Using Word2Vec method to improve the efficiency of word meaning understanding and memory in English teaching

Abstract

In order to improve the efficiency of word meaning understanding and memory in English teaching, this article studied a new vocabulary teaching method by applying Word2Vec, a neural network-based word embedding technology. Word2Vec technology can map vocabulary to high-dimensional space and represent semantic relationships between vocabulary in vector form, thereby capturing subtle semantic differences between vocabulary. By calculating the distance and direction between vectors to infer the relationship between vocabulary, this article also introduced a mobile application that integrated multiple functional modules such as vocabulary learning, memory games, learning progress tracking, and regular push notifications, providing students with a personalized learning experience. Through this application, students can learn vocabulary anytime and anywhere, and dynamically adjust their learning plans based on their learning progress and memory effects. The significance of this paper is that the accuracy and recall rates of word meaning comprehension and memory efficiency reached 96% and 98%, respectively, proving the effectiveness of this method in English teaching. This study is not only of great significance to improving the efficiency of primary school English vocabulary teaching, but also provides new ideas for the application of natural language processing and machine learning in the field of education.

Keywords

Word2Vec model English teaching understanding of word meanings memory efficiency natural language processing optimization of teaching methods

Introduction

In the process of English teaching, understanding and memorizing word meanings is a key link in student language learning. The traditional vocabulary teaching methods^1,2 mainly include direct translation, repetitive memory, and situational teaching. These methods have to some extent helped students master a large number of vocabulary, but there are also some obvious shortcomings. The method of direct translation³ can quickly convey the meaning of words, but it can easily lead to shallow understanding of vocabulary by students and cannot be flexibly applied to different contexts. Repetitive memory emphasizes mechanical memorization, which can make students feel dull and boring, and the learning effect is often not long-lasting. Situational teaching⁴ helps students understand vocabulary by constructing specific scenarios, but in actual teaching, it is limited by classroom time and resources, making it difficult to cover the learning needs of all vocabulary. With the development of information technology, computer-aided teaching has gradually been applied in the field of language teaching^5,6 to overcome the shortcomings of traditional methods.⁷ Nevertheless, how to effectively improve students’ understanding and memory efficiency of word meanings remains an urgent issue that needs to be addressed.

This article adopts the Word2Vec method to improve the efficiency of word meaning understanding and memory in English teaching. Word2Vec is a neural network-based word embedding technique that maps vocabulary to high-dimensional spaces and represents semantic relationships between words in vector form. Compared with traditional vocabulary representation methods, Word2Vec not only captures subtle semantic differences between words, but also infers relationships between words by calculating the distance and direction between vectors. Word2Vec includes two models: the Continuous Bag of Words (CBOW) model and the Skip-gram model. The former trains the model by predicting the context of a given vocabulary, while the latter trains the model by predicting the vocabulary in the given context. These two models have their own advantages and can perform excellently in different application scenarios. By applying Word2Vec technology, this article investigates a teaching method that enables students to improve memory efficiency and overall language proficiency while understanding word meanings.

Related work

In the past few decades, Natural Language Processing (NLP) and Machine Learning (ML) technologies have provided a rich technological foundation for the rapid development of educational applications. Johnson, S. Joshua⁸ et al. proposed methods for understanding and using word embeddings, summarized word embedding strategies such as Word2Vec from both theoretical and mathematical perspectives, and incorporated research results on standard word embedding techniques to help researchers quickly master word embedding techniques for more effective processing of text data. This study provided a solid theoretical foundation for the application of NLP^9,10 in education. Ning, Gelin¹¹ et al. proposed a biomedical named entity recognition model based on Glove-BiLSTM-CRF. The model utilized GloVe and Bidirectional Long Short-Term Memory (BiLSTM) networks^12,13 to train word vectors for semantic and character morphological features, and combined them into the final representation of words. The model used the BiLSTM-CRF deep learning model to recognize entity categories and achieved good results with an F1 value of 75.62% in the JNLPBA 2004 biomedical named entity recognition task. Compared with Word2Vec, GloVe can better capture global semantic information, making word vectors more applicable in different tasks. Ma, Kai¹⁴ et al. proposed a deep learning algorithm based on BERT-BiLSTM-CRF to automatically extract temporal information from social media messages. By combining BERT (Bidirectional Encoder Representations from Transformers)¹⁵ pre-trained model and BiLSTM-CRF model, the algorithm effectively captures long-range contextual information and achieves better performance than existing models on Chinese social media message datasets. The development of these technologies has enabled NLP-based educational applications to be realized, which not only improve teaching efficiency,^16,17 but also enhance students’ learning experience and effectiveness.

The application of word embedding technology in educational applications has significantly improved teaching effectiveness and student learning experience. Khomsah, Siti¹⁸ et al. compared the performance of sentiment analysis models using Word2Vec and FastText word embeddings^19,20 on the Indonesian hotel review dataset and found that FastText outperformed Word2Vec in accuracy when combined with random forests²¹ and additional tree classifiers,²² significantly improving the accuracy of sentiment analysis. Language learners often encounter different morphological changes and a large number of new words in language learning,^23,24 which requires word embedding techniques to help scholars better understand and remember vocabulary. This study hopes to improve learning efficiency by adjusting learning content and learning strategies in real time based on students’ learning behaviors and academic performance through NLP’s personalized learning system. This paper uses these technologies to further improve the efficiency of word meaning understanding and memory in education, thereby achieving the purpose of the study.

Methods to improve word meaning understanding and memory efficiency

Application of Word2Vec in teaching

Word2Vec uses neural network word embedding technology to map vocabulary to a high-dimensional vector space to capture semantic relationships between words. In large-scale corpus training, Word2Vec effectively captures semantic similarity between words with its efficient computing power. The process of generating word vectors during training is shown in Figure 1. The trained word vectors are used for text classification,²⁵ sentiment analysis, and machine translation²⁶ tasks.

Figure 1.

Context sensitive word vector generation process.

Traditional methods often rely on dictionary definitions and simple translations, which are difficult to capture subtle semantic differences of vocabulary in different contexts. Word2Vec uses neural network training, with backpropagation algorithm used to optimize the quality of word vectors. Backpropagation gradually reduces prediction errors by calculating error gradients and adjusting weights, thereby improving the accuracy and generalization ability of the model. By generating high-dimensional vector representations of vocabulary through Word2Vec, the similarity between words is calculated, providing a more intuitive and accurate understanding of word meanings. Text data is extracted from large-scale English corpora such as Wikipedia and news articles, and preprocessed to remove punctuation, stop words, and low-frequency vocabulary. Using the Skip-gram model, the processed corpus is trained to obtain vector representations for each vocabulary. During the training process, the vector dimension is set to 300, and the window size is 5. Negative sampling techniques are used to improve the training efficiency of the model. After training, the semantic relationship between words is determined by calculating the cosine similarity between word vectors.^27,28

The teaching design based on Word2Vec includes the following aspects: learning path planning dynamically adjusts according to students’ current vocabulary and learning goals; the vocabulary selection strategy is based on the semantic association of vocabulary and students’ interests; and the interactive design includes vocabulary matching, semantic association, and dynamic exercises to ensure that students strengthen their memory during the interaction.

This article designs many fun games to increase students’ interest in learning. The first game is “Word Matching,” which requires students to pair words with similar meanings. Another game is “Antonym Challenge,” where students need to find the antonyms of vocabulary within a limited time. There is also a “fill in the blank practice” game, in which the system provides a sentence with a blank, and students need to choose the most suitable word from the given vocabulary to fill in the blank. During the selection process, the system calculates the similarity between each option and the context, and provides corresponding prompts and feedback. There are also games such as “word relay” and “quick response” that reinforce memory by repeatedly using and connecting newly learned vocabulary.

Mobile software design

In order to facilitate students to learn vocabulary anytime and anywhere, this article develops a mobile application. Multi-threaded concurrent processing is implemented in the application to improve training efficiency. Negative sampling technology is used to optimize the algorithm, and memory-mapped files are used to manage large-scale datasets to ensure efficient and stable memory usage. This application includes several functional modules: vocabulary learning, memory games, learning progress tracking, and regular push notifications, as shown in Figure 2.

Figure 2.

Software main interface.

The vocabulary learning module utilizes Word2Vec technology to generate semantic association graphs between words. After students input a word, the system finds words with similar semantics based on the pre-trained Word2Vec model and displays them to the students. Each vocabulary comes with a detailed definition, usage examples, and pronunciation. The program provides multidimensional explanations of vocabulary sources, affixes, and root analysis to help students understand words from multiple perspectives. This module is equipped with dynamic exercise questions, through which students continuously strengthen their memory of new vocabulary. These exercise questions are dynamically generated based on the current learning progress and mastery of students, to ensure the targeted and effective nature of the exercises. The fun and interactivity in the learning process are very important for students to persist in learning. When designing memory games, it is important to pay attention to the diversity and challenges of the game. The core of the game is a vocabulary relationship network generated based on the Word2Vec model. In these games, the system dynamically adjusts the game difficulty based on the student’s performance, ensuring that each student can make progress in challenges that suit them.

The learning progress module records the detailed learning records of students, as shown in Figure 3.

Figure 3.

Learning progress tracking page.

These data enable students to have a comprehensive understanding of their learning situation. The system provides personalized learning suggestions for students based on this data, helping them optimize their learning strategies. The system regularly pushes review tasks and new vocabulary learning tasks suitable for the current learning stage of students by analyzing their learning data. When the system detects a low level of mastery of a certain word, it can promptly push a review reminder to ensure that students consolidate before forgetting. The push system also designs incentive mechanisms for learning challenges and reward tasks to stimulate students’ interest and motivation in learning. This application helps students improve the efficiency of vocabulary learning and develop good learning habits. This mobile application provides an efficient, convenient, and interesting vocabulary learning solution by cleverly integrating Word2Vec technology and multiple learning functions. Students can learn vocabulary anytime and anywhere, gradually improving their vocabulary understanding and memory efficiency. This innovative educational tool not only conforms to the development trend of modern education, but also provides students with a more personalized and intelligent learning experience. Through the organic combination of vocabulary learning module, memory game module, and learning progress tracking and regular push module, the application has shown great potential in improving the efficiency of word meaning understanding and memory in English teaching.

Model optimization

Although Word2Vec performs well in capturing semantic relationships between words, it is unable to handle contextual information, which may lead to insufficient understanding of word meanings in practical applications. In order to further improve the performance of the model, this article applies a Transformer-based BERT model to fine tune Word2Vec word vectors. The BERT model is proposed by Google and is a language representation model based on bidirectional Transformers.^29,30 Unlike traditional one-way language models, BERT aims to more accurately capture the meaning of vocabulary in different contexts by considering contextual information.

This article uses masking language model and next sentence prediction method to pre-train the model on the same corpus. The masking language model randomly masks some vocabulary in a sentence and requires the model to predict the masked vocabulary, thereby learning the contextual relationships of vocabulary. The next sentence prediction method captures the relationship between sentences by determining whether two sentences are related to each other. The fine-tuned BERT model is used to perform secondary training on Word2Vec word vectors. For the purpose of demonstrating more accurate semantic relationships in different contexts, the word vectors generated by Word2Vec are used as initial inputs during the training process, and the output word vectors are further adjusted through the BERT model.

Evaluation indicators

This article evaluates the improvement of word meaning understanding and memory efficiency in English teaching using Word2Vec and optimized models based on the use of multiple evaluation indicators. Accuracy (Acc) is used to evaluate the accuracy of students in vocabulary comprehension tests:

A c c = \frac{N_{c}}{N_{t}}

(1)

In Formula (1), $N_{c}$ represents the number of vocabulary correctly understood by the student, and $N_{t}$ is the total number of vocabulary in the test.

Recall (R) is an indicator used to evaluate the ability of a classification model to recognize positive samples. It represents the proportion of correctly recognized samples among all actual positive samples, and its calculation formula is as follows:

R = \frac{T P}{F N + T P}

(2)

Among them, TP is the number of samples correctly predicted as positive, and FN is the number of samples actually predicted as positive but incorrectly predicted as negative. The recall rate ranges from 0 to 1, and the higher the value, the stronger the model’s ability to recognize positive samples.

The calculation formula of Precision (P) is as follows:

P = \frac{T P}{F P + T P}

(3)

The F1 score (F1) combines precision and recall, and is their harmonic mean, which is used to balance the two aspects of model performance. Its calculation formula is as follows:

F 1 = 2 \times \frac{P \times R}{P + R}

(4)

The Recall Rate (R_r) is used to evaluate students’ performance in vocabulary memory tests:

R_{r} = \frac{N_{r}}{N_{t}}

(5)

N_{r}

is the number of vocabulary that students have successfully memorized.

Cosine Similarity (CS) is used to evaluate the semantic similarity between words. The similarity formula is as follows:

C S (A, B) = \frac{A \cdot B}{∥ A ∥ ∥ B ∥}

(6)

In this similarity formula, $A$ and $B$ are two word vectors, and the dot product of the two vectors is represented by $A \cdot B$ . $∥ A ∥$ and $∥ B ∥$ , respectively, represent the norm of the two vectors.

Improving vocabulary understanding and memory experiment

Data collection and preprocessing

For the purpose of improving the generalization ability of the model, this article collects a large amount of data with rich vocabulary and context from English Wikipedia, The New York Times news articles, scientific journals, and literary works, covering English texts from different fields and with different difficulties, which ensures that the Word2Vec model performs well in various contexts. The regular expressions in Python and the natural language processing library NLTK are used to remove HTML tags, special characters, and stop words from some text data. Cleaning makes the text more concise and unified, achieving the goal of improving the efficiency and effectiveness of model training.

Experimental design

This article selects 60 second year students majoring in English from a certain university to participate in the experiment, and evaluates their vocabulary and understanding level by completing a vocabulary test before the experiment. These 60 students are randomly divided into an experimental group and a control group. The experimental group uses a vocabulary learning tool based on the Word2Vec method for learning. The control group uses traditional vocabulary learning methods, using traditional vocabulary lists and accompanying example sentences for learning.

Results

Model performance

Model performance analysis is an important step in understanding and improving a model. By evaluating the performance during the training process, the learning ability, convergence speed, and optimization effects at different stages of the model are understood. The loss function reflects the prediction error of the model on the training data. This article visually observes the learning process and convergence of the model by recording the loss function values of each training round and drawing their change curves. Figure 4 shows the variation of the loss function of the model during the training process.

Figure 4.

Changes in loss function during model training process.

In Figure 4, as the number of training rounds increases, the loss function shows a significant downward trend. In the first 20 training rounds, the loss function value rapidly decreases from the initial value of 0.96 to 0.37, indicating that the model quickly learns the basic patterns of the data in the early stages, significantly reducing prediction errors. From the 20th training round to the 50th training round, the loss function continues to decrease, but the speed slows down, with a value dropping from 0.37 to approximately 0.1. After the 50th training round, the decline rate of the loss function further slows down, and finally stabilizes at around 0.03 after the 80th training round, indicating that the model reaches a better state. This trend of change indicates that the model gradually converges and continuously optimizes during the training process, ultimately achieving good training results.

The performance before and after model optimization is compared and analyzed to evaluate the effectiveness of optimization measures. Figure 5 shows the performance comparison of word meaning understanding and memory efficiency before and after model optimization.

Figure 5.

Performance comparison before and after model optimization.

Figure 5 shows the performance comparison of the Word2Vec model before and after BERT fine-tuning. The P of the Word2Vec model gradually increases from the initial 60% to nearly 80%, and the R increases from the initial 55% to about 75%, showing a good trend of improvement. The performance improvement of the BERT fine-tuning model is more significant, with an P increase from the initial 70%–96% and a R increase from 65% to 98%. The optimized F1 has also been significantly improved. It is evident that the BERT model has stronger performance in terms of word meaning comprehension and memory efficiency. The performance improvement of the BERT model is more stable and significant in the early and middle stages, indicating that it has higher efficiency and accuracy in capturing contextual information and understanding word meanings. Especially in the middle and later stages of training, the performance improvement of the BERT model is still relatively significant, while the Word2Vec model gradually tends to flatten, indicating its limitations in specific stages.

To further demonstrate the performance of the optimized Word2Vec model, this article compares the performance of Word2Vec, Word2Vec optimized by BERT, GloVe, and FastText, as shown in Figure 6.

Figure 6.

Performance comparison of each model.

Figure 6 shows that the Acc, R, and F1 of the optimized Word2Vec model are 96%, 98%, and 97%, respectively, which are the highest among the four models, reflecting the excellent performance of the optimized Word2Vec model.

Teaching optimization

Modern education requires advanced technology to enhance teaching effectiveness. This article applies the Word2Vec model to analyze the semantic relationships between vocabulary in order to better understand the importance of different vocabulary in student learning. Figure 7 shows the heatmap of word vector similarity, which helps students quickly grasp the correlation between words in a more intuitive way during the learning process, thereby improving learning efficiency.

Figure 7.

Word vector similarity.

Figure 7 shows the similarity between different vocabulary, where the color depth represents the level of similarity. Through the heatmap, it is found that the similarity between semantically similar words “book” and “library” is 0.85, appearing in a dark area, while the similarity between semantically distant words “book” and “car” is only 0.2 in a light area. This visualization method helps students better understand the relationships between vocabulary, establish semantic networks, and improve learning efficiency during the learning process. Based on this result, this study applies the Word2Vec model to actual teaching. In a vocabulary class, students use the word relationship map generated by Word2Vec to conduct word pairing and semantic association analysis, which significantly improves the depth of word meaning understanding.

This article compares the accuracy of student vocabulary comprehension between traditional teaching methods and Word2Vec-based methods. Research has shown that different teaching methods have a significant impact on the learning outcomes of students when understanding vocabulary. This article further evaluates the teaching effectiveness based on the Word2Vec method through comparative experiments. Figure 8 shows the comparison of vocabulary comprehension effects among students under different teaching methods, while Figure 9 shows the comparison of memory effects among students under different methods, in order to understand the differences in vocabulary teaching effects between traditional teaching methods and Word2Vec methods.

Figure 8.

Comparison of vocabulary comprehension effects among students.

Figure 9.

Comparison of student memory effects.

Figure 8 shows the comparison of vocabulary comprehension effects among students under different teaching methods. Students who use the Word2Vec method show significant advantages in understanding vocabulary. Under traditional teaching methods, students have an accuracy rate of only about 70% in understanding vocabulary, while students using the Word2Vec method have an accuracy rate of about 90%. The Word2Vec method improves the accuracy of students using vocabulary in practical applications. This indicates that in traditional teaching, students rely more on memory for their understanding of vocabulary rather than a profound understanding of its semantics. The application of Word2Vec-based methods in vocabulary teaching has improved students’ understanding accuracy, established a more solid vocabulary system, and enhanced their vocabulary application ability.

Figure 9 shows the comparison of student vocabulary memory rates between traditional teaching methods and Word2Vec-based methods at 1, 3, and 6 month time periods. The median memory rates of traditional methods are 65%, 60%, and 55%, respectively, while the median rates of Word2Vec method are 82%, 75%, and 70%. This indicates that the Word2Vec method significantly improves memory rate and reduces fluctuations in memory rate, exhibiting higher performance in both short-term and long-term memory. Through comparison, it can be seen that the teaching method based on Word2Vec not only improves memory rate, but also reduces the differences in memory effects among students, providing strong data support for optimizing teaching strategies.

In order to further evaluate the differences in long-term memory effects among different teaching methods, this study designs a six-month follow-up experiment. In this experiment, students use traditional methods and Word2Vec-based methods for vocabulary learning, and conduct memory tests after 1, 3, and 6 months. Long-term tracking experiments can more accurately evaluate the differences in improving long-term vocabulary memory effects among different teaching methods. By comparing the memory rates of different time periods, which method can more effectively help students maintain vocabulary memory is identified, providing more scientific teaching method improvement suggestions, and helping teachers optimize teaching strategies and improve student learning effectiveness, as shown in Figure 10.

Figure 10.

Student learning progress tracking chart.

Figure 10 shows the learning progress of students under different teaching methods. After using the Word2Vec method, students significantly accelerate their learning progress, increasing from 20 new vocabulary words per week to over 40 new vocabulary words. This indicates that the teaching method based on Word2Vec effectively helps students master new vocabulary and improve learning efficiency, with better retention of learned vocabulary.

This article also collects students’ feedback on the application, as shown in Figure 11.

Figure 11.

Student feedback.

This study finds through student feedback that 90% of students believe that the application improves their vocabulary learning interest and memory efficiency. Student feedback indicates that the ability to dynamically adjust study plans has significantly helped their learning progress.

In order to evaluate the impact of the Word2Vec method on students’ long-term vocabulary memory and language ability, this article designs and implements a six-month follow-up study. Figure 12 shows the comparison of vocabulary memory rates of students using the traditional method and the optimized Word2Vec method at different time periods.

Figure 12.

Comparison of vocabulary memory rates in different time periods.

From Figure 12, it can be seen that the memory rates of students using the optimized Word2Vec method at 1 month, 3 months, and 6 months are 82%, 75%, and 70%, respectively, significantly higher than those using the traditional method, with memory rates of 65%, 60%, and 55%, respectively. This indicates that the optimized Word2Vec method has significant advantages in improving long-term vocabulary memory.

Comparison of different methods

Comparative analysis of different teaching methods can provide scientific basis for optimizing teaching strategies. This article evaluates the teaching effectiveness of different teaching methods using learning records of 5 students. The learning records are presented in Tables 1–3, respectively, to demonstrate the students’ performance in vocabulary learning, comprehension, and application. The comparison of vocabulary learning quantity among students under different teaching methods is shown in Table 1, which can intuitively reflect the number of new vocabulary mastered by students within the same learning time.

Table 1.

Comparison of the number of new vocabulary learning by students under different teaching methods.

Students	Traditional	Word2Vec	Difference
Student A	20	35	+15
Student B	25	40	+15
Student C	18	30	+12
Student D	22	38	+16
Student E	19	33	+14

Table 2.

Comparison of vocabulary comprehension accuracy among students under different teaching methods.

Students	Traditional (%)	Word2Vec (%)	Increase amplitude (%)
Student A	70	85	15
Student B	75	90	15
Student C	68	82	14
Student D	72	88	16
Student E	69	84	15

Table 3.

Comparison of vocabulary application ability test scores of students under different teaching methods.

Students	Traditional	Word2Vec	Increase amplitude
Student A	65	80	+15
Student B	70	85	+15
Student C	62	75	+13
Student D	68	83	+15
Student E	64	78	+14

Table 1 records the number of new vocabulary learned by students under different teaching methods. Table 1 shows that students using the Word2Vec method on average master more new vocabulary within the same learning time. Student A learns 20 new vocabulary words through traditional methods and 35 new vocabulary words through Word2Vec method, with a difference of 15. This trend is reflected among all students, with differences ranging from 12 to 16 vocabulary words. Overall, the Word2Vec-based method significantly increases the amount of vocabulary learning for students, indicating that it has a significant advantage in improving student learning efficiency.

In the comparison of different teaching methods, not only is the amount of vocabulary learned an important indicator, but the accuracy of vocabulary understanding is also a key factor. The understanding accuracy of students after using traditional methods and Word2Vec-based methods for vocabulary learning is shown in Table 2. These data intuitively demonstrate the differences in improving students’ vocabulary comprehension abilities among different methods.

This article records the understanding accuracy of students after using traditional methods and Word2Vec-based methods for vocabulary learning in Table 2. The data shows that the Word2Vec-based method significantly improves students’ vocabulary comprehension accuracy. Student A achieves a comprehension accuracy of 70% when using traditional methods, while using Word2Vec method improves it to 85%, with an improvement of 15%. This also reflects in other students, with an increase of between 14% and 16%. These results conclude that the Word2Vec-based method significantly improves students’ vocabulary comprehension skills.

In order to comprehensively evaluate the impact of teaching methods on students’ comprehensive abilities, this article designs a vocabulary application ability test. According to Table 3, the application ability test scores of students after using traditional methods and Word2Vec-based methods for vocabulary learning are recorded.

Table 3 shows the application ability test scores of students after using traditional methods and Word2Vec-based methods for vocabulary learning. According to the recorded data, the Word2Vec-based method significantly improves students’ vocabulary application ability. Student A’s score is 65 when using traditional methods, but improves to 80 when using Word2Vec method, with an increase of 15. This trend is reflected among all students, with an increase range of 13–15. These results confirm the strong data support of Word2Vec-based methods for improving teaching strategies, and demonstrate that students’ vocabulary understanding and practical application can improve accordingly.

In order to compare the performance of Word2Vec with the latest technology in English teaching, this article applies technologies such as BERT and GPT and makes a detailed comparison. Figure 13 shows the performance of different embedding models in terms of accuracy, recall and F1 score.

Figure 13.

Performance of different technologies in English teaching.

As can be seen from Figure 13, BERT performs best in accuracy and F1 score, reaching 90% and 89%, while GPT performs well in the generation task with an accuracy of 85%. Although Word2Vec performs slightly worse than the latest technology, its training speed and computational efficiency still have advantages in practical applications.

This article compares the input and output of the traditional method and the optimized Word2Vec method in teaching. Figure 14 shows the cost and benefit comparison of the two methods.

Figure 14.

Cost-benefit analysis.

Figure 14 shows that although the initial investment of the Word2Vec method is higher at 8000 yuan, its income of 15,000 yuan is significantly higher than that of the traditional method, indicating that the Word2Vec method is more cost-effective in long-term teaching.

Conclusions

The effectiveness of English teaching is closely related to the teaching methods. This article experimentally verified the significant advantages of the Word2Vec method in improving word meaning understanding and memory efficiency in English teaching. This article applied Word2Vec technology to improve the effectiveness of English vocabulary teaching by converting vocabulary into high-dimensional vector representations. Through an application that includes vocabulary learning and memory game modules, it enhanced students’ learning interest and enables them to understand the semantic relationships between vocabulary more intuitively. When using student data to train the model, this article adopted data anonymization to ensure that the student’s identity cannot be identified, used encryption technology to protect data transmission and storage, and established strict data usage protocols to ensure that the data was only used for research purposes. An ethics committee was also established to oversee the data usage process to ensure compliance with ethical standards. In order to overcome the problems of long model training time, high computational resource requirements, and uneven corpus quality, this article adopted distributed computing, optimizes training algorithms, and established a high-quality corpus. In order to further optimize the vocabulary representation effect, this article applied the BERT model for fine-tuning. The results of the vocabulary application ability test for students showed that within the same learning time, students who used the Word2Vec learning method significantly improved their mastery of vocabulary and understanding accuracy, which further confirmed the effectiveness of the Word2Vec method in practical teaching applications. This article provides a new English teaching tool through the Word2Vec model, providing students with richer contextual information. Word2Vec technology is not only suitable for English teaching, but also has potential in teaching other languages. Especially in teaching English as a second language, Word2Vec can help students understand the multiple semantics of vocabulary. However, the quality of corpora in different languages and the ambiguity of vocabulary may pose challenges, requiring model optimization and adjustment for specific languages. Future research should explore more application scenarios and teaching strategies to further optimize the Word2Vec model and provide more scientific and effective solutions for English teaching.

Statements and declarations

Footnotes

Conflicting interest

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The authors received no financial support for the research, authorship, and/or publication of this article.

References

Permana

Yoga

. Teaching vocabulary for elementary school students. Art Teach Eng Foreign Lan 2020; 1(2): 1–4.

Schmitt

. Understanding vocabulary acquisition, instruction, and assessment: a research agenda. Lang Teach 2019; 52(2): 261–274.

Djauhar

. The grammar-translation method, the direct method, and the audio-lingual method. Langua: J Linguis Lit Lang Educ 2021; 4(1): 84–88.

Uleng

. Teaching preposition through oral situational language teaching method at the first year students of SMP Negeri 33 Makassar. Jurnal Andi Djemma| Jurnal Pendidikan 2021; 4(1): 1–7.

Turan

Akdag-Cimen

. Flipped classroom in English language teaching: a systematic review. Comput Assist Lang Learn 2020; 33(5-6): 590–606.

Magnusson

Godhe

A-L

. Multimodality in language education: implications for teaching. Des Learn 2019; 11(1): 127–137.

Bryant

. Effects of computer-assisted instruction on the mathematics performance of students with learning disabilities: a synthesis of the research. Exceptionality 2020; 28(1): 30–44.

Johnson

Ramakrishna Murty

Navakanth

. A detailed review on word embedding techniques with emphasis on word2vec. Multimed Tool Appl 2024; 83(13): 37979–38007.

Khurana

Koli

Khatter

, et al. Natural language processing: state of the art, current trends and challenges. Multimed Tool Appl 2023; 82(3): 3713–3744.

10.

Galassi

Lippi

Torroni

. Attention in natural language processing. IEEE Transact Neural Networks Learn Syst 2020; 32(10): 4291–4308.

11.

Ning

Bai

. Biomedical named entity recognition based on Glove-BLSTM-CRF model. J Comput Methods Sci Eng 2021; 21(1): 125–133.

12.

Sang

Yin

, et al. Short-term runoff prediction optimization method based on bgru-bp and blstm-bp neural networks. Water Resour Manag 2023; 37(2): 747–768.

13.

Sun

, et al. BLSTM and CNN stacking architecture for speech emotion recognition. Neural Process Lett 2021; 53(6): 4097–4115.

14.

Tan

Miao

, et al. Extraction of temporal information from social media messages using the BERT model. Earth Sci Inform 2022; 15(1): 573–584.

15.

Bhardwaj

Majumder

Poria

. Investigating gender bias in bert. Cognit Comput 2021; 13(4): 1008–1018.

16.

Jollibekova

. The role of communicative efficiency in teaching speaking. Galaxy Int Interdiscip Res J 2021; 9(11): 115–117.

17.

Tang

Abuhmaid

Olaimat

, et al. Efficiency of flipped classroom with online-based teaching under COVID-19. Interact Learn Environ 2023; 31(2): 1077–1088.

18.

Khomsah

Dias Ramadhani

Wijaya

. The accuracy comparison between Word2Vec and FastText on sentiment analysis of hotel reviews. Jurnal RESTI (Rekayasa Sistem dan Teknologi Informasi) 2022; 6(3): 352–358.

19.

Umer

Imtiaz

Ahmad

, et al. Impact of convolutional neural network and FastText embedding on text classification. Multimed Tool Appl 2023; 82(4): 5569–5585.

20.

Kang

Yang

. Performance comparison of Word2vec and fastText embedding models. J Dig Cont Soc 2020; 21(7): 1335–1343.

21.

Goulet

Philippe . The macroeconomy as a random forest. J Appl Econ 2024; 39(3): 401–421.

22.

Baby

Juliet Devaraj

Hemanth

, et al. Leukocyte classification based on feature selection using extra trees classifier: atransfer learning approach. Turk J Electr Eng Comput Sci 2021; 29(8): 2742–2757.

23.

Kohnke

Moorhouse

. Facilitating synchronous online language learning through Zoom. RELC J 2022; 53(1): 296–301.

24.

Lestari

Yudi Wahyudin

. Language learning strategies of undergraduate EFL students. J Eng Lan Teach Learn 2020; 1(1): 25–30.

25.

Kadhim

. Survey on supervised machine learning techniques for automatic text classification. Artif Intell Rev 2019; 52(1): 273–292.

26.

Dabre

Chu

Kunchukuttan

. A survey of multilingual neural machine translation. ACM Comput Surv 2020; 53(5): 1–38.

27.

Park

Seok Hong

Kim

. A methodology combining cosine similarity with classifier for text classification. Appl Artif Intell 2020; 34(5): 396–411.

28.

Liu

Chen

Peng

. Some cosine similarity measures and distance measures between q‐rung orthopair fuzzy sets. Int J Intell Syst 2019; 34(7): 1572–1587.

29.

Deepa

. Bidirectional encoder representations from transformers (BERT) language model for sentiment analysis task. Turkish J Comp Math Educ (TURCOMAT) 2021; 12(7): 1708–1721.

30.

Zhou

Liu

, et al. DNABERT: pre-trained bidirectional encoder representations from transformers model for DNA-language in genome. Bioinformatics 2021; 37(15): 2112–2120.