Word Clouds as a Learning Analytic Tool for the Cooperative

e-Learning Platform NeuroK

Fernando Calle-Alonso

, Vicente Botón-Fernández

, Jesús M. Sánchez-Gómez

Miguel A. Vega-Rodríguez

, Carlos J. Pérez

and Daniel de la Mata

Research & Development Department, ASPgems SL, Spain

Cátedra ASPgems, University of Extremadura, Spain

Department of Technologies of Computers & Communications, University of Extremadura, Spain

Department of Mathematics, University of Extremadura, Spain

Keywords: Word Cloud, Tag Cloud, e-learning, Online Learning, NeuroK, Evaluation, Engagement.

Abstract: Word cloud or tag cloud is very popular these days. It is a tool used to display text data summarization in a

visual way very easy to understand. However, it has not been extensively used in teaching, especially in e-

learning, where it would make a differential advantage. This research presents the definition and

implementation of a word cloud tool in a social network-based e-learning platform (NeuroK), which is based

on the principles of neurodidactics. The different features developed and the results are shown. Several options

to compare word clouds from students and teachers allow the teacher to follow the development of the course,

and they provide him more information to facilitate the evaluation process.

1 INTRODUCTION

Nowadays the use of e-learning platforms and

learning management systems is very common for

educational institutions and companies. Although

digital education is growing continuously some

problems still should be resolved to continue its

development, for example, the difficulty to evaluate

students (dos Santos and Favero, 2015), the

prevention of dropping out (Yukselturk, 2014), the

redesign of educator roles (Adams et al., 2017),

understanding automatically the natural language

(Aeiad and Meziane, 2016), the flaws in the

accreditation process of distance and flexible learning

programs (Reeves, 2003) or blending formal and

informal learning (Czerkawski, 2016).

Calle-Alonso et al., 2017 developed a new e-

learning platform based on the principles of

neurodidactics called NeuroK. This platform tries to

solve several of the current problems affecting online

learning environments.

Learning analytics (Baker and Inventado, 2014)

allows to track students’ performance by using the

data obtained from their activities (connections,

comments, evaluations, documents shared,

favorites...). With learning analytics and

visualizations, real-time analysis of the course could

be performed and future-tense adapting actions could

be carried out to anticipate the course drifting in the

wrong direction (see, e.g., Nevado-Maestre et al.

2017).

Word clouds or tag clouds can be used as learning

analytics tools. They are visual representations of a

group of words used by the participants, and based on

their frequency. These kinds of clouds give greater

prominence to the words appearing more frequently

and reflect on all the information from within the texts

written by students and teacher in a course. By

investigating the patterns of words or phrases, or lack

thereof, in textual student responses, instructors can

evaluate if students, as a whole, have grasped or

missed key concepts or have made common mistakes

(De Paolo and Wilkinson, 2014). Word clouds belong

to Natural Language Processing (NLP) field (Heimerl

et al., 2014). They are very easy to understand and

they can be included into any class, subject and age.

Although word clouds are very powerful, they are not

used very much in education (nor in online education

of course), but the use of word clouds could offer a

lot of benefits in e-learning for both, teachers and

students.

Nickell (2012) shows how word clouds work and

test them in mathematics classrooms. He showed how

508

Calle-Alonso, F., Botón-Fernández, V., Sánchez-Gómez, J., Vega-Rodríguez, M., Pérez, C. and de la Mata, D.

Word Clouds as a Learning Analytic Tool for the Cooperative e-Learning Platform NeuroK.

DOI: 10.5220/0006816505080513

In Proceedings of the 10th International Conference on Computer Supported Education (CSEDU 2018), pages 508-513

ISBN: 978-989-758-291-2

to promote engagement in online learning. Perry

(2012) presents a use case to teach speaking style with

word clouds. Miley and Read (2011) used Wordle

(see Feinberg, 2014) to introduce Word Clouds in a

classroom routine, obtaining remarkable results,

validated by the student’s opinions. Finally, also with

Wordle, McNaught and Lam (2010) applied word

cloud tool for content analysis in research. These are

some interesting results of the application of Word

Clouds in teaching and research.

NeuroK is meant to be a social network based

learning platform, and the learning process is

supported by discussions. These interactions motivate

students and engage them. Word cloud tool is

demonstrated to be engaging for the students, and to

motivate them. Kaptein et al. (2010) and Behar-

Horenstein and Niu (2011) established that visual

representations of students texts with a word cloud

improved both critical thinking and engagement,

especially in online discussions. It was tested by

deNoyelles and Reyes-Foster (2015), revealing that

students using word clouds reported moderately

higher scores on critical thinking and engagement, as

well as peer interaction.

In the next sections, NeuroK Word Cloud tool is

presented and the features and results are shown with

examples.

2 IMPLEMENTATION

Today, word clouds have been shown to be a

beneficial tool in many educational environments

(Miley and Read, 2011). They provide a fast visual

representation of the information contained in texts.

In this work, word clouds are built upon a cooperative

e-learning platform, Neurok, to provide an overview

of the main concepts used by the students in their

communications. Thereby, teachers can analyze what

is being talked about in a course and make decisions

on strategy to guide the learning process through the

texts the students write. And since it is a learning

methodology based on the content of

communications, the participation and activity of

students is promoted and supported.

The tool under consideration consists of three

different word clouds: the target cloud, with the

concepts that the teacher would like their students to

talk about; the real cloud, with the concepts most

commonly used by the students; and the mixture of

both, with the target concepts according to their use

by the students. These three word clouds provide a

summary at both topic and course level. Several

metrics to measure the concordance between the

target cloud and the real cloud have also been

implemented.

2.1 Word Cloud

Word clouds represent the concepts most frequently

used by students. In our word cloud representation,

the bigger and more colourful a concept is, the

larger the number of times it is referred to.

The data obtained from NeuroK to build the word

clouds and to compute the concordance are mainly:

 Remarks: these are opinions and statements

that a student publishes in a course or topic.

 Comments: these are the replies that other

students make of a certain remark.

 Rates: these are comments that a student makes

to justify their assessment of a given remark.

 Delivery documents: these are the documents

that a student encloses in a delivery. Valid

extensions are .doc, .docx, .pdf, .pptx and .txt.

Each time any of these contents are published in a

course or topic, the corresponding text is

automatically analysed. First of all, words are

extracted from the text and those that do not provide

useful information, such as e-mails, URLs, numbers,

acronyms… are removed from the analysis process.

Once the words have been extracted, they are

analysed one by one checking they are not stop words

and obtaining their stem. Stop words are the most

common words in a language (prepositions, articles,

adverbs, conjunctions, pronouns and some verbs),

whereas a stem is the root form of a word. Stop words

are filtered because they are not so relevant for natural

language processing purposes since they occur

frequently in a language and bring little semantic

value to the content. At present, there is no single

universal list of stop words available, so we have built

our own list. Furthermore, a stemming process based

on Porter Stemming Algorithm (Porter, 1997) is used

for getting the stem words. Other versions of the

algorithm have been included to support more

languages, such as Spanish (Porter, 2001).

Finally, both the analyzed word and those with

which it is related are mapped to the same stem, and

one of them is selected to represent them all. This

avoids repetition of words with the same meaning in

the resultant cloud. When choosing the most

representative word, several criteria are taken into

account: infinitive verb, shorter noun, etc. Words

frequency is counted by stem and only those with the

highest frequency are shown in the cloud.

Word Clouds as a Learning Analytic Tool for the Cooperative e-Learning Platform NeuroK

509

2.2 Target Cloud

The target cloud is a special type of cloud that

represents the ideal word cloud from the teacher’s

point of view. In this cloud, the teacher can add new

words and specify the weight of each word. By

default, this weight is set proportionally to the number

of target words. To add new words, the teacher can

either type in a new one or select one from the list of

words that the students have used. Once the target

cloud is defined, the teacher will be able to compare

it with the course word cloud and analyse the

concordance.

2.3 Interface

For the representation purpose, jQCloud plugin has

been used. This plugin has some advantages such as:

 Dynamic lightweight and customizable tag

cloud.

 Cloud shaped appearance.

 Vertical, elliptic and rectangular clouds

support.

 Custom tag’s links, styles and weights.

The teacher and students can interact with the

word cloud by doing any of the following actions:

 Choosing the course or topic they want to be

represented in the cloud.

 Filtering data by user: students and teacher/s,

only students or a specific student.

 Selecting the size of the cloud: 10 words, 25

words or 50 words.

 Comparing the current cloud with the target

cloud.

However, there are differences between students

and teachers in terms of permissions. On the one

hand, students can see highlighted in the mixture of

clouds those words they are talking about and that the

teacher wants them to use, but they cannot see the

target words they are not using yet. On the other hand,

the teacher can see all types of words: the target words

the students are talking about, the remaining target

words that have not been used, and the words the

students are using but are not part of the target cloud.

In addition, teachers are the only ones with enough

permissions to edit the target cloud.

3 A LEARNING ANALYTIC

TOOL

In this section, a typical scenario that covers the

concepts previously mentioned has been described in

order to better understand the system operation. We

have tested NeuroK through an existing topic called

“Machine Learning” to obtain a real dataset. There

are ten students enrolled in the course and two

teachers. In order to carry out the tag clouds service,

all the information from remarks, comments, delivery

documents and rates are saved into the NeuroK

database. This information is registered during a

period of one month. Once the dataset is ready, it is

time to navigate to the “Tag Cloud” view and set up

the filters. In this case, we establish the following

settings: Machine Learning topic, only students and

25 most relevant words. The rest of the filters remain

as default.

After running our tag cloud approach over the

previous dataset, it generates the word cloud that

appears in Fig. 1. This cloud consists of the 25 most

frequently used words by students of the Machine

Learning topic. In order to evaluate this word cloud it

is necessary to define the target cloud. To do this, the

teacher can click on the button “Manage target cloud”

and customize its own target cloud.

Figure 1: Word cloud of the “Machine Learning” topic.

A2E 2018 - Special Session on Analytics in Educational Environments

510

As shown in Fig. 2, a teacher is able to create a

target cloud per topic or course. They can type a new

term and add it to the cloud or select one of the words

they have already used during the topic or course and

insert it into the cloud. The words used by the teacher

are listed on the right column and sorted by frequency

of usage. The teacher can also remove a word from

the cloud at any time. In this case, we have built a

target cloud with ten concepts related to machine

learning: algorithm, classification, decision,

knowledge, learning, mining, model, network, pattern

and training. As it was mentioned before, if no weight

is specified, it is set proportionally to the number of

target words. Since the target cloud has 10 words,

each word will weigh 10%.

Figure 2: Target cloud customization view.

Once the target cloud is defined, the teacher can

compare it with the topic word cloud. Back to the

“Tag Cloud” view, we select the following options:

Machine Learning topic, only students, 10 most

relevant words and comparison with target cloud. It is

important to remark that the number of words

selected has a direct impact on the concordance

calculation. To compare two clouds as accurately as

possible, both must have the same size. That is why,

in this case, only the 10 most relevant words from the

topic word cloud have been selected.

Fig. 3 shows the mixture of word clouds generated

after the application of the above filters. This mixture

presents different colours for each word depending on

the cloud they belong to: dark slate blue for those

words belonging to the word cloud (i.e. Machine

Learning word cloud), light grey for those words

belonging to the target cloud and orange for those

words belonging to both clouds. Just like the rest of

the cloud representations, the size of each word is

proportional to its frequency of use. Keeping all this

in mind, let us analyse the concordance metric.

Concordance is measured through words and their

respective weights. The existence of each target word

in the word cloud is validated and then the deviation

between its weights is calculated. The smaller the

absolute deviation of all weights, the better the

concordance between the two clouds. The best case

scenario is when students use all words from the

target cloud at the same frequency/weight specified

by the teacher. In the case in hand, the concordance

between the two clouds is about 31.59%. There are

only four words that both clouds have in common:

learning, decision, model and classification; but their

frequency of use is pretty similar to that expected

from the teacher. That is the reason why, even though

students use only a few target words, the concordance

is acceptable.

Figure 3: Comparison of “Machine Learning” word cloud

with target cloud.

By modifying one of the previous filters, namely

the user filter, we can get a comparison between a

student word cloud and the target cloud. Fig. 4 and

Fig. 5 shows two examples of these comparisons. The

first one presents a peculiar case because: in this case

there are six words belonging to both clouds, but not

all of them have the expected frequency of use. That

explains why the concordance here is only of 18.82%.

Word Clouds as a Learning Analytic Tool for the Cooperative e-Learning Platform NeuroK

511

Figure 4: Comparison of “Machine Learning” word cloud

with target cloud for a specific student.

Figure 5: Comparison of “Machine Learning” word cloud

with target cloud for another student.

4 CONCLUSIONS AND FUTURE

WORK

The Word cloud provides a visual and intuitive

representation of knowledge from the student/s,

allowing to create a fast description from their

contribution in an online learning process. Education

analytics could overload the teacher, but the word

cloud makes it easier to understand what is going on

in a course with a single eye span.

Word cloud can be used in gap analyses, showing

what is missing and what is expected using the

different colours to identify if the words are in the

student cloud, the teacher cloud or both. With this

information, the teacher could redirect the learning

process introducing new materials and exercises to

reinforce the misrepresented subjects in the word

cloud.

In the future, we will expand these features also to

be available for the students. It could be very

interesting for them to know which concepts they

have missed, but that the teacher expects to be well

known. This could boost motivation to discover the

ideas proposed by the teacher that they have left

behind.

Also the distances from student’s word clouds to

the teacher one could be useful to provide an

automatic evaluation measure.

ACKNOWLEDGEMENTS

This research has been supported by Ministerio de

Economía y Competitividad (Centro para el

Desarrollo Tecnológico Industrial, Contract IDI-

20161039), Junta de Extremadura (Contract AA-16-

0017-1, and projects GR15106 and GR15011),

Cátedra ASPgems, and European Union (European

Regional Development Funds).

REFERENCES

Adams Becker, S., Cummins, M., Davis, A., Freeman, A.,

Hall Giesinger, C., and Ananthanarayanan, V. (2017).

NMC Horizon Report: 2017 Higher Education Edition.

Austin, Texas: The New Media Consortium.

Aeiad, E., & Meziane, F. 2016. Validating the learning

outcomes of an e-learning system using NLP. Natural

Language Processing and Information Systems, 9612,

292-300.

Baker, R. S., & Inventado, P. S. (2014). Educational data

mining and learning analytics. In Learning

analytics (pp. 61-75). Springer New York.

Behar-Horenstein, L. S., & Niu, L. (2011). Teaching critical

thinking skills in higher education: A review of the

literature. Journal of College Teaching and

Learning, 8(2), 25.

Calle-Alonso, F., Cuenca-Guevara, A., de la Mata Lara, D.,

Sánchez-Gómez, J. M., Vega-Rodríguez, M. A., &

Sánchez, C. P. (2017). NeuroK: A Collaborative e-

Learning Platform based on Pedagogical Principles

from Neuroscience. In Proceedings of the 9th

International Conference on Computer Supported

Education (CSEDU 2017)-Volume Vol. 1, pp. 550-555.

A2E 2018 - Special Session on Analytics in Educational Environments

512

Czerkawski, B. 2016. Blending formal and informal

learning networks for online learning. The international

review of research in open and distributed

learning, 17(3), 138-156.

dos Santos, J. C. A., & Favero, E. L. 2015. Practical use of

a latent semantic analysis (LSA) model for automatic

evaluation of written answers. Journal of the Brazilian

Computer Society, 21(1), 21.

De Paolo, C. A., & Wilkinson, K. 2014. Get your head into

the clouds: Using word clouds for analyzing qualitative

assessment data. TechTrends, 58(3), 38-44.

Feinberg, J. 2014. Wordle-Beautiful word

clouds. Wordle—Beautiful Word Clouds.

http://www.wordle.net/

DeNoyelles, A., & Reyes-Foster, B. (2015). Using Word

Clouds in Online Discussions to Support Critical

Thinking and Engagement. Online Learning, 19(4), n4.

Heimerl, F., Lohmann, S., Lange, S., & Ertl, T. (2014,

January). Word cloud explorer: Text analytics based on

word clouds. In System Sciences (HICSS), 2014 47th

Hawaii International Conference on (pp. 1833-1842).

IEEE.

Kaptein, R., Hiemstra, D., & Kamps, J. (2010, January).

How different are language models and word clouds.

In Advances in Information Retrieval: 32nd European

Conference on IR Research (ECIR 2010) (Vol. 5993,

pp. 556-568).

McNaught, C., & Lam, P. 2010. Using Wordle as a

supplementary research tool. The qualitative

report, 15(3), 630-643.

Miley, F., & Read, A. 2011. Using Word Clouds to Develop

Proactive Learners. Journal of the Scholarship of

Teaching and Learning, 11 (2), 91-110.

Nevado-Maestre, A., Vega-Rodriguez, M. A., Perez, C. J.,

Calle-Alonso, F., & Sanchez-Gomez, J. M. 2017 Social

Network Analysis and Educational Data Mining for

NeuroK e-Learning Platform. Innovative and Creative

Education and Technology, 23-27.

Nickell, J. 2012. Word clouds in math

classrooms. Mathematics Teaching in the Middle

School, 17(9), 564-566.

Perry, L. 2012. Using word clouds to teach about speaking

style. Communication Teacher, 26(4), 220-223.

Porter, M.F. 1997. An Algorithm for Suffix Stripping.

Readings in Information Retrieval, 313-316.

Porter, M.F. 2001. Snowball: A language for Stemming

Algorithms.

http://www.snowball.tartarus.org/texts/introduction.ht

Reeves, T. C. 2003. Storms clouds on the digital education

horizon. Journal of Computing in Higher Education,

15(1), 3.

Yukselturk, E., Ozekes, S., & Türel, Y. K. (2014).

Predicting dropout student: an application of data

mining methods in an online education

program. European Journal of Open, Distance and E-

learning, 17(1), 118-133

Word Clouds as a Learning Analytic Tool for the Cooperative e-Learning Platform NeuroK

513