阅读说明：本技术 文本纠错方法、装置、设备及可读存储介质 (Text error correction method, device and equipment and readable storage medium ) 是由 邓悦 郑立颖 徐亮 于 2020-12-18 设计创作，主要内容包括：本发明涉及人工智能技术领域,本发明公开了一种文本纠错方法、装置、设备及可读存储介质,该方法包括步骤：获取待纠错文本；将所述待纠错文本输入预设文本纠错模型,生成纠错编辑操作序列；所述预设文本纠错模型由预设标注编辑操作序列训练得到；所述预设标注编辑操作序列用于将预设错误文本转化为与所述预设错误文本对应的正确文本；基于所述纠错编辑操作序列对所述待纠错文本进行纠错,得到纠错后文本。本发明避免了由于编码器编码和解码器解码的交叉进行而产生的时间序列依赖的问题,即将文本纠错的问题转换为序列生成问题,使得生成纠错编辑操作序列和将错误文本转换为正确文本的过程可以并行,进而提高了文本纠错过程的纠错速度。(The invention relates to the technical field of artificial intelligence, and discloses a text error correction method, a device, equipment and a readable storage medium, wherein the method comprises the following steps: acquiring a text to be corrected; inputting the text to be corrected into a preset text error correction model to generate an error correction editing operation sequence; the preset text error correction model is obtained by training a preset labeling editing operation sequence; the preset labeling editing operation sequence is used for converting a preset error text into a correct text corresponding to the preset error text; and correcting the text to be corrected based on the error correction editing operation sequence to obtain the text after error correction. The invention avoids the problem of time sequence dependence caused by the cross proceeding of the encoding of the encoder and the decoding of the decoder, namely, the problem of text error correction is converted into the problem of sequence generation, so that the processes of generating an error correction editing operation sequence and converting an error text into a correct text can be parallel, and the error correction speed of the text error correction process is further improved.)

1. A text error correction method, characterized by comprising the steps of:

acquiring a text to be corrected;

inputting the text to be corrected into a preset text error correction model to generate an error correction editing operation sequence; the preset text error correction model is obtained by training a preset labeling editing operation sequence; the preset labeling editing operation sequence is used for converting a preset error text into a correct text corresponding to the preset error text;

and correcting the text to be corrected based on the error correction editing operation sequence to obtain a target text after error correction.

2. The method of claim 1, wherein obtaining the predetermined text correction model comprises:

acquiring a training data set and a model to be trained;

performing iterative training on the model to be trained based on the training data set to obtain an updated model to be trained, and determining whether the updated model to be trained meets a preset iteration ending condition;

if the updated model to be trained meets the preset iteration end condition, taking the updated model to be trained as the preset text error correction model;

if the updated model to be trained does not meet the iteration ending condition, the updated model to be trained is continuously subjected to iteration training and updating until the updated model to be trained meets the iteration ending condition.

3. The method of claim 2, wherein the obtaining the model to be trained comprises:

acquiring a bidirectional pre-training language model;

and carrying out adaptive adjustment on the bidirectional pre-training language model to obtain a model to be trained.

4. The method of claim 3, wherein said obtaining a bi-directional pre-trained language model comprises:

adding a self-attention mechanism to the bi-directional pre-trained language model.

5. The method of claim 4, wherein said adding a self-attention mechanism to said bi-directional pre-trained language model comprises:

and adding a multi-head self-attention mechanism to the bidirectional pre-training language model.

6. The method of claim 2, wherein the training data set includes one or more training samples and standard test results corresponding to each of the training samples, and wherein obtaining the training data set includes:

obtaining a training sample;

and marking the training sample to obtain a standard detection result.

7. The method according to any one of claims 1 to 6, wherein the correcting the text to be corrected based on the sequence of correction editing operations to obtain a target corrected text comprises:

correcting the text to be corrected based on the error correction editing operation sequence to obtain an initial corrected text;

inputting the initial corrected text into the preset text error correction model for iterative error correction to obtain an updated corrected text, and determining whether the updated corrected text meets the preset iteration ending requirement;

if the updated error-corrected text meets the preset iteration ending requirement, taking the updated error-corrected text as a target error-corrected text;

if the updated error-corrected text does not meet the preset iteration ending requirement, the updated error-corrected text is continuously subjected to iterative error correction updating until the updated error-corrected text meets the preset iteration ending requirement.

8. A text correction apparatus, characterized in that the text correction apparatus comprises:

the acquisition module is used for acquiring a text to be corrected;

the generating module is used for inputting the text to be corrected into a preset text error correction model and generating an error correction editing operation sequence; the preset text error correction model is obtained by training a preset labeling editing operation sequence; the preset labeling editing operation sequence is used for converting a preset error text into a correct text corresponding to the preset error text;

and the error correction module is used for correcting the text to be corrected based on the error correction editing operation sequence to obtain a target text after error correction.

9. A text correction device comprising a memory, a processor and a text correction program stored on the memory and executable on the processor, the text correction program when executed by the processor implementing the steps of the text correction method as claimed in any one of claims 1 to 7.

10. A computer-readable storage medium, having stored thereon a text correction program, which when executed by a processor, performs the steps of the text correction method of any one of claims 1 to 7.

Technical Field

The invention relates to the technical field of artificial intelligence, in particular to a text error correction method, a text error correction device, text error correction equipment and a readable storage medium.

Background

In the process of writing a official document or editing an article, the situations of multiple characters, wrong characters and missing characters occur frequently, manual time-consuming proofreading is often needed when a official document without the wrong or wrong characters is submitted, so that the office efficiency is reduced to a certain extent, and in order to solve the problem, the automation and the intellectualization of text error correction are very necessary.

At present, a modeling method for text error correction mainly relies on an attention-based sequence-to-sequence encoder-decoder framework, which takes an originally erroneous sentence as an input in a text error correction process, and decodes the corrected correct sentence one by using a decoder after encoding through an encoder. However, the decoding process of each step of the sequence-to-sequence model depends on the output of the decoder of the previous step, the decoding process is a decoding-by-decoding process, a problem of time sequence dependence is generated, a loss in running speed is caused, and the processes of encoder encoding and decoder decoding are difficult to be parallel, so that the running speed on the line is slow.

Therefore, the problem that the error correction speed is low in the text error correction process when the text error correction task is carried out at present is known.

Disclosure of Invention

The invention mainly aims to provide a text error correction method, a text error correction device, text error correction equipment and a readable storage medium, and aims to solve the technical problem that the error correction speed is low in the text error correction process when the existing text error correction task is carried out.

In order to achieve the above object, the present invention provides a text error correction method, including the steps of:

acquiring a text to be corrected;

inputting the text to be corrected into a preset text error correction model to generate an error correction editing operation sequence; the preset text error correction model is obtained by training a preset labeling editing operation sequence; the preset labeling editing operation sequence is used for converting a preset error text into a correct text corresponding to the preset error text;

and correcting the text to be corrected based on the error correction editing operation sequence to obtain a target text after error correction.

Optionally, the obtaining the preset text error correction model includes:

acquiring a training data set and a model to be trained;

performing iterative training on the model to be trained based on the training data set to obtain an updated model to be trained, and determining whether the updated model to be trained meets a preset iteration ending condition;

if the updated model to be trained meets the preset iteration end condition, taking the updated model to be trained as the preset text error correction model;

if the updated model to be trained does not meet the iteration ending condition, the updated model to be trained is continuously subjected to iteration training and updating until the updated model to be trained meets the iteration ending condition.

Optionally, the obtaining the model to be trained includes:

acquiring a bidirectional pre-training language model;

and carrying out adaptive adjustment on the bidirectional pre-training language model to obtain a model to be trained.

Optionally, after the obtaining the bidirectional pre-training language model, the method includes:

adding a self-attention mechanism to the bi-directional pre-trained language model.

Optionally, the adding a self-attention mechanism to the bidirectional pre-training language model includes:

and adding a multi-head self-attention mechanism to the bidirectional pre-training language model.

Optionally, the training data set includes one or more training samples and a standard detection result corresponding to each of the training samples, and the acquiring the training data set includes:

obtaining a training sample;

and marking the training sample to obtain a standard detection result.

Optionally, the correcting the text to be corrected based on the correction editing operation sequence to obtain a target text after correction, including:

correcting the text to be corrected based on the error correction editing operation sequence to obtain an initial corrected text;

inputting the initial corrected text into the preset text error correction model for iterative error correction to obtain an updated corrected text, and determining whether the updated corrected text meets the preset iteration ending requirement;

if the updated error-corrected text meets the preset iteration ending requirement, taking the updated error-corrected text as a target error-corrected text;

if the updated error-corrected text does not meet the preset iteration ending requirement, the updated error-corrected text is continuously subjected to iterative error correction updating until the updated error-corrected text meets the preset iteration ending requirement.

In addition, to achieve the above object, the present invention provides a text correction device, including:

the acquisition module is used for acquiring a text to be corrected;

the generating module is used for inputting the text to be corrected into a preset text error correction model and generating an error correction editing operation sequence; the preset text error correction model is obtained by training a preset labeling editing operation sequence; the preset labeling editing operation sequence is used for converting a preset error text into a correct text corresponding to the preset error text;

and the error correction module is used for correcting the text to be corrected based on the error correction editing operation sequence to obtain a target text after error correction.

Further, to achieve the above object, the present invention also provides a text correction device comprising a memory, a processor and a text correction program stored on the memory and operable on the processor, the text correction program when executed by the processor implementing the steps of the text correction method as described above.

Further, to achieve the above object, the present invention also provides a computer readable storage medium having stored thereon a text error correction program, which when executed by a processor, implements the steps of the text error correction method as described above.

The method comprises the steps of obtaining a text to be corrected; inputting the text to be corrected into a preset text error correction model to generate an error correction editing operation sequence; the preset text error correction model is obtained by training a preset labeling editing operation sequence; the preset labeling editing operation sequence is used for converting a preset error text into a correct text corresponding to the preset error text; and correcting the text to be corrected based on the error correction editing operation sequence to obtain the text after error correction. The method and the device realize the improvement of the text error correction process, so that the text conversion process is to generate an error correction editing operation sequence firstly, then directly convert the error text into the correct text according to the error correction editing operation sequence, but not to generate a part of error correction editing operation sequence and convert a part of error text into a part of correct text according to the part of error correction editing operation sequence at the same time, thereby avoiding the problem of time sequence dependence caused by the cross proceeding of encoder encoding and decoder decoding, namely converting the problem of text error correction into a sequence generation problem, and finally correcting the text to be corrected through the generated sequence, so that the processes of generating the error correction editing operation sequence and converting the error text into the correct text can be parallel, and further improving the error correction speed of the text error correction process.

Drawings

FIG. 1 is a flowchart illustrating a text error correction method according to a first embodiment of the present invention;

FIG. 2 is a schematic diagram of an implementation process of a multi-head attention mechanism in a bi-directional pre-training language model according to an embodiment of the present invention;

FIG. 3 is a flowchart illustrating a text error correction method according to a second embodiment of the present invention;

FIG. 4 is a functional block diagram of a text correction device according to a preferred embodiment of the present invention;

fig. 5 is a schematic structural diagram of a hardware operating environment according to an embodiment of the present invention.

The implementation, functional features and advantages of the objects of the present invention will be further explained with reference to the accompanying drawings.

Detailed Description

It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

The invention provides a text error correction method, and referring to fig. 1, fig. 1 is a flowchart illustrating a first embodiment of the text error correction method according to the invention.

While a logical order is shown in the flow chart, in some cases, the steps shown or described may be performed in an order different than presented herein. The text error correction method may be applied to a mobile terminal including, but not limited to, a mobile phone, a personal computer, etc., and various steps of performing the subject description text error correction method are omitted below for convenience of description. The text error correction method comprises the following steps:

step S110, acquiring a text to be corrected.

Specifically, a text to be corrected that needs to be corrected is acquired.

It should be noted that the task of correcting the error of the text to be corrected is a text error correction task, and for the text error correction task, it needs to correct some characters in the text to be corrected (that is, in most cases, the error sentence and the correct sentence are different only in a specific position), for example, when a news practitioner edits a newsfeed, the editing speed is generally fast in terms of time effectiveness, so that the editing errors caused by the errors include wrongly written characters, multiple characters, and missing characters are common. Thus, the text correction task only needs to modify certain locations of the text, rather than regenerate the text. It is understood that the text error correction task is a text conversion task.

For the text error correction task, the present embodiment adopts the idea of an edit distance (the edit distance is a quantitative measure of the difference between two character strings (e.g. english characters), and the measure is to determine how many times a process is required to change one character string into another character string), that is, for the process of converting the text E into the text F (the text E is different from the text F), a series of processes (at least one of adding a character at any position of the text E, deleting a character, and replacing a character) is required. For example, text E is "sun very big today" and text F is "sun too big today", and to convert text E to text F, the character "sun" needs to be added after the "sun" in text E.

Step S120, inputting the text to be corrected into a preset text error correction model to generate an error correction editing operation sequence; the preset text error correction model is obtained by training a preset labeling editing operation sequence; and the preset labeling editing operation sequence is used for converting a preset error text into a correct text corresponding to the preset error text.

Specifically, the text to be corrected is input into a preset text error correction model, and an error correction editing operation sequence is generated; the preset text error correction model is obtained by training a preset labeling editing operation sequence; the preset mark editing operation sequence is used for converting the preset error text into a correct text corresponding to the preset error text. It should be noted that the preset labeling editing operation sequence can be obtained by manually labeling the preset error text, that is, manually correcting the preset error text, and sorting the editing operation corresponding to the correction process into the preset labeling editing operation sequence.

It should be noted that the sequence of error correction editing operations includes at least one editing operation, and the editing operation includes at least one of the following: preserving the current character (C), deleting the current character (D), inserting a character or character string behind the current character (A (w)), wherein "w" is the character or character string. For example, text X is "really very big today's sun," text Y is "really very big today's sun," the process of converting text X to text Y may be: the reserved character "present", the character "day" inserted after the character "present", the reserved character "too", the reserved character "yang", the deleted character ", the reserved character" true ", the reserved character" no ", the reserved character" normal ", and the reserved character" large ".

It should be noted that the error correction editing operation sequence obtained from the text to be corrected needs to be implemented by an algorithm from a preset sequence to an editing operation, where the algorithm from the preset sequence to the editing operation may be a seq2 edge algorithm, and the implementation process specifically includes:

the erroneous text may be converted to the correct text by a series of editing operations, such as C, A, thereby generating a sequence of editing operations by each editing operation, for example, the error text is "i come from Shanghai", the correct text is "i come from Shanghai", only the word "is deleted and then modified to" self ", thus, the generated editing operation sequence is "CCDACC", which is optimized by the present embodiment, a new editing operation is proposed to replace the current character with a character or character string (r (w)), it will be appreciated that the "replace" editing operation can replace the combination of "delete" and "insert" editing operations, namely, the optimized editing operation sequence is 'CCRCC', and it can be understood that the optimized editing operation sequence is simplified, so that the efficiency of the preset text error correction model in generating the editing operation sequence is improved.

Further, the obtaining the preset text error correction model includes:

step a, acquiring a training data set and a model to be trained.

Specifically, a training data set and a model to be trained are obtained, so that the model to be trained is trained through the training data set.

The acquiring of the training data set includes:

step a11, obtaining a training sample;

and a12, labeling the training samples to obtain a standard detection result.

Specifically, the training data set includes one or more training samples and standard test results corresponding to each of the training samples. Specifically, a training sample is obtained, and then the training sample is labeled, so as to obtain a standard detection result.

Specifically, the training sample is an error text, the labeling process is to determine an editing operation required to be performed to convert the error text into a correct text and determine an editing operation sequence corresponding to the editing operation, and the editing operation sequence is a standard detection result.

The above-mentioned model to be trained that obtains includes:

step a21, obtaining a bidirectional pre-training language model.

Specifically, a bi-directional pre-trained language model is obtained. It should be noted that, for the bi-directional pre-trained language model, before the error text is input into the bi-directional pre-trained language model, the word sequence of the error text needs to be converted into an initial word vector, for example, the error text X ═ X (X)₁，x₂，...，x_n) Its corresponding initial word vector D ═ D₁，d₂，...，d_n]. In addition, in order to encode the position information of each character in the error text, the position vector P ═ P is needed to encode the position information in the error text₁，p₂，…，p_n]To represent the absolute position of each character in the error text, where n is the number of characters included in a preset lexicon (at least including all characters in the error text), and it should be noted that the position vector P can be used to represent the position of any character in the error text. For example, if the error text is "i am", where the position of the character "i" in the error text is 1 and the position of the character "i" in the preset lexicon is 32, then in the position vector P, the character "i" is P₃₂1. Finally, the initial word vector D ═ D₁，d₂，...，d_n]And position vector P ═ P₁，p₂，…，p_n]After addition, a target word vector H ═ H can be obtained₁，h₂，…，h_n]. For example, D is [2, 3, 4, 5, 6 ]]P is [0, 1, 0, 3, 2, 0, 4, 5 ]]Then H is [ (0, 0), (2, 1), (0, 0), (4, 3), (3, 2), (0, 0), (5, 4), (6, 5)]. Wherein for bi-direction, the bi-direction is corrected relative to a pre-trained language model that is corrected based only on the above informationThe pre-training language model uses the context information of a certain character in the error text during error correction, so that the accuracy of the output of the bidirectional pre-training language model is improved.

Step a22, performing adaptive adjustment on the bidirectional pre-training language model to obtain a model to be trained.

Specifically, the model to be trained is obtained by performing preset adjustment on the bidirectional pre-training language model, where the preset adjustment is adjustment for adapting to a use requirement, that is, performing adaptive adjustment on the bidirectional pre-training language model, including adjusting an input of the model, adjusting a loss function, and the like.

After the above-mentioned two-way pre-training language model is obtained, include:

step a23, adding a self-attention mechanism to the bi-directional pre-trained language model.

Specifically, a self-attention mechanism is added to the bi-directional pre-training language model. In order to improve the accuracy of the output of the bidirectional pre-training language model, the bidirectional pre-training speech model uses the self-attention mechanism to determine the target word vector H ═ H₁，h₂，…，h_n]Further encoding is performed. Specifically, the weight of each character of the above-described erroneous text with respect to the other characters is output by a self-attention mechanism.

The formula used in the encoding process is as follows:

wherein, Q, K and V all refer to the target word vector H ═ H₁，h₂，…，h_n]；D_kThe vector dimension of the target word vector is referred to.

The adding of the self-attention mechanism to the bidirectional pre-training language model includes:

step a24, adding a multi-headed self-attention mechanism to the bi-directional pre-trained language model.

Specifically, a multi-head self-attention mechanism is added to the bidirectional pre-training language model. It should be noted that, in order to extract multiple semantics from an error text and make the output of the bi-directional pre-training language model more accurate through the multiple semantics, the self-attention mechanism is a multi-head attention mechanism, and the formula thereof is as follows:

wherein, Q, K and V all refer to the target word vector H ═ H₁，h₂，…，h_n]；W_i ^q、W_i ^kAnd W_i ^vAnd updating parameters required in the process of training the bidirectional pre-training language model.

Then outputting result head to the multi-head self-attention mechanism_iEach head (e.g. head)₁、head₂) And splicing to obtain the text characteristic representation of the error text, wherein the formula corresponding to the splicing process is as follows:

MultiHead＝concat(head₁，head₂，…，head_o)；

and after the splicing result is obtained, carrying out full-connection processing on the splicing result to realize the mixing of the output results of the multi-head attention mechanism, and then obtaining the output result of the bidirectional pre-training language model.

The specific implementation process of the multi-head attention mechanism in the bidirectional pre-training language model can be referred to fig. 2, where the elements in the "circle" represent nodes in the next layer network in the bidirectional pre-training language model, the elements in the "square" represent nodes in the previous layer network in the bidirectional pre-training language model, and the arrows therein represent attention information of the multi-head attention mechanism, such as in the calculation of the elementsWhen the attention of the node is paid, the node in the previous layer network is pointed to by an arrowAnd (4) calculating.

And b, performing iterative training on the model to be trained based on the training data set to obtain an updated model to be trained, and determining whether the updated model to be trained meets a preset iteration ending condition.

Specifically, iterative training is performed on the model to be trained based on the training data set to obtain an updated model to be trained, and whether the updated model to be trained meets a preset iteration end condition is determined. It should be noted that the preset iteration ending condition may be a loss function convergence.

Specifically, when training the model to be trained, the sequence R is (R)₁，r₂，...，r_n) And the sequence A ═ a₁，a₂，...，a_n) And error text X ═ X₁，x₂，...，x_n) Splicing to obtain a target Input sequence Input ═ (r)₁，r₂，...，r_n，x₁，x₂，...，x_n，a₁，a₂，...，a_n). Wherein, the sequence R is an operation sequence for replacing the current character with a character or a character string, the sequence A is an operation sequence for inserting the character or the character string behind the current character, R_i＝[M，p_i]，a_i＝[M，(p_i+p_i+1)/2]Where M is a MASK character [ MASK ]]Corresponding word vectors of the erroneous text X, it will be understood that the sequence R and the sequence a are related to the position vector P and not to the word vector D or the target word vector H.

It should be noted that the target Input sequence Input emphasizes the position information, and it is understood that the target Input sequence Input already contains the content of the error text X, and in order to avoid the repeated appearance of the content of the error text X, the sequence R and the sequence a are related to the absolute position of each character in the error text X and are not related to the content of the error text X.

The resulting target output sequence is (w)₁₁，w₁₂，…，w_1n，e₁，e₂，…，e_n，w₂₁，w₂₂，…w_2n) Wherein (w)₁₁，w₁₂，…，w_1n) For characters that need to be replaced, (w)₂₁，w₂₂，…，w_2n) The character that needs to be inserted.

Thus, the corresponding editing operation e of each character in the error text X can be obtained_iThe probability of (c) can be calculated by:

P(e_i|Input)＝softmax(logit(e_i|Input))；

wherein

The corresponding cross entropy loss function is thus calculated:

L(e，x)＝-∑_i log(P(e_i|x))；

and then updating the relevant parameters of the model to be trained by minimizing the cross entropy loss function so as to obtain the preset text error correction model.

Step c, if the updated model to be trained meets the preset iteration end condition, taking the updated model to be trained as the preset text error correction model;

and d, if the updated model to be trained does not meet the iteration ending condition, continuing to perform iteration training updating on the updated model to be trained until the updated model to be trained meets the iteration ending condition.

Specifically, if the updated model to be trained meets the preset iteration end condition, namely the model training is finished, taking the updated model to be trained as a preset text error correction model; and if the updated model to be trained does not meet the iteration ending condition, namely the model is not trained, continuing to perform iterative training and updating on the updated model to be trained until the updated model to be trained meets the iteration ending condition.

The embodiment obtains the text to be corrected; inputting the text to be corrected into a preset text error correction model to generate an error correction editing operation sequence; the preset text error correction model is obtained by training a preset labeling editing operation sequence; the preset labeling editing operation sequence is used for converting a preset error text into a correct text corresponding to the preset error text; and correcting the text to be corrected based on the error correction editing operation sequence to obtain the text after error correction. The method and the device realize the improvement of the text error correction process, so that the text conversion process is to generate an error correction editing operation sequence firstly, then directly convert the error text into the correct text according to the error correction editing operation sequence, but not to generate a part of error correction editing operation sequence and convert a part of error text into a part of correct text according to the part of error correction editing operation sequence at the same time, thereby avoiding the problem of time sequence dependence caused by the cross proceeding of encoder encoding and decoder decoding, namely converting the problem of text error correction into a sequence generation problem, and finally correcting the text to be corrected through the generated sequence, so that the processes of generating the error correction editing operation sequence and converting the error text into the correct text can be parallel, and further improving the error correction speed of the text error correction process.

Referring to fig. 3, a second embodiment is proposed based on the first embodiment of the text error correction method of the present invention, where the correcting the text to be corrected based on the error correction editing operation sequence to obtain the target text after error correction includes:

step S131, correcting the text to be corrected based on the error correction editing operation sequence to obtain an initial corrected text.

Specifically, an error correction editing operation is performed on the text to be corrected based on the error correction editing operation sequence to complete error correction of the text to be corrected, so as to obtain an initial text after error correction, where a certain gap may exist between the initial text after error correction and a correct text, that is, the initial text after error correction is not necessarily the correct text, for example, the initial text after error correction needs to be converted into the correct text after one or more editing operations, and it can be understood that the accuracy of the preset text error correction model generally does not reach 100%.

Step S132, inputting the initial corrected text into the preset text error correction model for iterative error correction to obtain an updated corrected text, and determining whether the updated corrected text meets the preset iteration ending requirement.

Specifically, in order to solve the problem that a difference exists between the initial corrected text and the correct text, the implementation proposes that the initial corrected text is input into a preset text error correction model for iterative error correction to obtain an updated corrected text, and whether the updated corrected text meets a preset iteration ending requirement is determined. It should be noted that the preset iteration ending requirement may be that the accuracy of the updated error-corrected text meets the requirement that the text does not need to be updated again, or that the number of times of the iteration update reaches a preset threshold, where the preset threshold may be set according to a specific situation, and this embodiment is not limited specifically.

Step S133, if the updated error-corrected text meets the preset iteration end requirement, taking the updated error-corrected text as a target error-corrected text;

step S134, if the updated error-corrected text does not meet the preset iteration end requirement, continuing to perform iterative error correction updating on the updated error-corrected text until the updated error-corrected text meets the preset iteration end requirement.

Specifically, if the updated error-corrected text meets the preset iteration end requirement, taking the updated error-corrected text as a target error-corrected text; and if the updated error-corrected text does not meet the preset iteration ending requirement, continuously performing iterative error correction updating on the updated error-corrected text, stopping iterative error correction until the updated error-corrected text meets the preset iteration ending requirement, and taking the updated error-corrected text as the target error-corrected text.

In this embodiment, the corrected text is input into the preset text error correction model for error correction again, so that the preset text error correction model is improved on the more "correct" corrected text each time, thereby improving the accuracy of the output of the preset text error correction model, and further solving the problem of error propagation in the text error correction process in the prior art.

In addition, the present invention also provides a text correction apparatus, as shown in fig. 4, the text correction apparatus including:

the first obtaining module 10 is configured to obtain a text to be corrected;

the generating module 20 is configured to input the text to be corrected into a preset text error correction model, and generate an error correction editing operation sequence; the preset text error correction model is obtained by training a preset labeling editing operation sequence; the preset labeling editing operation sequence is used for converting a preset error text into a correct text corresponding to the preset error text;

and the error correction module 30 is configured to correct the error of the text to be corrected based on the error correction editing operation sequence to obtain a target text after error correction.

Further, the text correction apparatus further includes:

the second acquisition module is used for acquiring a training data set and a model to be trained;

the iterative training module is used for performing iterative training on the model to be trained based on the training data set to obtain an updated model to be trained;

the determining module is used for determining whether the updated model to be trained meets a preset iteration ending condition; if the updated model to be trained meets the preset iteration end condition, taking the updated model to be trained as the preset text error correction model; if the updated model to be trained does not meet the iteration ending condition, the updated model to be trained is continuously subjected to iteration training and updating until the updated model to be trained meets the iteration ending condition.

Further, the first obtaining module 10 includes:

the first acquisition unit is used for acquiring a bidirectional pre-training language model;

and the adjusting unit is used for carrying out adaptive adjustment on the bidirectional pre-training language model to obtain a model to be trained.

Further, the first obtaining module 10 further includes:

and the adding unit is used for adding a self-attention mechanism to the bidirectional pre-training language model.

Further, the adding unit includes:

and the adding subunit is used for adding a multi-head self-attention mechanism to the bidirectional pre-training language model.

Further, the first obtaining module 10 further includes:

the second acquisition unit is used for acquiring a training sample;

and the marking unit is used for marking the training sample to obtain a standard detection result.

Further, the error correction module 30 includes:

the error correction unit is used for correcting the text to be corrected based on the error correction editing operation sequence to obtain an initial corrected text;

the iterative error correction unit is used for inputting the initial error corrected text into the preset text error correction model for iterative error correction to obtain an updated error corrected text;

a determining unit, configured to determine whether the updated error-corrected text meets a preset iteration end requirement; if the updated error-corrected text meets the preset iteration ending requirement, taking the updated error-corrected text as a target error-corrected text; if the updated error-corrected text does not meet the preset iteration ending requirement, the updated error-corrected text is continuously subjected to iterative error correction updating until the updated error-corrected text meets the preset iteration ending requirement.

The specific implementation of the text error correction device of the present invention is basically the same as that of the above text error correction method, and is not described herein again.

In addition, the invention also provides text error correction equipment. As shown in fig. 5, fig. 5 is a schematic structural diagram of a hardware operating environment according to an embodiment of the present invention.

It should be noted that fig. 5 is a schematic structural diagram of a hardware operating environment of the text error correction apparatus.

As shown in fig. 5, the text correction apparatus may include: a processor 1001, such as a CPU, a memory 1005, a user interface 1003, a network interface 1004, a communication bus 1002. Wherein a communication bus 1002 is used to enable connective communication between these components. The user interface 1003 may include a Display screen (Display), an input unit such as a Keyboard (Keyboard), and the optional user interface 1003 may also include a standard wired interface, a wireless interface. The network interface 1004 may optionally include a standard wired interface, a wireless interface (e.g., WI-FI interface). The memory 1005 may be a high-speed RAM memory or a non-volatile memory (e.g., a magnetic disk memory). The memory 1005 may alternatively be a storage device separate from the processor 1001.

Optionally, the text error correction device may further include RF (Radio Frequency) circuits, sensors, audio circuits, WiFi modules, and the like.

Those skilled in the art will appreciate that the configuration of the text correction device shown in FIG. 5 does not constitute a limitation of the text correction device and may include more or fewer components than shown, or some components may be combined, or a different arrangement of components.

As shown in fig. 5, a memory 1005, which is a kind of computer storage medium, may include therein an operating system, a network communication module, a user interface module, and a text correction program. The operating system is a program for managing and controlling hardware and software resources of the text correction device, and supports the operation of the text correction program and other software or programs.

In the text error correction apparatus shown in fig. 5, the user interface 1003 is mainly used for connecting a terminal, and performing data communication with the terminal, for example, acquiring an error text sent by the terminal; the network interface 1004 is mainly used for the background server and performs data communication with the background server; the processor 1001 may be configured to invoke a text correction program stored in the memory 1005 and perform the steps of the text correction method as described above.

The specific implementation of the text error correction device of the present invention is basically the same as the embodiments of the text error correction method, and is not described herein again.

In addition, an embodiment of the present invention further provides a computer-readable storage medium, where a text error correction program is stored on the computer-readable storage medium, and when being executed by a processor, the text error correction program implements the steps of the text error correction method described above.

The specific implementation of the computer-readable storage medium of the present invention is substantially the same as the embodiments of the text error correction method, and is not described herein again.

It should be noted that, in this document, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising an … …" does not exclude the presence of other like elements in a process, method, article, or apparatus that comprises the element.

The above-mentioned serial numbers of the embodiments of the present invention are merely for description and do not represent the merits of the embodiments.

Through the above description of the embodiments, those skilled in the art will clearly understand that the method of the above embodiments can be implemented by software plus a necessary general hardware platform, and certainly can also be implemented by hardware, but in many cases, the former is a better implementation manner. Based on such understanding, the technical solutions of the present invention may be embodied in the form of a software product, which is stored in a storage medium (such as ROM/RAM, magnetic disk, optical disk) and includes instructions for enabling a terminal device (such as a mobile phone, a computer, a server, a device, or a network device) to execute the method according to the embodiments of the present invention.

The above description is only for the preferred embodiment of the present invention and is not intended to limit the scope of the present invention, and all modifications of equivalent structures and equivalent processes, which are made by the contents of the present specification and the accompanying drawings, or directly or indirectly applied to other related technical fields, are included in the scope of the present invention.