A Study of Knowledge Mapping Technology-Assisted Student Learning Path Design in Group Piano Lessons in Colleges and Universities

College piano education is a product of the popularization of modern university music and art education, its main teaching mode is in the multi-piano classroom, or collective electric piano classroom, by a teacher at the same time on several students teaching, students unified practice [1]. The main teaching object for music majors, music teacher training students, pre-school education students, the course is to cultivate students’ basic skills of piano playing, improve the comprehensive ability of music [2]. In the era of science and technology, the society develops rapidly, the demand for talents is more diversified, and more emphasis is placed on the students’ professional comprehensive ability and learning efficiency [3]. The teaching efficiency and quality of traditional piano course education can not meet the demand of the society for application-oriented talents. The limitations of traditional piano education have gradually appeared, and constrained the development of piano education. For example, the teaching mode is single, the teaching quality is difficult to improve, the lack of personalized teaching, etc. [4].

Learning is a process of continuous development and growth in one’s life. For students, choosing appropriate learning paths and high-quality learning resources can greatly enhance the learning effect and quality. The university stage is the stage of cultivating students’ professional knowledge and practical ability [5]. In the teaching of collective courses, there exists a lack of effective development of students’ professional knowledge and practice, which is due to the lack of students’ initiative on the one hand, and the mismatch of teaching resources and teaching modes on the other. Therefore, the prerequisite for improving student learning is the need to choose an appropriate learning path.

Knowledge graph technology is a hot topic in the field of artificial intelligence, this technology is able to structure, organize and integrate large-scale data, and it is one of the most ideal technical means for intelligent management and utilization of big data [6]. It mainly uses structuring and semanticization to better describe and manage the relationship between entities, and is able to eliminate the heterogeneity, duplicity and incompleteness of data sources, thus making the data more accurate, reliable and efficient for value mining [7-10]. Therefore, the research on the application of knowledge graph technology in the education industry is getting more and more attention.

Knowledge graph technology is widely used in the field of re-education. Literature [11] used knowledge graph to apply curriculum design, personalized learning path planning, innovative teaching models, and multidimensional teaching evaluation to the teaching of higher art and design by collecting multidimensional teaching resources. Literature [12] utilized knowledge graph to construct a teaching evaluation system in order to implement tailored teaching and precise teaching. Literature [13] explored the path recommendation for personalized learning by semantically integrating the teaching data in online learning through knowledge graph to accurately match the learning objectives. Literature [14] developed an intelligent Q&A system using knowledge graph, which accurately answered questions on the one hand, and helped students consolidate their knowledge focus on the other. Literature [15] constructed a knowledge graph for a database course, integrated knowledge fragments, systematized and intelligentized teaching resources, and improved teaching quality. Literature [16] developed a knowledge graph-supported management system for automatic course assignment, which improved the efficiency and flexibility of school course allocation. Literature [17] used knowledge graph to create a new path for teachers to prepare lessons by constructing the semantics of learning materials and recommending teaching resources to teachers, semi-automatically creating teaching materials, and improving the efficiency of lesson preparation.

Overall, knowledge graph technology can help students achieve personalized learning path planning, intelligent quizzes to review knowledge points, and assist teachers in student learning evaluation, textbook creation, course assignment, and course design. Among them, students as the core of teaching, their learning path should not be limited to the teacher’s teaching, especially similar to the piano this kind of art courses and group courses, student learning there may be a teacher to more than one student there is a deviation in the degree of attention, resulting in the decline in the academic performance of some students, or students do not understand the key points of the knowledge, but the lack of understanding of the pathway. For this reason, there is a need for an effective pathway that can assist students in their learning both in and out of the classroom.

In this paper, a typical multimodal piano knowledge graph is constructed, and the construction process of a knowledge graph is sorted out, including knowledge acquisition, knowledge fusion, knowledge retrieval and inference, and visualization display. And a method for designing student learning paths based on knowledge graphs is proposed. The learner’s cognitive level and other characteristics are mapped to the piano knowledge map as the starting point of path planning, and the learning goal is taken as the end point of path planning, and the optimal knowledge point learning path is obtained through path planning. The recommended utility of the knowledge point learning path planning model is verified by comparing the order degree of knowledge points in the path and the learning efficiency.

2

Method

Piano group teaching refers to the way teachers teach music theory, improvisation and accompaniment, sight-reading and ear-training, and suggested compositions of piano to students through group class teaching, thus promoting the improvement of students’ music literacy and ability in all aspects.

Knowledge mapping is an important tool in the information age to demonstrate the evolution of knowledge and its internal structural relationships. Through a series of well-designed graphics and advanced visualization technology, it vividly depicts knowledge resources and their carriers, and deeply explores, carefully analyzes, skillfully constructs, accurately draws, and intuitively displays the knowledge entities and the intricate and interdependent connections between them. Knowledge mapping not only promotes the effective organization and presentation of knowledge, but also provides a strong support for people to understand the development of knowledge and grasp the cutting-edge dynamics of the discipline.

In the field of piano education, the potential of knowledge mapping is particularly prominent, especially when it comes to the great challenges faced by teachers in the information age, and it is expected to become an important force in promoting the innovation of the teaching mode of piano group lessons and realizing accurate and personalized teaching.

2.1

Construction of the Piano Knowledge Graph

In this paper, we design to build a typical multimodal piano knowledge graph, processing information covering auditory, visual, and text, piano knowledge from audio, sheet music, metadata, and other kinds of data sources, and fusion knowledge types including external description information and piano content information. It can support knowledge retrieval and discovery based on piano content, as well as a visualization display combining multiple media forms.

Piano knowledge acquisition is the first step in the construction of the atlas, which requires the acquisition of piano knowledge, the basic building element of the piano knowledge system. After clarifying the scope of the atlas knowledge, the data sources are selected and collected, and then the piano knowledge embedded in various data sources is extracted, so that the piano knowledge covering the topic can be obtained at last.

Piano knowledge fusion is the second step of atlas construction, which includes three processes: ontology construction, entity alignment, and entity linking. Ontology construction addresses the unification of the conceptual layer of the model, while entity alignment addresses the unification of the instance layer. Together, they can complete the integration of piano knowledge from different sources. Entity linking links entities in the knowledge graph with external data sources to achieve a wider scope of knowledge integration.

Piano knowledge retrieval and reasoning is the third step in the construction of the graph, which mainly solves the problems of piano knowledge discovery and self-generation in the application phase of the knowledge graph.

The final step of the construction process is the visualization of the piano knowledge graph, which unites piano information from different sources and the results of piano content analysis into a visual form for presentation.

2.1.1

Piano knowledge acquisition

Knowledge acquisition is the process of acquiring and extracting the required knowledge from multiple data sources. The basic task of piano knowledge acquisition is to acquire knowledge about the piano domain and build a robust, complete, and effective piano knowledge map to meet the knowledge needs of the piano domain. When constructing a piano knowledge system, the main data sources for knowledge acquisition include piano literature, piano works of various representation types (audio, video, sheet music), and piano information recorded in databases, knowledge bases, and web pages.

2.1.2

Integration of piano knowledge

Knowledge fusion is a high-level knowledge organization that enables knowledge from different knowledge sources to achieve heterogeneous data integration under the same framework specification, and the steps to achieve this include ontology construction, entity alignment, entity linking, and ultimately the fusion of data, information, methods, experiences, and ideas to form a high-quality knowledge graph.

Ontology construction is a key step in piano knowledge integration, which requires the completion of abstract modeling and a structured definition of piano domain knowledge. At present, the knowledge information involved in piano knowledge mapping can be divided into three main categories: descriptive information of piano resources or works, piano event information, and piano content recording and analysis information. The first two types of information are commonly found in textual knowledge graphs, while the third type of information needs to be obtained after analyzing piano content data, which is an important type of information for multimodal knowledge graphs and a necessary information support for semantic-based piano analysis systems. Piano Knowledge Graph should choose to reuse the ontology or extend the definition of a new ontology according to the range of information it contains.

Entity alignment is the process of corresponding entities from different data sources to the same entity to which they jointly refer. An important task of knowledge fusion in the piano domain is to complete entity alignment of core entities such as pianists, piano works, and musical instruments.

Ontology construction and entity alignment complete the internal knowledge fusion of the knowledge graph, while entity linking is the process of linking disambiguating entities to external authoritative knowledge bases to realize the knowledge fusion between the knowledge graph and external data sources.

2.1.3

Piano Knowledge Retrieval and Reasoning

In the Piano Knowledge Graph, piano knowledge retrieval can be directly realized by constructing query statements using SPARQL language, or it can also be realized through natural language forms and example-based knowledge retrieval. Among them, relevance and similarity retrieval based on piano examples is a unique knowledge discovery method in the piano domain, which belongs to the retrieval based on piano content. This type of retrieval requires the use of knowledge graphs to perform deep semantic processing on piano content data. For audio data, it is necessary to use audio feature extraction technology to obtain content feature data, and then organize and store it according to the audio analysis class ontology. For the encoding of score data, it is necessary to perform RDF transformation or semantic annotation according to the score-related ontology. Through the deep integration of piano knowledge of different representation types in the construction process, the retrieval and discovery of examples and target entities across resource types can be realized. For example, the characterization of audio examples leads to the discovery of target entities with the same or similar feature values, which can be audio or sheet music. Piano knowledge reasoning involves inferring new relationships between entities or new properties of entities from existing relationships with piano entities.

2.1.4

Piano Knowledge Graph Visualization

Visualization studies of knowledge graphs include the visual representation of different types of information in the graph, as well as the analysis of visual data to infer new relationships and discover potential patterns or problems. [18] For the visual representation of piano metadata-type information in the piano knowledge graph, node-link diagrams can be used to achieve a direct visual presentation of entities and inter-entity relationships. For the visualization of piano content class information in the atlas, because it involves entities (e.g., notes, pitches, tones) that are more abstract, direct visual representation is not practical. Therefore, the knowledge graph requires customized visual analysis tools based on the type of information. In order to flexibly meet the user’s demand for piano content analysis, Knowledge Graph adopts query-based visual analysis technology, which drives visualization generation based on the type of information and internal structure of the query results, and combines information filtering technology to optimize the visual representation in an interactive form.

2.2

Knowledge mapping technology to assist student learning path design

A learning path is a collection of sequences of learning resources that a learner selects or is selected in the learning process. The process of learning path recommendation is to map the learner’s cognitive level and other features into the knowledge graph as the starting point of path planning, take the learning goal as the end point of path planning, and generate the set of all knowledge points to be sequenced by combining the relationship between knowledge points in the knowledge graph. Finally, the optimal learning path for knowledge points is obtained through path planning [19]. The learning path recommendation model based on knowledge graph is shown in Figure 1.

The specific contents are as follows: 1)

Learner model construction stage. Extract the basic information and learning behavior data of learners in the system and construct a three-dimensional learner model based on the cognitive level of learners.

2)

Generate a collection of knowledge points to be learned. Based on the basic information of learners, we can obtain the core work area knowledge and skill requirements for their specialties, as well as the professional talent training knowledge and skill objectives. Based on the learner’s goals/needs and cognitive level, get the collection of knowledge and skill points to be sorted.

3)

Knowledge point path planning stage. The knowledge learning path planning problem can be regarded as a directed graph traversal problem, the constraints are the rationality of the order of the knowledge points on the learning path, and the optimal path can be found to enable the learner to complete the learning of all the knowledge points on the learning path.

Learner models can be constructed in various ways, which aim to truly reflect learner characteristics, such as knowledge and skill objectives, preferences and cognitive level, and convert them into data that can be understood by computers, which can be used as a basis to serve the practice of upper-level applications. In this paper, we construct a learner model from the basic attributes of learners, learning styles and cognitive levels.

Define learner S, and express the learner model as: (1) $S = (B (S), P (S), C (S))$

Where B(S) represents the basic attributes of the learner, the basic attributes are descriptions of some basic properties such as the learner’s name, grade, etc. Considering the group tendency that the learner carries in the learning process, the learners at this stage are analyzed in common, which is used to form the framework of the basic learning path. Refine the big data of learners’ learning process and analyze the learning style characteristics such as learners’ learning style preference and cognitive characteristics.

According to the learning style theory, learning styles can be categorized into four types: concrete-experiential, reflective-observational, abstract-conceptual, and active-experimental, and the same individual may have a tendency to have multiple learning styles. The learning style of learner S is expressed as: (2) $P (S) = {C E, R O, A C, A E}$

CE, RO, AC, AE indicates the degree of tendency of learner S to belong to the four learning styles respectively. Extracting the big data of learners’ learning results to accurately analyze the learners’ mastery of basic knowledge, which is used to form personalized learning resource recommendations. Use C(S) = {c_i|i ∈ [1, N]} to quantitatively represent learner S ‘s cognitive level in N knowledge and skill points in the subject area, the value of which is calculated during the learning process based on learner pre-tests, learning behavior data, and other aspects, and the use of continuous values from 0 to 1 to distinguish between different levels of cognition.

2.2.1

Generation of knowledge point paths

The goal of this phase is to sequentialize the knowledge points to be sequenced to ensure that learners master all the a priori knowledge points of a knowledge point before learning it, and to arrange the knowledge points according to the principle of ascending to descending importance so as to minimize the overall learning cost. 1)

Knowledge point learning path planning model

The objective function of this problem is defined as the knowledge point path connectivity (KPC): (3) $K P C = \sum_{i \neq j} x_{i j} d_{i j}$

where X denotes the knowledge point learning path selection matrix and is set: (4) $x_{i j} = {\begin{matrix} 1 & k_{i} to k_{j} is on the optimal path \\ 0 & others \end{matrix}$

D_K denotes the knowledge point relationship matrix and matrix element d_ij denotes the degree of connectivity between knowledge points k_i and k_j and its value is defined as: (5) $d_{i j} = \frac{D e g r e e | k_{j} |}{S h o r t e s t P a t h (k_{i}, k_{j})}$

The shortest path distance from knowledge point k_i to k_j is Shortest Path (k_i, k_j), the shorter the distance, the closer the connection between the two knowledge points. If two knowledge points belong to the same knowledge area, it is considered that there is no order between the two knowledge points and their shortest path distance is set to 1.

If there is no path between two knowledge points or two knowledge points are synonymous, their shortest path is assigned to the penalty parameter [20]. The formula is set: (6) $\begin{array}{l} ShortestPath (k_{i}, k_{j}) \\ = {\begin{matrix} ShortestPath (k_{i}, k_{j}) \\ 1 (if k_{i} \in K S_{n} and k_{j} \in K S_{n}) \\ A (Penalty argument, if k_{i} = k_{j} or there is no path between k_{i} and k_{j}) \end{matrix} \end{array}$

The degree of knowledge point Degree|k_i| = inf f|k_i| + out|k_i|, where in|k_i| indicates the in-degree of knowledge point k_i and out|k_i| indicates the out-degree of knowledge point k_i, the higher the degree indicates the more important the knowledge point is in the graph. 2)

Knowledge point learning path planning model solution

The steps for solving the knowledge learning path are as follows: first, according to the cognitive level in the learner model, mark the set of mastered knowledge points in the piano knowledge map as the starting point of the path. Then the target knowledge points are taken as the endpoints of the path, and the knowledge points that have connected paths between the start point and the end point set are added into the solution space. Finally the algorithm searches for the optimal path in the solution space.

In this paper, the particle swarm optimization (PSO) algorithm is used to solve the optimal path, and each particle of the particle swarm represents a n-dimensional learning path. The flow of the algorithm is as follows: (1)

Initialize the velocity and position of each particle in the population.

(2)

Take the objective function of the knowledge point learning path planning problem as the fitness function, and calculate the fitness value fitness (the length of each path) according to the current position. Record the best position p_best and fitness of each particle and select the best particle position of fitness from the population as the position of the population g_besto

(3)

Update the velocity and position of the particles.

(4)

Calculate the fitness of each particle after the position update and update the individual extreme position p_best.

(5)

Find the global extreme position g_best by comparing the fitness of each particle with the fitness corresponding to the previous g_best.

(6)

If the number of iterations is reached or degradation behavior occurs, stop the iteration otherwise return to (3).

(7)

Output the current optimal solution as a knowledge learning sequence path_k = {k_p1, k_p2, ⋯, k_pm}.

3

Results and discussion

3.1

Experimental validation and analysis

This chapter verifies the performance of the knowledge point learning path planning model through model solving experiments, experiments comparing the order degree of knowledge points in the path, and experiments comparing learning efficiency. 1)

In order to verify the performance of the knowledge point learning path planning model constructed in the previous paper, learners from the MOOC platform of Chinese universities are randomly selected, and the piano courses in their history logs are added into the solution space, and the particle swarm optimization algorithm is used to solve the optimal path. In order to make the model can better serve the research purpose of this paper - learning path planning, the parameters commonly used in the particle swarm optimization algorithm are referred to for solving, and the number of iterations is shown in Figure 2. It can be seen that the optimal solution of the problem is found by this model after about 110 iterations.

2)

Path knowledge point order degree is the probability that two neighboring knowledge points in a recommended learning path have a successive relationship. Assuming that a recommendation model recommends learning path $α = {a_{1}, a_{2}, a_{3}, a_{4}, \dots, a_{s u m - 1}, a_{s u m}}$ for the piano course learnt by learner S, then the formula for calculating the knowledge point order degree of the path is as follows: (7) $Path knowledge point order degree = \frac{\sum_{i = 1}^{n - 1} f (a_{i}, a_{i + 1})}{n - 1}$

Where f(a_i, a_i+1) indicates whether the two neighboring knowledge points a_i, a_i+1 in the path have a sequential relationship, and if they do, then 3 and vice versa. If there is, then f(a_i, a_i+1) = 1, and vice versa f(a_i, a_i+1) = 0. The greater the degree of order of the knowledge points, the more in line with the daily learning habits of the learners.

This section compares the order degree of knowledge points in the learning path recommended by the Trans-CF algorithm model (Model 1) and the Ontology-CF algorithm model (Model 2). The maximum length (max_pre) of the recommended path of the current knowledge point in the learning path is selected as an important parameter that affects the performance of this model.Since the Trans-CF algorithm model and the Ontology-CF algorithm model do not take into account the relationship between the sequential order of the knowledge points, the value of max_pre will not affect their recommendation results.

When max_pre is 2, 4, 6, 8, and 10, we calculate the knowledge point order degree of the recommended paths of this paper’s model and compare it with the knowledge point order degree of the recommended paths of the Trans-CF algorithm and the Ontology-CF algorithm, respectively. The results of the knowledge point order degree comparison experiments are shown in Fig. 3.The order degree of the learning path recommended by Trans-CF algorithm and Ontology-CF algorithm is 0.0148 and 0.0117 respectively, which is almost zero. The learning path order degree recommended by the model in this paper is significantly higher than that of Trans-CF algorithm and Ontology-CF algorithm regardless of the value of max_pre, which can better enhance the learning experience, user satisfaction and the interpretability of the learning path.

3)

In this paper, we measure the learning efficiency by comparing the improvement of the learner’s mastery of knowledge points in the same learning time through the paths recommended by each model. For learner S, let the learner’s own knowledge point learning path be: (8) $α = {a_{1}, a_{2}, a_{3}, a_{4}, \dots, a_{s u m - 2}, a_{s u m - 1}, a_{s u m}}$

However after the learning of the first k% knowledge points in this sequence, according to the knowledge point learning path planning model constructed in this paper, the learning path strain is: (9) $β = {a_{1}, a_{2}, a_{3}, a_{4}, \dots, a_{s u m - 2}^{'}, a_{s u m - 1}^{'}, a_{s u m}^{'}}$

According to the Trans-CF recommendation model, the learning path strain is: (10) $χ = {a_{1}, a_{2}, a_{3}, a_{4}, \dots, a_{s u m - 2}^{n}, a_{s u m - 1}^{n}, a_{s u m}^{n}}$

According to the Ontology-CF recommendation model, the learning path strain is: (11) $δ = {a_{1}, a_{2}, a_{3}, a_{4}, \dots, a_{s u m - 2}^{m}, a_{s u m - 1}^{m}, a_{s u m}^{m m}}$

and the learning duration of each sequence is the same. Then the model of this paper is used to calculate the mean value of the learners’ mastery of all the knowledge points of the piano course after learning through Sequence α, β, χ, δ. By comparing the mean values of the mastery of the knowledge points of the course, we can measure whether the learning paths given by each recommendation model have improved the learning efficiency of the learners.

After the learners’ learning records for all the knowledge points of the piano course were sorted in chronological order, K was set as the learners’ current learning progress, i.e., the proportion of the learning records for the knowledge points currently completed. The comparison experiment is conducted when K is equal to 50, 60, 70, 80 and 90 respectively. Indicates that when the learner completes 50%, 60%, 70%, 80%, 90% of all the learning records of the piano course, statistically the mean value of the degree of mastery of the knowledge points of the piano course without follow-up learning, as well as after learning according to the paths recommended by this paper’s model, the Trans-CF model, the Ontology-CF model, and according to the learners’ own paths during the same follow-up learning time, the The mean value of the mastery level of all knowledge points of the piano course obtained.

The learning efficiency comparison experiment is shown in Figure 4. It can be seen that this paper’s model, the Trans-CF model and the Ontology-CF model are all better than the learner’s original learning path to the degree of mastery of knowledge, indicating that the above three learning path recommendation algorithms can improve the learning efficiency of the learner. Among them, the model in this paper has the greatest improvement in the mastery of knowledge points, up to 0.352. The Trans-CF model and Ontology-CF model are essentially identical. This shows that the model presented in this paper can maximize the help of learners in mastering more knowledge points and improving learning efficiency simultaneously.

3.2

Analysis of learning path planning results

In order to verify the reasonableness of this paper’s model for students’ learning path recommendation in college piano group lessons, 50 students with solid basic knowledge of piano major and clear mastery of knowledge structure in School A’s School of Music were selected to conduct a questionnaire survey on the effect of model use. It mainly evaluates the three aspects of target knowledge point selection, path recommendation, and improvement of learning efficiency. A total of four evaluation levels (very compliant, more compliant, less compliant, and very non-compliant) were designed, and five questions including learning goals are reasonable (A), meeting learning needs (B), improving learning efficiency (C), promoting learning transfer (D), and improving learning performance (E).

The survey results are shown in Figure 5. The survey results show that more than 92% of the learners think that the target knowledge point selection is reasonable, indicating that it is feasible to use the model of this paper to select the target knowledge points.90% of the learners think that the learning path of the recommended path can satisfy the learning needs.88% of the learners think that the learning efficiency improves when they learn according to the learning paths recommended by the model, which indicates that the system has some social value.86% of the learners think that the learning paths recommended by the model can promote the transfer of learning knowledge points, indicating that the model can basically promote the transfer of learning. Only 10% of the learners did not improve their learning performance. Overall, the practicality and feasibility of the application of knowledge graph-based learning path recommendations are high.

4

Conclusion

For the personalized teaching of piano group lessons in colleges and universities, this paper constructs a piano knowledge graph for student learning path recommendation. 1)

The order of the learning paths recommended by the Trans-CF model and the Ontology-CF model is almost zero, and the knowledge points are cluttered and unstructured, which is not conducive to learners’ systematic learning. Compared with the Trans-CF model and Ontology-CF model, the learning path planning model constructed in this paper is more in line with the learning habits of the learners and can improve the learning efficiency, up to 0.352, which shows the feasibility of this paper’s method.

2)

After applying the model of this paper to recommend students’ learning paths in piano group lessons, more than 86% of the students agree with the three aspects of target knowledge point selection, path recommendation and learning efficiency improvement, which has a more ideal practical application effect.

Lingua:: Inglese

Frequenza di pubblicazione:: 1 volte all'anno
Argomenti della rivista:: Scienze biologiche, Scienze della vita, altro, Matematica, Matematica applicata, Matematica generale, Fisica, Fisica, altro

Feed RSS della rivista

A Study of Knowledge Mapping Technology-Assisted Student Learning Path Design in Group Piano Lessons in Colleges and Universities

Ya Li

Jie Guo

Pubblicato online: 24 mar 2025

Ricevuto: 12 ott 2024

Accettato: 06 feb 2025

DOI: https://doi.org/10.2478/amns-2025-0795

Parole chiaveKnowledge Graph, Learning Path Recommendation, Particle Swarm Optimization, Piano Group Lessons

© 2025 Ya Li et al., published by Sciendo

This work is licensed under the Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License.

Parole chiave
Knowledge Graph, Learning Path Recommendation, Particle Swarm Optimization, Piano Group Lessons