Method for checking Chinese homophone wrongly written characters and similar characters based on artificial intelligence
Through methods based on deep learning and natural language processing, a fine Chinese typo verification process is constructed, which solves the problem of low accuracy of homophone typo verification in the existing technology, and realizes efficient and accurate Chinese text verification to adapt to the needs of different contexts.
Patent Information
- Application Number
- CN202510199461.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-02-24
- Publication Date
- 2025-07-25
AI Technical Summary
The existing Chinese typo verification methods have low accuracy, especially homophone typos are difficult to accurately verify in different contexts, and are costly to maintain and lack flexibility.
Using deep learning-based models combined with natural language processing technology, through steps such as text segmentation, model analysis, and obfuscation verification, we construct a fine typo verification method, including file reception, preprocessing, text segmentation, model analysis and error correction, obfuscation verification and file annotation, and use advanced technologies such as MacBERT to improve the verification accuracy.
It realizes efficient and accurate verification of homophone typos and similar characters in Chinese text. The model has strong learning ability and generalization ability, adapts to different context needs, and significantly improves the accuracy and efficiency of calibration.
Smart Images
Figure CN120373291A_ABST
Abstract
Description
Technical Field
[0001] The invention belongs to the field of information technology, and particularly relates to a method for checking Chinese homophone typos and similar characters based on artificial intelligence. Background Art
[0002] In the context of the profound and complex Chinese character culture, Chinese characters have the dual characteristics of both phonetic and semantic representation. This characteristic makes it difficult to avoid frequent typos in the use of Chinese characters. Typos are mainly divided into two categories: homophonic typos and similar typos. They not only affect the accuracy and readability of the text, but also may cause misunderstanding and confusion of information. Although with the rapid progress of artificial intelligence technology, some verification methods for Chinese character typos have emerged, these methods still have obvious shortcomings. Most of the Chinese character typo verification methods in the prior art rely on building a typo library for matching and retrieval. However, this method is limited by factors such as the coverage and update speed of the typo library, as well as the accuracy of sentence decomposition, resulting in low verification accuracy, especially for homophonic typos. Since there are many homophones in Chinese characters and the frequency and habits of using homophones in different contexts are different, it is difficult to accurately verify them through a simple typo library. In addition, existing methods often lack flexibility, are difficult to adapt to the typo verification needs in different contexts, and have high maintenance costs. Summary of the invention
[0003] 1. Technical issues to be resolved In order to solve the problems raised by the prior art in the above-mentioned background technology, the present invention proposes a method for checking Chinese homophones, typos and similar characters based on artificial intelligence. The method splits the text, combines the model to analyze the probability of character occurrence, and constructs a confusion set for special processing to increase the recall rate and accuracy, thereby achieving accurate verification of homophones, typos and similar characters in Chinese sentences. The method combines natural language processing and deep learning technology to achieve accurate verification of homophones, typos and similar characters in Chinese texts.
[0004] (II) Technical solution The present invention is implemented by the following technical scheme: The present invention proposes a method for checking Chinese homophone typos and similar characters based on artificial intelligence, and the method comprises the following steps: a. Document reception: Receive document files to be processed; If the document file is in editable text format, go directly to the next step; If the document file is in image format, perform optical character recognition (OCR) to convert it into editable text format and proceed to the next step; b. Preprocessing: Use natural language processing techniques to preprocess the editable text format to form pure text; c. Text segmentation: Natural language processing techniques use text segmentation methods to initially split the pure text into sentences or phrases based on punctuation marks and semantic information. The sentences or phrases include homophonic characters or similar-looking characters and their respective complete sentence meanings; d. Model analysis and error correction: Input the sentence or phrase into a pre-trained model for analysis. Based on deep learning techniques, through learning a large amount of text data, the model can accurately judge the probability of character occurrence and perform error correction accordingly; For homophonic characters or similar-looking characters that appear, the model will give a prompt and mark them in the original text; e. Confusion set verification: Construct a confusion set that contains common homophonic and similar-looking misspelled characters, as well as their usage frequencies and habits in different contexts; During the process of model analysis and error correction, refer to the information in the confusion set to further verify and correct the determined possible errors; f. Loop processing and result integration: The model checks whether the currently processed sentence or phrase is the last sentence in the document; If not, return to the text segmentation step to continue processing the next sentence or phrase; If so, integrate all processing results and enter the next step; g. File annotation and correction: The model directly annotates the identified misspelled characters in the editable text format, marks the error positions, gives correct suggestions, and the user directly corrects the misspelled characters according to the suggestions h. End: Display a processing completion prompt and provide the processed document for the user to view and save.
[0005] Preferably, the text segmentation method includes the following steps: Establish a basic database, including a basic vocabulary database of daily expressions, poems, commonly used network words, and special name words, as well as a Chinese character word formation priority principle judgment database that can judge the priority order principle of the combination of individual Chinese characters into words; Match and compare the sentence or phrase with the basic vocabulary database, and use the judgment database to split the sentence into sentences or phrases that can express complete meanings; Perform part-of-speech classification processing on the split sentences or phrases for subsequent verification of homophonic characters or similar-looking characters and characters with their respective complete sentence meanings.
[0006] Preferably, the model is based on deep learning technology, especially a deep learning model in the field of natural language processing, and the model includes but is not limited to recurrent neural network (RNN), long short-term memory network (LSTM), Transformer, and MacBERT.
[0007] Preferably, the training process of the deep learning model includes collecting a large amount of text data, preprocessing the data, constructing the model structure, setting the loss function and optimization algorithm, and making the model converge through iterative training.
[0008] Preferably, in step III) of the text segmentation method, a classification processing database is also established, and the classification processing database is used to store sentences or phrases after part-of-speech classification or other related classification processing.
[0009] Preferably, in step b, the preprocessing includes: Removing redundant information: deleting irrelevant characters, special symbols, and extra spaces in the text to make the text cleaner and more standardized; Text normalization: converting non-standard characters in the text into standard characters to ensure the consistency of the text.
[0010] Preferably, the preprocessing further includes: Word segmentation preparation: For Chinese text, it may be necessary to perform preparatory work before word segmentation, identify and process numbers, English words, and proper nouns to make the subsequent word segmentation steps more accurate; Format adjustment: Adjust the format of the text according to the requirements of subsequent processing steps to convert the text into a format suitable for analysis.
[0011] Preferably, in step g, the model annotates the identified typos, and the annotation also includes the frequency of occurrence of typos.
[0012] The present invention also provides a system, and the system is implemented based on the artificial intelligence-based Chinese homophone typo and similar character verification method described in any one of the above, and the system includes: A file receiving module for receiving a document file to be processed; A preprocessing module that uses natural language processing technology to preprocess the received text to form pure text; A text segmentation module that uses the text segmentation method to split the pure text into sentences or phrases; A model analysis and error correction module that inputs the sentences or phrases into a pre-trained deep learning model for analysis and error correction; A confusion set verification module that constructs a confusion set and further verifies and corrects the model error correction results with reference to the information in the confusion set; The loop processing and result integration module processes each sentence or phrase in the document in a loop and integrates all processing results; The document annotation and correction module annotates the recognized typos in the text, marks the error positions, gives correct suggestions, and records the frequency of occurrence of the typos; The user interface module is used to display the processing results, receive user input, and provide user interaction functions; The storage module is used to store the preprocessed text, segmented sentences or phrases, model analysis results, confusion set verification results, and final processing results; The network connection module is used to communicate with other systems or servers to achieve remote processing or data sharing.
[0013] Preferably, the system further includes: a learning module for training the deep learning model, including collecting a large amount of text data, preprocessing the data, constructing the model structure, setting the loss function and optimization algorithm, and making the model converge through iterative training; The update module is used to update the basic database, judgment database, classification processing database, and confusion set regularly or according to user feedback to improve the accuracy and adaptability of the system; The security module is used to protect the system from malicious attacks and data leakage, and ensure user privacy and document security.
[0014] The present invention also provides a computer-readable storage medium, on which a computer program is stored. When the program runs on a computer, it executes the verification method described in any one of the above.
[0015] The present invention also provides a computer for executing the verification method described in any one of the above, including: A processor for executing an instruction set stored in a non-volatile memory to implement the functions of file reception, preprocessing, text segmentation, model analysis and error correction, confusion set verification, loop processing and result integration, document annotation and correction, and ending the processing flow; A non-volatile memory for storing an operating system, application programs, and an instruction set of the method described in claims 1 to 8; An input / output device for receiving a document file input by a user, displaying the processing results, and providing a user interaction interface; A network interface for communicating with other computers or servers to achieve remote data exchange or collaborative work; Wherein, by executing the instruction set, the processor can automatically complete the intelligent verification of Chinese homophonic typos and similar-shaped characters, and improve the accuracy and efficiency of document processing.
[0016] (III) Beneficial effects Compared with the prior art, the present invention has the following beneficial effects: The core of the present invention lies in its model based on deep learning technology, which has powerful learning ability and generalization ability. Through learning a large amount of text data, the model can continuously optimize and improve its error correction ability, and flexibly cope with the needs of spelling mistake verification in different contexts. Whether it is a one-to-one or one-to-many verification mode, the present invention can handle it with ease; In the process of model construction, the present invention adopts advanced natural language processing technologies such as MacBERT, which significantly improves the accuracy of the model. At the same time, the present invention also constructs a fine confusion set, which includes common homophonic and similar-looking spelling mistakes, as well as their usage frequencies and habits in different contexts, providing an important reference basis for the model to analyze and correct errors. BRIEF DESCRIPTION OF THE DRAWINGS
[0017] By reading the following detailed description of non-limiting embodiments with reference to the accompanying drawings, other features, objects, and advantages of the present invention will become more apparent: Figure 1 It is a flowchart of the method of the present invention.
[0018] Figure 2 It is a framework diagram of the system of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS
[0019] In this technical solution: In order to make the purpose, technical solution and advantages of the present invention clearer, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention, and are not used to limit the present invention.
[0020] Referring to Figure 1 As shown, the present invention is realized through the following technical solutions: The present invention proposes a method for verifying homophonic and similar-looking Chinese characters based on artificial intelligence, and the method includes the following steps: a. File reception: Receive the document file to be processed; If the document file is in an editable text format, directly proceed to the next step; If the document file is in an image format, perform optical character recognition (OCR) to convert it into an editable text format and proceed to the next step; b. Preprocessing: Use natural language processing technology to preprocess the editable text format to form pure text; c. Text segmentation: The natural language processing technology uses a text segmentation method to preliminarily segment the pure text into sentences or phrases according to punctuation marks and semantic information. The sentences or phrases include homophonic or similar-looking characters and their respective complete sentence meanings; d. Model analysis and error correction: Input the sentence or phrase into a pre-trained model for analysis. This model is based on deep learning technology and can accurately judge the probability of character appearance and correct errors accordingly through learning a large amount of text data. For homophonic or similar-looking characters that appear, the model will give a prompt and mark them in the original text. e. Confusion set verification: Construct a confusion set that contains common homophonic and similar-looking misspelled characters, as well as their usage frequencies and habits in different contexts. During the process of model analysis and error correction, refer to the information in the confusion set to further verify and correct the determined possible errors. f. Loop processing and result integration: The model checks whether the currently processed sentence or phrase is the last sentence in the document. If not, return to the text segmentation step to continue processing the next sentence or phrase. If so, integrate all processing results and enter the next step. g. File annotation and correction: The model annotates the identified misspelled characters directly in the text in an editable text format, marking the error location and giving correct suggestions, and the user directly corrects the misspelled characters according to the suggestions. h. End: Display a prompt indicating that the processing is completed and provide the processed document for the user to view and save.
[0021] Among them, the text segmentation method includes the following steps: Establish a basic database, including a basic word library of daily expressions, poems, commonly used network words, and special name words, as well as a database for judging the priority principle of word formation between individual Chinese characters to determine the priority order of word combination. Match and compare the sentence or phrase with the basic word library, and use the judgment database to split the sentence into sentences or phrases that can express complete meanings. Perform part-of-speech classification processing on the split sentences or phrases for subsequent verification of homophonic or similar-looking characters and characters that express complete sentence meanings respectively.
[0022] Among them, the model is based on deep learning technology, especially a deep learning model in the field of natural language processing. The model includes but is not limited to recurrent neural network (RNN), long short-term memory network (LSTM), Transformer, and MacBERT.
[0023] Among them, the training process of the deep learning model includes collecting a large amount of text data, preprocessing the data, constructing the model structure, setting the loss function and optimization algorithm, and making the model converge through iterative training.
[0024] Among them, in step III) of the text segmentation method, a classification processing database is also established, which is used to store sentences or phrases after part-of-speech classification or other relevant classification processing.
[0025] Among them, in step b, the preprocessing includes: Removing redundant information: deleting irrelevant characters, special symbols, and extra spaces in the text to make the text cleaner and more standardized; Text normalization: converting non-standard characters in the text into standard characters to ensure the consistency of the text.
[0026] Among them, the preprocessing also includes: Word segmentation preparation: for Chinese text, it may be necessary to perform preparatory work before word segmentation, identify and process numbers, English words, and proper nouns to make the subsequent word segmentation steps more accurate; Format adjustment: adjusting the format of the text according to the requirements of subsequent processing steps to convert the text into a format suitable for analysis.
[0027] Among them, in step g, the model annotates the identified typos, and the annotation also includes the frequency of occurrence of typos.
[0028] Refer to Figure 2 As shown, the present invention also provides a system, which is implemented based on the artificial intelligence-based Chinese homophone typo and similar character verification method described in any one of the above, and the system includes: A file receiving module, which is used to receive the document file to be processed; A preprocessing module, which uses natural language processing technology to preprocess the received text to form a pure text; A text segmentation module, which uses the text segmentation method to split the pure text into sentences or phrases; A model analysis and error correction module, which inputs the sentences or phrases into a pre-trained deep learning model for analysis and error correction; A confusion set verification module, which constructs a confusion set and further verifies and corrects the model error correction results with reference to the information in the confusion set; A loop processing and result integration module, which loops through each sentence or phrase in the document and integrates all processing results; A file annotation and correction module, which annotates the identified typos in the text, marks the error positions, gives correct suggestions, and records the frequency of occurrence of typos; A user interface module, which is used to display the processing results, receive user input, and provide user interaction functions; A storage module for storing preprocessed text, segmented sentences or phrases, model analysis results, confusion set verification results, and final processing results; A network connection module for communicating with other systems or servers to enable remote processing or data sharing.
[0029] Among them, the system further includes: a learning module for training a deep learning model, including collecting a large amount of text data, preprocessing the data, constructing a model structure, setting a loss function and an optimization algorithm, and making the model converge through iterative training; An update module for periodically or according to user feedback updating the basic database, judgment database, classification processing database, and confusion set to improve the accuracy and adaptability of the system; A security module for protecting the system from malicious attacks and data leakage, ensuring user privacy and document security.
[0030] The present invention also provides a computer-readable storage medium having a computer program stored thereon, and when the program runs on a computer, it executes the verification method described in any one of the above.
[0031] The present invention also provides a computer for executing the verification method described in any one of the above, including: A processor for executing an instruction set stored in a non-volatile memory to implement the functions of file reception, preprocessing, text segmentation, model analysis and error correction, confusion set verification, loop processing and result integration, file annotation and correction, and ending the processing flow; A non-volatile memory for storing an operating system, application programs, and an instruction set of the method described in claims 1 to 8; An input / output device for receiving a document file input by a user, displaying a processing result, and providing a user interaction interface; A network interface for communicating with other computers or servers to enable remote data exchange or collaborative work; Among them, by executing the instruction set, the processor can automatically complete the intelligent verification of Chinese homophonic misspelled words and similar-shaped words, improving the accuracy and efficiency of document processing The present invention proposes an innovative method for a computer to verify Chinese homophonic misspelled words and similar-shaped words. This method deeply integrates natural language processing (NLP) and deep learning technologies, aiming to achieve efficient and accurate recognition and correction of homophonic misspelled words and similar-shaped words in Chinese texts. Compared with traditional misspelled word verification means, the present invention demonstrates significant advantages and many beneficial effects: The core of the present invention lies in its model based on deep learning technology, which has powerful learning ability and generalization ability. Through learning a large amount of text data, the model can continuously optimize and improve its error correction ability, and flexibly meet the needs of typo checking in different contexts. Whether it is a one-to-one or one-to-many checking mode, the present invention can handle it with ease; In the process of model construction, the present invention adopts advanced natural language processing technologies such as MacBERT, which significantly improves the accuracy of the model. At the same time, the present invention also constructs a fine confusion set, which includes common homophonic and shape-similar typos, as well as their usage frequencies and habits in different contexts, providing an important reference basis for the model to analyze and correct errors; The processing flow of the present invention includes file reception, preprocessing, text segmentation, model analysis and error correction, loop processing and result integration, file annotation and correction, and the end stage; First of all, through the file reception step, the present invention can flexibly process document files in different formats and convert them into editable text formats, providing convenience for subsequent processing; In the preprocessing stage, natural language processing technology is used to clean and standardize the text, remove redundant information, and ensure the cleanliness and consistency of the text. At the same time, through word segmentation preparation and format adjustment, a solid foundation is laid for subsequent text segmentation and model analysis and error correction; The text segmentation step is one of the keys of this method. By establishing a basic database and a database based on the principle of giving priority to Chinese character word formation, the present invention can accurately split long texts into short sentences that contain homophonic or shape-similar characters and express complete sentence meanings. This step provides strong support for subsequent homophonic or shape-similar character checking; The model analysis and error correction step makes full use of the advantages of deep learning technology. Through a pre-trained deep learning model, it analyzes sentences or phrases, accurately judges the probability of character occurrence, and corrects errors accordingly. For homophonic or shape-similar characters that appear, the model will give a prompt and mark them in the original text, greatly improving the accuracy and efficiency of error correction; The loop processing and result integration step ensures the coherence and integrity of the entire processing flow. The model checks whether the currently processed sentence or phrase is the last sentence in the document. If not, it returns to the text segmentation step to continue processing the next sentence or phrase; if so, it integrates all processing results and enters the next step; In the file annotation and correction stage, the model directly annotates the identified typos in the text, marks the error positions and gives correct suggestions. Users can directly correct the typos according to the suggestions, greatly improving the efficiency and accuracy of document processing; Finally, at the end stage, the method will display a prompt indicating that the processing is complete and provide the processed document for the user to view and save. The entire processing flow is both efficient and convenient, providing great convenience for users. To sum up, the artificial intelligence-based Chinese homophone typos and similar-looking characters verification method proposed in the present invention, through a series of carefully designed steps and technical means, realizes the efficient and accurate recognition and correction of homophone typos and similar-looking characters in Chinese documents. In the future, with the continuous advancement of technology and the continuous expansion of application scenarios, the present invention is expected to play an important role in document processing, education and guidance, publishing and printing and other fields, and contribute to improving text quality and efficiency.
[0032] Example 1: Intelligent verification of Chinese homophone typos and similar characters in news articles Target audience: A news article about the latest developments in science and technology, in .doc format; Specific text content (excerpt): "With the rapid development of science and technology, artificial intelligence (AI) technology has penetrated into every aspect of our lives. Among them, the application of AI in text processing is particularly prominent, such as the intelligent proofreading system, which can effectively reduce typos and grammatical errors in texts and improve the quality of texts. However, due to the complexity of Chinese characters, the problem of homophones and similar characters still troubles many writers;" Parameter setting and process: Document reception: Receive news releases in .doc format; Preprocessing: remove redundant information (such as advertising links, irrelevant images, etc.), normalize text (convert non-standard characters to standard characters), prepare for word segmentation (recognize and process numbers, English words, and proper nouns); Text segmentation: Using the basic vocabulary and the principle of Chinese character grouping priority, the text is segmented into sentences or phrases, such as "With the rapid development of science and technology", "Artificial intelligence (AI) technology has penetrated into every aspect of our lives", etc. Model analysis and error correction: Use a pre-trained deep learning model (such as Transformer) and input the segmented sentences or phrases for analysis; for example, the model will recognize that the word "飞" in "飞发展" is a typo and should be "速" in "飞速发展"; Confusion set verification: Build a confusion set containing common homophones and similar characters, such as "速" and "飞", "辩" and "辨", and further verify and correct the model results based on the context; Loop processing and result integration: Perform the above processing on each sentence or phrase in the document and integrate all processing results; File annotation and correction: Mark typos and correct suggestions in the original text. For example, mark "flying development" as incorrect and give the correct suggestion "rapid development"; at the same time, indicate the frequency of the occurrence of the typo. For example, the character "flying" appears incorrectly 1 time here; End: Display a prompt indicating that the processing is completed, and provide the processed document for the user to view and save.
[0033] Example 2: Intelligent verification of Chinese homophonic typos and similar-looking characters in online novels Target: A chapter of an online novel about the grudges and feuds in the martial arts world, in.txt format; Specific text content (excerpt): "In the martial arts world, there is a knight named Li Xunhuan. He holds a long sword and travels the martial arts world, upholding justice; One day, he came to a small town and met a woman named Lin Shiyin. The two fell in love at first sight; However, Lin Shiyin has a childhood fiancé named Long Xiaoyun, which puts Li Xunhuan in a dilemma;" Parameter settings and process: File reception: Receive a chapter of an online novel in.txt format; Preprocessing: Remove redundant information (such as online advertisements, emoticons, etc.), and retain Chinese punctuation marks and special symbols to maintain the original style; In terms of text normalization, since online novels may contain some non-standard vocabulary, no traditional-simplified conversion is performed; When preparing for word segmentation, identify and process online buzzwords, character names, place names, etc.; Text segmentation: Use a basic vocabulary library and a commonly used online vocabulary library to cut the text into sentences or phrases, such as "In the martial arts world, there is a knight named Li Xunhuan", "He holds a long sword and travels the martial arts world", etc.; Model analysis and error correction: Select a pre-trained deep learning model (such as BERT), and input the cut sentences or phrases for analysis; For example, the model will identify that the character "travel" in "travel the martial arts world" is not incorrect here, but if the character "travel" in "uphold justice" is miswritten as the homophone "form", it will be marked as incorrect by the model; Confusion set verification: Build a confusion set containing common homophonic typos and similar-looking characters in online novels, such as "travel" and "form", "knight" and "coerce", etc., and further verify and correct the model results in combination with the context; Loop processing and result integration: Perform the above processing on each sentence or phrase in the document, and integrate all processing results; Pay attention to the accuracy when processing online buzzwords and special usages; File annotation and correction: Mark typos and correct suggestions in the original text. For example, mark "form knight and uphold justice" as incorrect and give the correct suggestion "travel knight and uphold justice"; at the same time, indicate the frequency of the occurrence of the typo. For example, the character "form" appears incorrectly 1 time here; End: Displays a prompt that the processing is complete and provides the processed document for the user to view and save; the user can revise the original text according to the suggestions to improve the text quality; The above two embodiments demonstrate the application of the artificial intelligence-based Chinese homophone typos and similar characters verification method in different text types (news articles and online novels); through preprocessing, text segmentation, model analysis and error correction, confusion set verification and other steps, the method can effectively identify and correct typos in the text.
[0034] It is obvious to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the present invention can be implemented in other specific forms without departing from the spirit or essential features of the present invention. Therefore, from any point of view, the embodiments should be regarded as exemplary and non-restrictive, and the scope of the present invention is defined by the appended claims rather than the above description, and it is intended that all changes falling within the meaning and scope of the equivalent elements of the claims are included in the present invention. Any figure mark in the claims should not be regarded as limiting the claims involved. In addition, it is obvious that the word "comprising" does not exclude other units or steps, and the singular does not exclude the plural. Multiple units or devices stated in the device claim can also be implemented by one unit or device through software or hardware. The words first, second, etc. are used to indicate names, and do not indicate any particular order.
[0035] Finally, it should be noted that the above is only a specific example of the present invention. Obviously, the present invention is not limited to the above example, and there are many variations. All variations that can be directly derived or associated with the content disclosed by a person skilled in the art should be considered as the protection scope of the present invention.
Claims
1. A method for verifying Chinese homophonic misspelled words and similar-looking characters based on artificial intelligence, characterized in that: The method includes the following steps: a. Document reception: Receive the document file to be processed; If the document file is in an editable text format, directly proceed to the next step; If the document file is in an image format, perform optical character recognition (OCR) to convert it into an editable text format and proceed to the next step; b. Preprocessing: Use natural language processing technology to preprocess the editable text format to form pure text; c. Text segmentation: Use natural language processing technology and text segmentation methods to initially split the pure text into sentences or phrases based on punctuation marks and semantic information. The sentences or phrases include homophonic characters or characters with similar forms and each expresses a complete sentence meaning; d. Model analysis and error correction: Input the sentence or phrase into a pre-trained model for analysis. The model is based on deep learning technology and can accurately judge the probability of character occurrence through learning a large amount of text data and perform error correction accordingly; For the homophonic characters or characters with similar forms that appear, the model will give a prompt and mark them in the original text; e. Confusion set verification: Construct a confusion set that contains common homophonic and shape-similar misspelled characters, as well as their usage frequencies and habits in different contexts; During the process of model analysis and error correction, refer to the information in the confusion set to further verify and correct the determined possible errors; f. Loop processing and result integration: The model checks whether the currently processed sentence or phrase is the last sentence in the document; If not, return to the text segmentation step to continue processing the next sentence or phrase; If so, integrate all processing results and proceed to the next step; g. File annotation and correction: The model directly annotates the identified misspelled characters in the editable text format, marks the error positions, gives correct suggestions, and the user directly corrects the misspelled characters according to the suggestions h. End: Display a prompt indicating that the processing is completed and provide the processed document for the user to view and save.
2. The method for verifying Chinese homophonic misspelled words and similar-shaped characters based on artificial intelligence according to claim 1, characterized in that: The text segmentation method includes the following steps: Establish a basic database, including a basic vocabulary database of daily expressions, poems, commonly used network words, and special name words, and a Chinese character word formation priority principle judgment database that can judge the priority order of the combination of individual Chinese characters into words; Match and compare the sentence or phrase with the basic vocabulary database, and use the judgment database to split the sentence into sentences or phrases that can express complete meanings; Perform part-of-speech classification processing on the split sentences or phrases for subsequent verification of homophonic characters or characters with similar forms and characters that each express a complete sentence meaning.
3. The method for verifying Chinese homophonic misspelled words and similar-shaped characters based on artificial intelligence according to claim 1, wherein: The model is based on deep learning technology, especially a deep learning model in the field of natural language processing. The model includes but is not limited to recurrent neural network (RNN), long short-term memory network (LSTM), Transformer, and MacBERT.
4. The method for checking Chinese homophonic misspelled words and similar-shaped characters based on artificial intelligence according to claim 3, characterized in that: The training process of the deep learning model includes collecting a large amount of text data, preprocessing the data, constructing the model structure, setting the loss function and optimization algorithm, and making the model converge through iterative training.
5. The Chinese homophone misspelling and similar - form character verification method based on artificial intelligence according to claim 1, wherein: In step III) of the text segmentation method, a classification processing database is also established, which is used to store sentences or phrases after part-of-speech classification or other related classification processing.
6. The method for verifying Chinese homophonic misspelled words and similar-shaped characters based on artificial intelligence according to claim 1, wherein: In step b), the preprocessing includes: Removing redundant information: deleting irrelevant characters, special symbols, and extra spaces in the text to make the text cleaner and more standardized; Text normalization: converting non-standard characters in the text into standard characters to ensure the consistency of the text.
7. A method for verifying Chinese homophonic misspellings and similar-shaped characters based on artificial intelligence according to claim 6, characterized in that: The preprocessing also includes: Word segmentation preparation: for Chinese text, it may be necessary to perform preparatory work before word segmentation, identify and process numbers, English words, and proper nouns to make the subsequent word segmentation steps more accurate; Format adjustment: adjusting the format of the text according to the requirements of subsequent processing steps to convert the text into a format suitable for analysis.
8. A Chinese homophone misspelling and similar character verification method based on artificial intelligence according to claim 1, characterized in that: In step g), the model annotates the identified misspelled words, and the annotation also includes the frequency of occurrence of the misspelled words.
9. A system, characterized in that: The system is used to implement the artificial intelligence-based Chinese homophone and similar form misspelling verification method described in any one of claims 1-8. The system includes: A file receiving module, which is used to receive the document file to be processed; A preprocessing module, which uses natural language processing technology to preprocess the received text to form a pure text; A text segmentation module, which uses the text segmentation method to split the pure text into sentences or phrases; A model analysis and error correction module, which inputs the sentences or phrases into a pre-trained deep learning model for analysis and error correction; A confusion set verification module, which constructs a confusion set and further verifies and corrects the model error correction results with reference to the information in the confusion set; A loop processing and result integration module, which loops through each sentence or phrase in the document and integrates all processing results; A file annotation and correction module, which annotates the identified misspelled words in the text, marks the error positions, gives correct suggestions, and records the frequency of occurrence of the misspelled words; A user interface module, which is used to display the processing results, receive user input, and provide user interaction functions; A storage module, which is used to store the preprocessed text, segmented sentences or phrases, model analysis results, confusion set verification results, and final processing results; A network connection module, which is used to communicate with other systems or servers to achieve remote processing or data sharing.
10. A system according to claim 9, wherein: The system also includes: a learning module, which is used to train the deep learning model, including collecting a large amount of text data, preprocessing the data, constructing the model structure, setting the loss function and optimization algorithm, and making the model converge through iterative training; An update module, which is used to update the basic database, judgment database, classification processing database, and confusion set regularly or according to user feedback to improve the accuracy and adaptability of the system; A security module, which is used to protect the system from malicious attacks and data leaks, and ensure user privacy and document security.