Chinese word simplification method and system based on complex word change decoding

Abstract

The application discloses a Chinese word simplification method based on complex word change decoding, comprising the following steps: identifying the word difficulty level based on the Chinese Hanyu Shuiping Ceshi Cengji Biao (Chinese Proficiency Test), and taking high-difficulty words as target complex words; constructing a Chinese paraphrase corpus and training a Chinese paraphrase model; using the Chinese paraphrase model and based on a complex word change decoding method, generating paraphrase sentences with candidate simplified words; obtaining candidate simplified alternative words from the generated paraphrase sentences; and using an open source tool and word frequency to sort the candidate simplified alternative words and obtaining final simplified words. The application constructs a large-scale Chinese paraphrase corpus, uses a paraphrase model for Chinese vocabulary simplification tasks, proposes a decoding method based on complex word change, improves the simplification accuracy, and adds BARTscore in candidate word sorting to consider the change of alternative words to the original meaning of the sentence.

Description

Technical Field

[0001] This invention relates to the field of vocabulary simplification and decoding technology, and in particular to a method for simplifying Chinese words based on complex word variation decoding. Background Technology

[0002] Word simplification first requires identifying complex words, then providing appropriate simplified alternatives while ensuring the meaning remains unchanged. Word simplification can help people with reading difficulties, second language learners, and others; it is also significant for tasks such as text simplification. Currently, research on word simplification mainly focuses on English, while other languages ​​have received less attention due to a lack of sufficient corpus data.

[0003] Recent research utilizes pre-trained models like BERT to replace complex words. BERT employs a Transformer architecture, with pre-training tasks including a masked language model (MLM) and a next-sentence relation prediction (NSP). During MLM training, parts of the words are randomly masked, prompting the model to predict these masked parts. The NSP task aims to teach BERT relationships between sentences. Therefore, when the masked part is a complex word, BERT, based on the MLM pre-training task, provides candidate replacement words for the masked part, thus generating candidate replacement words for the complex word. Since Chinese pre-trained language models segment words using individual characters, BERT needs to set multiple masking tags to generate replacement words containing multiple characters when predicting complex words. Some studies have used autoregressive methods to predict masking tags sequentially, but BERT's performance still lags significantly behind synonym search methods, exhibiting lower accuracy when predicting replacement words with multiple characters. Furthermore, BERT focuses more on associating replacement words with context, neglecting semantic shifts, potentially altering the original meaning of the generated simplified words. Summary of the Invention

[0004] The purpose of this section is to outline some aspects of embodiments of the present invention and to briefly describe some preferred embodiments. Simplifications or omissions may be made in this section, as well as in the abstract and title of this application, to avoid obscuring the purpose of these documents; however, such simplifications or omissions should not be construed as limiting the scope of the invention.

[0005] In view of the aforementioned existing problems, the present invention is proposed.

[0006] Therefore, this invention provides a Chinese word simplification method based on complex word change decoding to solve the problems in the prior art where the accuracy is low when predicting replacement words of multiple Chinese characters; and where the simplified words generated cannot take into account changes in sentence meaning, and may change the original sentence meaning.

[0007] To solve the above-mentioned technical problems, the present invention provides the following technical solution, including:

[0008] In a first aspect, the present invention provides a method for simplifying Chinese words based on decoding complex word variations, including:

[0009] Based on the Chinese vocabulary level table for Chinese language exams, the difficulty level of words is identified, and high-difficulty words are identified as target complex words.

[0010] Construct a Chinese paraphrase corpus and train a Chinese paraphrase model;

[0011] Using the aforementioned Chinese paraphrasing model and based on the complex word variation decoding method, paraphrased sentences containing candidate simplified words are generated. The complex word variation decoding method applies bundle decoding based on the length of the complex word until a corresponding decoding length is reached, including:

[0012] The length of the complex word is defined as ;

[0013] Define the current decoding length Define restatement set The length of the restatement set is defined as ; This indicates the length of the preparatory sentence generated after preprocessing the prefixes of complex words using a Chinese paraphrasing model;

[0014] like For restatement sets De-marking and retaining the one with the highest probability Each restatement sentence returns a set of restates. ;

[0015] Will Updated to ;

[0016] Decoding using a Chinese paraphrasing model yields the following probability distribution for the current sentence:

[0017] =

[0018] Obtain from the current probability distribution The candidate sentence with the highest probability ;

[0019] Traversal ,like The last mark is ,Add to In the set of restates middle;

[0020] if , for restatement sets De-marking and retaining the one with the highest probability Each restatement sentence returns a set of restates. Otherwise, return to the judgment. Is it greater than or equal to? ;

[0021] Extract candidate simplified replacement words from the generated paraphrased sentences;

[0022] The candidate simplified replacement words are sorted using open-source tools and word frequency analysis to obtain the final simplified words, including:

[0023] The open-source tools mentioned are BERTscore and BARTscore;

[0024] The BARTscore is used to calculate the degree of semantic change of the restatement to the original sentence;

[0025] The similarity between the restatement and the original sentence is calculated using the BERT score.

[0026] The complexity of the substitution word is calculated based on the word frequency.

[0027] The sorting includes:

[0028] Define the final simplified word set CC, and define variables. Define the length of CC as ;

[0029] make ,like Proceed to the next step; otherwise, proceed to the step of setting up the CC set based on the score. Sort;

[0030] The candidate simplified word set C Alternative Complex words , to get the sentence ;

[0031] Calculate the substitution score The definition is as follows:

[0032] ;

[0033] in , , These are the corresponding weights;

[0034] Will Add to set CC;

[0035] Return Order Steps;

[0036] The CC set will be based on the score. Sort the results and return the final set of simplified words.

[0037] As a preferred embodiment of the Chinese word simplification method based on complex word change decoding described in this invention, the construction of the Chinese paraphrasing corpus utilizes an open-source machine translation model to obtain translated sentences, and combines the translated sentences and the original target sentences to construct the Chinese paraphrasing corpus;

[0038] The training of the Chinese paraphrasing model includes training and adjusting the Chinese paraphrasing corpus using the BART model to generate multiple candidate sentences.

[0039] As a preferred embodiment of the Chinese word simplification method based on complex word change decoding described in this invention, wherein: generating a restatement sentence containing candidate simplified words includes:

[0040] definition To determine the number of sentences to restate, the input sentence is defined as follows: , The number of words in the input sentence;

[0041] Using a Chinese paraphrasing model, prefixes of complex words are preprocessed to generate corresponding preparatory sentences, defined as follows: Length is defined as ;

[0042] Define the decoding space:

[0043] After the sentence is input into the Chinese paraphrasing model for encoding, it is decoded, specifying the preceding... Step, decoded content is prefix ;

[0044] Based on the prefix space, the probability distribution of the current word is obtained by decoding. ;

[0045]

[0046] in, , For activation function, for Decode step state information at any time. This represents the information of the original sentence;

[0047] Based on the current word probability distribution, obtain The candidate token with the highest probability is merged into the current decoding space.

[0048] Secondly, the present invention provides a system for Chinese word simplification based on complex word change decoding, including an identification module for identifying word difficulty levels based on a Chinese vocabulary level table for Chinese language exams, and identifying high-difficulty words as target complex words;

[0049] The training module is used to construct Chinese paraphrase corpora and train Chinese paraphrase models.

[0050] The decoding module is used to generate a paraphrased sentence using the Chinese paraphrasing model and based on the complex word change decoding method, wherein the complex word change decoding method is based on the length of the complex word and applies bundle decoding until the corresponding decoding length is reached;

[0051] The simplification module is used to extract candidate simplified alternatives from the generated paraphrased sentences;

[0052] The sorting module is used to sort the candidate simplified replacement words using open-source tools and word frequency to obtain the final simplified words. The sorting includes using open-source tools to calculate the degree of semantic change of the restatement to the original sentence.

[0053] Thirdly, the present invention provides a computing device, comprising:

[0054] Memory and processor;

[0055] The memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions. When the computer-executable instructions are executed by the processor, they implement the steps of the Chinese word simplification method based on complex word change decoding.

[0056] Fourthly, the present invention provides a computer-readable storage medium storing computer-executable instructions that, when executed by a processor, implement the steps of the Chinese word simplification method based on complex word variation decoding.

[0057] Compared with the prior art, the beneficial effects of the present invention are as follows: The present invention constructs a large-scale Chinese paraphrasing corpus; it uses a paraphrasing model to perform Chinese vocabulary simplification tasks and proposes a decoding method based on complex word changes to improve simplification accuracy; and it incorporates BARTscore into the candidate word ranking to consider the changes in the original meaning of the sentence caused by the substitute word. Attached Figure Description

[0058] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort. Wherein:

[0059] Figure 1 This is a schematic diagram of the overall process of a Chinese word simplification method based on complex word change decoding according to an embodiment of the present invention;

[0060] Figure 2This is a schematic diagram illustrating the specific process of a Chinese word simplification method based on complex word change decoding according to an embodiment of the present invention. Detailed Implementation

[0061] To make the above-mentioned objects, features, and advantages of the present invention more apparent and understandable, specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the protection scope of the present invention.

[0062] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0063] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.

[0064] This invention is described in detail with reference to the schematic diagrams. When describing the embodiments of this invention, the schematic diagrams are merely examples and should not be used to limit the scope of protection of this invention.

[0065] Example 1

[0066] Reference Figure 1-2 As an embodiment of the present invention, a method for simplifying Chinese words based on complex word change decoding is provided, including:

[0067] S1: Based on the Chinese vocabulary level table for Chinese language exams, identify the difficulty level of words and designate high-difficulty words as target complex words;

[0068] Preferably, the vocabulary level table published by the Chinese Proficiency Test (HSK) is used, which has a unified evaluation standard and is universally applicable.

[0069] S2: Construct a Chinese paraphrase corpus and train a Chinese paraphrase model;

[0070] Furthermore, a Chinese paraphrasing corpus was constructed by using an open-source machine translation model to obtain translated sentences, combining the translated sentences with the original target sentences;

[0071] It should be noted that the open-source machine translation model HelsinkiNLP can be used for translation.

[0072] Training the Chinese paraphrasing model involves training and adjusting the BART model on a Chinese paraphrasing corpus to generate multiple candidate sentences.

[0073] It should be noted that the BART model is trained using Denoising Auto-Encoding and performs outstandingly in various natural language processing tasks. By fine-tuning the BART model on the constructed Chinese paraphrasing corpus, it can modify the vocabulary and grammatical structure to generate multiple candidate sentences without changing the meaning of the sentences.

[0074] S3: Using the Chinese paraphrase model CNpara and based on the complex word change decoding method, generate paraphrased sentences with candidate simplified words. The complex word change decoding method is based on the length of the complex word and applies bundle decoding until the corresponding decoding length is reached.

[0075] Furthermore, generate paraphrased sentences containing candidate simplified words, including:

[0076] A3.1 Definition To determine the number of sentences to restate, the input sentence is defined as follows: Complex words are The target word length is , The number of words in the input sentence;

[0077] A3.2 Using the Chinese Paraphrasing Model prefixes of complex words = Preprocessing is performed to generate corresponding preparatory sentences, defined as follows: Length is defined as ;

[0078] A3.3 defines the decoding space as follows: ;in This is the current decoding step size. ;

[0079] A3.4 After encoding the sentence into the Chinese paraphrasing model, it is decoded, specifying the preceding... Step, decoded content is prefix The current decoding space is ;

[0080] A3.5 Based on Prefix Space Decoding yields the probability distribution of the current word. ;

[0081]

[0082] in , For activation function, for Decode step state information at any time. This represents the information of the original sentence;

[0083] A3.6 Based on the current probability distribution, it is possible to obtain... The candidate labels with the highest probability Merge candidate tags into the current decoding space; the current decoding space is ;

[0084] Furthermore, this also includes:

[0085] A3.7 The complex word variation-based decoding method is based on the length of the complex word and applies bundle decoding until the corresponding decoding length is reached. The length of the complex word is defined as follows: ;

[0086] Specifically, it includes:

[0087] A3.7.1 Define the current decoding length Define restatement set The length of the restatement set is defined as ;

[0088] A3.7.2 If ,make For restatement set De-marking and retaining the one with the highest probability Each restatement sentence returns a set of restates. ;

[0089] A3.7.3 will Updated to ;

[0090] A3.7.4 uses a Chinese paraphrasing model for decoding to obtain the probability distribution of the current sentence as follows:

[0091] ;

[0092] Obtain from the current probability distribution The candidate sentence with the highest probability ;

[0093] A3.7.5 Traversal ,like The last mark is ,Add to In the set of restates middle;

[0094] A3.7.6 If For restatement sets De-marking and retaining the one with the highest probability Each restatement sentence returns a set of restates. Otherwise, return to A3.7.2 for judgment. Is it greater than or equal to? .

[0095] It should be noted that this decoding method, through step A3.4, first specifies the preceding... The decoded content is a prefix. This ensures that the next decoding step must be at a complex word position; step A3.7.4 uses bundle decoding to ensure that the candidate sentences selected at each step are accurate. It is the set of sentences with the highest probability score, that is, the most accurate; A3.7.2 determines whether to end decoding based on the length of complex words, which reduces the generation of noise words and improves the accuracy of extracting simplified replacement words from sentences.

[0096] S4: Obtain candidate simplified replacement words from the generated paraphrased sentences;

[0097] Furthermore, candidate simplified alternatives are obtained from the generated paraphrased sentences, including:

[0098] A4.1: Define the set of candidate simplified words as Define variables ;

[0099] A4.2: Order ,like Returns a set of candidate words. ;

[0100] A4.3: Using Regular Expressions Greedy matching If a match is found and the result is in the Chinese level vocabulary list, add it to the set. Return to A4.2, and let Steps;

[0101] A4.4: Delete prefix ;

[0102] A4.5: If the target word length ,and Add it to the set in the Chinese level vocabulary list. Return to A4.2, and let Steps;

[0103] A4.6: Defining Variables ;

[0104] A4.7: Order ,like Return to A4.2, and let Steps;

[0105] A4.8: If Length is The prefix is ​​added to the set in the Chinese level vocabulary list. Return to A4.2, and let Steps;

[0106] A4.9: Return to A4.7, let The steps.

[0107] S5: Rank the candidate simplified replacement words using open-source tools and word frequency to obtain the final simplified words. The ranking includes using open-source tools to calculate the degree of semantic change of the restatement to the original sentence.

[0108] Furthermore, open-source tools and word frequency were used to rank the candidate simplified replacement words to obtain the final simplified words. The open-source tools used were BERTscore and BARTscore.

[0109] The degree of semantic change of the restatement to the original sentence is calculated using BARTscore;

[0110] The similarity between the restatement and the original sentence is calculated using BERTscore;

[0111] Word frequency calculation replaces word complexity.

[0112] Furthermore, the sorting specifically includes:

[0113] A5.1: Define the final simplified word set CC, and define variables. Define the length of CC as ;

[0114] A5.2: Order ,like If not, proceed to the next step A5.3; otherwise, jump to A5.7 and set the CC according to the score. Sort;

[0115] A5.3: Select the candidate simplified words from set C. Alternative Complex words , to get the sentence ;

[0116] A5.4: Calculate the substitution score The definition is as follows:

[0117] ;

[0118] in , , These are the corresponding weights;

[0119] A5.5: Will Add to collection CC;

[0120] A5.6: Return to A5.2, let Steps;

[0121] A5.7: Calculate the set CC based on the score. Sort the results and return the final set of simplified words.

[0122] It should be noted that adding BARTscore to the candidate word ranking process can take into account the changes that substitute words make to the original meaning of the sentence. It can work in conjunction with BERTscore to ensure that the meaning of the sentence is not changed while maintaining similarity.

[0123] The above is an illustrative scheme of a Chinese word simplification method based on complex word change decoding in this embodiment. It should be noted that the technical solution of the Chinese word simplification device based on complex word change decoding belongs to the same concept as the technical solution of the Chinese word simplification method based on complex word change decoding described above. Details not described in detail in the technical solution of the Chinese word simplification push device based on complex word change decoding in this embodiment can be found in the description of the technical solution of the Chinese word simplification push method based on complex word change decoding described above.

[0124] This embodiment of the Chinese word simplification system based on complex word change decoding includes:

[0125] The recognition module is used to identify the difficulty level of words based on the Chinese vocabulary level table for Chinese language exams, and to identify high-difficulty words as target complex words.

[0126] The training module is used to construct Chinese paraphrase corpora and train Chinese paraphrase models.

[0127] The decoding module is used to generate paraphrased sentences using a Chinese paraphrasing model and a complex word change decoding method. The complex word change decoding method is based on the length of the complex word and applies bundle decoding until the corresponding decoding length is reached.

[0128] The simplification module is used to extract candidate simplified alternatives from the generated paraphrased sentences;

[0129] The sorting module is used to sort candidate simplified replacement words using open-source tools and word frequency to obtain the final simplified words. The sorting includes using open-source tools to calculate the degree of semantic change of the restatement to the original sentence.

[0130] This embodiment also provides a computing device suitable for Chinese word simplification based on complex word change decoding, including:

[0131] The memory and processor are used to store computer-executable instructions and execute the computer-executable instructions to implement the Chinese word simplification method based on complex word change decoding as proposed in the above embodiments.

[0132] This embodiment also provides a storage medium on which a computer program is stored. When the program is executed by a processor, it implements the Chinese word simplification method based on complex word change decoding as proposed in the above embodiments.

[0133] The storage medium proposed in this embodiment and the Chinese word simplification method based on complex word change decoding proposed in the above embodiments belong to the same inventive concept. Technical details not described in detail in this embodiment can be found in the above embodiments, and this embodiment has the same beneficial effects as the above embodiments.

[0134] Based on the above description of the implementation methods, those skilled in the art can clearly understand that the present invention can be implemented using software and necessary general-purpose hardware, and of course, it can also be implemented using hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as a computer floppy disk, read-only memory (ROM), random access memory (RAM), flash memory, hard disk, or optical disk, etc., including several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods of the various embodiments of the present invention.

[0135] Example 2

[0136] Referring to Table 1, an embodiment of the present invention provides a Chinese word simplification method based on complex word variation decoding. To verify its beneficial effects, the superiority of the method is demonstrated through automatic index evaluation and actual output evaluation.

[0137] This method is evaluated using the HanLS dataset: the only current Chinese vocabulary simplification dataset; HanLS contains 524 samples; each sample contains one sentence and one complex word, with an average of 8.51 substitute words per sample; among them, there are 9 complex words of length 1, 472 complex words of length 2, 13 complex words of length 3, and 30 complex words of length 4.

[0138] This method uses Potential, Pre, Rec, and F1 scores for comparison of candidate alternatives: assuming the test set contains... For each sample, Potential, Pre, and Rec are defined as follows:

[0139]

[0140]

[0141]

[0142] in Indicates the first The label words of each sample Indicates the generated first A set of candidate words for each sample; and Represent and The number of words in the text; Indicates when generating words There exists The value is 1 if the word is correct, and 0 otherwise.

[0143] Compare this method with the latest methods: (Synonyms) (Word vectors) (Yoshihara) and BERT are compared:

[0144] As shown in Table 1, our method achieved the highest Potential score, Pre score, and F1 score. The BERT method based on the masked language model is weaker than our method in all metrics. This is because our method is the only one that preserves the original sentence meaning while generating alternative words based on complex word changes by fine-tuning the paraphrased corpus.

[0145] Table 1: Comparative Experiment Results

[0146]

[0147] Due to the characteristics of pre-trained model word segmentation, BERT-based methods require setting different numbers of mask tags for autoregressive generation. Because of BERT's mask training, a lot of noise is inevitably generated during the decoding process. However, autoregressive paraphrasing models do not have similar problems.

[0148] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A method for simplifying Chinese words based on decoding complex word variations, characterized in that, include: Based on the Chinese vocabulary level table for Chinese language exams, the difficulty level of words is identified, and high-difficulty words are identified as target complex words. Construct a Chinese paraphrase corpus and train a Chinese paraphrase model; Using the aforementioned Chinese paraphrasing model and based on the complex word variation decoding method, paraphrased sentences containing candidate simplified words are generated. The complex word variation decoding method applies bundle decoding based on the length of the complex word until a corresponding decoding length is reached, including: The length of the complex word is defined as ; Define the current decoding length Define restatement set The length of the restatement set is defined as ; This indicates the length of the preparatory sentence generated after preprocessing the prefixes of complex words using a Chinese paraphrasing model; like For restatement sets De-marking and retaining the one with the highest probability Each restatement sentence returns a set of restates. ; Will Updated to ; Decoding using a Chinese paraphrasing model yields the following probability distribution for the current sentence: Obtain from the current probability distribution The candidate sentence with the highest probability ; Traversal ,like The last mark is ,Add to In the set of restates middle; if For restatement sets De-marking and retaining the one with the highest probability Each restatement sentence returns a set of restates. Otherwise, return to the judgment. Is it greater than or equal to? ; Extract candidate simplified replacement words from the generated paraphrased sentences; The candidate simplified replacement words are sorted using open-source tools and word frequency analysis to obtain the final simplified words, including: The open-source tools mentioned are BERTscore and BARTscore; The BARTscore is used to calculate the degree of semantic change of the restatement to the original sentence; The similarity between the restatement and the original sentence is calculated using the BERT score. The complexity of the substitution word is calculated based on the word frequency. The sorting includes: Define the final simplified word set CC, and define variables. Define the length of CC as ; make ,like Proceed to the next step; otherwise, proceed to the step of setting up the CC set based on the score. Sort; The candidate simplified word set C Alternative Complex words , to get the sentence ; Calculate the substitution score The definition is as follows: ； in , , These are the corresponding weights; Will Add to set CC; Return Order Steps; The CC set will be based on the score. Sort the results and return the final set of simplified words.

2. The Chinese word simplification method based on complex word change decoding as described in claim 1, characterized in that, The construction of the Chinese paraphrasing corpus utilizes an open-source machine translation model to obtain translated sentences, and combines the translated sentences with the original target sentences to construct the Chinese paraphrasing corpus. The training of the Chinese paraphrasing model includes training and adjusting the Chinese paraphrasing corpus using the BART model to generate multiple candidate sentences.

3. The Chinese word simplification method based on complex word change decoding as described in claim 1 or 2, characterized in that, The generation of paraphrased sentences containing candidate simplified words includes: definition To determine the number of sentences to restate, the input sentence is defined as follows: , The number of words in the input sentence; Using a Chinese paraphrasing model, prefixes of complex words are preprocessed to generate corresponding preparatory sentences, defined as follows: Length is defined as ; Define the decoding space: After the sentence is input into the Chinese paraphrasing model for encoding, it is decoded, specifying the preceding... Step, decoded content is prefix ; Based on the prefix space, the probability distribution of the current word is obtained by decoding. : in, , For activation function, for Decode step state information at any time. This represents the information of the original sentence; Based on the current word probability distribution, obtain The candidate token with the highest probability is merged into the current decoding space.

4. A system for simplifying Chinese words based on complex word variation decoding, employing the method described in any one of claims 1-3, characterized in that, include: The recognition module is used to identify the difficulty level of words based on the Chinese vocabulary level table for Chinese language exams, and to identify high-difficulty words as target complex words. The training module is used to construct Chinese paraphrase corpora and train Chinese paraphrase models. The decoding module is used to generate a paraphrased sentence using the Chinese paraphrasing model and based on the complex word change decoding method, wherein the complex word change decoding method is based on the length of the complex word and applies bundle decoding until the corresponding decoding length is reached; The simplification module is used to extract candidate simplified alternatives from the generated paraphrased sentences; The sorting module is used to sort the candidate simplified replacement words using open-source tools and word frequency to obtain the final simplified words. The sorting includes using open-source tools to calculate the degree of semantic change of the restatement to the original sentence.

5. An electronic device, comprising: Memory and processor; The memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions. When the computer-executable instructions are executed by the processor, they implement the steps of the Chinese word simplification method based on complex word change decoding as described in any one of claims 1 to 3.

6. A computer-readable storage medium storing computer-executable instructions that, when executed by a processor, implement the steps of the Chinese word simplification method based on complex word variation decoding as described in any one of claims 1 to 3.

Images