows: OD i =∑ j A ij ; Wherein, OD i represents the out-degree of node i; i and j represent the index of the node, and ∀i∈[1,n], that is, it is applicable to all nodes with index 1 to n in the network; when calculating the out-degree of node i, node i is fixed, and node j needs to traverse all nodes, that is, from 1 to n to sum; n represents the total number of nodes in the network; A ij represents the element of the adjacency matrix, if there is a propagation edge from node i to node j, then A ij =1, otherwise A ij =0; the circular dependency path is identified by traversing the directed propagation chain network, which refers to a closed path that returns to the starting node after passing through several propagation edges from a certain node; S350, determine the node whose out-degree value exceeds a preset threshold or which is located at the starting point of the circular dependency path as a root cause error node that triggers multiple secondary errors, and the conditions for determining the root cause error node are as follows: When the out-degree of node i is greater than the out-degree threshold θ, that is, OD i >θ, the root cause error node determination condition is satisfied, the out-degree threshold θ is determined based on the out-degree statistical data of historical error nodes in a standard document sample library, and a dynamic threshold is set by statistically analyzing the out-degree distribution characteristics of historical root cause error nodes; for adapting to different complexity of document scenarios; When node i belongs to the set of starting nodes V of the circular dependency path c , i.e. i∈V c , the root cause error node determination condition is met; the set of starting nodes of the circular dependency path is identified by traversing the directed conductive chain network, and contains all elements that are starting nodes of the circular dependency path; The node is determined to be a root cause error node when it meets any of the above conditions, and the two conditions have the same determination effect.

[0010] According to the above content, step S400 includes: S410, generating error positioning information, the error positioning information including the root cause error node and detailed information of all secondary error nodes directly associated and indirectly associated with the root cause error node in the defect conductive chain network, the detailed information including error content, error type, location information and belonging context type; S420, generating a priority correction strategy for the root cause error, the priority correction strategy including determining the priority order of the correction operation based on the attributes of the root cause error node and the centrality of the root cause error node in the defect conductive chain network, and describing the secondary error nodes associated with the root cause error node that will be automatically eliminated or transformed into a pending state after the root cause error is corrected; the centrality of the root cause error node is calculated based on its out-degree, conductive path coverage and associated error node number in the defect conductive chain network, and the higher the centrality of the root cause error node, the higher the correction priority; S430, generating a global correction suggestion, the global correction suggestion generating systematic logical defects and structural problems existing in the document based on the analysis of the topological structure of the defect conductive chain network, and providing modification directions for overall document quality optimization.

[0011] According to the above content, step S500 includes: S510, receiving a revised new version of the document submitted by the user, and aligning the new version of the document with the original version of the document; based on the text difference algorithm, comparing the new version of the document with the original version of the document line by line, identifying and positioning the text change area in the document where the increase, deletion and modification operations occur; S520, in the text change area, re-calculate the format specification features and text quality statistical features, and re-execute error detection to obtain the error state and quality deviation value after revision; S530, compare the error state of the revised change area with the quality deviation value with the state of the corresponding area in the original version document, and calculate the quality difference; the quality difference includes repaired errors, newly introduced errors and the change amount of the quality deviation value; the repaired errors are determined by comparing the difference between the error states before and after revision, the newly introduced errors are obtained by identifying the newly added error instances after revision, and the change amount of the quality deviation value is calculated by subtracting the deviation values of the corresponding areas of the new version and the original version.

[0012] According to the above content, step S600 includes: S610, generating an incremental proofreading report, the incremental proofreading report including a newly introduced error set represented by the quality difference and an unsolved root cause error node, and accurate positioning information of the root cause error node in the document; S620, store the quality difference data and the corresponding document revision record in association to construct a standard document sample library for context type library optimization; S630, based on the standard document sample library, incrementally train the context classification model used for context type labeling to optimize the classification accuracy of the context classification model; S640, based on the instances of error propagation in the standard document sample library, optimize the defect propagation rule library, add new propagation rules or adjust the confidence of existing rules; S650, based on the determination result of the root cause error in the standard document sample library, optimize the out-degree threshold and the recognition sensitivity of the cyclic dependency path in the root cause error determination logic.

[0013] An artificial intelligence-based document proofreading system, the system comprising: a document analysis module, an error detection module, a proofreading report module, a version comparison module and a model optimization module; The document analysis module includes a semantic analysis unit and a feature extraction unit; the semantic analysis unit is used for analyzing the document, dividing semantic units, and labeling the semantic unit types based on the context type library to generate document data with context labels; the feature extraction unit is used for extracting global format specification features and text quality statistical features from the document; The error detection module includes an initial detection unit, a defect propagation chain unit and a root cause error unit; the initial detection unit is used for calling corresponding rule sets based on the context labels for error detection, and at the same time comparing the extracted features with the standard sample vector space to calculate the quality deviation and obtain the initial error set; the defect propagation chain unit is used for taking the initial error set as a node and constructing a defect propagation chain network according to the defect propagation rule library; the root cause error unit is used for locating the root cause error node that causes multiple secondary errors by tracing the defect propagation chain network; The proofreading report module comprises a report generation unit and an incremental report generation unit; the report generation unit is configured to generate an intelligent proofreading report containing error positioning, priority correction strategy and global correction suggestion based on the root cause error node and the defect propagation chain; and the incremental report generation unit is configured to output an incremental proofreading report containing newly introduced errors and unsolved root cause errors based on the quality difference; The version comparison module comprises a change positioning unit and a quality difference unit; the change positioning unit is configured to receive the revised new version document, compare it with the original version document and locate the changed area; and the quality difference unit is configured to calculate the quality difference of the changed area; The model optimization module comprises a sample library unit and an optimization unit; the sample library unit is configured to store the quality difference data and the document revision record, and build a standard document sample library; and the optimization unit is configured to continuously optimize the context type labeling, the defect propagation rule library and the root cause error determination logic by using the quality difference data.

[0014] Compared with the prior art, the present application has the following beneficial effects: 1. The present application can accurately locate the root cause of multiple secondary errors by constructing a defect propagation chain network and analyzing the causal relationship between different errors, thereby providing a correction scheme that both treats the symptoms and treats the root cause, and greatly improving the document correction efficiency; 2. The present application combines rule-based error detection and vector space-based quantitative evaluation to comprehensively evaluate the document quality from multiple dimensions such as grammar, format, logic and data consistency, thereby improving the accuracy and comprehensiveness of proofreading work; 3. The present application focuses on newly introduced problems and unsolved root cause errors in the new round of proofreading by version comparison and quality difference calculation, thereby providing accurate incremental proofreading reports for users and optimizing the workflow of multiple rounds of proofreading. BRIEF DESCRIPTION OF DRAWINGS

[0015] Fig. 1 The step flowchart of the document proofreading method based on artificial intelligence of the present application; Fig. 2 The structural schematic diagram of the document proofreading system based on artificial intelligence of the present application. DETAILED DESCRIPTION

[0016] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. 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 skilled in the art without creative labor fall within the scope of protection of the present application.

[0017] Embodiment: As Figs. 1-2As shown, the present application provides a technical solution, a document proofreading method based on artificial intelligence, which comprises: S100, analyzing the document, dividing semantic units, and labeling semantic unit types based on a context type library to obtain document data with context labels; the context type library is obtained by training based on a standard document sample library; the training process is to correlate the text features and semantic functions of the semantic units in the standard document sample library to establish a mapping relationship from text features to context types; Specifically, step S100 comprises: S110, performing syntax structure and semantic dependency analysis on the input document to identify and divide independent semantic units that constitute the logical structure of the document; S120, extracting text features of each semantic unit, including syntax structure mode, keyword distribution, semantic role, and logical association relationship with adjacent units; syntax structure mode refers to the sentence structure and component collocation form of the semantic unit; keyword distribution refers to the position and frequency characteristics of core terms in the semantic unit; semantic role refers to the functions of each component in the semantic unit, such as statement, modification, and limitation; logical association relationship refers to the association form of cause and effect, progression, and parallelism between adjacent semantic units; For example: analyzing the document to be proofread and dividing it into 128 semantic units; extracting the text feature vector of each semantic unit, wherein: syntax structure mode feature dimension: 15, keyword distribution feature dimension: 20, semantic role feature dimension: 8, and logical association relationship feature dimension: 6; this is only an example and is not limited; S130, inputting the text features into a pre-trained context classification model; the context classification model learns from a large number of standard document samples through a machine learning algorithm to establish a mapping relationship from text features to context types; the pre-training process of the context classification model uses supervised learning, with the manually labeled semantic unit types in the standard document sample library as labels to optimize model parameters to improve classification accuracy; S140, based on the output of the context classification model, assigning a specific context type label to each semantic unit to generate document data with context labels; the context types include technical definition statement, data argumentation statement, argumentation statement, instance explanation statement, and background introduction statement; For example: inputting a 49-dimensional feature vector into a pre-trained context classification model, the model outputs the type probability distribution of each semantic unit; for example, the 25th semantic unit has a probability distribution of: technical definition statement: 0.12, data argumentation statement: 0.08, argumentation statement: 0.65, instance explanation statement: 0.10, and background introduction statement: 0.05; based on the maximum probability principle, the unit is labeled as an argumentation statement type; this is only an example and is not limited.

[0018] S200, based on the context label, using the corresponding rule set for initial error detection, while comparing the format and text quality features of the document with the standard sample vector space, calculating the quality deviation of each dimension, and obtaining the initial error set; Specifically, step S200 includes: S210, according to the context type label of the semantic unit, calling the special error checking rule set pre-bound with the context type label, and performing syntax rule compliance check and semantic logic consistency check on the semantic unit; Each context type label corresponds to a unique special error checking rule set, and the checking dimension and judgment standard of the special error checking rule set are customized according to the semantic function and quality requirement of the corresponding context type, to ensure the pertinence and accuracy of the check; For example: the 25th semantic unit, argumentation statement type, calling the rule set configured for the argumentation statement type, finding that there is an incomplete argument chain error, and the severity score is 0.75; This is only for illustration and does not limit; S220, extracting global format specification features and text quality statistical features from the document, the format specification features including title hierarchy structure, numbering sequence consistency and chart label specification, and the text quality statistical features including term usage consistency, paragraph length distribution and logical connection word density; S230, mapping the extracted format specification features and text quality statistical features to the multi-dimensional vector space learned and constructed based on a large number of standard document samples, calculating the Euclidean distance between the feature vector of the document to be corrected and the clustering center of the standard sample feature vector in each dimension as the quality deviation of each dimension; The multi-dimensional standard sample vector space is constructed by clustering analysis on the feature data of all samples in the standard document sample library, and the clustering center is the standard reference value of each feature dimension, to ensure the objectivity and uniformity of quality evaluation; S240, integrating the rule violation errors identified by the special error checking rule set and the statistical abnormal errors identified by the quality deviation calculation to generate the initial error set; During the integration process, the error location and error type are used for deduplication processing, and the feature dimension data and context label corresponding to each error are recorded.

[0019] For example: calculating the quality deviation of the document in the term usage consistency dimension: standard sample feature vector clustering center: 12.5, 8.3, 15.7, …, 9.2; The feature vector of the document to be corrected: 13.1, 7.8, 14.9, …, 8.7 (20 dimensions); The calculation of this dimension quality deviation D=2.34, which exceeds the threshold value 1.5, and is recorded as a statistical abnormal error; This is only for illustration and does not limit; Further, the context types include technical definition statements, data argument statements, reasoning argument statements, instance explanation statements, and background introduction statements; the special error checking rule set is configured based on different context type tags; the rule set configured for the technical definition statement type focuses on the uniqueness, completeness, and unambiguity of the core term definition; the rule set configured for the data argument statement type focuses on the unit uniformity of the numerical data, the normativity of the data source citation, and the rationality of the support relationship between the data and the conclusion; the rule set configured for the reasoning argument statement type focuses on the logical completeness of the cause-effect relationship in the argument chain and the effectiveness of the reasoning premise; the rule set configured for the instance explanation statement type focuses on the relevance of the instance to the abstract concept or rule being explained; and the rule set configured for the background introduction statement type focuses on the objectivity of the historical fact statement and the basis grammar compliance.

[0020] S300, taking the initial error set as nodes, analyzing the cause-effect relationship between the error nodes according to the defect propagation rule library, constructing a defect propagation chain network, and locating the root error node causing multiple secondary errors by tracing the network; Specifically, step S300 includes: S310, abstracting each error instance in the initial error set as a network node, and recording the error type, the semantic unit to which the network node belongs, and the context tag attribute of the network node; each network node has a unique identifier, and the attribute information thereof corresponds to the error details in the initial error set one by one; S320, calling the defect propagation rule library, which defines the cause-effect relationship logic between different error type nodes in a specific context; the defect propagation rule library includes error type association rules, context adaptation rules, and cause-effect strength determination rules, the error type association rules explicitly define the propagation possibility between different error types, the context adaptation rules limit the semantic scenarios in which the propagation relationship is established, and the cause-effect strength determination rules quantify the propagation association degree between error nodes; S330, based on the defect propagation rule library, traversing all error node pairs, if the node pair satisfies the cause-effect relationship condition defined in the defect propagation rule library, establishing a directed propagation edge between the two nodes, and constructing a directed defect propagation chain network; the direction of the propagation edge is from the cause error node to the effect error node, and the rule basis and the cause-effect strength value corresponding to the propagation relationship are recorded at the same time; S340, calculating the out-degree of each node in the defect propagation chain network, and identifying the circular dependency path in the network; the out-degree of the node is as follows: OD i =∑ j A ij ; Wherein, OD idenotes the out-degree of node i; i and j denote the indices of nodes, and for all i∈[1, n], that is, it applies to all nodes with indices from 1 to n in the network; when calculating the out-degree of node i, node i is fixed, and node j needs to traverse all nodes, that is, sum from 1 to n; n denotes the total number of nodes in the network; A ij denotes an element of the adjacency matrix, and A ij =1 if there is a conductive edge from node i pointing to node j, otherwise A ij =0; the circular dependency path is identified by traversing the directed conductive chain network, which refers to a closed path starting from a certain node and returning to the starting node after passing through several conductive edges; S350, the nodes with out-degree values exceeding a preset threshold or located at the starting points of the circular dependency paths are determined as root cause error nodes causing multiple secondary errors, and the conditions for determining the root cause error nodes are as follows: When the out-degree of node i is greater than the out-degree threshold θ, that is, OD i > θ, the root cause error node determination condition is met, the out-degree threshold θ is determined based on the out-degree statistical data of historical error nodes in a standard document sample library, and a dynamic threshold is set by statistically analyzing the out-degree distribution characteristics of historical root cause error nodes; for adapting to document scenarios of different complexities; When node i belongs to the starting node set V c of the circular dependency path, that is, i∈V c , the root cause error node determination condition is met; the starting node set of the circular dependency path is identified by traversing the directed conductive chain network, and contains all elements that are starting nodes of the circular dependency path; The node is determined as a root cause error node if it meets any of the above conditions, and the two conditions have the same determination effect.

[0021] For example: 35 error instances are abstracted as network nodes to construct a 35*35 adjacency matrix A; wherein: node 17, the core concept definition is fuzzy, pointing to node 25, the argument chain is incomplete; node 25 points to node 31, the conclusion lacks support; corresponding matrix elements: A 17,25 =1, A 25,31 =1; The out-degree of node OD 17 =4, OD 25 =2; identify the circular dependency path: node 8-node 12-node 19-node 8, forming a circular path; this is only for illustration and is not limiting.

[0022] S400, based on the root cause error node and the defect conductive chain, an intelligent proofreading report is generated, and the intelligent proofreading report includes error positioning, a priority correction strategy for the root cause error, and a global correction suggestion; Specifically, step S400 includes: S410, generate error positioning information, the error positioning information includes the root cause error node and detailed information of all secondary error nodes directly associated and indirectly associated with the root cause error node in the defect conduction chain network, the detailed information includes error content, error type, location information and belonging context type; S420, generate a priority correction strategy for the root cause error, the priority correction strategy includes determining the priority order of the correction operation based on the attributes of the root cause error node and the centrality of the root cause error node in the defect conduction chain network, and describing the secondary error nodes associated with the root cause error node that will be automatically eliminated or transformed into a pending state after the root cause error is corrected; the centrality of the root cause error node is calculated based on its out-degree, conduction path coverage range and associated error node number in the defect conduction chain network, the higher the centrality of the root cause error node, the higher the correction priority; S430, generate a global correction suggestion, the global correction suggestion is based on the analysis of the topological structure of the defect conduction chain network to generate systematic logical defects and structural problems existing in the document and provide modification direction for overall document quality optimization.

[0023] For example: generate error positioning information, including root cause error node 17 and its associated 8 secondary error nodes; calculate the centrality index of node 17: out-degree: 4, conduction path coverage range: 12 nodes, associated error node number: 9, centrality score: 8.5, correction priority: highest; this is only an example and is not limiting.

[0024] S500, receive the new version of the document revised by the user based on the intelligent proofreading report, compare the new version of the document with the original version, locate the changed area, and calculate the quality difference of the changed area in the related quality dimension; Specifically, step S500 includes: S510, receive the revised new version of the document submitted by the user, and align the new version of the document with the original version; based on the text difference algorithm, compare the new version of the document with the original version line by line, identify and locate the text change area in the document which has addition, deletion and modification operations; S520, in the text change area, recalculate the format specification features and text quality statistical features, and re-execute error detection to obtain the error state and quality deviation value after revision; S530, compare the error state of the revised change area with the quality deviation value with the state of the corresponding area in the original version document, and calculate the quality difference; the quality difference includes repaired errors, newly introduced errors and the change amount of the quality deviation value; the repaired errors are determined by comparing the difference between the error states before and after revision, the newly introduced errors are obtained by identifying the newly added error instances after revision, and the change amount of the quality deviation value is calculated by subtracting the deviation values of the corresponding areas of the new version and the original version.

[0025] For example: after the user revises, the quality indicators of the changed area are recalculated: 15 errors such as nodes 17, 25 and 31 are repaired, 2 format errors are newly introduced, and the term usage consistency dimension is reduced from 2.34 to 1.12.

[0026] S600, based on the quality difference, output an incremental proofreading report, and the incremental proofreading report includes newly introduced errors and unsolved root cause errors; the quality difference data is used to continuously optimize the context type labeling, the defect transmission rule library and the root cause error judgment logic; Specifically, step S600 includes: S610, generating an incremental proofreading report, and the incremental proofreading report includes a newly introduced error set represented by the quality difference and an unsolved root cause error node, and accurate positioning information of the root cause error node in the document; S620, store the quality difference data and the corresponding document revision record in association to construct a standard document sample library for optimizing the context type library; S630, based on the standard document sample library, incrementally train the context classification model used by the context type labeling to optimize the classification accuracy of the context classification model; S640, based on the error transmission instances in the standard document sample library, strongly optimize the defect transmission rule library, add new transmission rules or adjust the confidence of the existing rules; S650, based on the judgment result of the root cause error in the standard document sample library, optimize the out-degree threshold and the recognition sensitivity of the cyclic dependency path in the root cause error judgment logic.

[0027] The application provides another technical scheme, a document proofreading system based on artificial intelligence, which comprises a document analysis module, an error detection module, a proofreading report module, a version comparison module and a model optimization module. The document analysis module comprises a semantic analysis unit and a feature extraction unit; the semantic analysis unit is used for analyzing the document, dividing semantic units, labeling the types of the semantic units based on the context type library, and generating document data with context labels; and the feature extraction unit is used for extracting global format specification features and text quality statistical features from the document. The error detection module comprises an initial detection unit, a defect propagation chain unit and a root cause error unit; the initial detection unit is configured to perform error detection based on the context label to call a corresponding rule set, and to calculate a quality deviation degree by comparing the extracted features with a standard sample vector space to obtain an initial error set; the defect propagation chain unit is configured to take the initial error set as a node, and to construct a defect propagation chain network according to a defect propagation rule library; and the root cause error unit is configured to locate a root cause error node causing multiple secondary errors by tracing the defect propagation chain network; The proofreading report module comprises a report generation unit and an incremental report generation unit; the report generation unit is configured to generate an intelligent proofreading report containing error positioning, a priority correction strategy and a global correction suggestion based on the root cause error node and the defect propagation chain; and the incremental report generation unit is configured to output an incremental proofreading report containing newly introduced errors and unsolved root cause errors based on the quality difference; The version comparison module comprises a change positioning unit and a quality difference unit; the change positioning unit is configured to receive a revised new version document, and to compare the new version document with an original version document to position a change area; and the quality difference unit is configured to calculate a quality difference of the change area; The model optimization module comprises a sample library unit and an optimization unit; the sample library unit is configured to store quality difference data and document revision records, and to construct a standard document sample library; and the optimization unit is configured to continuously optimize context type labeling, a defect propagation rule library and root cause error determination logic by using the quality difference data.

[0028] It will be obvious to a person skilled in the art that the application is not limited to the details of the above-described exemplary embodiments, and the application can be implemented in other concrete forms without departing from the spirit or essential characteristics of the application. Therefore, the embodiments should be considered as exemplary and non-limiting, and the scope of the application is defined by the appended claims rather than the above description, and all changes falling within the meaning and range of the equivalent elements of the claims are intended to be embraced by the application. Any reference signs in the claims should not be considered as limiting the claims involved.

Claims

1. An artificial intelligence-based document proofreading method, characterized by: The method comprises: S100, parsing the document, dividing semantic units, and obtaining a document with context labels based on context type library labeling; the context type library is obtained based on a standard document sample library training; S200, based on the context labels, using the corresponding rule set for initial error detection, and comparing the format and text quality features of the document with the standard sample vector space to obtain an initial error set; S300, taking the initial error set as a node, analyzing the causal relationship between error nodes, constructing a defect propagation chain network, and locating the root cause error node that causes multiple secondary errors by tracing the defect propagation chain network; S400, based on the root cause error node and the defect propagation chain network, generating an intelligent proofreading report, the intelligent proofreading report including error positioning, priority correction strategy for root cause errors and global correction suggestions; S500, receiving a new version of the document revised by the user, comparing the new version of the document with the original version, positioning the changed area, and analyzing the quality difference of the changed area in the related quality dimension; S600, based on the quality difference, outputting an incremental proofreading report, the incremental proofreading report including newly introduced errors and unsolved root cause errors; using quality difference data to optimize the context type library, the defect propagation rule library and the root cause error judgment logic. 2.The method of claim 1, wherein: Step S100 comprises: S110, performing syntax structure and semantic dependency analysis on the input document, identifying and dividing independent semantic units constituting the logical structure of the document; S120, extracting text features of each semantic unit, the text features including syntax structure mode, keyword distribution, semantic role and logical association relationship with adjacent units; S130, inputting the text features into a pre-trained context classification model, the context classification model learning a large number of standard document samples through a machine learning algorithm to establish a mapping relationship from text features to context types; S140, based on the output of the context classification model, giving each semantic unit a specific context type label to generate a document with context labels; the context types include technical definition statement, data argument statement, argument argument statement, instance explanation statement and background introduction statement. 3.The method of claim 1, wherein: Step S200 comprises: S210, according to the context type label of the semantic unit, calling the special error checking rule set pre-bound with the context type label to perform syntax rule compliance check and semantic logic consistency check on the semantic unit; S220, extracting global format specification features and text quality statistical features from the document, the format specification features including title hierarchy structure, numbering sequence consistency and chart labeling specification, and the text quality statistical features including term usage consistency, paragraph length distribution and logical connection word density; S230, mapping the extracted format specification features and the text quality statistical features to a multi-dimensional vector space constructed based on a large number of standard document samples, calculating the Euclidean distance between the feature vector of the document to be proofread and the clustering center of the feature vector of the standard sample in each dimension as the quality deviation degree of each dimension; S240, integrating the rule violation errors identified by the special error checking rule set and the statistical abnormal error identified by the quality deviation degree calculation to generate an initial error set.

4. The document proofreading method based on artificial intelligence according to claim 3, characterized in that: the context types include technical definition statements, data argument statements, reasoning argument statements, instance description statements, and background introduction statements; and the special error checking rule set is configured based on different context type labels.

5. The method of claim 1, wherein: Step S300 includes: S310, abstracting each error instance in the initial error set as a network node, and recording the error type, the semantic unit to which the network node belongs, and the context label attribute of the network node; S320, calling a defect propagation rule library, the defect propagation rule library defining the causal relationship logic between different error type nodes in a specific context; S330, based on the defect propagation rule library, traversing all error node pairs, and if a node pair satisfies the causal relationship condition defined in the defect propagation rule library, establishing a directional propagation edge between the two nodes to construct a directed defect propagation chain network; S340, calculating the out-degree of each node in the defect propagation chain network, and identifying the circular dependency path in the network; the out-degree of the node is calculated as follows: OD i =∑ j A ij ; where OD i represents the out-degree of node i; i and j represent the indices of nodes, and V i e [1, n] applies to all nodes with indices from 1 to n in the network; n represents the total number of nodes in the network; A ij represents an element of the adjacency matrix, and A ij = 1 if there is a conductive edge from node i pointing to node j, otherwise A ij = 0. S350, determining a root cause error node that triggers multiple secondary errors as a node whose out-degree value exceeds a preset threshold or is located at the starting point of a circular dependency path; the determination conditions include: a first determination condition that the out-degree value of the node is greater than a preset out-degree threshold, the out-degree threshold being determined based on the out-degree statistical data of historical error nodes in a standard document sample library; and a second determination condition that the node belongs to a circular dependency path starting node set, the circular dependency path starting node set being identified by traversing the directed propagation chain network and containing all elements that are starting nodes of the circular dependency path; wherein, the node is determined as a root cause error node if it satisfies any of the determination conditions, and the two determination conditions have equal effect.

6. The method of claim 1, wherein: Step S400 includes: S410, generating error positioning information, the error positioning information including the root cause error node and detailed information of all secondary error nodes directly associated and indirectly associated with the root cause error node in the defect propagation chain network, the detailed information including error content, error type, location information, and context type; S420, generating a priority correction strategy for the root cause error, the priority correction strategy including determining the priority order of correction operations based on the attributes of the root cause error node and the centrality of the root cause error node in the defect propagation chain network, and describing the secondary error nodes associated with the root cause error node that will be automatically eliminated or converted to a pending state after the root cause error is corrected. S430, generating a global correction suggestion, the global correction suggestion is generated based on the topology analysis of the defect propagation chain network, and a systematic logical defect and a structural problem existing in the document are generated, and a modification direction for optimizing the overall document quality is provided.

7. The method of claim 1, wherein: Step S500 includes: S510, receiving a user-submitted revised new version document, and version aligning the new version document with the original version document; based on a text difference algorithm, comparing the new version document with the original version document line by line, identifying and locating the text change area in which the addition, deletion and modification operations occur in the document; S520, in the text change area, re-calculate the format specification features and the text quality statistical features, and re-execute error detection to obtain the error state and quality deviation value after revision; S530, comparing the error state and quality deviation value of the revised change area with the state of the corresponding area in the original version document, and calculating the quality difference; the quality difference includes repaired errors, newly introduced errors and the change amount of quality deviation value. 8.The method of claim 1, wherein: Step S600 includes: S610, generating an incremental proofreading report, the incremental proofreading report includes a newly introduced error set represented by quality difference and an unsolved root cause error node, and accurate positioning information of the root cause error node in the document; S620, storing the quality difference data and the corresponding document revision record in association to construct a standard document sample library for context type library optimization; S630, based on the standard document sample library, incrementally training the context classification model used for context type labeling to optimize the classification accuracy of the context classification model; S640, based on the error propagation instances in the standard document sample library, strengthening and expanding the defect propagation rule library, adding new propagation rules or adjusting the confidence of existing rules; S650, based on the determination result of the root cause error in the standard document sample library, optimizing the out-degree threshold value in the root cause error determination logic and the recognition sensitivity of the circular dependency path.

9. An artificial intelligence-based document proofing system, characterized by: The system includes a document analysis module, an error detection module, a proofreading report module, a version comparison module and a model optimization module; The document analysis module includes a semantic analysis unit and a feature extraction unit; the semantic analysis unit is used for analyzing the document, dividing semantic units, and labeling the semantic unit types based on the context type library to generate document data with context labels; the feature extraction unit is used for extracting global format specification features and text quality statistical features from the document; The error detection module comprises an initial detection unit, a defect propagation chain unit and a root cause error unit; the initial detection unit is configured to call a corresponding rule set based on the context label to perform error detection, and to calculate a quality deviation degree by comparing the extracted features with a standard sample vector space, thereby obtaining an initial error set; the defect propagation chain unit is configured to take the initial error set as a node, and to construct a defect propagation chain network according to a defect propagation rule library; and the root cause error unit is configured to locate a root cause error node causing a plurality of secondary errors by tracing the defect propagation chain network. The proofreading report module comprises a report generation unit and an incremental report generation unit; the report generation unit is configured to generate an intelligent proofreading report containing error positioning, a priority correction strategy and a global correction suggestion based on the root cause error node and the defect propagation chain; and the incremental report generation unit is configured to output an incremental proofreading report containing newly introduced errors and unsolved root cause errors based on a quality difference. The version comparison module comprises a change positioning unit and a quality difference unit; the change positioning unit is configured to receive a revised new version document, and to compare the new version document with an original version document to position a change region; and the quality difference unit is configured to calculate a quality difference of the change region. The model optimization module comprises a sample library unit and an optimization unit; the sample library unit is configured to store quality difference data and document revision records, and to construct a standard document sample library; and the optimization unit is configured to continuously optimize context type labeling, a defect propagation rule library and root cause error determination logic by using the quality difference data.

Images