1. bookAHEAD OF PRINT
Detalles de la revista
License
Formato
Revista
eISSN
2444-8656
Primera edición
01 Jan 2016
Calendario de la edición
2 veces al año
Idiomas
Inglés
access type Acceso abierto

Research on the mining of ideological and political knowledge elements in college courses based on the combination of LDA model and Apriori algorithm

Publicado en línea: 20 May 2022
Volumen & Edición: AHEAD OF PRINT
Páginas: -
Recibido: 22 Mar 2022
Aceptado: 10 Apr 2022
Detalles de la revista
License
Formato
Revista
eISSN
2444-8656
Primera edición
01 Jan 2016
Calendario de la edición
2 veces al año
Idiomas
Inglés
Abstract

In recent years, mapping of knowledge domain and political knowledge has developed rapidly and gradually penetrated into many practical applications. The construction of an ideological and political knowledge framework for colleges has become one of the important applications. Therefore, commencing with the theory of mapping of knowledge domain, and aiming at resolving the difficulty involved in effectively extracting and analysing ideological and political knowledge, a mining method for ideological and political knowledge elements is constructed in this paper, based on LDA model and Apriori algorithm. By setting a three-dimensional matrix of keywords and association rules, an algorithm for mining of ideological and political knowledge elements is proposed, where LDA model is used to acquire ideological and political subject words, and Apriori algorithm is used to discover tacit ideological and political knowledge. It is helpful to solve the problem of excavation and presentation of ideological and political elements in college curriculum, serve the ideological and political construction of curriculum, which is of great significance to dredge students’ learning paths and reduce learners’ learning cost.

Keywords

Introduction

With the gradual enhancement of China's political status, activities connected with government affairs and also various domestic construction measures have increased correspondingly, resulting in the accumulation of a steadily increasing amount of political knowledge. In the face of new activities connected with state affairs that are constantly emerging, the mainstream way to obtain ideological and political information is to browse the political news from the current political web portal, which can enable visitors to gain certain information in a short time. However, because the ideological and political news is mostly text-based, each article basically contains more proper nouns. At the same time, China is traversing the road of socialism with Chinese characteristics, which are different from the national conditions of other countries, political activities are numerous and targeted, and ideological and political knowledge is fragmented, so it is difficult for people to establish a clear structure of knowledge in their minds [1,2,3]. At present, this method is time-consuming and cannot provide a clear and effective understanding of the main links that characterise ideological and political information, and moreover lacks a structured browsing platform.

Mapping of knowledge domain is a semantic relation network that connects nodes with relation lines. It uses the structure of “map” to express various relations in the colourful world, and is intuitive, natural and efficient [4, 5]. Using knowledge mapping to represent the knowledge prevailing in the ideological and political field can effectively incorporate these complicated and numerous aspects of knowledge into a system and display them in a custom-structured presentation format, and thus serve as an improvement over the original mode of showing the inquiry on web pages. In addition, because of the structural characteristics of knowledge mapping itself, it can also make ideological and political knowledge inquiry simple and fast, which improves the efficiency and quality of ideological and political learning.

For specialised courses in colleges and universities, in addition to professional knowledge, the ideological elements implied by professional knowledge are equally important. As a kind of ideological element, ideological and political elements play an important role in subject education [6]. Moreover, the ideological and political construction of curriculum is the focus of the current task involving mainstream curriculum construction, and the ideological and political knowledge elements of specialised curriculum have become an important part of specialised curriculum knowledge system, where the effective extraction of ideological and political elements is of vital importance. Therefore, the mining of knowledge element in the ideological and political field can aggregate scattered knowledge in a scientific and effective way, and can greatly improve the accuracy and efficiency of ideological and political knowledge retrieval. At the same time, if we can further study the knowledge mapping integrated with ideological and political elements and expand ideological and political knowledge from the depth level, it can be used as a reference for the ideological and political construction of Chinese colleges and universities.

Mining method of ideological and political knowledge elements
Data processing

The data sources of ideological and political knowledge can be divided according to different characteristics. Some data in the acquisition process are in a structured state and generally have a fixed format, and these data are known as structured data [7]. Others are scattered or having no structure, which can be referred to as the semi-structured and unstructured states, respectively. This kind of data needs to be processed separately. First, the noise and inconsistency in the initial data are eliminated, data from various sources are integrated, and relevant data are extracted and analysed from the preliminary data. Then, according to the data mining form that needs to be adopted, the demand data is transformed, such as segmentation of ideological and political words. After data acquisition, pre-processing is needed to effectively extract knowledge from it. Data pre-processing [8, 9] mainly includes data cleaning, data filtering, data deduplication and integration, etc. Data cleaning is to remove unnecessary information from the extracted data, such as redundant characters and random codes, which, if present, would have a direct adverse impact on the accuracy of knowledge acquisition in the next step. The samples of ideological and political knowledge used in this paper are all structured data, which are relatively simple to adopt. Generally, they can be directly used only by converting or calling into the map according to the corresponding rules.

Construction of ontology

Ontology is the explicit formal specification of shared conceptualisation [10]. Therefore, ontology can be considered as a model specification for the concept of specific domain entities, and their relationship, characteristics and laws. By summarising the entity concepts, the hierarchical representation of conceptual abstraction knowledge can be obtained, which has advantages in the representation and management of knowledge concepts [11]. Based on the difference of automation degree, methods of ontology construction can be divided into manual construction, semi-automatic construction and automatic construction. The manual construction method is to determine the content of knowledge and the relationship between knowledge by experts where the process can be controlled and the quality can be guaranteed. Semi-automation completes the construction by combining manual operation with machine works; on the other hand, the automation mode is that the content and relationship are completely acquired by the machine. In this paper, the artificial construction method is adopted, and its construction process is shown in Figure 1.

Fig. 1

Process of ontology construction

Keywords extraction

Keywords extraction of knowledge elements is an important direction and trend of research on knowledge mapping [12]. The extraction can be constructed from top-down or bottom-up according to different objects [13, 14]. The top-down method is to define the data schema for the knowledge mapping, that is, to define the ontology, and then start with the top-level conceptual ontology and refine it layer by layer, so as to form a better hierarchical structure of taxonomy. Finally, entities are filled in the defined data schema. Compared with the top-down method, the bottom-up method construction process is exactly opposite to it. First, it starts with entity processing, where the entity is induced and organised, and the class hierarchy and relationships between classes are automatically extracted from the knowledge mapping, so as to form the bottom ontology class, and then gradually abstract upwards to form the top ontology class. Since the accuracy of the ideological and political knowledge selected in this paper is related to the learners’ knowledge reserve and application, which is deterministic or easy-control, it often needs high precision. Therefore, the top-down method is selected for keyword construction, and the process is shown in Figure 2.

Fig. 2

Process of keyword extraction

Establishment of basic model
LDA model

LDA model is an unsupervised machine learning technology [15, 16] that can be used to identify potential subject information in a large number of document sets or corpora. When LDA model is used in keyword extraction, it is difficult for common words to become keywords, and the obtained keywords cannot completely and accurately cover the subject information of documents that contributes to the improvement of accuracy in keyword extraction [17].

LDA model is a three-level Bayesian probability model, and its basic assumption is that [18]: Each document is a polynomial distribution about subjects, and each subject is a polynomial distribution about vocabulary. Since LDA contains a Bayesian framework, the subject distribution and word distribution of LDA need to be given by prior knowledge, and this process is random. The overall task of LDA model is to infer the subject distribution of each document and the word distribution in each subject according to M documents and a priori parameters α and β. The establishment process of ideological and political documents is as follows:

The number of documents to be generated in a given corpus m piece is m, which is called d1, d2, … dm, assuming that there is K subject T1, T2, … Tk.

Based on experience, we give a priori parameters α and β; α indicates the document-subject density; the higher the α, the more subjects the document contains, and vice versa; β indicates subject-word density; the higher the β, the more words the subject contains.

We assume that the subject prior distribution corresponding to each document is the Dirichlet distribution, and so its subject distribution ϑi is: ϑi=Dirichlet(α) {\vartheta _i} = Dirichlet(\alpha )

We assume that the prior distribution of words corresponding to each subject is the Dirichlet distribution, and so its word distribution φk is: φk=Dirichlet(β) {\varphi _k} = Dirichlet(\beta )

The sample subject distribution of the ith document di is that ϑi = (ϑi1, ϑi2, … ϑik);

For the ith document di, under the subject distribution of ϑi, its specific subject number can be determined as Zij = k, k ∈ [1,K], j ∈ [1,Ni], and Ni represents the number of words in each document.

According to step (6), the subject Zij is obtained, sampling a word distribution from several words corresponding to the subject φk = (φk1, φk2φkv), v ∈ [1,V], and V represents number of words contained under Zij.

The sample from the previous step φk is used to finally generate the words wij, which are the jth words of the ith document.

We repeat steps (5) to (8), and finally get all words in all ideological and political documents.

In which Beta is distributed in Dirichlet distribution. According to Beta distribution, Eq. (3) can be derived as shown below: f(x)={1B(α,β)χα1(1χ)β1χ[0,1]0others f(x) = \left\{ {\matrix{ {{1 \over {B(\alpha ,{\kern 1pt} \beta )}}{\chi ^{\alpha - 1}}{{(1 - \chi )}^{\beta - 1}}} \hfill & {\chi \in [0,1]} \hfill \cr 0 \hfill & {{\rm{others}}} \hfill \cr } } \right. Among them: B(α,β)=Γ(α)Γ(β)Γ(α+β) B(\alpha ,{\kern 1pt} \beta ) = {{\Gamma (\alpha )\Gamma (\beta )} \over {\Gamma (\alpha + \beta )}}

The specific expression of the Dirichlet distribution is shown in Eq. (4): f(p|α)={1Δ(α)k=1KPkαk1Pk[0,1]0others f(\vec p|\vec \alpha ) = \left\{ {\matrix{ {{1 \over {\Delta (\vec \alpha )}}\prod\nolimits_{k = 1}^K P_k^{{\alpha _k} - 1}} \hfill & {{P_k} \in [0,1]} \hfill \cr 0 \hfill & {{\rm{others}}} \hfill \cr } } \right. Dir(p|α)=1Δ(α)Πk=1Kpkαk1 {\rm{Dir}}(\vec p|\vec \alpha ) = {1 \over {\Delta (\vec \alpha )}}\Pi _{k = 1}^Kp_k^{{\alpha _k} - 1} where Δ(α)=Πk=1KΓ(αk)Γ(Σk=1Kαk) \Delta (\vec \alpha ) = {{\Pi _{k = 1}^K\Gamma \left( {{\alpha _k}} \right)} \over {\Gamma \left( {\Sigma _{k = 1}^K{\alpha _k}} \right)}}

From the above steps, learning estimation can be made on the variables: Zij, ϑi and φk. In the original paper of LDA, the variational EM algorithm was used to estimate unknown parameters [19]. However, Gibbs Sampling has a better effect when estimating unknown parameters of LDA [20].

The main idea underlying Gibbs sampling is Bayesian estimation, and unknown parameters can be solved through the posterior distribution results of subject distribution and word distribution. For the required target distribution, the conditional probability distribution of each feature dimension of subject distribution and word distribution needs to be obtained [21]. The specific steps are as follows:

According to the above analysis of LDA, the joint distribution of all variables can be given as shown in Eq. (5): p(wm,Zm,vecϑm,Φ|α,β)=n=1Nmp(wmn|φzm)p(Zmn|ϑm)p(ϑm|α)p(Φ|β) p\left( {{{\vec w}_m},{{\vec Z}_m},\;vec{\vartheta _m},\Phi |\vec \alpha ,\vec \beta } \right) = \prod\limits_{n = 1}^{{N_m}} p\left( {{w_{mn}}|{{\vec \varphi }_{{z_m}}}} \right)p\left( {{Z_{mn}}|{{\vec \vartheta }_m}} \right) \cdot p\left( {{{\vec \vartheta }_m}|\vec \alpha } \right) \cdot p(\Phi |\vec \beta ) Note: In Eqs (5) and (6), Zmn is equivalent to the Zij that has been defined above, wmn is equivalent to the wij that has been defined above and φZmm {\vec \varphi _{{Z_{mm}}}} is equivalent to the defined φk.

According to the generation process of LDA subject document, prior parameter α generates subject distribution θ, subject distribution θ determines the specific subject and β generates word distribution φ. The joint probability distribution is shown in Eq. (6): p(w,z|α,β)=p(w|z,β)p(z|α) p\left( {\vec w,\vec z|\vec \alpha ,\vec \beta } \right) = p\left( {\vec w|\vec z,\vec \beta } \right)p\left( {\vec z|\vec \alpha } \right)

According to the above step, the probability distribution of subject feature zi corresponding to the word wi is calculated as p(zi=k|z¬i,w)=p(w,z)p(w,z¬i)=p(w,z)p(w¬i,z¬i)p(wi)p(z)p(z¬i)Δ(nz+β)Δ(nz,¬i+β)Δ(nm+α)Δ(nm,¬+α)nk,¬it+βtt=1vnk,¬it+βt(nm,¬tk+αk) \matrix{ {p\left( {{z_i} = k|{{\vec z}_{\neg i}},\vec w} \right)} \hfill & { = {{p(\vec w,\vec z)} \over {p\left( {\vec w,{{\vec z}_{\neg i}}} \right)}} = {{p(\vec w,\vec z)} \over {p\left( {{{\vec w}_{\neg i}},{{\vec z}_{\neg i}}} \right)p\left( {{w_i}} \right)}} \cdot {{p(\vec z)} \over {p\left( {{{\vec z}_{\neg i}}} \right)}}} \hfill \cr {} \hfill & { \propto {{\Delta \left( {{{\vec n}_z} + \vec \beta } \right)} \over {\Delta \left( {{{\vec n}_{z,\neg i}} + \vec \beta } \right)}} \cdot {{\Delta \left( {{{\vec n}_m} + \vec \alpha } \right)} \over {\Delta \left( {{{\vec n}_{m,\neg }} + \vec \alpha } \right)}}} \hfill \cr {} \hfill & { \propto {{n_{k,\neg i}^t + {\beta _t}} \over {\sum\nolimits_{t = 1}^v n_{k,\neg i}^t + {\beta _t}}}\left( {n_{m,\neg t}^k + {\alpha _k}} \right)} \hfill \cr } where nkt=nk,¬it n_k^t = n_{k,\neg i}^t or nkt=nk,¬it+1 n_k^t = n_{k,\neg i}^t + 1 . Among them, Z¬i represents the subject distribution that removes the word subscript i corresponding to the subject, and nk(t) n_k^{(t)} is the number of occurrences of the word t in subject K. Finally, by Gibbs sampling, the subjects of all words and the word distribution under each subject can be obtained. Then, we can count the number of subjects of words corresponding to each document to get the subject distribution of each document.

Knowledge element association model

The definition of association rule is [22]: hypothesis I = {I1,I2 …, Im} is a collection of items, given a transaction database D, where each (transaction) t is a nonempty subset of I, namely TI. Association rules are usually measured by confidence and support.

The support degree (Sup) of association rules in D is the probability that transactions in D contain XY at the same time, as shown in Eq. (8): Sup(X,Y)=p(XY)=number(XY)num(AllSamples) {\rm{Sup}}(X,Y) = p(XY) = {{{\rm{number}}{\kern 1pt} (XY)} \over {{\rm{num}}{\kern 1pt} ({\rm{AllSamples}})}}

If the probability of simultaneous occurrence of X, Y is small, there is little relationship between them; but if probability is high, it means that X, Y are correlated. Confidence (Con) is that the transaction in D has contained Y, including conditional probability of X, as shown in Eq. (9): Con(XY)=p(X|Y)=p(XY)/p(Y) {\rm{Con}}(X \Leftarrow Y) = p(X|Y) = p(XY)/p(Y)

If the confidence is 100%, it means that when Y appears, X must also appear; then X, Y can be deduced together, and otherwise, X, Y is irrelevant. When the above indicators are used to measure association rules, if the thresholds of minimum support and minimum confidence are satisfied, a strong association rule can be inferred. Among them, the two thresholds are artificially set according to specific applications.

Mining process of ideological and political knowledge elements
Identification of knowledge elements

Common methods of knowledge element extraction include the co-occurrence-based and semantic-based methods [23]. The former means that the potential relationship between entities is obtained by analysing the co-occurrence relationship between entities, which is simple and easy to calculate. Semantic-based method solves the disadvantage of ignoring semantics based on the co-occurrence method, where the relationship between entities is determined by semantic predicates. But it needs more human participation. On the one hand, both of these two methods focus on finding the hidden relationship between various elements of knowledge, without acquiring new hidden knowledge entities. On the other hand, the knowledge discovered exclusively by the co-occurrence method is too simple, and it doesn’t pay attention to the co-occurrence between different types of words.

Therefore, a method of combining LDA model with Apriori association rules is proposed to extract knowledge elements. Since LDA subject model is a method of text clustering, the same type of subject words are clustered under the same subject by Gibbs sampling iteration, so as to obtain different types of subject words, the purpose of this exercise being to find candidate hidden knowledge entities. In order to further screen out accurate tacit knowledge entities from candidate tacit knowledge entities, first, a statistical matrix of co-occurrence frequency is constructed for subject words under different subjects, that is, different types of words by the proposed statistical method of co-occurrence frequency. Since there are more than two subjects, a three-dimensional matrix is constructed, which is different from the co-occurrence matrix. The latter is the co-occurrence statistics of words of the same category, and solves the similarity problem between words. Finally, in order to verify the correlation between words, Apriori algorithm is used to analyse the correlation of knowledge elements.

Correlation between knowledge elements

By setting keywords such as ‘data structure’ and ‘ideological and political education’, 200 related documents integrating data structure and ideological and political education were crawled from Baidu Academic Knowledge Network and other academic platforms as the main source of data. In addition, the ideological attributes hidden in the knowledge points of professional courses were discovered, and the method of associating knowledge that was employed in this paper was adopted to realise the knowledge mining of professional courses in colleges with the attribute of connecting thoughts.

Aiming at the ideological and political education of professional courses in colleges, this paper mainly studies the ideological and political knowledge elements of courses in a major. Adopting the subject model is equivalent to clustering the contents of the documents, and through subject clustering, we can obtain different subjects such as the curriculum subject and the ideological and political subject of each curriculum document, where various elements of useful knowledge can be better separated from the text data. However, in LDA model, all words in corpus are regarded as word bags without context, and only the probability distribution of words–documents–subjects is obtained, which ignores the semantic information of the words themselves and that between words in documents [24]. Therefore, after mining the subject words, in order to find the related subject words under different subjects more accurately, analysis of association rule is needed, and in the calculation of support and confidence, the frequency of common occurrence of subject words under different subjects is needed. However, because the subject words obtained under different subjects are different after subject classification, the algorithm of association rule cannot be directly used to count the frequency of co-occurrence of subject words.

Implementation process
Algorithm flow

According to the above model and its extraction rules, the specific algorithm flow is as follows:

According to LDA model, the pre-processed literature data are clustered into k subjects to get m subject words under each subject.

Define a two-dimensional array (subject_word) for subject words [k][m]. The set of subject words under the ith subject is (subject_word) [i].

Establish (m + 1)* (m + 1) two-dimensional matrix of X, make its [0][0] blank.

The values of the first row and the first column of the matrix are, respectively, two different subjects, that are i and j set of subject words (subject_word) [i] and (subject_word) [j].

Obtain (k(k − 1)/2)-layer three-dimensional matrix.

Set the distance between two subject words in the document, that is, the window size, L.

Traverse (subject_word)[i] and (subject_word)[j], let the checked out (subject_word) [i] [a] and (subject_word) [j][b] (among them, a ∈ [0, m], b ∈ [0, m]), the subject words are combined and then the query is made according to the original data. If the two words based on the original data are within the set window threshold, the combination is made, +1.

Obtain the frequency of the co-occurrence of pairwise subject words under different subjects through the matrix.

Substitute the above results into the formulas of support and confidence, which are shown in Eqs (10) and (11): S(wij,wpq)=N(wijwpq)/N(All) S\left( {{w_{ij}},{w_{pq}}} \right) = N\left( {{w_{ij}}{w_{pq}}} \right)/N({\rm{All}}) C(wijwpq)=S(wij,wpq)/P(wpq) C\left( {{w_{ij}} \Leftarrow {w_{pq}}} \right) = S\left( {{w_{ij}},{w_{pq}}} \right)/P({w_{pq}})

Among them, S (wij, wpq) represents two subject words. wij, wpq represents the degree of support, C (wijwpq) represents the confidence of wijwpq, wij represents the ith subject words under the jth subject and wpq represents pth subject words under the qth subject.

Establishment of a three-dimensional matrix on keywords

First, set the window size, which determines the number of related keywords. The window sizes L are 15, 20, 25 and 30, respectively. When the value of L is 25, the maximum number of related subject words is obtained, and so the window size of 25 is finally selected.

In the three-dimensional matrix X, dth rows of layers represent subject words wij of subject Zi(i ∈ [1,K]) (where j represents that the subject Zi containing j subject words), and the line indicates the subject. Set Xdrc as the elements of row r and column c of layer d of the three-dimensional matrix X, whose value is the frequency of documents in which wij and wpq appear together. As an example, the co-occurrence frequency of subject words of one layer in Xdrc three-dimensional matrix is selected. Each layer is a two-dimensional matrix, as shown in Table 1.

Three-dimensional matrix of Keywords

X Y

Queue Sort Recursion Traverse Binary tree Tree Linear table

Seeking truth 0 0 0 10 0 0 3
Innovate 0 0 7 0 0 0 0
Teaching 2 0 0 0 0 8 5
Diligent 0 0 0 0 0 13 26
Course 17 0 1 0 20 0 20
Craftsman 14 32 9 28 2 11 1
Order 42 0 16 16 0 0 0

From Table 1, it can be seen that if the co-occurrence frequency of ‘queue’ and ‘order’ is higher, then ‘queue’ and the thought attribute ‘order’ are closely related. Next, the association is further confirmed by association rules, and the co-occurrence frequency among subject words obtained by statistics is used as the input of association rule analysis.

Keyword Association Analysis

Set the minimum support to 0.03 and the minimum confidence to 0.27. The association rule obtained by the algorithm represents the association between two subject words under different subjects. Calculate the support and confidence of the co-occurrence frequency matrix obtained in the previous step to screen the rules, and judge according to the threshold value to obtain association rules, as shown in Table 2.

Association analysis of keyword

Association rule Support degree Confidence level

Queue-Diligence 0.2180 0.1992
Queue-Craftsman 0.3556 0.2300
Queue-Ordered 0.1212 0.8273

From Table 2, it can be seen that the confidences of the ‘Queue-Diligence’ and ‘Queue-Artisan’ rules do not satisfy the threshold setting, but are greater than the support threshold, and so these are related but not strong association rules. Finally, through the threshold screening of support and confidence, 31 association rules are obtained, that is, 31 thought attributes can be obtained.

Evaluation of the effect on mining ideological and political knowledge elements

Since the discovery of ideological and political knowledge elements is unknown and uncertain, it is not possible to make an evaluation and come to a confirmatory conclusion as to whether the obtained knowledge is correct or wrong; also, the knowledge elements expressed visually and explicitly have not yet been certified, and it is thus difficult to verify the validity and correctness of the results of knowledge element mining. Reproduction of known classic examples is a common verification method for knowledge element; however, ideological and political knowledge elements, as new knowledge that has not been acquired by curriculum, cannot be verified by appropriate examples [25]. Therefore, the methods of literature verification and expert evaluation of ideological and political work are used to verify the excavated knowledge elements.

Based on the method of literature verification, the results were compared for 200 literature elements, where 67 out of the 72 pairs of knowledge elements related to the data structure course and humanistic literacy obtained in the final experiment are correct; that is, the accuracy of the verification can reach 93.1%. Ideological and political experts have been involved as researchers in this field for many years, or have been employed as front-line teachers in colleges and universities. The results of literature verification are fed back to 10 professional teachers, and the results are generally good. However, owing to the diversity of ideological attributes, each teacher has a different understanding of the ideological attributes of knowledge elements.

Conclusion

Ideological and political knowledge, as an important part of college students’ study in school, is applied to the construction of professional curriculum knowledge in colleges. By mining and extracting the elements of ideological and political knowledge, the inquiry of ideological and political knowledge can become simple and fast. In this paper, 200 cases of ideological and political knowledge obtained from website crawling and literature research are taken as sample data. Aiming at the deficiency of deep knowledge extraction in ideological and political teaching, a mining method of ideological and political knowledge elements is constructed based on LDA model and Apriori algorithm. First, different subjects are obtained by clustering data by LDA model, then the co-occurrence frequency of keywords under different subjects is counted to build a three-dimensional matrix, and finally the three-dimensional matrix is taken as the input of Apriori algorithm, where the ideological and political knowledge elements can be obtained by setting a threshold to judge the validity of association rules. The results show that the ideological and political knowledge elements extracted by this method can achieve an accuracy rate of 93.1%.

Fig. 1

Process of ontology construction
Process of ontology construction

Fig. 2

Process of keyword extraction
Process of keyword extraction

Three-dimensional matrix of Keywords

X Y

Queue Sort Recursion Traverse Binary tree Tree Linear table

Seeking truth 0 0 0 10 0 0 3
Innovate 0 0 7 0 0 0 0
Teaching 2 0 0 0 0 8 5
Diligent 0 0 0 0 0 13 26
Course 17 0 1 0 20 0 20
Craftsman 14 32 9 28 2 11 1
Order 42 0 16 16 0 0 0

Association analysis of keyword

Association rule Support degree Confidence level

Queue-Diligence 0.2180 0.1992
Queue-Craftsman 0.3556 0.2300
Queue-Ordered 0.1212 0.8273

Liu Xiaoxia, a new exploration of ideological and political teaching in colleges and universities in the “internet plus” era. Frontier of higher education, 2021, 4(12) (in Chinese) LiuXiaoxia a new exploration of ideological and political teaching in colleges and universities in the “internet plus” era Frontier of higher education 2021 4 12 (in Chinese) Search in Google Scholar

Hu Xiang, He Hongjuan analyze the importance of teachers and students’ dual-subject status in the teaching practice of ideological and political courses. Education World, 2021, 3(10) (in Chinese) HuXiang HeHongjuan analyze the importance of teachers and students’ dual-subject status in the teaching practice of ideological and political courses Education World 2021 3 10 (in Chinese) Search in Google Scholar

Dai Ying-fu, Research on the Fusion Path of Chinese Excellent Traditional Culture and Ideological and Political Teaching in Secondary Vocational Schools Frontier of Higher Education, 2020, 3(10) (in Chinese) DaiYing-fu Research on the Fusion Path of Chinese Excellent Traditional Culture and Ideological and Political Teaching in Secondary Vocational Schools Frontier of Higher Education 2020 3 10 (in Chinese) Search in Google Scholar

Alina Zhang. Research on the organization and search technology of curriculum knowledge map. Wuhan University, 2016. (in Chinese) AlinaZhang Research on the organization and search technology of curriculum knowledge map Wuhan University 2016 (in Chinese) Search in Google Scholar

Pan JZ, Vetere G, Gomez-Perez JM, et al. Exploiting Linked Data and Knowledge Graphs in Large Organisations. Cham: Springer, 2017. PanJZ VetereG Gomez-PerezJM Exploiting Linked Data and Knowledge Graphs in Large Organisations Cham Springer 2017 10.1007/978-3-319-45654-6 Search in Google Scholar

Ma liya, the exploration and reflection on the “five in one” reform path of ideological and political courses in colleges and universities Creativity and Innovation, 2020, 4(10) (in Chinese) Maliya the exploration and reflection on the “five in one” reform path of ideological and political courses in colleges and universities Creativity and Innovation 2020 4 10 (in Chinese) Search in Google Scholar

Peng Ganhui. Automatic construction of domain knowledge map. Chongqing University, 2017. (in Chinese) PengGanhui Automatic construction of domain knowledge map Chongqing University 2017 (in Chinese) Search in Google Scholar

Xu Zenglin, Sheng Yongpan, He Lirong et al. Summary of knowledge mapping technology. Journal of University of Electronic Science and Technology of China, 2016, 45(04): 589–606. (in Chinese) XuZenglin ShengYongpan HeLirong Summary of knowledge mapping technology Journal of University of Electronic Science and Technology of China 2016 45 04 589 606 (in Chinese) Search in Google Scholar

Zhongwei Zhang, Cao Lei, Chen Xiliang et al. A summary of knowledge reasoning based on neural network. Computer Engineering and Application, 2019, 55(12): 8–19. (in Chinese) ZhongweiZhang CaoLei ChenXiliang A summary of knowledge reasoning based on neural network Computer Engineering and Application 2019 55 12 8 19 (in Chinese) Search in Google Scholar

Katy Bömer, Chaomei Chen, Kevin W. Boyack. Visualizing knowledge domains[J]. Wiley Subscription Services, Inc. A Wiley Company, 2003, 37(1): 1–9. BömerKaty ChenChaomei BoyackKevin W. Visualizing knowledge domains [J] Wiley Subscription Services, Inc. A Wiley Company 2003 37 1 1 9 Search in Google Scholar

M J. Cobo, A.G. Lopez-Herrera, E. Herrera-Viedma, et al. Science mapping software tools: Review analysis, and cooperative study among tools[J]. Journal of the American Society for Information Science & Technology, 2012, 62(7): 1382–1402. CoboM J. Lopez-HerreraA.G. Herrera-ViedmaE. Science mapping software tools: Review analysis, and cooperative study among tools [J]. Journal of the American Society for Information Science & Technology 2012 62 7 1382 1402 10.1002/asi.21525 Search in Google Scholar

Li T, Wang C, Li H. Development and Construction of Knowledge Graph. Journal of Nanjing University of Science andTechnology.2017.41(1): 122–34. LiT WangC LiH Development and Construction of Knowledge Graph Journal of Nanjing University of Science andTechnology 2017 41 1 122 34 Search in Google Scholar

ZHU M. PAO B. XU C. Research Progress on Development and Construction of Knowledge Graph. Journal of Nanjing University of Information Science and Technology (Natural Science Edition). 2017. 9(6):575–582. ZHUM. PAOB. XUC. Research Progress on Development and Construction of Knowledge Graph Journal of Nanjing University of Information Science and Technology (Natural Science Edition) 2017 9 6 575 582 Search in Google Scholar

Sheng Tian. Research on the method of tacit knowledge discovery in biomedical literature. Dalian University of Technology, 2019. (in Chinese) TianSheng Research on the method of tacit knowledge discovery in biomedical literature Dalian University of Technology 2019 (in Chinese) Search in Google Scholar

Xu Ge, Wang Houfeng. The Development of Subject Models in Natural Language Processing. Chinese Journal of Computers, 2011, 34(8): 1423–1436. XuGe WangHoufeng The Development of Subject Models in Natural Language Processing Chinese Journal of Computers 2011 34 8 1423 1436 10.3724/SP.J.1016.2011.01423 Search in Google Scholar

Zhu Zede, Li Miao, Zhang Jian, Zeng Weihui, Zeng Xinhua. A keyword extraction method based on LDA model. Journal of Central South University (Natural Science Edition), 2015, 46(06): 2142–2148. (in Chinese) ZhuZede LiMiao ZhangJian ZengWeihui ZengXinhua A keyword extraction method based on LDA model Journal of Central South University (Natural Science Edition) 2015 46 06 2142 2148 (in Chinese) Search in Google Scholar

Wang Yajing, Guo Qiang, Deng Chunyan, et al. Research on user feature prediction based on LDA subject model Complex Systems and Complexity Science, 2020, 17(4): 9–15 (in Chinese) WangYajing GuoQiang DengChunyan Research on user feature prediction based on LDA subject model Complex Systems and Complexity Science 2020 17 4 9 15 (in Chinese) Search in Google Scholar

Griffiths T L, Steyvers M. Finding scientific subjects. Proc Natl Acad Sci U S A, 101(Supplement 1): 5228–5235. GriffithsT L SteyversM Finding scientific subjects Proc Natl Acad Sci U S A 101 Supplement 1 5228 5235 10.1073/pnas.030775210138730014872004 Search in Google Scholar

Mikolov T, Chen K, Corrado G, et al. Efficient Estimation of Word Representations in Vector Space. Computer Science, 2013, 3: 104–208. MikolovT ChenK CorradoG Efficient Estimation of Word Representations in Vector Space Computer Science 2013 3 104 208 Search in Google Scholar

Gao Yongbing, Xu Qingrui. Research on user interest mining in Weibo based on improved LDA model. Journal of Inner Mongolia University of Science and Technology, 2019, 38(3): 272–276 (in Chinese) GaoYongbing XuQingrui Research on user interest mining in Weibo based on improved LDA model Journal of Inner Mongolia University of Science and Technology 2019 38 3 272 276 (in Chinese) Search in Google Scholar

Egara E S, Triadi D, Andryani R. Subject modelling Twitter data with latent Dirichlet allocation method //2019 International Conference on Electrical Engineering and Computer Science (ICECOS), 2019: 386–390. EgaraE S TriadiD AndryaniR Subject modelling Twitter data with latent Dirichlet allocation method 2019 International Conference on Electrical Engineering and Computer Science (ICECOS) 2019 386 390 10.1109/ICECOS47637.2019.8984523 Search in Google Scholar

Zhu Zede, Li Miao, Zhang Jian, Zeng Weihui, Zeng Xinhua. A keyword extraction method based on LDA model. Journal of Central South University (Natural Science Edition), 2015, 46(06): 2142–2148. (in Chinese) ZhuZede LiMiao ZhangJian ZengWeihui ZengXinhua A keyword extraction method based on LDA model Journal of Central South University (Natural Science Edition) 2015 46 06 2142 2148 (in Chinese) Search in Google Scholar

Wu Aihua, Chen Chuxin. Mining positive and negative association rules of relational data in distributed database. Computer Simulation, 2021,38(09):344–347+352. (in Chinese) WuAihua ChenChuxin Mining positive and negative association rules of relational data in distributed database Computer Simulation 2021 38 09 344 347+352 (in Chinese) Search in Google Scholar

Mikolov T, Sutskever I, Chen K, et al. Distributed representations of words and phrases and their compositionality. Advances in Neural Information Processing Systems, 2013, 26: 3111–3119. MikolovT SutskeverI ChenK Distributed representations of words and phrases and their compositionality Advances in Neural Information Processing Systems 2013 26 3111 3119 Search in Google Scholar

Wang Xinyu, Pan Yuqing. Design and Practice of Ideological and Political Education in Data Structure Course. Computer Education, 2021(01): 97–100. (in Chinese) WangXinyu PanYuqing Design and Practice of Ideological and Political Education in Data Structure Course Computer Education 2021 01 97 100 (in Chinese) Search in Google Scholar

Artículos recomendados de Trend MD

Planifique su conferencia remota con Sciendo