Application of Knowledge Chain in the Construction of Online Tourism Education Resources

— Online tourism education occupies an important position in tourism education. Building a knowledge chain can clearly demonstrate the logical relationship between knowledge points of the entire discipline of tourism education, and enable the students to better grasp and understand the relationship between relevant knowledge points through online tourism education. This paper explores the application of knowledge chain in the construction of online tourism education resources (OTERs). Specifically, the associations between knowledge units in OTERs were modeled, the associations between knowledge units were computed, the structure of knowledge chain was optimized, and the OTER knowledge chain was analyzed in details. Through experiments, an example knowledge chain was constructed, and the students’ satisfaction with the application of OTER knowledge chain was evaluated thoroughly.


Introduction
In recent years, tourism has been developing rapidly, and occupied an important position in the tertiary industry, making significant contributions to politics, economy, culture, and ecology [1][2][3][4][5]. The demand for high-quality tourism talents continues to grow, owing to the structural adjustment and upgrading of tourism, and the integration between culture and tourism [6][7][8][9][10][11][12][13][14][15]. Online tourism education plays a pivotal role in tourism education. With the surging amount of data on the Internet, online tourism education resources (OTERs) become increasingly diverse and colorful [16]. Building a knowledge chain can clearly demonstrate the logical relationship between knowledge points of the entire discipline, and enable the students to better grasp and understand the relationship between relevant knowledge points through online tourism education.
With the aim to develop a teaching model based on mobile technology and elearning environment, Vuojärvi et al. [17] introduced the theoretical bases for the model design, and tried to build a work-based teaching model for mobile learning in tourism education. Perez-Valle and Sagasti [18] summarized the previous research 2 Construction of OTER knowledge chain

Knowledge unit association modeling
One knowledge unit in OTERs may depend on, fall on the same level, or be superior/inferior to the other knowledge unit. This association can be represented by a vector space model based on knowledge units. Table 1 illustrates the structure of knowledge chain. Let P(Li, Lj) be the association between two knowledge units Li and Lj. Then, the associations Ep between knowledge units can be expressed as: (1)  17, No. 07, 2022 By the improved term frequency-inverse document frequency (IF-IDF) algorithm, this paper determines the weights of word vectors of knowledge units, and those of knowledge chain, in OTER text set. The attributes of OTERs are illustrated in Table  2. The improved algorithm can select OTER text features based on the thesaurus of synonyms. Let Ql, WFl, and IFFl be the feature weight, number of occurrences in OTER text set E, and IDF of word vector l, respectively; SFl, and SFFl be the number of occurrences in OTER text set E, and IDF of synonyms of word vector l, respectively; M be the number of texts in E; ml be the number of texts containing word vector l. Then, we have: To facilitate the similarity computing between knowledge units and our knowledge chain, this paper vectorizes the OTERs corresponding to each knowledge unit and the knowledge chain. Let N1 and N2 be the knowledge unit information and knowledge chain information to be matched, respectively; Lmi and Qmi be the word vector and feature weight of word vector i in N1, respectively; Lnj and Qmj be the word vector and feature weight of word vector of N2, respectively. Then, we have: After N1 and N2 are extracted, the similarity Sim(N1, N2) between them can be solved by cosine similarity: Based on the results of formula (6), the M knowledge units with relatively high association weights are selected as the OTER knowledge units for association generation. The similarity calculation flow is given in Figure 1.

Calculation of associations between OTER knowledge units
The associations between OTER knowledge units are calculated based on path and depth. The similarity between two knowledge units depends on the commonality and individuality of these two units. Let log(COM(X, Y)) be the volume of commonality information between X and Y; log(IND(X, Y)) be the denominator representing the information volume required by X and Y. From the angle of information theory, the similarity between any two knowledge units can be calculated by: The association between two knowledge units can be obtained by: (8) Let COM(X, Y) and DIF(X, Y) be the commonality and difference between knowledge units X and Y in the knowledge chain, respectively; RD be the root node of the knowledge chain; CN be the common node between X and Y; RX and RY be the distance from X and Y to the nearest common node, respectively; SJ be the depth from RD to the nearest common node; R(A, B)=RX+RY be the shortest path between X and Y; β be the depth adjustment parameter; SJ(CN(X, Y)) be the depth of the nearest common node CN of X and Y. In the OTER knowledge chain, the commonality and difference between any two knowledge units X and Y can be respectively calculated by: The association weight between any two knowledge units X and Y can be given by: Let NSN be the direct secondary nodes of the nearest main public node of two knowledge units; BS be the branch spacing of two knowledge units in the nearest main public node. Then, the spacing between two knowledge units in the branch layer can be defined as a path adjustment parameter α: (12) Let Q(BSX, BSY) be the association weight between knowledge units X and Y. Then, the association weight set Eq between knowledge units can be established as: (13) The next is to compute the associations between hierarchical knowledge chains. The formation flow of such a chain is given in Figure 2. Suppose knowledge chain LO1 contains m knowledge units {BS11, BS12, ..., BS1m}, and knowledge chain LO2 contains n 个 knowledge units {BS11, BS12, ..., BS1m}. The associations between knowledge units of the two chains can be calculated by: (14) The proportionality coefficient, association weight relationship, and association weight between knowledge units of LO1 and LO2 can be respectively calculated by: , ,

Structural optimization of OTER knowledge chain
In each OTER knowledge chain, every knowledge unit is associated to a number of OTERs, which have association weights between them. The association weight between OTERs P1 and P2 can be calculated by: (18) Similarly, the association weights between P1 and P4, P3 and P2, as well as P4 and P2 can be obtained as AWP1, P4, AWP2, P3, and AWP2, P4, respectively. Let i(i=0, 1. ..., I) and (j=0, 1, ..., J) be the resources associated with knowledge chains LO1 and LO2, respectively. Then, the association weight of the OTERs between LO1 and LO2 can be calculated by: (19) Let Q(X, Y) be the association weight between knowledge units X and Y; Ql(X, Y) be the associations between the OTERs correlated with knowledge units X and Y If the association structure between the knowledge points in two knowledge chains is correct, the positive correlation between Q(X, Y) and Ql(X, Y) can be expressed as: (20) Based on the positive correlation Ω of association distribution, Q(X, Y) and Ql(X, Y) can be analyzed by: If threshold λ is greater than β, then the associated knowledge units X and Y are reasonable; if threshold λ is smaller than β, then at least one of the associated knowledge units X and Y is reasonable. Figure 4 shows the structural optimization of knowledge chain.

Analysis of OTER knowledge chain
In each OTER knowledge chain, the learners realize knowledge activities by learning OTERs. This paper visualizes the dependence between learners, knowledge activities, and learning resources. Let L_ACH be a knowledge activity; CAR be the combination of learner-activity-resource; γx and γ be the dependences of learners' knowledge activities on knowledge unit X, and on knowledge unit Y, respectively; L_ACH(lX, lY) be the degree of influence of the variation in knowledge activity lX in L_ACH. Then, the dependence I_ACHI(X, Y) between knowledge units X and Y can be calculated based on L_ACH and CAR: From OTER P_ACH and learner-activity-resource combination CAR, it is possible to derive the dependence I_ACH2(X, Y) between knowledge units X and Y: Let αX and αY be the dependences of knowledge units X and Y on OTERs, respectively; P-ACH(pX, pY) be the degree of influence of resource pY over resource pX in P_ACH.
From knowledge activity L_ACH, OTER P_ACH, and learner-activity-resource combination CAR, the dependence between knowledge units X and Y can be calculated by: Then, the knowledge units of each OTER knowledge chain are classified by a clustering algorithm. Let θ1 and θ2 be two weights satisfying θ1+θ2=1; I_ACH(X, Y) be the dependence between knowledge units X and Y; l be the class of the selected OTER knowledge unit; υl and μl be the first and last knowledge units in class l, respectively; MKE be the total number of knowledge units; Σ υl X=υlI_ ACH(X, υl) and Σ υl Y=υlI_ ACH(υl, Y) be the sum of nonzero influencing factors in the row and column of the newly added knowledge unit υl in a class, respectively; Σ MKE X=1I_ ACH(X, υl) and Σ MKE j=1I_ACH(υl, Y) be the total influence of the selected OTER knowledge unit over all the other knowledge units in the knowledge chain, and the total influence of all the other knowledge units in the knowledge chain over the selected OTER knowledge unit, respectively. Then, the additional ratio of intra-class dependence to betweenclass dependence can be calculated by: Experiments and results analysis Figure 5 exemplifies the construction of knowledge system-knowledge chain for online tourism education. The example mainly involves OTERs like hotel management, tourism management, cooking skills and nutrition, western food skills, exhibition planning and management, catering management, leisure service and management, wine marketing and management, as well as Chinese and western pastry skills; travel English, travel Japanese, applied Korean, as well as applied Spanish; financial management, accounting, e-commerce, media marketing, as well as flight attendance. The effectiveness and scientific nature of the knowledge chain were empirically analyzed based on the core courses of information analysis in library and information discipline. The experimental results were demonstrated and analyzed with examples. According to the sequence of courses, the relevant OTERs need to be adjusted according to the contents of the earlier courses, making the education more in-depth. If a student notices any interesting knowledge unit or educational resource, he/she could quickly find the associated courses and teaching units via the knowledge chain. Figure  8 provides a complete knowledge chain for Tourism Psychology.  The Independent Samples T-Test was employed to verify whether there are significant differences in the knowledge chains of different courses and teaching units. The test results are displayed visually as a P-P plot in Figure 9. It can be seen that the sample points were all close to the diagonal of the first quadrant, and approximate the normal distribution. This means the application of knowledge chain to online tourism education greatly facilitates the construction of the knowledge logic system for relevant courses and teaching units.   Table 3 provides the statistics on student satisfaction with the proposed knowledge chain of online tourism education. The overall satisfaction was measured by three indices: the satisfaction with knowledge system, that with the implementation of courses or teaching units, and that with resource recommendation. The total mean stood at 3.62, suggesting that the students are satisfied with the learning or query of learning resources, using the proposed knowledge chain of online tourism education. Except for the overall satisfaction with resource recommendation, Groups I and II differed in the mean satisfaction with knowledge system and that with the implementation of courses or teaching units. The overall satisfaction of Group I was slightly higher than that of Group II. In addition, although the overall satisfactions of the two groups with resource recommendation were not very different, the two groups had a certain difference in the dispersion of satisfaction scores on resource recommendation. The slightly lower dispersion of Group I indicates that the students in that group have less consistent views of resource recommendation.

Conclusions
This paper explores the application of knowledge chain in OTERs. After modeling the associations between knowledge units in OTERs, the authors computed these associations, optimized the structure of knowledge chain, and analyzed OTER knowledge chain. The relevant experimental results were demonstrated and analyzed with examples. Through experiments, the following issues of online tourism education were exemplified: the construction of knowledge system-knowledge chain, the construction of core online courses-knowledge chain, the construction of teaching units-knowledge chain, and the construction of the knowledge chain of Tourism Psychology. The Independent Samples T-Test was employed to verify whether there are significant differences in the knowledge chains of different courses and teaching units. The test results are displayed visually as a P-P plot, which demonstrates that the application of knowledge chain to online tourism education greatly facilitates the construction of the knowledge logic system for relevant courses and teaching units. Finally, the students' satisfaction with the application of OTER knowledge chain was obtained.

7 Authors
Chunyan Chen Obtained master of science in education from Zhejiang Normal University. She is an associate professor in Zhejiang Business College. Her research interests include tourism management, online education in hospitality and tourism (email: 00637@zjbc.edu.cn).
Tong Wu is a Ph.D. candidate of tourism management and commerce in James Cook University. And he is a professor in Ma'anshan Teachers' College. His research experience has mainly focused on regional destination management and planning, the tourism education (email: tong.wu3@my.jcu.edu.au).