Construction and Effectiveness Evaluation of Virtual Reality-based Immersive Learning Environments in International Chinese Language Teaching and Learning

VR technology integrates a variety of high-tech, is the highest level of multimedia applications. This technology is based on computer technology and head-mounted display technology as the core fusion sensors, with eye-tracking, data gloves and big data and other multi-disciplinary technologies to generate a three-dimensional virtual world, through the handle or data gloves to achieve human-computer interaction, to achieve the effect of expanding the perceptual boundaries of human beings, it is a kind of computer simulation technology that can create virtual worlds and provide simulation experience with a very high degree of experience, combined with the network that can satisfy the requirements of VR applications Combined with 5G technology that can meet the network and latency requirements of VR applications, the immersive nature of virtual situations can be truly realized.

The VR system needs to realize the generation of virtual environment, the natural interaction between the experiencer and the virtual environment, the recognition of various forms of input and the generation of corresponding feedback information in real time, as well as the establishment of the environment model, the generation of virtual sound, the establishment of a database to manage the information of all objects in the entire virtual environment. Key technologies include dynamic environment modeling technology, real-time 3D graphics generation technology, real-time drawing technology, stereoscopic display and sensor technology and system integration technology. 1)

Dynamic environment modeling technology

Creating a virtual situation is the core content of VR technology, and different tools and technologies are used for dynamic modeling according to the actual environment. Relatively regular environment using CAD technology from the actual environment to obtain three-dimensional data, irregular environment needs to use non-contact visual modeling technology to obtain three-dimensional data, in the actual application, you can combine the two ways of data acquisition, you can improve the accuracy and efficiency of the actual environment of three-dimensional data acquisition, and then according to the application requirements according to the acquired data to build the corresponding Virtual environment model.

VR technology has the “3I” characteristics, including immersion, interactivity and conceptualization. Immersion, also known as sense of presence, is the performance scale of the VR system, the ideal VR should cover a variety of perceptual functions that people have, such as visual, auditory, tactile, gustatory, olfactory and physical, etc., a high degree of simulation of the learner feels as the protagonist exists in the simulation of the environment in the real reduction, so that the learner is difficult to distinguish between the real and the fake, through the use of a number of devices, immersed in the virtual situation. Interactivity refers to the learner’s real-time and operability of objects in the virtual context. If learners use their hands to contact the objects in the virtual context, they can feel the tactile sensation of contact with the objects and the weight of the objects, and the visual experience can see the objects being grasped as the hands move in the qualitative and quantitative organic combination of technology-constructed environments to realize the double understanding of rationality and sensibility, play the subjective initiative, and rely on perception and cognitive ability to obtain knowledge from the multi-dimensional information space of the virtual context in an all-round way, inspire We can realize both rational and perceptual understanding in a qualitative and quantitative organic environment, give full play to our subjectivity, rely on our perceptual and cognitive abilities to acquire knowledge from the multi-dimensional information space of the virtual context, inspire new ideas, and master the learning content at a deeper level.

The VR-based immersive Chinese learning environment is an artificial environment used to meet the immersive learning needs of learners in international Chinese language teaching, and its design is mainly guided by the immersive learning theory, the constructivist learning theory and the contextual cognitive theory.

The VR-based immersive learning environment should firstly be able to provide an authentic learning situation, and the authenticity of the situation is conducive to stimulating students’ learning motivation. Learners can carry out specific experience and independent inquiry learning according to certain learning purposes in the context, and gain specific experience through experience and inquiry activities. The learning context should provide rich interactive functions and appropriate collaborative conversation mechanisms to facilitate communication between learners and the learning environment as well as between learners, and to form abstract concepts through communication and reflection activities, so as to complete the construction of the meaning of knowledge. Assessment should be based on students’ self-evaluation, summarizing and reflecting on the learning process and learning outcomes at the end of learning. It should also provide opportunities to test the knowledge acquired by the learner in order to validate the learning outcomes.

1)

Principle of authenticity. Make full use of three-dimensional modeling and visualization technology to provide real or near-real environments and situations to produce a strong sense of immersion and stimulate learner motivation, and authenticity is conducive to improving the degree of environmental fidelity, manipulation credibility and user experience.

In order for the learning environment to fulfill its educational value, it is necessary to coordinate and plan the design elements and clarify the functions and interrelationships of each element. The design elements are shown in Figure 1.

Figure 1.

Design elements of VR-based immersive learning environment

The first step is to consider the characteristics and application needs of the learners. There is a preliminary analysis of the learners, clearly defining the object of use, such as the user’s grade, interest, specialty, learning style, familiarity with VR and ability to use it.

The second element is the application occasion. This element focuses on the application occasions and conditions of support, including whether it is applied in the classroom or outside the classroom, independent learning or group learning, whether the application form is desktop or mobile, online or offline, whether it is used in the school or at home, and the length of time it is used, etc. The analysis of the application occasions is conducive to the selection of the technical elements.

The third element is learning content and learning objectives. The immersive learning environment should be designed according to the learning content. On the basis of determining the learning content, the type and level of learning objectives should be further clarified.

The last element to be considered is the VR technology element. This includes hardware support, interaction methods and types, the learner’s observation perspective, whether to use avatars, and the way of switching scenes, etc. In addition, development tools and engines, technical feasibility, development costs and cycles should be clarified.

In the VR-based immersive Chinese learning environment, four learning scenarios are designed: “experience situation”, “reflection situation”, “concept formation situation” and “verification situation”, and each situation can provide necessary cognitive and communication auxiliary tools. Contexts are designed to enable learners to identify problems during exploration, to have the opportunity to generate problems, to formulate hypotheses for solving problems, to be supported by a variety of resources in solving authentic problems, and to provide other rich examples and similar problems for conceptualization and transfer.

1)

Design student activities

In the “Context of Experience”, students’ activities are mainly based on exploration and experience, and students can interact with virtual objects to perceive information, and can also observe objects from multiple angles. In “Reflective Situations” and “Conceptual Situations”, student activities focus on reflection, observation, discussion, sharing, dialogue, and practice. In the “Verified Situation”, students’ activities are represented by solving similar problems, testing and reinforcing what they have learned through quizzes, decision-making, trial and error, and other activities.

The design of the VR-based immersive Chinese learning environment is centered on student activities and supported by teacher activities, constituting a design framework with teacher-student interaction activities as the kernel as shown in Figure 2. Students are integrated into the learning context through proactive activities, and achieve learning goals through experience, observation, reflection, collaboration and sharing. Teachers help learners complete the learning tasks through activities such as goal guidance, explaining the context, motivating the engine, and assisting reflection. In the interaction between teacher activities and student activities, communication and cognitive aids should be provided, as well as validation and testing functions. All these modules and tools should be designed around student and teacher activities.

Figure 2.

Design framework of immersive learning environment based on VR

First, the test questions (genes) are encoded. In this paper, segmented decimal real number encoding is used to replace the traditional binary encoding to generate a number of non-repeating random number arrays, with each real number representing the question number of a test question. In order to simplify the subsequent processing of crossover and mutation, each series is divided into five small segments, representing five types of questions in a set of standardized test papers: single-choice, multiple-choice, fill-in-the-blanks, judgmental, and short-answer questions, which are used as initial populations for evolution.

Next, the size of the population is determined; too small a population and a lack of sampling points for the algorithm will limit its performance. Larger population size can effectively avoid the algorithm from falling into local optimization, but it will inevitably increase the amount of computation, resulting in slower convergence of the algorithm. Generally, the population size is set in the range of 20~100, and in this paper, the value is 60.

Selection is used to simulate the “survival of the fittest” by natural selection, whereby individuals with high fitness values have a better chance of surviving, while those with low fitness values are difficult to pass on to the next generation, and the population evolves from generation to generation. Common selection methods include roulette and optimal individual preservation strategy. Roulette method is based on the proportion of the individual fitness value in the population fitness value, the higher the fitness value of the individual is more likely to be selected, and the lower fitness value of the individual has a certain chance to be hereditary, which can better maintain the genetic diversity of the population, but with a greater degree of randomness, or even discard the optimal individual, resulting in the degradation of the population [22].

The probability that the i st individual is selected to survive in the roulette method, Equation (10): 10Pi=Fi∑i=1mFi$${P_i} = \frac{{{F_i}}}{{\sum\limits_{i = 1}^m {{F_i}} }}$$

Where F_i is the fitness value of the ind individual and ∑i=1mFi$$\sum\limits_i = {1^m}{F_i}$$ is the total fitness value of the m individuals in the sum group.

In order to guarantee that the genetic algorithm can converge, but also to avoid too fast convergence into the local optimum. In this paper, the improved selection operator combines the roulette method with the optimal individual preservation strategy, which preserves 30% of the optimal individuals with the highest fitness, does not participate in the cross mutation, and the rest of the solutions in the solution set are selected by the roulette method, and finally replaces the portion of the optimal individuals selected by the roulette method which is lower than 30% with the 30% of the optimal individuals that are preserved.

Traditional genetic algorithms use fixed parameters as crossover probability and mutation probability, and this fixed way of taking values does not take into account the gradual optimization of the population in iteration, and cannot dynamically adapt to the needs of the population in different periods of evolution. In view of such problems, some scholars have proposed an adaptive genetic algorithm that dynamically adjusts the crossover and mutation probabilities, and its basic idea is: for the inferior individuals below the average fitness of the population, higher crossover and mutation probabilities are used in order to improve the fitness of the individual. For individuals above the average fitness of the population, a smaller crossover probability and mutation probability should be used in order to maintain their superior genotype [23]. The formula of this algorithm is expressed as: 11Pc={ k1(fmax−f)fmax−favg f≥favg k2 f<favg$${P_c} = \left\{ {\begin{array}{*{20}{c}} {\frac{{{k_1}({f_{\max }} - f)}}{{{f_{\max }} - {f_{avg}}}}}&{f \ge {f_{avg}}} \\ {{k_2}}&{f < {f_{avg}}} \end{array}} \right.$$ 12Pm={ k3(fmax−f)fmax−favg f≥favg k4 f<favg$${P_m} = \left\{ {\begin{array}{*{20}{c}} {\frac{{{k_3}({f_{\max }} - f)}}{{{f_{\max }} - {f_{avg}}}}}&{f \ge {f_{avg}}} \\ {{k_4}}&{f < {f_{avg}}} \end{array}} \right.$$

where k₁ ~ k₄ is a control parameter with value in the range (0, 1]. f_max is the maximum fitness of the current population. f_avg is the average fitness of the current population. f is the fitness of the better individual in the group of crossed individuals or the fitness of the individual with mutation.

However, this adaptive genetic algorithm also has more obvious drawbacks: firstly, P_c and P_m are strongly influenced by the control parameter k, while the value of k is more random, which affects the individual quality of the population. Second, when f = f_max, P_c and P_m are both 0, i.e., the optimal individuals no longer cross over and mutate, which can lead to stagnation of population evolution.

In contrast, the logistic function converges smoothly at both ends of the interval and tends to a fixed value, which can better describe the phenomenon of bounded growth and can balance the balance between linear and nonlinear changes, and the function value of the logistic function ranges from 0 to 1 [24].

A simplified logistic function is shown in equation (13): 13y=11+ex$$y = \frac{1}{{1 + {e^x}}}$$

In this paper, this good property of logistic function is utilized to improve for the crossover and variance operator, which is embedded in the adjustment formula of crossover and variance probability by adjusting the coefficients, and the improved adaptive crossover and variance probability is shown in Eq. (14) and Eq. (15): 14Pc={ fmax−ffmax−farg×k11+ek2α+k3 f≥favg k4 f<farg$${P_c} = \left\{ {\begin{array}{*{20}{c}} {\frac{{{f_{\max }} - f}}{{{f_{\max }} - {f_{\arg }}}} \times \frac{{{k_1}}}{{1 + {e^{\frac{{{k_2}}}{\alpha }}}}} + {k_3}}&{f \ge {f_{avg}}} \\ {{k_4}}&{f < {f_{\arg }}} \end{array}} \right.$$ 15Pm={ fmax−ffmax−farg×k51+ek6α+k7 f≥favg k8 f<farg$${P_m} = \left\{ {\begin{array}{*{20}{c}} {\frac{{{f_{\max }} - f}}{{{f_{\max }} - {f_{\arg }}}} \times \frac{{{k_5}}}{{1 + {e^{\frac{{{k_6}}}{\alpha }}}}} + {k_7}}&{f \ge {f_{avg}}} \\ {{k_8}}&{f < {f_{\arg }}} \end{array}} \right.$$

where α denotes the current display coefficient of the population, defined as equation (16): 16α=EX+1DX$$\alpha = \frac{{EX + 1}}{{\sqrt {DX} }}$$

Where expectation EX reflects the average fitness of the current population as shown in equation (17). Variance DX reflects the degree of dispersion of the current population’s fitness, as shown in Equation (18): 17EX=favg=f1+f2+⋯+fmm$$EX = {f_{avg}} = \frac{{{f_1} + {f_2} + \cdots + {f_m}}}{m}$$ 18DX=f12+f22+⋯+fm2m−favg2$$DX = \frac{{f_1^2 + f_2^2 + \cdots + f_m^2}}{m} - f_{avg}^2$$

Obviously, with the increase in the number of iterations, the population gradually evolves, the average fitness of the population continues to increase, converging to the superior genotypes, the similarity between individuals is increasing, the optimal solution tends to converge, i.e., the degree of discretization decreases again and again, and the similarity coefficient α increases consequently.

In order to better assess the quality of the intelligent grouping of papers, the experiment chooses difficulty and knowledge point coverage as evaluation indexes respectively. Difficulty, i.e., the threshold value of the test paper that distinguishes the students’ mastery of the knowledge taught by the teacher, through which the overall difficulty of the test paper is judged. Its specific formula is: 19N=∑i = 1nNi⋅fiZ$$N = \frac{{\sum\limits_{i\: = \:1}^n {{N_i}} \: \cdot {f_i}}}{Z}$$

In Eq. (19), N_i and f_i denote the difficulty and score of Question i, respectively. Z represents the overall score of the paper and n represents the total number of questions in the paper. The larger the value of N, the lower the difficulty level of the paper. The smaller the value of N, the greater the degree of difficulty of the paper.

In order to better distinguish the degree of difficulty of the test paper, it is proposed to categorize the difficulty of the test paper, and the results of the specific difficulty grading are shown in Table 1.

Knowledge point coverage is the occurrence of the knowledge required to be tested in the assessment syllabus in the examination paper. The greater the occurrence of the knowledge points examined, the higher the knowledge point coverage. The specific formula is: 20K=mk$$K = \frac{m}{k}$$

In Eq. (20), K represents the knowledge point coverage. m and k indicate the chapter to which the selected knowledge point belongs and all chapters of the examination subject, respectively.

1)

Question paper question types and their percentages

In order to verify whether the proposed optimized genetic algorithm is effective, the experiment will use 150 test paper question types as experimental data. The question types mainly include five types of single-choice, multiple-choice, judgment, fill-in-the-blank and short-answer questions, and the number of each type of question is 45, 15, 45, 30 and 15, respectively, and the corresponding allocation ratio is 30%, 10%, 30%, 20% and 10%, respectively.

Figure 4.

Time comparison of genetic algorithm before and after optimization

Figure 5.

Comparison of the overall difficulty of the two test papers

Before conducting the teaching, students in the experimental group and the control group were respectively given a pre-test of Chinese language proficiency, which included four items: vocabulary, sentence modification, reading and composition. Then, SPSS software was used to conduct independent samples t-test on the test results to determine whether there was a significant difference between the two groups. The results of the pre-test for each dimension of the experimental group and the control group are shown in Table 2.

The data collected through the pre-test found that the mean scores of the two groups of students in the four dimensions of words, sentence correction, reading and composing were not much different before the study, and the corresponding Sig. values were 0.972, 0.473, 0.497, and 0.294, which were all greater than 0.05, indicating that there was almost no significant difference in Chinese language proficiency between the two groups of students before the experiment, which made the two groups suitable for carrying out the teaching experiment.

This teaching experiment was conducted between August and September 2024, during which the VR immersive learning environment-supported teaching mode and the traditional teaching mode were implemented for the two groups of students. In order to control for extraneous variables in the experiment and to ensure the validity of the results, factors such as content, course pace, and class time were kept consistent between the two groups, except for differences in teaching methods and approaches. Specifically, 50 students received instructional interventions in the VR immersive learning environment, while the other 52 students received traditional teaching methods.

In the teaching process, in order to promote students’ interactive communication and cooperative learning, this paper adopts a heterogeneous grouping method, dividing the students in the experimental group and the control group into 8 groups each, with each group containing approximately 6-7 students. When grouping students, factors such as their academic performance, interests and personalities were fully considered to ensure the diversity of students within each group. This grouping arrangement is conducive to promoting discussion and cooperation among students, improving the effect of classroom learning and interaction, and also catering to the usual habit of students learning in groups in class. In the teaching process of the control group, multimedia courseware is mainly used for teaching, and rich teaching links are designed. In the teaching process of the experimental group, the immersive teaching environment based on VR technology is fully utilized, and the same multimedia courseware is used to design rich and colorful classroom teaching sessions.

After conducting the teaching, post-tests need to be conducted for the experimental and control groups in order to assess the teaching and learning effects. By distributing the Questionnaire on Inquiry Interest, Attitude and Participation and conducting a test on the four dimensions of Chinese language proficiency, in order to get a comprehensive understanding of the actual role of the VR-based immersive learning environment in the enhancement of students’ Chinese language proficiency and interest effects. After the test, the author analyzed the collected data with a view to drawing objective and scientific conclusions to provide a basis for further optimizing the design strategy of the VR-based immersive learning environment.

1)

Analysis of differences in Chinese language proficiency between the experimental group and the control group

Using SPSS software to conduct independent samples t-test on the Chinese proficiency post-test data of the experimental group and the control group, the comparison of the results of the Chinese proficiency post-test of the two groups of students is shown in Table 3.

After careful analysis of the posttest data, it can be observed that the mean scores of the two groups of students in the four competency dimensions of vocabulary, sentence rewriting, reading, and composing after the experiment are significantly different (Sig. < 0.05), especially in the reading dimension, the Sig value is as low as 0.000, indicating that the difference between the two groups of students in Chinese reading ability after the experiment is extremely significant. Overall, there were large differences in the data of the two groups of students, and this result proves that after using the VR-based immersive learning environment, the students of the experimental group demonstrated a significant advantage in Chinese proficiency compared with the students in the traditional teaching mode.

In international Chinese language teaching, it is important to enable students to improve their Chinese language proficiency, as well as to cultivate their interest in learning Chinese, change their learning attitudes and increase their participation. In order to test the effectiveness of the teaching, the Questionnaire on Learning Interest, Learning Attitude and Participation was distributed to the students to find out the situation of each aspect. After the teaching experiment, the questionnaire was distributed to the experimental group and the control group, a total of 102 copies were distributed and 102 valid data were successfully recovered. In order to further analyze the differences between the data of the two groups, an independent sample t-test was conducted using SPSS software. The results of the significance analysis of learning interest, attitude, and participation are shown in Table 5.

By analyzing the data, it can be seen that there is a significant difference in the mean score value of both learning interest and learning attitude of the experimental group compared with the control class (Sig.<0.05). As for the participation, the average score value of the participation of the experimental group is not significantly different from the control class for the time being (Sig.=0.052>0.05), which means that the VR immersive learning environment has less influence on the students’ participation. Meanwhile, according to the statistical results of the questionnaire, as many as 96% of the students in the experimental group indicated that they were willing to continue to use the VR immersive learning environment for their learning activities, which further verified that the students’ attitudes towards learning were stronger in the immersive learning environment. In summary, the application of VR-based immersive learning environment in the international Chinese classroom effectively promotes students’ learning interests and attitudes.