Volume 2022, Issue 1 4422754

Research Article

Open Access

Analyzing the Application of Multimedia Technology Assisted English Grammar Teaching in Colleges

Feng Wu,

Feng Wu

Mathematics and Computer Department, Fuzhou Preschool Education College, FuZhou 344000, China

Search for more papers by this author

Shaohai Huang,

Shaohai Huang

Mathematics and Computer Department, Fuzhou Preschool Education College, FuZhou 344000, China

Search for more papers by this author

Lijuan Cheng,

Corresponding Author

Lijuan Cheng

[email protected]

orcid.org/0000-0002-8036-1812

School of Humanities and Law, Hebei University of Engineering, Handan 056000, China hebeu.edu.cn

Search for more papers by this author

Feng Wu,

Feng Wu

Mathematics and Computer Department, Fuzhou Preschool Education College, FuZhou 344000, China

Search for more papers by this author

Shaohai Huang,

Shaohai Huang

Mathematics and Computer Department, Fuzhou Preschool Education College, FuZhou 344000, China

Search for more papers by this author

Lijuan Cheng,

Corresponding Author

Lijuan Cheng

[email protected]

orcid.org/0000-0002-8036-1812

School of Humanities and Law, Hebei University of Engineering, Handan 056000, China hebeu.edu.cn

Search for more papers by this author

First published: 2022

https://doi.org/10.1155/2022/4422754

Academic Editor: Fazli Wahid

Share a link

Email
Wechat
Bluesky

Abstract

To address the problem that the traditional English composition is manually reviewed, which leads to low efficiency of grammar error checking and heavy teaching burden for teachers and seriously affects the quality of English teaching, a verb grammar error checking algorithm based on the combination of a multilayer rule model and a deep learning model is proposed with current multimedia technology. First, the basic principles and network structure of the long short-term memory network LSTM are specifically analyzed; then, a multilayer rule model is constructed; after that, an attention model-based LSTM verb grammar checker is designed, and a word embedding model is used to encode the text and map it to the status space to retain the textual information. Finally, the proposed algorithm is experimentally validated using the corpus dataset. The experimental results revealed that the accuracy, recall, and F1 values of the proposed algorithm in verb error detection were 45.51%, 28.77%, and 35.5%, respectively, which were higher than those of the traditional automatic review algorithm, indicating that the proposed algorithm has superior performance and can improve the accuracy and efficiency of grammar error detection.

1. Introduction

In recent years, with the rapid development of multimedia technology, the learning of English in university gradually tends to be automated and intelligent. Among them, English composition is gradually converted from manual marking to automatic marking, which gradually improves marking efficiency and reduces teachers’ teaching pressure. However, the traditional English composition error detection system has a low accuracy rate of grammar detection and a high rate of error detection for words such as verbs, prepositions, and coronals, which cannot meet the current requirements of college English grammar-assisted teaching [1–3]. In response to this problem, a large number of studies have been proposed by scholars and experts. Krzysztof pajak et al. proposed a seq2seq error correction method based on denoising automatic coding and obtained an F1 score of 56.58%[4]. Zhaoquan Qiu proposed a two-stage Chinese grammar error correction algorithm, which is characterized by providing reference for Chinese grammar error correction and introducing a deep learning algorithm [5]; Ailani Sagar summarizes the current mainstream syntax error correction algorithms, which provide a reference for a comprehensive understanding of syntax error correction [6]. Chanjun Park et al. put forward the syntax error correction method of a neural network and the evaluation index of correction and over correction [7]; Bo Wang et al. proposed marking the statistical machine translation results to improve the accuracy of error correction [8]; Lichtage Jared et al. improved grammar error correction from the perspective of data training so as to provide a way to improve training and improve the accuracy of grammar error correction [9]. Crivellari Alessandro et al. proposed a direction-finding algorithm with error detection and fault tolerance for longer statements in English, and the LSTM network model with attention mechanism is effective in learning longer statements in both directions and preserving the features of long and short statement sequences [10–13]. Based on the above, the LSTM network in deep learning is applied to English grammar error detection to improve the accuracy and efficiency of English grammar error detection and to assist English teachers in instructing students.

2. Basic Methods

2.1. LSTM Fundamentals

Long short term memory (LSTM) is a recurrent neural network with memory function to process the next period of time sequence. Its main role is to solve the dependent memory decline in data modeling, which can avoid the gradient disappearance and explosion problem during model training. To solve the problem that the RNN has only h_t states in the hidden layer at moment t, the unit state c_t, i.e., the information storage, will be added. The values to be updated are selected in the input gate sigmoid function, and then a candidate vector ${\tilde{c}}_{t}$ obtained from the tanh layer is added to c_t. As a result, LSTM can effectively solve nonlinear time series data associated with specific memory functions. With the control of the learning rates of input gate, output gate, and forget gate, the large data sequences can be better processed [14–16].

At time t, the input to the LSTM consists mainly of the input value x_t of the neural network at time t, the output value h_t−1 at time t − 1, and the cell state c_t−1 at time t-1; The resulting LSTM outputs are as follows: the output value h_t of the neural network at moment t and the cell state c_t at moment t. The LSTM neuron structure is shown in Figure 1.

Details are in the caption following the image — Open in figure viewer PowerPoint

The formulas for each structure are expressed as

\begin{matrix} f_{t} = σ (ω_{x f} x_{t} + ω_{h t} h_{t - 1} + b_{f}) = σ (ne t_{f, t}), \\ i_{t} = σ (ω_{x f} + ω_{h t} h_{t - 1} + b_{i}) = σ (ne t_{f, t}), \\ {\tilde{c}}_{t} = \tanh (ω_{x c} x_{t} + ω_{h c} h_{t - 1} + b_{c}) = \tanh (ne t_{\tilde{c}, t}), \\ c_{t} = f_{t} c_{t - 1} + i_{t} {\tilde{c}}_{t}, \\ o_{t} = σ (ω_{x o} x_{t} + ω_{h o} h_{t - 1} + b_{o}) = σ (ne t_{o, t}), \\ h_{t} = o_{t} \tanh (c_{t}) . \end{matrix}

(1)

In the above equation, i_t, f_t, and o_t denote the states of the input, forget, and output gate at moment t, respectively; σ and tanh are the sigmoid function and hyperbolic tangent activation function, respectively; w and b denote the corresponding weight coefficient matrix and bias terms; c_t denotes the state of the memory cell at moment t; h_t denotes the full output of the LSTM network unit at time point t.

The sigmoid function and hyperbolic tangent activation function expressions are [17, 18]

\begin{matrix} σ (x) = \frac{1}{1 + e^{- x}}, \\ \tanh (x) = \frac{e^{x} - e^{- x}}{e^{x} + e^{- x}} . \end{matrix}

(2)

2.2. Back Propagation Process of LSTM

LSTM model training is mainly implemented by the BPTT algorithm, which calculates all output values according to equations (3) to (9); the error term δ is calculated from the spatiotemporal direction; according to δ, the weight gradient is derived and a suitable gradient optimization algorithm is selected to update the weights. diag denotes the diagonal matrix with diagonal elements as vector elements; o denotes the corresponding elements multiplied by each other, and the calculation process is

Based on the time back propagation error term δ, we find the error δ_t−1 at time t − 1 [19].

\begin{matrix} δ_{t}^{T} \frac{\partial h_{t}}{\partial h_{t - 1}} = δ_{o, t}^{T} \frac{\partial n e t_{o, t}}{\partial h_{t - 1}} + δ_{f, t}^{T} \frac{\partial n e t_{f, t}}{\partial h_{t - 1}} + δ_{i, t}^{T} \frac{\partial n e t_{i, t}}{\partial h_{t - 1}} δ_{\tilde{c}, t}^{T} \frac{\partial n e t_{\tilde{c}, t}}{\partial h_{t - 1}}, \end{matrix}

(3)

\begin{matrix} \frac{\partial h_{t}}{\partial o_{t}} = diag [\tanh (c_{t})], \end{matrix}

(4)

where i_t, f_t,

{\tilde{c}}_{t}

, and o_t are all functions of h_t−1, which can be found by the total derivative formula:

\begin{matrix} \frac{\partial h_{t}}{\partial c_{t}} = diag [o_{t} \circ (1 - q \tanh {(c_{t})}^{2})] . \end{matrix}

(5)

According to

\nabla_{w} E = \sum_{k = 1}^{t} \nabla w_{k} E

, it is possible to calculate

\begin{matrix} \frac{\partial c_{t}}{\partial f_{t}} = diag [c_{t - 1}], \frac{\partial c_{t}}{\partial i_{t}} = diag [{\tilde{c}}_{t}], \frac{\partial c_{t}}{\partial {\tilde{c}}_{t}} = diag [i_{t}] . \end{matrix}

(6)

It is easy to derive the following formula:

\begin{matrix} \frac{\partial o_{t}}{\partial n e t_{o, t}} = diag [o_{t} \circ (1 - o_{t})], \end{matrix}

(7)

\begin{matrix} \frac{\partial e t_{o, t}}{\partial h_{t - 1}} = W_{o h}, \end{matrix}

(8)

\begin{matrix} \frac{\partial f_{t}}{\partial n e t_{f, t}} = diag [f_{t} • (1 - f_{t})], \end{matrix}

(9)

\begin{matrix} \frac{\partial n e t_{f, t}}{\partial h_{t - 1}} = W_{f h}, \end{matrix}

(10)

\begin{matrix} \frac{\partial i_{t}}{\partial n e t_{i, t}} = diag [i_{t} • (1 - i_{t})], \end{matrix}

(11)

\begin{matrix} \frac{\partial n e t_{i, t}}{\partial h_{t - 1}} = W_{i h}, \end{matrix}

(12)

\begin{matrix} \frac{\partial {\tilde{c}}_{t}}{\partial n e t_{{\tilde{c}}_{t}, 1}} = diag [1 - {\tilde{c}}_{t}^{2}], \end{matrix}

(13)

\begin{matrix} \frac{\partial n e t_{\tilde{c}, t}}{\partial h_{t - 1}} = W_{c h} . \end{matrix}

(14)

Bring the above partial derivatives into equation (3) for calculation, we can find [20]

\begin{matrix} δ_{t - 1} = δ_{o, t}^{T} \frac{\partial n e t_{o, t}}{\partial h_{t - 1}} + δ_{f, t}^{T} \frac{\partial n e t_{f, t}}{\partial h_{t - 1}} + δ_{i, t}^{T} \frac{\partial n e t_{i, t}}{\partial h_{t - 1}} + δ_{o, t}^{T} \frac{\partial n e t_{\tilde{c}, t}}{\partial h_{t - 1}} \\ = δ_{o, t}^{T} W_{o h} + δ_{f, t}^{T} W_{f h} + δ_{i, t}^{T} W_{i h} + δ_{\tilde{c}, t}^{T} W_{c h} . \end{matrix}

(15)

From δ_o,t, δ_f,t, δ_i,t, and

δ_{\tilde{c}, t}

, it can be calculated that:

\begin{matrix} δ_{o, t}^{T} = δ_{t}^{T} • \tanh (c_{t}) • o_{t} • (1 - o_{t}), \\ δ_{f, t}^{T} = δ_{t}^{T} • o_{t} • (1 - \tanh {(c_{t})}^{2}) • c_{t - 1} • f_{t} • (1 - f_{t}), \\ δ_{i, t}^{T} = δ_{t}^{T} • o_{t} • (1 - \tanh {(c_{t})}^{2}) • {\tilde{c}}_{t} • i_{t} • (1 - i_{t}), \\ δ_{\tilde{c}, t}^{T} = δ_{t}^{T} • o_{t} • (1 - \tanh {(c_{t})}^{2}) • i_{t} • (1 - {\tilde{c}}_{t}^{2}) . \end{matrix}

(16)

It is possible to obtain the equation that passes the error term forward to any k moments as [16–18]

\begin{matrix} δ_{k}^{T} = \prod_{j = k}^{t - 1} δ_{o, j}^{T} W_{o h} + δ_{f, j}^{T} W_{f h} + δ_{i, j}^{T} W_{i h} + δ_{\tilde{c}, j}^{T} W_{c h} . \end{matrix}

(17)

If the current is the l th layer, the error term in the l − 1 layer is the derivative of the error function to the weighted input in the l − 1 layer, which can be expressed as

\begin{matrix} δ_{t}^{l - 1} \overset{d e f}{=} \frac{\partial E}{n e t_{t}^{l - 1}} . \end{matrix}

(18)

The input x_t to the LSTM at the current moment can be found by the following equation [21, 22]:

\begin{matrix} x_{t}^{l} = f^{l - 1} (n e t_{t}^{l - 1}) . \end{matrix}

(19)

In the above equation, f^l−1 is the activation function of l − 1. Since $ne t_{f, t}^{l}$ , $ne t_{l, t}^{l}$ , $ne t_{\tilde{c}, t}^{l}$ , and $ne t_{o, t}^{l}$ are all functions of x_t, and x_t is a function of $n e t_{t}^{l - 1}$ , the error can be transferred to the l − 1 layer by the total derivative formula.

3. English Grammar Checker Design

3.1. Syntax Checking Based on Multilayer Rule Model

3.1.1. Overall Module Design

In English grammar error checking, firstly, we obtain the rules in the grammar, build an error checking model based on chopped rules, match the rules with the database through the input text, find out the error points, and modify them according to the corpus information. The module design is obtained as follows which is referring to Xu et al. [23] (Figure 2):

3.1.2. Part-of-Speech Tagging

English statements are mainly composed of various words, including nouns, verbs, and coronals. Since the meaning of each word is different and has an impact on the later rule matching, part-of-speech tagging will be performed before grammar error checking for better error correction of the statement, thus making the rule base better to discriminate the wrong text. At present, the commonly used annotation method is the marking convention, which is useful for different part of speech. Based on the information characteristics of the English corpus, Stanford parser will be used for annotation.

3.1.3. Rule Base Design

To improve the accuracy of grammar error detection, referring to the results of Xu et al., a rule base will be designed to match the input statements, and its design structure is shown in Figure 3 [24]. It is mainly ranked according to the specific characteristics of the grammar and thus stored, and is arranged in accordance with the name of the file so that multiple rules can be found for subsequent queries.

In the rule base, its rules are mainly adjusted according to the corresponding algorithm. The adjustment is done from top to bottom, and its specific process is as follows which refers to the results of Xu et al. (Figures 4–6).

First, open the index table in the rule base, check whether it is blank, if so, reopen for operation; conversely, select an index item according to the order, and then the system number it, check whether its number is consistent with the index item, if so, get its rule base name, end the index, otherwise, go back to the second step for operation.

3.1.4. Multilayer Rule Syntax Error Correction

After completing the above design, the syntax error checking is tested according to the criteria of the rule base, and a certain English statement is input to the error checking module for multiple hierarchical matching to find out the wrong statement. The error checking steps are divided into three steps, which are represented as follows:

(1)
After inputting the statements, appropriate deletions are made, and the statements with important information are preserved
(2)
The above algorithm is used to check for errors, and if an error is found, it goes to the next step, and if not, it ends the calculation and returns to the above steps for recalculation until the end of the rule base
(3)
If the system finds an error, it goes back to step (2) to search for the error location

The overall flow of the system is shown in Figure 5 [25].

3.2. LSTM Syntax Checker Design Based on Attention Model

In order to improve the accuracy of English syntax error detection, the LSTM network model is applied to English syntax error detection by encoding the utterance, extracting features through it, and later classifying the features using the attention mechanism, thus realizing syntax error detection. The structure of the syntax error checker is shown in Figure 6 [26].

In this model, there are three main steps, which are analyzed as follows:

(1)
Word vector representation: the English statements are represented using the word embedding model. If the input utterance is X = {x₁, x₂, x₃, ⋯, x_T}, where x_i belongs to K-dimension, the summation is averaged over all utterances, after which the output is X^′.
(2)
Feature extraction: the statement features are extracted and encoded using the LSTM model.X and X^′ are input into the system as X^′. h₁, h₂, h₃, ⋯, h_t and x₁, x₂, x₃, ⋯, x_t corresponding to each other, and they are the invisible corresponding values.
(3)
Error classification: an attention model is used to construct a classifier, and English statements are input into this classifier to classify them according to the above extracted features, thus achieving error classification.

Among them, the word embedding method can make the statements shorter and improve the consistency of statements, which can effectively solve the grammatical error correction problem and is applicable to shorter English texts for grammatical error correction.

3.3. LSTM Model Based on Attention Model with Combination of Positive and Reverse Order

The above error detection model is mainly for short English statements, but when longer statements with more complex semantic information are input, the model is unable to detect errors and cannot take care of the contextual connections, resulting in poor error detection.

Therefore, based on the above model, the positive and negative sequences are added to combine the information in both directions to achieve multilayer feature extraction and encoding [27]. The specific structure is shown in Figure 7.

In the above model structure, x₁, x₂, x₃, ⋯, x_t are the sequences of the system being input; h₁, h₂, h₃, ⋯, h_t are the potential nodes of the layer with forward transmission direction; $h_{3}^{'}, h_{3}^{'}, h_{3}^{'}, \dots, h_{t}^{'}$ indicate the potential nodes of the layer with backward transmission direction; Y₁, Y₂, Y₃, ⋯, Y_t indicate the sequences of the final obtained output.

The model is divided into three main parts, advance processing, extraction of features, and classification, and the specific process is as follows:

(1)
Two directional data are input into the model, and x₁, x₂, x₃, ⋯, x_t and X^′ are both backward sequences, and the rest are forward sequences. The two data are represented by the word embedding method, after which they are input into the error detection model.
(2)
Feature extraction is performed using the LSTM in the error detection model, and the association information of the two sequences is encoded.
(3)
After completing the above operations, the attention model is used for classification, and the classification of erroneous statements is thus obtained.

The error detection model designed in the study can fully take into account the length and direction of the input data, which can retain more text information for better text extraction and classification, and finally realize the error detection of long and short statements.

3.4. Design and Implementation of Verb Syntax Checking

Based on the above LSTM model, it is combined with the rule set to design an English grammar checker with the following overall architecture [28–30] (Figure 8).

As can be seen from the above figure, the checker mainly consists of a rule model and an LSTM model. First, the rule model is used to perform operations such as information matching on the input statements to find out the location of the wrong syntax and give a preliminary modification plan; then, the LSTM error checking module based on the attention model is used to detect, compare it with the standard statements of the corpus to find out the statement errors, mark and locate the errors, and finally give the corresponding revision strategy.

4. Experimental Results and Analysis

4.1. Evaluation Contents

There are a variety of grammatical problems in English statements, the more common ones being grammatical errors, word morphology errors, missing words, and other aspects of application. The study will identify the more typical common errors, which are described in Table 1.

1. Common English grammatical errors.

Type	Content
Vt	Verb tense error
Vo	Verb missing error
Vform	Verb form error
SVA	Subject-verb agreement error
ArtOrDet	Article error
Prep	Preposition error
Nn	Form error of singular and plural nouns
Others	Other errors

In order to enhance the evaluation accuracy of the experimental results, NUCLE-releaSe2.2 will be selected as the experimental corpus, and the types and numbers of errors it contains are shown in Table 2.

2. Evaluation contents of grammatical errors.

Type	Quantity
Vt	3211
Vo	411
Vform	1453
SVA	1527
ArtOrDet	6658
Prep	2404
Nn	3779
Others	16865

4.2. Evaluation Criteria

In order to better evaluate the error detection results of the proposed error detection system, F-Measure will be used as the evaluation index in this experiment, and its expression is

\begin{matrix} F_{β} = \frac{(β^{2} + 1) P R}{β^{2} P + R} . \end{matrix}

(20)

In equation (17), P denotes the precision; R denotes the recall, and the expression of the precision is

\begin{matrix} P = \frac{A}{A + C}, \end{matrix}

(21)

where P denotes the proportion of marked correct statements among all statements, A and C denote the number of marked correct statements and the number of statements that are actually correct but marked incorrectly, respectively. The solution formula of R is

\begin{matrix} R = \frac{A}{A + B} . \end{matrix}

(22)

In the above formula, R represents the proportion of statements with errors but marked correctly in all wrong statements. B represents the total number of unmarked errors and sentences in the text.

F1 − Measure is as follows:

\begin{matrix} F_{1} = \frac{P * R}{P + R} . \end{matrix}

(23)

The effectiveness of English syntax detection is verified by F₁. If F₁ takes a higher value, it means that the detection effect of this syntax detection system is better.

4.3. Experimental Results and Analysis

4.3.1. Verification of Article Detection

In English grammar, article errors mainly include three forms of article redundancy, absence, and misuse, and their error ratios are shown in (Tables 3 and 4).

3. Proportions of the three types of article errors in the corpus.

Type	Article redundancy (%)	Article absence (%)	Article misuse (%)
Error ratio	26.8	57.5	15.7

4. Statistics of article corpus.

	Training set	Test set
Number of sentences	57151	1381
Number of participles	1161567	29207
Number and proportion of article errors	14.9%	19.9%

The information of the experimental corpus of article errors was obtained from the NUCLE English composition of a university, and the training and test sets of article class errors were classified after manual annotation, as shown in Table 4.

The above datasets are applied to the error detection model individually; the model is trained and predicted, respectively, and the flowchart of the article error detection system is shown in Figure 9.

The experiment compares the results of the above three evaluation metrics on the test set before and after pretreatment, and the results are shown in Table 5.

As can be seen from Table 5, the precision, recall, and F1 values of this corpus information increased after preprocessing the dataset, by 24.57%, 56.74%, and 34.08%, respectively, and the model training is significantly faster and better.

5. Comparison of the test set results before and after preprocessing the corpus.

	Precision (%)	Recall (%)	F₁ (%)
Before pretreatment	22.64	54.90	33.13
After pretreatment	24.57	56.74	34.08

In English syntax error detection, the error detection model belongs to the maximum entropy model, and the number of iterations seriously affects the error detection effect, and the relationship between the number of iterations and F-Measure is obtained as shown in Table 6.

6. The relationship between the number of iterations and F-Measure.

Number of iterations	1 (%)	2 (%)	3 (%)	4 (%)	5 (%)	10 (%)	100 (%)
F₁	34.08	34.08	34.11	34.12	34.18	32.24	35.48

From Table 6, we can see that the F-Measure increases with the increasing number of iterations, and the increase is small. The process increases the training time, which will affect the subsequent syntax error detection results; therefore, it is most reasonable to set the number of iterations to 1.

To verify the effect of the maximum entropy model on the processing of grammatical information, the model was compared with the Naive Bayesian model using the test set data, and the experimental results are shown in Table 7.

7. Comparison results of the two models in the test set.

	Precision (%)	Recall (%)	F₁ (%)
Naive bayesian model	16.13	41.59	23.25
Maximum entropy model	24.42	56.38	34.08

As can be seen from Table 7, the precision, recall, and F-Measure of the maximum entropy model are 24.42%, 56.38%, and 34.08%, respectively, which are better than the Naive Bayesian model for classification recognition and faster training.

Since the article corpus dataset is small, a partial wiki corpus will be added to this dataset for comparison. The test results after adding the corpus are obtained as shown in Table 8.

From Table 8, it can be seen that the precision, recall, and F-Measure of the training corpus are improved after adding different amounts of corpus, but the improvement is small.

8. Effect comparison of test set after adding 100,000 and 200,000 Wikipedia corpus.

	Precision (%)	Recall (%)	F₁ (%)
NUCLE	24.42	56.38	34.08
NUCLE + Wiki (l0w strip)	24.27	56.81	34.01
NUCLE + Wiki (20w strip)	24.54	57.39	34.38

In order to further test the superiority of the article error detection system, this experiment compares the error detection results of this system and the CoNLL 2013 shared task system, and the comparison results are obtained as shown in Table 9.

From Table 9, we can find that the P value, R value and F value of the article error detection system are 24.54%, 57.39%, and 34.38%, which are 1.35%, 9.87%, and 1.19% higher than the CoNLL 2013 system, respectively, thus indicating that the proposed article error detection system can achieve better error detection results and superior system performance.

9. Comparison of the error detection results of the two systems.

	Precision (%)	Recall (%)	F₁ (%)
End results	24.54	57.39	34.38
Shared task	25.65	47.84	33.40

4.3.2. Preposition Detection Verification

Preposition errors in English statements can be mainly classified as the same as article. The proportions of incorrect statements are shown in Table 10.

10. Proportion of three types of prepositional errors.

Type	Preposition redundancy (%)	Preposition missing (%)	Preposition misuse (%)
NUCLE	24.42	56.38	34.08
Error ratio	18.12	24.03	57.91

Preposition data were obtained from the NUCLE-release2.2 corpus, and 28 types were obtained, where prep represents the number of errors. The preposition corpus data statistics are shown in Table 11.

11. Preposition corpus data statistics.

	Training set	Test set
Number of statements	57151	1381
Number of participles	1161567	29207
Number of prep errors	133665	3332

The data preprocessing of the prepositional corpus according to the experimental evaluation indexes yielded the following test results before and after preprocessing.

As can be seen from Table 12, the preprocessing of this dataset improved all the indicators of the prepositional dataset, indicating that the preprocessing was carried out with better results, which was beneficial to the subsequent experiments.

12. Test comparison results before and after corpus preprocessing.

	Precision (%)	Recall (%)	F₁ (%)
Before preprocessing	68.13	13.47	21.82
After preprocessing	68.87	14.20	23.54

In order to test the effect of different values of the error difference C factor on the F1 value of the prepositional error detection system, the experiment will set the C value, as shown in Table 13.

From Table 13, it can be seen that when C takes the value of 0.95, the F₁ value is 26.74, which takes the highest value among the five C values. Therefore, the C value was set to 0.95, and the Wike corpus was added to the original dataset, respectively, and placed into the prepositional dataset for testing, and the test results were obtained, as shown in Table 14.

13. Variation of values with different values.

C	0.98 (%)	0.95 (%)	0.9 (%)	0.8 (%)	0.7 (%)
F₁	25.27	26.74	23.47	18.54	10.78

As can be seen from Table 14, the expansion of the preposition training set resulted in a significant increase in its F1 value up to 27.68%. To further test the error detection effectiveness of the preposition detection model, the model was compared with Naive Bayes and shared task, and the results are shown in Table 15.

14. Changes in values after expanding the training set.

	F₁ (%)
UNCLE	23.54
UNCLE + Wiki (10wstrip)	24.57
UNCLE + Wiki (50wstrip)	27.68

As can be seen from Table 15, the F1-value of this model is 27.68, which is 16.19 and 0.17 higher than the other two models, respectively, indicating that this model is more effective in error detection.

15. Comparison results of F₁ values of the three models.

	F₁ (%)
NB-priors	11.87
Maxent	27.68
Shared task	27.51

4.3.3. Validation of Detection System Effect

At present, the common grammatical errors of verbs in college English composition mainly include four types of errors, verb missing errors, modal errors, tense errors, and subject-verb agreement errors. According to the statistics of the corpus, the results of verb error classification are shown in Table 16.

16. Proportion of the four verb errors.

Error type	Verb missing error (%)	Verb mood error (%)	Verb tense error (%)	Subject-verb agreement error (%)
Error percentage	6	19	36	39

The test materials were obtained from the language material library CLEC, from which 100 English compositions were arbitrarily selected for the experiment, which contained a total of 1,128 sentences and 1,083 mislabeled sentences. This experiment, referring to the process of proposed by Xu et al., uses the above dataset to test the effect of grammar error detection, as shown in Table 17 [23].

From Table 16, the overall accuracy of the grammar error detection system is 62.14%, the recall rate is 37.00%, and the F1-value is 46.40%. The overall error detection rate is relatively high, and the overall error detection accuracy is further improved and compared with the results of articles, prepositions, and verbs.

17. Effect of grammar error detection.

	Precision (%)	Recall (%)	F1 (%)
Verb error detection	45.47	29.67	34.95
Subject-verb agreement error detection	39.69	20.98	32.06
Error detection of the singular and plural forms of noun	86.03	53.97	63.04
Article error detection	68.14	43.05	52.57
Preposition error detection	71.98	38.06	50.18
Overall error detection	62.26	37.14	46.56

5. Conclusion

In summary, the English syntax error detection model based on the combination of multilayer rules and LSTM model proposed in the study can improve the accurate error detection results of English syntax with significant error detection effects and can be applied in the teaching of English grammar. The experimental results suggest that the overall error detection effect of the proposed model is significantly improved compared with the traditional error detection model, and better error detection effect can be achieved in the grammar of articles, prepositions, and verbs. The pressure of English teachers at work can be reduced by higher efficiency of error detection. The innovation of this study is to correct grammar by combining the advantages of LSTM in temporal change data processing so as to greatly improve the accuracy of grammar translation.

Conflicts of Interest

The authors declared that they have no conflicts of interest regarding this work.

Open Research

Data Availability

The experimental data used to support the findings of this study are available from the corresponding author upon request.

References

1 Jin Y., Yu L., Li G., and Fei S., A 6-DOFs event-based camera relocalization system by CNN-LSTM and image denoising, Expert Systems with Applications. (2020) 170, 114535, https://doi.org/10.1016/j.eswa.2020.114535.
10.1016/j.eswa.2020.114535
Web of Science® Google Scholar
2 Li W., Gao X., and Liu H., Efficient prediction of transonic flutter boundaries for varying Mach number and angle of attack via LSTM network, Aerospace Science and Technology. (2020) 110, 106451.
10.1016/j.ast.2020.106451
Web of Science® Google Scholar
3 Kang H., Yang S., Huang J., and Oh J., Time series prediction of wastewater flow rate by bidirectional LSTM deep learning, International Journal of Control, Automation and Systems. (2020) 18, no. 12, 3023–3030, https://doi.org/10.1007/s12555-019-0984-6.
10.1007/s12555-019-0984-6
Web of Science® Google Scholar
4 Pajak K. and Adam G., Grammatical error correction with denoising autoencoder, International Journal of Advanced Computer Science and Applications. (2021) 12, no. 8, https://doi.org/10.14569/ijacsa.2021.0120893.
10.14569/IJACSA.2021.0120893
Web of Science® Google Scholar
5 Qiu Z. and Qu Y., A two-stage model for Chinese grammatical error correction, IEEE Access. (2019) 7, 146772–146777, https://doi.org/10.1109/access.2019.2940607.
10.1109/ACCESS.2019.2940607
Web of Science® Google Scholar
6 Sagar A., Ashwini D., and Irfan S., Grammatical error correction (GEC): research approaches till now, International Journal of Computer Application. (2019) 178, no. 40, 1–3.
10.5120/ijca2019919275
Google Scholar
7 Park C., Yang Y., Lee C., and Lim H., Comparison of the evaluation metrics for neural grammatical error correction with o, IEEE Access. (2020) 8, 106264–106272, https://doi.org/10.1109/access.2020.2998149.
10.1109/ACCESS.2020.2998149
Web of Science® Google Scholar
8 Wang B., au F., Hirota K., Liu C., Dai Y., and Jia Z., An approach to NMT Re-ranking using sequence-l for grammatical error correction, Journal of Advanced Computational Intelligence and Intelligent Informatics. (2020) 24, no. 4, 557–567, https://doi.org/10.20965/jaciii.2020.p0557.
10.20965/jaciii.2020.p0557
Web of Science® Google Scholar
9 Jared L., Chris A., and Shankar K., Data weighted training strategies for grammatical error correction, Transactions of the Association for Computational Linguistics. (2020) 8, 634–646.
10.1162/tacl_a_00336
Google Scholar
10 Alessandro C. and Euro B., Forecasting spatially-distributed urban traffic volumes via multi-target LSTM-based neural network regressor, Mathematics. (2020) 8, no. 12.
Web of Science® Google Scholar
11 Sun L., Wang Y., He J., Li H., Peng D., and Wang Y., A stacked LSTM for atrial fibrillation prediction based on multivariate ECGs, Health Information Science and Systems. (2020) 8, no. 1, https://doi.org/10.1007/s13755-020-00103-x.
10.1007/s13755-020-00103-x
PubMed Web of Science® Google Scholar
12 Mario T., Miguel R., and Felipe R., PV power prediction, using CNN-LSTM hybrid neural network model. Case of study: temixco-morelos, méxico, Energies. (2020) 13, no. 24.
Web of Science® Google Scholar
13 Ren B., The use of machine translation algorithm based on residual and LSTM neural network in translation teaching, Plos One. (2020) 15, no. 11, e0240663, https://doi.org/10.1371/journal.pone.0240663.
10.1371/journal.pone.0240663
CAS PubMed Web of Science® Google Scholar
14 Sur C., CRUR: coupled-recurrent unit for unification, conceptualization and context capture for language representation - a generalization of bi directional LSTM, Multimedia Tools and Applications. (2020) 80, 1–43.
Web of Science® Google Scholar
15 Zhu A., Wu Q., Cui R., Wang T., Hang W., Hua G., and Snoussi H., Exploring a rich spatial-temporal dependent relational model for skeleton-based action recognition by bidirectional LSTM-CNN, Neurocomputing. (2020) 414, 90–100, https://doi.org/10.1016/j.neucom.2020.07.068.
10.1016/j.neucom.2020.07.068
Web of Science® Google Scholar
16 Fang Z., Wang Y., Peng L., and Hong H., Predicting flood susceptibility using long short-term memory (LSTM) neural network model, Journal of Hydrology. (2020) 594, https://doi.org/10.1016/j.jhydrol.2020.125734.
10.1016/j.jhydrol.2020.125734
Web of Science® Google Scholar
17 Yaqub M., Asif H., Kim S., and Lee W., Modeling of a full-scale sewage treatment plant to predict the nutrient removal efficiency using a long short-term memory (lstm) neural network, Journal of Water Process Engineering. (2020) 37, 101388, https://doi.org/10.1016/j.jwpe.2020.101388.
10.1016/j.jwpe.2020.101388
Web of Science® Google Scholar
18 Yang H., Efficiency of online grammar checker in English writing performance and students’ perceptions, Korea Journal of English Language and Linguistics. (2018) 18, no. 3, 328–348.
10.15738/kjell.18.3.201809.328
Google Scholar
19 Potter R. and Fuller D., My new teaching partner? Using the grammar checker in writing instruction, English Journal. (2008) 98, no. 1, 36–41.
10.58680/ej20086723
Google Scholar
20 Choi A., Jung H., Lee K. Y., Lee S., and Mun J. H., Machine learning approach to predict center of pressure trajectories in a complete gait cycle: a feedforward neural network vs. LSTM network, Medical, & Biological Engineering & Computing. (2019) 57, no. 12, 2693–2703, https://doi.org/10.1007/s11517-019-02056-0.
10.1007/s11517-019-02056-0
PubMed Web of Science® Google Scholar
21 Dai Y., Lu Z., Zhang H., Zhan T., Lu J., and Wang P., A correction method of environmental meteorological model based on long-short-term memory neural network, Earth and Space Science. (2019) 6, no. 11, 2214–2226, https://doi.org/10.1029/2019ea000641.
10.1029/2019EA000641
Web of Science® Google Scholar
22 Lee S., Xie K., Ngoduy D., and Keyvan-Ekbatani M., An advanced deep learning approach to real-time estimation of lane-based queue lengths at a signalized junction, Transportation Research Part C: Emerging Technologies. (2019) 109, no. C, 117–136, https://doi.org/10.1016/j.trc.2019.10.011, 2-s2.0-85074159922.
10.1016/j.trc.2019.10.011
Web of Science® Google Scholar
23 Xu X., Exploration of English composition diagnosis system based on rule matching, International Journal of Emerging Technologies in Learning (iJET). (2018) 13, no. 7, https://doi.org/10.3991/ijet.v13i07.8603, 2-s2.0-85049194069.
10.3991/ijet.v13i07.8603
Google Scholar
24 Zhu Y., Zhang W., Chen Y., and Gao H., A novel approach to workload prediction using attention-based LSTM encoder-decoder network in cloud environment, EURASIP Journal on Wireless Communications and Networking. (2019) 2019, no. 1, 7–18, https://doi.org/10.1186/s13638-019-1605-z.
10.1186/s13638-019-1605-z
Web of Science® Google Scholar
25 Saumil M., Agarwal A., Shukla A., and Tiwari R., A comprehensive evaluation for the prediction of mortality in intensive care units with LSTM networks: patients with cardiovascular disease, Biomedizinische Technik. Biomedical engineering. (2019) 65, no. 4, 435–446, https://doi.org/10.1515/bmt-2018-0206.
10.1515/bmt-2018-0206
Web of Science® Google Scholar
26 Kim M. and Kim J., A short-term prediction method of the IGS RTS clock correction by using LSTM network, Journal of Positioning, Navigation, and Timing. (2019) 8, no. 4, 209–214.
Google Scholar
27 Autonomous A., Networks - neural networks; new findings on neural networks described by investigators at beijing technology and business university (Bi-directional lstm model with symptoms-frequency position attention for question answering system in medical domain), Journal of Robotics & Machine Learning. (2019) 20.
Google Scholar
28 Cho Y.-B., Kim K.-J., Ku J.-A., and Woo S.-H., Change of malicious code API call pattern extraction using RNN and LSTM, International Conference On Future Information & Communication Engineering. (2019) 11, no. 1, 277–280.
Google Scholar
29 Kim H. and Bae H., Hyper parameter tuning method based on sampling for optimal LSTM model, Journal of the Korea Society of Computer and Information. (2019) 24, no. 1, 137–143.
Google Scholar
30 Aslam M. and Osama A. H., A control chart for gamma distribution using multiple dependent state sampling, Industrial Engineeering & Management Systems. (2017) 16, no. 1, 109–117.
10.7232/iems.2017.16.1.109
Web of Science® Google Scholar

All articles

Analyzing the Application of Multimedia Technology Assisted English Grammar Teaching in Colleges

Abstract

1. Introduction

2. Basic Methods

2.1. LSTM Fundamentals

2.2. Back Propagation Process of LSTM

3. English Grammar Checker Design

3.1. Syntax Checking Based on Multilayer Rule Model

3.1.1. Overall Module Design

3.1.2. Part-of-Speech Tagging

3.1.3. Rule Base Design

3.1.4. Multilayer Rule Syntax Error Correction

3.2. LSTM Syntax Checker Design Based on Attention Model

3.3. LSTM Model Based on Attention Model with Combination of Positive and Reverse Order

3.4. Design and Implementation of Verb Syntax Checking

4. Experimental Results and Analysis

4.1. Evaluation Contents

4.2. Evaluation Criteria

4.3. Experimental Results and Analysis

4.3.1. Verification of Article Detection

4.3.2. Preposition Detection Verification

4.3.3. Validation of Detection System Effect

5. Conclusion

Conflicts of Interest

Open Research

Data Availability

References

Figures

References

Information

About Wiley Online Library

Help & Support

Opportunities

Connect with Wiley

Analyzing the Application of Multimedia Technology Assisted English Grammar Teaching in Colleges

Abstract

1. Introduction

2. Basic Methods

2.1. LSTM Fundamentals

2.2. Back Propagation Process of LSTM

3. English Grammar Checker Design

3.1. Syntax Checking Based on Multilayer Rule Model

3.1.1. Overall Module Design

3.1.2. Part-of-Speech Tagging

3.1.3. Rule Base Design

3.1.4. Multilayer Rule Syntax Error Correction

3.2. LSTM Syntax Checker Design Based on Attention Model

3.3. LSTM Model Based on Attention Model with Combination of Positive and Reverse Order

3.4. Design and Implementation of Verb Syntax Checking

4. Experimental Results and Analysis

4.1. Evaluation Contents

4.2. Evaluation Criteria

4.3. Experimental Results and Analysis

4.3.1. Verification of Article Detection

4.3.2. Preposition Detection Verification

4.3.3. Validation of Detection System Effect

5. Conclusion

Conflicts of Interest

Open Research

Data Availability

References

Figures

References

Related

Information