Open Access

A Study on the Effectiveness of Designing and Applying Computer-Assisted Grammar Error Correction System in College English Writing Teaching

,  and   
Sep 29, 2025

Cite
Download Cover

Figure 1.

Transformer model structure based on replication mechanism
Transformer model structure based on replication mechanism

Figure 2.

Overall system framework
Overall system framework

Figure 3.

Syntax correction process
Syntax correction process

Figure 4.

User feedback process
User feedback process

Figure 5.

Error correction results of the method for four different syntax errors
Error correction results of the method for four different syntax errors

Figure 6.

Arranged by precision from lowest to highest
Arranged by precision from lowest to highest

Figure 7.

Arranged by recall from lowest to highest
Arranged by recall from lowest to highest

The correct precision rate of various grammatical errors is segmented statistics

Precision of error correction Syntax error class
60~70% Trans, Wci, WOinc
70~80% Wform, WOadv, Pform, Pref
80~90% SVA, Ssub, Npos, Vt, Rloc-, Prep, Vm, Others, Vform
>90% ArtOrDet, Nn, V0

Pretest and posttest paired sample t-test for the second round of action experiments

Pairing difference t df Sig. (2-tailed)
Mean value Standard deviation Standard error mean 95% confidence interval of the difference
Lower limit Upper limit
Pairing 2 Posttest-Pretest 13.27% 13.425% 4.436% 3.418% 23.064% 3.138 9 0.012

Test results of two syntax error correction models

Model Precision /% Recall /% F0.5 /%
CAMB 20.84 22.37 21.13
Ours 37.52 42.75 38.46

Article and preposition error correction results

Error type Method comparison P R F0.5
Article error UIUC 0.5246 0.3922 0.4914
Corpus GEC 0.5873 0.4084 0.5400
Ours 0.6148 0.5982 0.6114
Preposition error UIUC 0.4072 0.3348 0.3903
Corpus GEC 0.2685 0.2752 0.2698
Ours 0.4214 0.4563 0.4280

Pretest and posttest paired sample t-test for the first round of action experiments

Pairing difference t df Sig. (2-tailed)
Mean value Standard deviation Standard error mean 95% confidence interval of the difference
Lower limit Upper limit
Pairing 1 Posttest-Pretest 0.74% 19.502% 5.974% -13.024% 14.122% 0.079 10 0.952

Comparative statistics of scoring rate between pretest and posttest

Item number Examine the part of speech Pretest score rate Posttest score rate Fluctuating value
1 Nouns 83.52% 100% 16.48%
2 Article 83.52% 92.25% 8.73%
3 Article 75.14% 66.43% -8.71%
4 Nouns 27.15% 61.98% 34.83%
5 Adverb 88.02% 100% 11.98%
6 Adjective 66.34% 88.05% 21.71%
7 Article 44.29% 70.48% 26.19%
8 Preposition 57.83% 70.48% 12.65%
9 Preposition 62.26% 92.72% 30.46%
10 Non-predicate verb 62.26% 88.41% 26.15%
11 Pronoun 35.94% 70.74% 34.80%
12 Non-predicate verb 70.64% 92.48% 21.84%
13 Pronoun 79.35% 96.79% 17.44%
14 Predicate verb 22.86% 40.35% 17.49%
15 Predicate verb 70.64% 88.07% 17.43%
16 Coordinate conjunction 79.35% 92.43% 13.08%
17 Subordinating conjunctions 79.35% 100% 20.65%
18 Coordinate conjunction 57.43% 79.24% 21.81%
19 Predicate verb 70.64% 85.87% 15.23%
20 Subordinating conjunctions 5.47% 27.28% 21.81%

Pre-test score rate statistics

Item number Examine the part of speech Scoring rate Item number Examine the part of speech Scoring rate
1 Nouns 83.52% 11 Pronoun 35.94%
2 Article 83.52% 12 Non-predicate verb 70.64%
3 Article 75.14% 13 Pronoun 79.35%
4 Nouns 27.15% 14 Predicate verb 22.86%
5 Adverb 88.02% 15 Predicate verb 70.64%
6 Adjective 66.34% 16 Coordinate conjunction 79.35%
7 Article 44.29% 17 Subordinating conjunctions 79.35%
8 Preposition 57.83% 18 Coordinate conjunction 57.43%
9 Preposition 62.26% 19 Predicate verb 70.64%
10 Non-predicate verb 62.26% 20 Subordinating conjunctions 5.47%

Error correction results of Chinese students’ compositions by this paper model

Error type The number of errors marked Number of errors corrected Correct the number of errors P R F1
ArtOrDet 42 37 34 0.9189 0.8095 0.8608
Nn 69 64 59 0.9220 0.8551 0.8872
Npos 7 6 5 0.8333 0.7143 0.7692
Pform 18 13 10 0.7692 0.5556 0.6452
Pref 31 25 20 0.8000 0.6452 0.7143
Prep 32 28 24 0.8571 0.7500 0.8000
Rloc- 59 52 44 0.8462 0.7458 0.7928
Ssub 145 125 104 0.8320 0.7172 0.7704
SVA 63 58 47 0.8103 0.7460 0.7769
Trans 6 5 3 0.6000 0.5000 0.5455
V0 61 53 49 0.9245 0.8033 0.8597
Vform 46 30 26 0.8667 0.5652 0.6842
Vm 28 21 18 0.8571 0.6429 0.7347
Vt 35 32 27 0.8438 0.7714 0.8060
Wci 7 5 3 0.6000 0.4286 0.5000
Wform 6 4 3 0.7500 0.5000 0.6000
WOadv 14 8 6 0.7500 0.4286 0.5455
WOinc 20 15 9 0.6000 0.4500 0.5143
Others 89 79 68 0.8608 0.7640 0.8095
Total 778 660 559 0.8470 0.7185 0.7775

Test results before and after the first round of action experiments

Item number Before the first round of action experiments test (scoring rate) After the first round of action experiments test (scoring rate) Fluctuating value
Preposition 67.74% 77.52% 9.78%
Article 60.24% 88.45% 28.21%
Pronoun 83.91% 89.36% 5.45%
Adjective 35.02% 64.37% 29.35%
Adverb 73.23% 66.74% -6.49%
Nouns 86.17% 94.02% 7.85%
Predicate verb 58.93% 91.51% 32.58%
Non-predicate verb 54.47% 76.24% 21.77%
Coordinate conjunction 72.05% 73.13% 1.08%
Subordinating conjunctions 42.39% 45.52% 3.13%
Average score 6.41 6.85 0.44

Paired sample T-test for different question types

Pairing difference t df Sig. (2-tailed)
Mean value Standard deviation Standard error mean 95% confidence interval of the difference
Items Lower limit Upper limit
A Posttest-Pretest 5.164 3.364 0.526 4.115 6.172 9.864 46 0.000
B 3.285 4.128 0.612 1.954 4.376 5.201 46 0.000
C 3.447 3.902 0.579 2.283 4.519 5.945 46 0.000
D 1.841 3.017 0.474 0.814 2.754 3.822 46 0.000
E 3.423 4.725 0.923 1.532 4.308 3.487 46 0.000
F 1.586 2.413 0.361 0.341 0.797 4.378 46 0.000

Test data set

Error type Number Proportion /%
Article 652 11.65
Preposition 516 9.21
Nouns 461 8.23
Principal consensus 347 6.20
Verb form 265 4.73
Error total 2241 40.02
Total amount 5600 100

The correct recall rate of various grammatical errors is segmented statistics

Precision of error correction Syntax error class
<60% Wci, WOadv, WOinc, Trans, Wform, Pform, Vform
60~70% Vm, Pref
70~80% Npos, Ssub, Rloc-, SVA, Prep, Others, Vt
>80% V0, ArtOrDet, Nn
Language:
English