Research on business English autonomous learning based on artificial intelligence and improved BP network model

Abstract

Due to various factors, the learning process of business English is mostly autonomous learning. However, the traditional autonomous learning model is difficult to effectively improve the learning effect of business English. In order to improve the business English learning model, based on artificial intelligence and improved BP network model, this paper builds a business intelligence autonomous learning system with certain intelligence. Moreover, this paper designs functional modules for the characteristics of business English learners, and combines the self-learning needs to facilitate the processing of structural functions, so that students can complete the operation independently. The system sets up multiple functional modules, conducts guided recommendation learning according to the characteristics of the self-learning process, and combines the feedback system to correct the shortcomings in students’ autonomous learning. Through this system, teachers can perform a variety of operations offline and eliminate restrictions on location and teaching time. In addition, in order to verify the performance of the model, the experimental study was conducted by setting up a control group and an experimental group. The research results show that the model constructed in this paper has good performance.

Keywords

Artificial intelligence improved algorithm BP neural network business English

1 Introduction

Language barriers have become a bottleneck for corporate managers to communicate directly with foreign businessmen and understand their real intentions. There are many unrealistic factors for business people learning English in rural areas:The first is that there is no suitable training institution. The relatively scattered distribution of enterprises in rural areas determines that training institutions cannot carry out business English training;The second is that business people’s free time and working hours are not fixed. Some business people have flexible working hours, but non-working hours are often occupied. The limitations of space and time prevent them from learning English according to their own ideas [1].

With the rapid development of network equipment across the country, some colleges and universities have gradually adopted the form of classroom-based education and gradually increased online teaching links, that is, some learning materials are uploaded to the Internet for students to download and read. The rapid development of the Internet has promoted the transformation of teaching from a single classroom to multiple online learning and has encouraged students to change from passively accepting teachers’ course content to autonomously learning corresponding knowledge. Students even need an independent learning mode with no restrictions on time, location, etc., and can contact teachers and classmates to answer questions, exchange ideas, and share information through the network communication platform. At the same time, teachers can provide guidance for students’ autonomous learning and conduct flexible testing forms such as online testing. However, at present, some colleges and universities still lack a autonomous learning system that integrates all learning resources and provides a platform for teachers and students to communicate freely. The Internet age connects everything, and people can help answer some of their knowledge points online. Everyone’s knowledge is different, and knowledge is not geographically divided. Only by putting their own knowledge online can people make progress together. The Internet in the future will definitely develop into a network that can provide a knowledge base for anyone. Each university will launch its own online teaching, put its own teaching resources online, and learn knowledge through registration application. At that time, netizens can learn the university-related courses they like at any time and any place. In short, the Internet should provide students with abundant resources and effective ways for autonomous learning and interactive learning. Therefore, in order to help students solve the above problems more conveniently, the design and implementation of the software learning autonomous learning system is proposed. It provides online learning, online communication, online testing, special seminars and other functions for teachers and students of the college to facilitate students to obtain learning resources and enhance interactive communication with teachers and classmates [2].

In recent years, multimedia technology has developed rapidly, and the country has paid more and more attention to vocational education. Many vocational English education has extensively built multimedia classrooms, and more and more classrooms have moved from traditional classrooms to multimedia classrooms. Teachers can use multimedia equipment to bring students a multi-sensory learning experience such as vision and hearing and turn the traditional classroom into a classroom based on multimedia teaching equipment. Students can receive information in all directions and have achieved good teaching results. However, this teaching mode is essentially to allow teachers to have more auxiliary tools and has not changed the knowledge model that teachers actively teach and students passively accept. For schools, although they have made some achievements in building multimedia classrooms, there is no essential difference from traditional teaching. In this mode, teachers cannot leave the classroom to impart knowledge, and students cannot leave the classroom to receive knowledge [3].

2 Related work

The key technology of autonomous learning system is collaborative filtering technology. In simple terms, collaborative filtering utilizes groups with similar interests, hobbies, experiences, etc. to integrate their resource information, and score similar resources to share learning resources [4].

The enlightenment of autonomous learning abroad can be traced back to the “Knowledge Midwife” advocated by Socrates in ancient Greece, which was intended to inspire students to search for answers to their own questions. With the continuous deepening of research in the field of psychology, autonomous learning has become a research hotspot in teaching and psychology [5]. By observing the progress of foreign autonomous learning research, the following characteristics can be found:First, there are many theoretical schools of foreign autonomous learning, and the system is mature. Different schools stand on different theoretical positions and put forward different viewpoints. In particular, in the fields of education and psychology, many scholars are very wise in their thinking. In addition, scholars also have insights in behavior, sociology, and logic. The most representative and widely concerned abroad are the humanistic thought, the social cognitive school and the constructivist school [6]. The literature [7] believedthat autonomous learning requires students to be active participants in metacognition, behavior, and motivation, and proposed a systematic six-dimensional research framework for autonomous learning, which has been widely recognized by the academic community. The second is to carry out a lot of exploration at the practical level to train students’ creativity and creativity. With the improvement of information technology, education and teaching practice has also been constantly changing and deepening with the progress of technology. In the 1980 s, countries such as the United States, Japan, and South Korea began to attach importance to cultivating students’ creativity and creativity, and education that originally used memory as the main method of knowledge acquisition was gradually eliminated [8]. From the 1990 s to the beginning of this century, countries around the world have given new missions to higher education centering on cultivating students’ independent innovation capabilities and implementing their national science and technology development strategies.” The United States has launched the “Innovation America Program”, the United Kingdom has established a “Science and Innovation Investment Program”, Canada has launched a national innovation strategy, and Russia has developed innovative university evaluation standards. Among them, the cultivation of autonomous learning ability is listed as one of the important teaching indicators” [9]. As the application of information technology in network technology education is more and more extensive, foreign countries have conducted further research and practice on the autonomous learning model under the network environment, and carried out a large number of empirical studies in combination with basic theories to develop and construct many autonomous learning application centers or platforms. Many colleges and universities offer autonomous learning courses to cultivate students’ awareness and ability of independent access to information resources and autonomous learning ability in various ways [10].

Autonomous learning has multiple research perspectives and a wide range of fields [11]. The research perspective also extends to the teacher-student relationship under the autonomous learning model [12], the internal psychological factors of autonomous learning [13], the improvement of autonomous learning strategies [14], the status of autonomous learning [15], and the evaluation research of autonomous learning [16]. The angles are diverse, and the system is becoming more and more perfect. At the practical level, with the advancement of teaching reform and the guidance of the national quality online courses, domestic universities have carried out many teaching experiments and reforms with the main goal of training students’ autonomous learning ability. Moreover, educators have proposed many effective autonomous learning methods, which have promoted the deepening of domestic theoretical research on autonomous learning. Second, online autonomous learning has become a hot topic, and new media topics have drawn attention. With the development of Internet technology, the construction of online courses and learning platforms, online autonomous learning has once become a research hotspot [17]. With the development of new media technology, the study of new media and autonomous learning of college students has triggered the thinking of scholars. One is to study the impact of new media environment on college students’ autonomous learning ability. The literature [18] analyzed the advantages and disadvantages of new media factors for autonomous learning ability of college students by analyzing the internal and external conditions of autonomous learning. The literature [19] calls for actively responding to the opportunities and challenges brought about by the new media for autonomous learning of college students, and to explore and play the important role of new media in the education of autonomous learning of college students. The second is the research on autonomous learning approach, ability training and system design of college students in the new media environment. The literature [20] analyzes the autonomous learning process by investigating the status quo of practical operations and case studies. The literature [21] established micro-groups through Weibo, and took “new media research” as a course, and introduced in detail the research ideas, course settings, teaching process, interactive response, evaluation, etc. Moreover, the literature believed that students generally accept and love teaching methods that rely on new media for teaching, which plays a certain role in promoting college students’ autonomous learning. The literature [23] talks about the construction of directed acyclic graph for video coding algorithms for motion estimation in parallel reconfigurable computing systems. The partitioning algorithm also plays a key role in optimizing the encoding of images. The literature [24] dealt with the exploitation of IoT and BigData Analytics using the Hadoop ecosystem in real-time environments. The implementation of IoT-based Smart City is accomplished through the above-mentioned processes. The article [25] centers around IoT and its noteworthy work in sophisticating the human hones and endeavors. This paper moreover overseen the combination of diverse data from distinctive resources that are related with the web. The article [26] talks approximately the different issues within the vehicular communication field with the proposition of agreeable centralized and distributed spectrum detecting model. Due to the execution of the agreeable cognitive model, obstructions and different hidden issues are minimized. The article [27] discusses the problem, such as the tremendous amount of big data, and introduces the SmartBuddy idea of a smart and intelligent world using individual activities and human resources [28, 29].

3 Research on solving linear equations based on gradient descent

The gradient descent method is also known as the steepest descent method, which is one of the classic algorithms for solving unconstrained optimization problems. The gradient descent method uses the negative gradient direction to determine the new search direction of each iteration, so that each iteration can gradually reduce the objective function to be optimized. The gradient descent method is an iterative method, and its basic formula is [22]: $a_{k + 1} = a_{k} + ρ_{k} s^{(k)}$ (1)

In this formula, s^(k) represents the negative direction of the gradient, and ρ_k represents the search step (along the gradient direction). In this formula, the negative direction s^(k) of the gradient can be obtained by differentiating the function, and the key lies in how to determine the step size, because the step size cannot be too large or too small. If the step size is set too large, the result will be divergent, and if the step size is too small, the convergence rate will be too slow, and it is difficult to obtain accurate values. Usually a reasonable step value is established by a linear search algorithm, that is, the coordinate a_k+1 of the next point is regarded as a function of ρ_k, and then the minimum value ρ_k that satisfies f (a_k+1) can be obtained.

The research on the gradient descent method has a long history. Although it has some shortcomings (such as the very harsh selection of the starting point; the convergence speed will be slower near the minimum value), as a classic algorithm, its function cannot be ignored, and many new algorithms can be obtained by modifying and improving on the basis of it. For example, Liu Yingchao and others studied the gradient descent method. Chen Haibo et al. applied the gradient descent method to the fitting problem of sediment particle size distribution. In the research on the quality diagnosis of the combined classifier proposed by Ye Rui, the gradient descent method can effectively improve the efficiency. Wang Xiaohua et al. combined the gradient descent method with the neural network model and applied it to the design of the type 4 FIR digital filter, and obtained a neural network model with superior performance [22].

The system of linear equations is set to: $AX = b$ (2)

Among them, $\begin{matrix} A = [\begin{matrix} a_{11} & a_{12} & \dots & a_{1 n} \\ a_{21} & a_{22} & \dots & a_{2 n} \\ ⋮ & ⋮ & ⋮ \\ a_{n 1} & a_{n 2} & \dots & a_{nn} \end{matrix}] \end{matrix}$ $\begin{matrix} X = {[x_{1}, x_{2}, \dots, x_{n}]}^{T} \\ b = {[b_{1}, b_{2}, \dots, b_{n}]}^{T} \end{matrix}$ (3)

Formula (2) can be rewritten as: $\sum_{j = 1}^{n} a_{kj} x_{j} = A (k, :) X = b_{k} (k = 1, 2, \dots, n)$ (4)

Among them, A (k, :) = [a_k1, a_k2, ⋯ , a_kn].

We set $y (k) = \sum_{j = 1}^{n} a_{kj} x_{j} = A (k, :) X$ (5)

If A (k, :) = [a_k1, a_k2, ⋯ , a_kn] is the input of the neural network, X = [x₁, x₂, ⋯ , x_n] ^T is the weight of the neural network training, and y (k) is the output of the neural network. Meanwhile, {A (k, :) , b_k|k = 1, 2, ⋯ , n } is the training sample of the neural network, the neural network model is shown in Fig. 1.

Fig. 1

Neural network model.

Neural network algorithm description:

(1) The neural network output is given as: $y (k) = \sum_{j = 1}^{n} a_{kj} x_{j} = A (k, :) X$ (6)

(2) The error function is given as: $e (k) = b_{k} - y (k)$ (7)

(3) The performance index is defined as: $J = \frac{1}{2} \sum_{k = 1}^{n} {| e (k) |}^{2}$ (8)

(4) The gradient descent method is used to adjust the weights: $Δ x_{j} = - η \frac{\partial J}{\partial x_{j}} = - η \frac{\partial J}{\partial e (k)} \frac{\partial e (k)}{\partial y (k)} \frac{\partial y (k)}{\partial x_{j}} = η e (k) a_{kj}$ (9)

Therefore, the weight is adjusted to: $x_{j}^{k + 1} = x_{j}^{k} + η e (k) a_{kj}$ (10)

It is written in matrix form as: $x^{k + 1} = x^{k} + η e (k) A^{T} (k, :)$ (11)

Among them, η is the learning rate, and 0 < η < 1.

As we all know, there is a close relationship between the learning rate and the convergence of the algorithm. If the learning rate is large, the algorithm cannot converge (due to oscillation). If the learning rate is small, the algorithm convergence speed is very slow. Therefore, choosing a learning rate of appropriate size is extremely important for the implementation of the algorithm. The following provides the value theorem about the learning rate η and gives a proof.

Theorem 1 When η is set to the learning rate, only if the learning rate satisfies [23]: $0 < η < \frac{2}{max_{k = 1, 2, \dots, n} ({∥ A (k, :) ∥}_{2}^{2})}$ (12)

The neural network algorithm is convergent. In the formula, ${∥ A (k, :) ∥}_{2}^{2}$ is the square of the Euclidean norm of the vector of the k-th row of the coefficient matrix A.

Proof: The Lyapunov function is: $V (k) = \frac{1}{2} e^{2} (k)$ (13)

Then there is $Δ V (k) = \frac{1}{2} e^{2} (k + 1) - \frac{1}{2} e^{2} (k)$ (14) Because $e (k + 1) = e (k) + Δ e (k) = e (k) + {(\frac{\partial e (k)}{\partial X})}^{T} Δ X$ (15) and $Δ X = - η e (k) \frac{\partial e (k)}{\partial X}$ (16) Therefore, there are: $\begin{matrix} Δ V (k) = Δ e (k) [e (k) + \frac{1}{2} Δ e (k)] = \\ {∥ \frac{\partial e (k)}{\partial X} ∥}_{2}^{2} e^{2} (k) (- η + \frac{1}{2} η^{2} {∥ \frac{\partial e (k)}{\partial X} ∥}_{2}^{2}) \end{matrix}$ (17)

The following results are obtained: $\frac{\partial e (k)}{\partial X} = \frac{\partial e (k)}{\partial y (k)} \frac{\partial y (k)}{\partial X} = - A^{T} (k, :)$ (18)

In order to ensure the convergence of the neural network algorithm, the following conditions must be met: $- η + \frac{1}{2} η^{2} {∥ \frac{\partial e (k)}{\partial X} ∥}_{2}^{2} < 0$ (19)

Because of η > 0, there is $0 < η < \frac{2}{{∥ \frac{\partial e (k)}{\partial X} ∥}_{2}^{2}} = \frac{2}{{∥ A (k, :) ∥}_{2}^{2}}$ . In order for the neural network algorithm to meet the conditions of absolute convergence, the learning rate is: $0 < η < \frac{2}{max_{k = 1, 2, \dots, n} ({∥ A (k, :) ∥}_{2}^{2})}$ , proof is completed.

According to experience, when the algorithm converges fastest, the learning rate is usually $η = \frac{1.0 \sim 1.2}{max_{k = 1, 2, \dots, n} ({∥ A (k, :) ∥}_{2}^{2})}$ .

1.1.4 Algorithm flow

The weight vector X = [x₁, x₂, ⋯ , x_n] ^T is randomly generated, the coefficient matrix A and the vector b = [b₁, b₂, ⋯ , b_n] ^T are given, and any small positive real number Tol is set. We set J = 0 and determine the learning rate $η = \frac{1.2}{max_{k = 1, 2, \dots, n} ({∥ A (k, :) ∥}_{2}^{2})}$ .

The neural network output y (k) is calculated.

The error function e (k) and performance index J are calculated.

The weight is adjusted.

Whether the performance index meets J < Tol is judged, if it meets, the training is terminated, otherwise continue to adjust the weight [24].

4 Research on solving linear equations based on conjugate gradient method

In the early 1950 s, the mathematician Hesteness and the geologist Stiefel first proposed the conjugate gradient method, which effectively solves the problem of solving the linear system AX = b (the coefficient matrix A is a symmetric positive definite matrix). The famous paper with basic significance for the conjugate gradient method was completed by the two of them in cooperation. Before long, Fletcher et al. used this algorithm in a large-scale nonlinear optimization model, and a nonlinear conjugate gradient method was born [25].

The essence of the conjugate gradient method is to combine its conjugation with the gradient descent method as the basic starting point, use the gradient at the known point to construct a set of conjugate directions, and then search along this direction until the minimum value of the objective function is obtained. Any search direction is conjugated to each other. It is a typical conjugate direction method, and its essence is to use the orthogonal system theorem and iteratively perform residual correction. The conjugate gradient method is an iterative method whose stability is related to the condition number of the coefficient matrix. Compared with the gradient descent method, the conjugate gradient method has the advantages of simple algorithm, fast convergence and iteration, and does not require storage matrix, so it is very suitable for computer programming. The conjugate gradient method has gone through nearly 60 years of research. Among them, many famous scholars have appeared, and many remarkable results have also been achieved. At present, there are still many studies on the conjugate gradient method, and many new and improved methods of the conjugate gradient method are proposed. Zheng Li studied several conjugate gradient algorithms. Duan Qiufeng and Zhang Ying both discussed the conjugate gradient method in detail. Zhou Wenjuan combined the conjugate gradient method with neural network to spectrum analysis.

The system of linear equations is set $AX = b$ (20)

Then the initial approximation X₀ is given by $p_{0} = - r_{0} = b - {Ax}_{0}$ (21)

Calculation: $α_{k} = - \frac{r_{k}^{T} p_{k}}{p_{k}^{T} {Ap}_{k}}$ (22) $x_{k + 1} = x_{k} + α_{k} p_{k}$ (23) $r_{k + 1} = {Ax}_{k + 1} - b = r_{k} + α_{k} {Ap}_{k}$ (24) $β_{k} = - \frac{r_{k + 1}^{T} {Ap}_{k}}{p_{k}^{T} {Ap}_{k}}$ (25) $p_{k + 1} = - r_{k + 1} + β_{k} p_{k}$ (26)

Among them, α_k, β_k is the gain coefficient, r_k is the residual vector, p_k is the conjugate vector of r_k, and r_k = Ax_k - b, k = 0, 1, 2, ⋯.

Precautions:

Generally speaking, when the conjugate gradient method is in a non-bad condition coefficient matrix, the order n of the coefficient matrix A is much larger than the required number of iterations.

The conjugate gradient method is especially suitable for solving linear equations of large sparse matrices. And it is not only convenient to solve real symmetric positive definite linear equations, but also has fast convergence speed. Because it is unnecessary to select any iteration parameters, it is not required to predict the upper and lower bounds of the eigenvalues of the coefficient matrix A of the system of equations in advance.

The conjugate gradient method is only suitable for the case where the coefficient matrix A is a real symmetric positive definite matrix. Therefore, when the coefficient matrix A is an asymmetric matrix, the following conversion is required: when A is a non-singular matrix, then A^TA is a symmetric positive definite matrix, and the linear equation system Ax = b is converted to A^TAx = A^Tb, which can be calculated by this method.

The condition number of the coefficient matrix A often affects the convergence rate of the linear system Ax = b.

The weight vector X = rand (n, 1) is randomly generated, the coefficient matrix A and the vector b = [b₁, b₂, ⋯ , b_n] ^T are set, $p_{0} = - r_{0} = b - {Ax}_{0} .$ (27)

The weight of the neural network is adjusted.

Whether the performance index meets J < Tol is judged, if it meets, the training is terminated, and the weight vector (that is, the solution of the required linear equations) is obtained. Otherwise, the weight is adjusted continuously.

5 System design

The Business English Online Learning Platform is a learning website for business people to learn English and communicate. It enables students to achieve zero-based self-learning under the guidance of online teachers. It has powerful functions, scientific design of courses and teaching methods, simple operation, friendly interface, and excellent visual effects, so it has very good practicality.

According to different users, the learning platform is divided into two parts: teacher administrator user and student user. One part is for the teacher administrator, and the teacher user is also the system administrator. The teacher user’s functional content includes course (including courseware) production and uploading, course management, management of various information on the network platform, publishing and reviewing assignments and exams, management of interactive modules, user account and information management, and exit functions. The second part is for student users, and student users are members of the platform. Student users can view announcements, course studies, courseware downloads, online answering questions, online exams, questions, and evaluations. The functional structure of the system is shown in the system functional structure diagram of Fig. 2.

Fig. 2

System function structure diagram.

According to the analysis of the functions that the above system should implement, the detailed functional requirements of the e-learning platform system are summarized as follows:

(1) Login function

In order to meet the needs of system security and make some valuable information and usage functions of the website open to specific (authorized) users, there must be a user login function module that can distinguish permissions. The module should have four functions: distinguish user identity, password identification and access, error message prompt, and password change.

The system login module does not accept new user registration. For new students, teachers are responsible for adding and assigning account passwords in the management background to manage, review and maintain the student accounts.

(2) Course online publishing and learning function

Teachers use this module to create courses, including recording videos and courseware, and uploading them. Students can use this model for online learning. The teacher uploads the video and courseware in the background of the teacher, and the student can open a webpage link, and the teaching video and courseware are played simultaneously. In terms of design, the teaching videos and teaching courseware of this platform can be downloaded by students to facilitate future learning. The module needs to have the function of downloading teaching videos and courseware. The online publishing and learning function of the course is the most basic and important function of this business English online learning platform.

(3) Upload function

Teacher users have the right to upload files. The teacher background management system has the function of uploading files in the format of videos, documents, compressed files and so on. Different modules of the system can call the upload function module to upload files of different types and different uses. The upload function is temporarily not available to student users. After the system matures, the system can add the functions of uploading handwritten homework pictures and uploading shared documents.

(4) Accessibility function

After the user logs in, the user can use the function. After the teacher user logs in, he can manage courses including uploading, modifying, and deleting functions, can manage interactive information, and can manage assignments, including posting assignments, modifying assignments, and posting results. At the same time, it must have a good exit function. Student users can interact after logging in, including asking and answering questions, evaluating courses and teachers, publishing articles, doing homework exercises and taking exams online, and viewing results. At the same time, there must also be a good exit system function.

(5) Announcement notification function

Teachers can post announcements or news and manage them. Teacher users write, adjust, publish, modify, and delete announcements in the teacher management background system. The announcement is displayed on the homepage of the platform and is open to everyone.

(6) Voice training room function

It is similar to a network video conference system. Students can follow the original sound, train spoken language, and teachers can explain and correct students’ pronunciation. In addition, teachers can simulate dining, phone calls, international business negotiations and other scenarios to conduct simulated conversations and business negotiations with students.

Because the platform has high data requirements and needs to perform large-scale data processing, the cloud database is selected as the system database. Database design has a very important position in this business English online learning platform. It is related to whether the functions of the platform can run correctly and whether the teacher and student users can log in normally and have corresponding permissions. Scientifically and rationally designing the data logical relationship of the database of the platform can improve the storage and reading efficiency of the platform, and ensure the accuracy, real-time, consistency, integrity, sharing, and independence of the data platform. Moreover, it contributes to better realization of platform functions. In this example, the design of the database basically meets the requirements of convenient operation, reasonable layout, meticulous logic, good maintenance and upgradeability, less redundant data, and small system occupation.

The E-R diagram clearly describes the relationship between the various entities in the system and the attributes of each entity. Through the E-R diagram, we can clearly understand the overall processing method of the entire system. The E-R diagram of this business English online learning platform is shown in Fig. 3.

Fig. 3

System E-R diagram.

The system is divided into ordinary member login and administrator teacher login. The user needs to select the role identity when logging in. The user login field is set on the left side of the home page, and the user selects the identity and then enters the account password to log in. If the account password is entered incorrectly, or there is a text box with no input, the system will have a relative prompt dialog box. If the input content matches the account password in the database, the system will transfer to the corresponding work interface. The teacher user transfers to the teacher background management platform, and the student user still stays on this interface, but has various rights of the student user.

After the user enters the account password, the system performs the following processing: (1) The system checks the validity of the characters entered and checks whether the account password entered by the user meets the requirements of the input type. First, the system checks whether the input is empty, and then checks the type of characters entered by the user. (2) The system compares the characters entered by the user with the account password information in the database. The page prompts: The login is successful, the user name cannot be empty, the password cannot be empty, and the user name or password is incorrect. Figure 4 is a flowchart of the program processing of the login module:

Fig. 4

Flow chart of program processing of login module.

Different business needs are directly displayed to the corresponding users through the business logic layer. Therefore, corresponding to different business needs, the system uses different pages to effectively display and list, so that the user presentation layer can be truly effective and convenient for the corresponding interaction with the user. When the user presentation layer needs to insert or modify data, the data flow of the three-tier architecture is shown in Fig. 5.

Fig. 5

Schematic diagram of the three-tier architecture.

The overall system module is divided as shown in Fig. 6.

Fig. 6

The overall system module diagram.

The division of teacher modules is shown in Fig. 7:

Fig. 7

Teacher module division.

The division of student modules is shown in Fig. 8.

Fig. 8

Student module division.

The division of the system learning material display module is shown in Fig. 9.

Fig. 9

System learning materials display.

Through the above analysis, the overall system structure is constructed. In order to verify the performance of the business English autonomous learning model constructed in this paper, this study designed a controlled experiment to conduct research. In the research, we set up a control class to analyze the results, and start from the aspects of vocabulary, grammar, spoken language, translation, and e-mail of independent study of business English. Before the experiment, the academic performance is analyzed. There are 50 people in the test group and the control group. The results are shown in Table 1 and Fig. 10.

Table 1

Comparison table of comprehensive scores of test group and control group before experiment

	Control group	Test group		Control group	Test group
1	84	84	26	59	58
2	82	82	27	56	57
3	82	81	28	55	57
4	81	80	29	55	56
5	80	78	30	54	54
6	80	78	31	52	53
7	79	77	32	51	51
8	78	77	33	51	51
9	78	76	34	49	50
10	77	76	35	47	48
11	77	76	36	47	47
12	76	75	37	46	47
13	75	74	38	45	46
14	73	73	39	45	46
15	73	72	40	45	45
16	73	71	41	43	45
17	72	71	42	39	45
18	70	70	43	37	44
19	69	67	44	37	43
20	68	66	45	37	43
21	67	64	46	37	42
22	65	61	47	37	42
23	65	60	48	36	41
24	62	60	49	36	40
25	60	58	50	35	36

It can be seen from Fig. 10 that the experimental group and the control group had similar results before the test, so it can be considered that the comprehensive scores before the test are the same. On this basis, a full range of scores are compared, and the English scores of the four months after the experiment are analyzed through experiments. The results are shown in Table 2 and Fig. 11.

Fig. 10

Comparison diagram of comprehensive scores of test group and control group before experiment.

Table 2

Comparison table of comprehensive scores of test group and control group after experiment

	Control group	Test group
Vocabulary	86	75
Grammar	84	80
Spoken language	80	75
Translation	88	83
mail	93	88

Fig. 11

Comparison diagram of comprehensive scores of test group and control group after experiment.

As shown in Fig. 11, after the experiment, the test group led the control group in terms of business English vocabulary, grammar, spoken language, translation, and email. It can be seen that the students in the test group of the system achieved better learning results in autonomous learning. The two groups of students are sorted separately from the first to the 50th, and the difference in scores is calculated. The experimental group score is subtracted from the control group score. The results obtained are positive, indicating that all the rankings of the experimental group are higher than the control group. The results are shown in Table 3 and Fig. 12.

Table 3

Statistical table of the difference between the comprehensive English scores of the test group and the control group

NO.	Calculation results	NO.	Calculation results	NO.	Calculation results
1	13	19	17	35	13
2	11	20	18	36	16
3	12	21	18	37	17
4	14	22	18	38	16
5	15	23	17	39	15
6	15	24	18	40	15
7	13	25	18	41	15
8	12	26	17	42	14
9	12	27	17	43	16
10	12	28	17	44	16
11	13	29	17	45	20
12	13	30	17	46	21
13	13	31	17	47	21
14	14	32	16	48	19
15	17	33	13	49	19
16	16	34	13	50	20
17	14

Fig. 12

Statistical table of the difference between the comprehensive English scores of the test group and the control group.

As shown in Fig. 12, the scores of the test group minus the scores of the control group are all positive, and the score difference is more than 10. It can be seen that the students in the test group have significantly higher business English scores than the control group, and this also confirms that the autonomous learning system applied by the students in the experimental group has better performance and is more scientific.

6 Conclusion

“Business English e-learning platform” is an independent learning system for business professionals who have built professionally for business English. In terms of function, this platform has realized online teaching: teaching in the form of lesson frequency and courseware; online practice: consolidating knowledge of residence learning; stage test: testing the user’s knowledge mastery level; online interaction: students can ask questions and can also evaluate teachers and courses;download resources; students download relevant learning materials and tools; voice training room; students can rely on other software for online video communication; teacher background management: teachers manage users, upload courses, assign homework, and answer questions online, etc. The overall function of the system is powerful, and the interface is simple, with good practicality. The application of network information technology to education has great advantages. Once the software is successfully written, it can be used multiple times and used by many people without being restricted by time and region. In addition, the successful development of the English teaching platform will have a good demonstration effect on network teaching in other disciplines.

References

Bach

N.X.

, Hai

N.D.

and Phuong

T.M.

, Personalized recommendation of stories for commenting in forum-based social media, Information Sciences 352(3) (2016), 48–60.

Costa

and Ortale

, Model-based collaborative personalized recommendation on signed social rating networks, Acm Transactions on Internet Technology 16(3) (2016), 1–21.

Dazhi

X.U.

, Research on music culture personalized recommendation based on factor decomposition machine, Personal and Ubiquitous Computing 24(2) (2020), 247–257.

Guan

, Wei

and Chen

, Deep learning based personalized recommendation with multi-view information integration, Decision Support Systems 118(3) (2019), 58–69.

Guilin

, Xuzhen

, Zhao

, et al., Personalized recommendation via suppressing excessive diffusion, Mathematical Problems in Engineering 20(7) (2017), 1–10.

Guo

, Zhu

, Xu

, et al., How to find appropriate automobile exhibition halls: Towards a personalized recommendation service for auto show, Neurocomputing 213(12) (2016), 95–101.

and Lee

, Item-network-based collaborative filtering: A personalized recommendation method based on a user’s item network, Information Processing & Management 53(5) (2017), 1171–1184.

and Tan

, Study on SINA micro-blog personalized recommendation based on semantic network, Expert Systems with Applications 42(10) (2015), 4797–4804.

Intayoad

, Becker

and Temdee

, Social context-aware recommendation for personalized online learning, Wireless Personal Communications 97(1) (2017), 1–17.

10.

Kassak

, Kompan

and Bielikova

, User preference modeling by global and individual weights for personalized recommendation, Acta Polytechnica Hungarica 12(8) (2015), 27–41.

11.

, Yuan

and Xu

, An approach to task-oriented knowledge recommendation based on multi-granularity fuzzy linguistic method, Kybernetes 44(3) (2015), 460–474.

12.

Liji

, Chai

and Chen

, Improved personalized recommendation based on user attributes clustering and score matrix filling, Computer Standards & Interfaces 57(13) (2018), 59–67.

13.

Lin

, Wang

, Hu

, et al., Multiobjective personalized recommendation algorithm using extreme point guided evolutionary computation, Complexity 20(8) (2018), 1–18.

14.

Mao

, Lu

, Li

, et al., Profiling users with tag networks in diffusion-based personalized recommendation, Journal of Information Science 42(5) (2016), 711–722.

15.

Qiu

, Han

T.Y.

, Zhong

L.X.

, et al., Redundant correlation effect on personalized recommendation, Computer Physics Communications 185(2) (2014), 489–494.

16.

Subramaniyaswamy

, Manogaran

, Logesh

, et al., An ontology-driven personalized food recommendation in IoT-based healthcare system, Journal of Supercomputing (2) (2018), 1–33.

17.

Viktoratos

, Tsadiras

and Bassiliades

, Combining community-based knowledge with association rule mining to alleviate the cold start problem in context-aware recommender systems, Expert Systems with Application 101(17) (2018), 78–90.

18.

Xia

, Li

, Cao

, et al., Personalized exercise recommendation algorithm combining learning objective and assignment feedback, Journal of Intelligent & Fuzzy Systems 35(3) (2018), 2965–2973.

19.

Xia

, Li

and Cao

, Personalized exercise recommendation algorithm combining learning objective and assignment feedback, Journal of Intelligent & Fuzzy Systems 35(4) (2018), 1–9.

20.

Yang

, Fan

, Wang

, et al., RepoLike: a multi-feature-based personalized recommendation approach for open-source repositories, Frontiers of Information Technology & Electronic Engineering 20(2) (2019), 86–101.

21.

Zhou

and Han

, Personalized recommendation via user preference matching, Information Processing & Management 55(3) (2019), 955–968.

22.

Zuo

, Gong

, Zeng

, et al., Personalized recommendation based on evolutionary multi-objective optimization [Research Frontier], Computational Intelligence Magazine IEEE 10(1) (2015), 52–62.

23.

Mukdasai

, Sutrak

, Suwannatrai

and Kaewbanjak

, Delay-dependent exponential stability criteria for certain neutral time-delay equation via model transformations approach, Dynamic Systems and Applications 28 (2019), 651–672.

24.

Luo

Hongying

, Liu

Jun

and Li

Xuebin

, Analysis about a class of second order damping dynamic model based on the Runge-Kutta algorithm, Dynamic Systems and Applications 29(2) (2020), 270–280.

25.

Paul

, Jiang

Y.C.

, Wang

J.F.

and Yang

J.F.

, Parallel reconfigurable computing-based mapping algorithm for motion estimation in advanced video coding, ACM Transactions on Embedded Computing Systems (TECS) 11(S2) (2012), 1–18.

26.

Rathore

M.M.

, Paul

, Hong

W.H.

, Seo

H.C.

, Awan

and Saeed

, Exploiting IoT and big data analytics: Defining smart digital city using real-time urban data, Sustainable Cities and Society 40 (2018), 600–610.

27.

Paul

, “Internet of Things: A primer’, R Jeyaraj Human Behavior and Emerging Technologies 1(1) (2019), 37–47.

28.

Paul

, Daniel

, Ahmad

and Rho

, Cooperative cognitive intelligence for internet of vehicles, IEEE Systems Journal 11(3) (2017), 1249–1258.

29.

Anand Paul

, Ahmad

Awais

, Mazhar Rathore

and Jabbar

Sohail

, Smartbuddy: defining human behaviors using big data analytics in social internet of things, IEEE Wireless Communications 23(5) (2016), 68–74.