=Paper=
{{Paper
|id=Vol-1601/CrossLAK16Paper12
|storemode=property
|title=Web-based Interactive and Visual Data Analysis for Ubiquitous Learning Analytics
|pdfUrl=https://ceur-ws.org/Vol-1601/CrossLAK16Paper12.pdf
|volume=Vol-1601
|authors=Benjamin Weyers,Christian Nowke,Torsten W. Kuhlen,Mouri Kousuke,Hiroaki Ogata
|dblpUrl=https://dblp.org/rec/conf/lak/WeyersNKMO16
}}
==Web-based Interactive and Visual Data Analysis for Ubiquitous Learning Analytics==
https://ceur-ws.org/Vol-1601/CrossLAK16Paper12.pdf
Web-based Interactive and Visual Data Analysis for Ubiquitous
Learning Analytics
Benjamin Weyers, RWTH Aachen University, Virtual Reality and Immersive Visualization Group,
weyers@vr.rwth-aachen.de
Christian Nowke, RWTH Aachen University, Virtual Reality and Immersive Visualization Group,
nowke@vr.rwth-aachen.de
Torsten W. Kuhlen, RWTH Aachen University, Virtual Reality and Immersive Visualization Group,
kuhlen@vr.rwth-aachen.de
Mouri Kousuke, Kyushu University, Faculty of Arts and Sciences, mourikousuke@gmail.com
Hiroaki Ogata, Kyushu University, Faculty of Arts and Sciences, ogata@artsci.kyushu-u.ac.jp
Abstract: Interactive visual data analysis is a well-established class of methods to gather
knowledge from raw and complex data. A broad variety of examples can be found in literature
presenting its applicability in various ways and different scientific domains. However, fully
fledged solutions for visual analysis addressing learning analytics are still rare. Therefore, this
paper will discuss visual and interactive data analysis for learning analytics by presenting best
practices followed by a discussion of a general architecture combining interactive
visualization employing the Information Seeking Mantra in conjunction with the paradigm of
coordinated multiple views. Finally, by presenting a use case for ubiquitous learning analytics
its applicability will be demonstrated with the focus on temporal and spatial relation of
learning data. The data is gathered from a ubiquitous learning scenario offering information
for students to identify learning partners and provides information to teachers enabling the
adaption of their learning material.
Keywords: interactive analysis; web-based visualization; learning analytics
Introduction
Interactive visual data analysis is a well-established class of methods to gather knowledge from raw and
complex data. Information visualization approaches and tools have shown high impact in various fields of
research. For instance, the use of visual data analysis enables high performance computing experts to analyze
code running on NUMA architectures (Weyers et al., 2014). In neuroscience, one major challenge is the analysis
and interpretation of heterogeneous data resulting from simulations as well as biological experiments. Tools
such as VisNEST provide coordinated multiple views which offer various perspectives on the data. Time-
varying data is presented in such a way that domain scientists are able to navigate as well as analyze it (Nowke
et al., 2013). From the data perspective, various types of visualization concepts can be found in literature, e.g.,
bar charts or pie charts (Spence, 2001) or representation concepts for relational data such as classic tables or
graphs (Battista et al., 1999).
In learning analytics, only a few works can be found which address the benefits of visualization:
Figure 1: Left: three-layered representation of ubiquitous learning logs as generated in SCROLL. Right: Web-
based visulaization architecture
65
Copyright © 2016 for this paper by its authors. Copying permitted for private and academic purposes.
presenting raw and complex data in an easy to understand and interpretable manner. In context of learning
dashboards, Verbert et al. (2013) present results how visualization can be used in learning analytics. Their work
concentrates on a learning analytics process model in which information visualization is used to enrich learning
dashboards to support students and teachers in their daily work. A similar approach has been presented by
Leony et al. (2012). They present a web-based tool called GLASS that is based on a four-layered architecture.
The architecture addresses the use of classic visualization components such as bar charts and the application of
filters for pre-processing data. However, both works do not focus on interactive data analysis and omit the
discussion of how to create suitable visualizations for learning analytics. The work of Liddo et al. (2011)
presents a successful use of visualization techniques for learning analytics by using graph visualizations to
represent the course of a discussion. Nonetheless, they constrain their work on a very specific use case. For
location-based and mobile learning, some works can be found such as this presented by Melero et al. (2015).
They present a set of visualizations for abstract information (including location) for a specific scenario, which
do not offer interactive analysis features but show a relevant set of visualizations that can be used in location-
based learning.
To our knowledge there is no work on the support of ubiquitous learning analytics by using interactive
visualization techniques. Hence, the main contribution of this paper consists in the presentation of an
architecture for ubiquitous learning analytics which is based on well-established concepts in information
visualization. The applicability of this architecture will be demonstrated by means of a use case for the
visualization of ubiquitous learning logs (ULLs). ULLs are gathered from learning scenarios where students use
ubiquitous devices, e.g., mobile devices to track learning progress in the wild or in the class room. Logging
generates spatiotemporal data that represents what the learner has acquired at which point in time (Ogata et al.,
2011). ULLs can be interpreted as a four dimensional space where the first three dimensions represent the
learner, the acquired knowledge, and the location. The fourth dimension represents time at which a student
created the log. Beside the creation of these logs, further aspects can be included into the analysis, such as how
often the ULL has been used to recall knowledge. In general, the first three dimensions can be interpreted and
visualized as a graphical structure such that these dimensions can be interpreted as three layers which are related
to each other as shown in Figure 1.
The paper is structured as follows. The next section presents an analysis of ULL and their specific
requirements regarding interactive data analysis tools and the role of users. This discussion is followed by the
presentation of an architecture for visual ubiquitous learning analytics which facilitates the information seeking
mantra and the concept of coordinated multiple views. A use case is discussed which presents the feasibility of
the architecture in real world scenarios. The paper concludes with a short summary and the discussion of future
work.
VISUAL DATA ANALYSIS FOR UBIQUITOUS LEARNING ANALYTICS
Ubiquitous learning data (ULD) in general and ULLs in specific present a special challenge for interactive
visual data analysis approaches and tools. First, such data is heterogeneous comprising spatial, temporal as well
as learning specific data (D1). For instance, ULLs contain a unique identifier for the student, data addressing the
time and place where the student acquired knowledge. Second, ULD datasets can be large due to the number of
users and stored data items (D2). Last, ULD has a tendency to be incomplete or to contain corrupted data by
faulty entries (D3).
Beside the data requirements, a visualization architecture (VA) has to consider the analysis tasks of the
user. These tasks are therefore user centric. This paper concentrates on two user roles: the teacher and the
student. A user who is a teacher has certain requirements for visual analysis such as to include the obtained
information contained in the ULD in the preparation of courses, extend and change course material or
specifically include the ubiquitous learning infrastructure into the course. The latter could address to use of the
system during the course or in between the course as well as for the definition of the final grade for a student.
The student has a slightly different requirement on visual analysis system. A student can be interested in
tracking achievements, in the learning progress or finding learning partners who have the same learning goals.
In addition, she could be interested in where to learn best or where to find relevant learning material in her
nearby surroundings. In summary, the following requirements can be identified:
V1: The VA has to offer an overview of the data as well as specific details in a certain context
depending on user roles
V2: The VA has to offer various perspectives on the data, which reflect different requirements on the
data depending on the user role
V3: The VA has to offer presentations which consider different combinations of data dimensions of
the underlying dataset to address the individual needs of a user
66
Most ubiquitous learning systems are implemented based on web technologies. To seamlessly integrate the
analysis with the ubiquitous use of such learning technologies, the VA should be also based on web
technologies leading to the last requirement:
V4: The VA should be implemented using web-based technologies
ULD Visualization Architecture
To address the identified requirements, the realization of a VA has to follow two methodical paradigms which
are already well-established in the visualization community: the information seeking mantra and the concept of
coordinated multiple views. The information seeking mantra introduced by Shneiderman (1996) is based on
interactive visualization which has been shown as successful for visual data analysis (Fuchs & Hauser, 2009).
Interactive visual data analysis is understood as the analysis of data using visualizations which are customizable
during runtime by the user. This customization can be defined by various types of manipulations such as the
application of filters, the selection of data items in a visualization, details of this selection, or the change of the
visualization technique used to display a dataset. The information seeking mantra proposes and specifies a
general workflow for the visual analysis process: overview first, zoom and filter, and details on demand. Thus,
as a first step, a visualization tool should present an overview of the dataset. Following this, the user should be
able to explore the dataset’s representation interactively by zooming into it, e.g., selecting a subset of data items,
and apply filters, e.g., restricting the datasets dimensions. An architecture realizing the information seeking
mantra addresses in particular D2 as well as V1 and V3.
Roberts (2007) presents an overview on coordinated multiple views. Coordinated multiple views “is a
specific exploratory visualization technique that enables users to explore their data. In fact, the overall premise
for the technique is that users understand their data better if they interact with the presented information and
view it through different representations” (Roberts, 2007, p. 1). This is specifically true when the interpretation
task benefits from different perspectives on the data by utilizing various visualization techniques and the data is
multi-dimensional or heterogeneous as it is the case with ULD. The coordinated multiple view paradigm
combines different types of visualizations of the same data with a coordination mechanism of views that react to
user’s interaction intents accordingly. For instance, this can be a selection of a data item in one view which is
then propagated to all coupled views displaying this subset in their perspective on the data (i.e., a coordination
mechanism termed brushing). Analogously, the zoom and filter step of the information seeking mantra can be
coordinated between views by applying the same zooming and filtering operation to all other views. A
visualization architecture implementing coordinated multiple views addresses the requirements D1, V2 as well
as V3. V1 is implicitly addressed because an overview of the data can be obtained by various views showing
different perspectives but in combination present the whole dataset at one glance. V4 is addressed by
implementing the visualization architecture as web-based components as presented in Figure 1, right. In a web-
based environment, the complete dataset must be accessible by the server. By a combination of a server for the
communication with the client-side visualization and the data pre-processing, a reduction of the data size can be
considered to make the communication more efficient. The server-side data preprocessing should be able to pre-
compute data structures, such as graph-based representations from a table-based dataset. The client-side
application should offer a coordination component that provides interfaces and communication logic for the
propagation of interaction events between the views. A controller view should be provided to offer a graphical
user interface for general control operations, e.g., triggering the loading of datasets and the application of client-
or server-side filtering operations.
Use Case – Learning Analytics for Ubiquitous Learning Logs
The applicability of the ULD visualization architecture will be shown by means of two perspectives on an
analysis use case: (a) a student tries to improve her learning workflow by analyzing her own ULLs and (b) a
teacher who would like to extend her learning material and activities. The gathered data originates from the
SCROLL system (Ogata et al., 2011), a system for gathering ubiquitous learning logs which collects words a
student observed in their everyday surrounding and which they learned for their individual vocabulary. The
current implementation consists of three major visualization designs: two force-directed-layout-based interactive
graph visualizations, a circular graph visualization using edge bundling, and a Google earth-based visualization
of the position the ULLs have been captured, thus showing in which spatial context students learned words (see
Figure 3, left). The presented visualizations are based on D3.js, a JavaScript library capable of visualizing data
in web-based applications. We used different open source extensions for D3 to build these visualizations. Figure
2 shows two graph-based visualizations which represent the relation between students according to similar
learned words. The left representation shows the words as light blue circles where students are represented in
dark blue. The right representations only shows the relation of one specific student (here Sophie) to other
67
Figure 2: Left: Overview of the dataset showing the relation of students (dark blue) to learned words (light
blue). Right: Graph visualization of the nearest neighbors of student “Sophie”.
Figure 3: Left: Google Earth based visualizatoin of the geographic position of ULLs. Right: A circular graph
view based on an edge bundling approach showing the relation between studens by means of learned words.
students and omits information on words they have in common. The latter information is also presented in
Figure 3 on the right. Here, all students are visible and can be interactively selected such that the relations of
students are highlighted. In the following, an informal description of use cases regarding the two user roles will
be discussed and the potential benefit of the ULD visualization architecture identified.
Use Case - Student’s Perspective
A student (Sophie) would like to find collaborators to learn new words in order to extend her vocabulary. Her
first step is to find other students which have some overlap in their already learned words. Therefore, she uses
the graph view presented in Figure 2, right. In a second step, she wants to find a student who knows as many
words as possible that Sophie does not know yet. As both visualizations can be assumed to be coordinated
linked, the selection of a student in the right view of Figure 2 is propagated to the left graph view. Sophie is now
able to see which person fits to the previously defined criteria by inspecting the number of words connected to
the selected student. Following this, Sophie inspects the Google earth representation which highlights the
selected student’s location she is interested in to learn with. This will inform her whether the matching student is
in a reachable distance. Finally, Sophie can identify the hot spots which students are visiting to learn
vocabularies. This process follows the Information Seeking Mantra as Sophie first gets an overview and then
selecting certain entities she is interested in. This can be interpreted as “zooming in” into the data set. Finally,
by checking the location information, details are included into Sophie’s analysis.
68
Use Case - Teacher’s Perspective
The teacher’s use case is in various points similar to this of the student’s but follows slightly different
objectives. A teacher could be interested in the interpretation and analysis of ULLs to extend or to adapt the
content and the pedagogical approach she is following in class. For instance, the teacher could define learning
groups for the classes along the same analysis process as described in the student use case above. On top, the
teacher can identify new vocabulary to be taught in class based on prior knowledge of students and identify
words unknown to most students. By zooming in and highlighting details on demand she is able to identify well-
known words and topics. Finally, she can use the circular graph visualization in Figure 3 (right) to identify
which students have comparable vocabulary skills.
Conclusion and Future Work
This paper explored the application of well-established visual analysis methods into the field of Ubiquitous
Learning Analytics. We proposed a web-based architecture for interactive and visual ubiquitous learning
analytics following two main concepts: the information seeking mantra and coordinated multiple views. By
means of a simple use case, the applicability of the architecture has been shown.
For future work, the implementation and evaluation of such a framework is planned. It is planned to
realize the proposed architecture in the context of e-book based instructions in various scenarios, such as
MOOCs. Furthermore, we plan to analyze further visualization approaches and techniques to offer various types
of analysis workflows in the future.
Acknowledgement
The research have been supported by “Research and Development on Fundamental and Utilization Technologies
for Social Big Data” (178A03), the Commissioned Research of the National Institute of Information and
Communications Technology (NICT), Japan.
References
Battista, D., Eades, P., Tollis, I. G., & Tamassia, R. (1999). Graph drawing: algorithms for the visualization of
graphs. Upper Saddle River: Prentiece Hall.
De Liddo, A., Shum, S. B., Quinto, I., Bachler, M., & Cannavacciuolo, L. (2011). Discourse-centric learning
analytics. In Proceedings of the 1st International Conference on Learning Analytics and Knowledge
(pp. 23-33). ACM.
Fuchs, R., & Hauser, H. (2009). Visualization of multivariate scientific data. Computer Graphics Forum, 28,
1670-1690.
Hänel, C., Weyers, B., Hentschel, B., & Kuhlen, T.W. (2016). Visual Quality Adjustment for Volume
Rendering in a Head-Tracked Virtual Environment. Transactions on Visualization and Computer
Graphics, in press.
Melero, J., Hernández‐Leo, D., Sun, J., Santos, P., & Blat, J. (2015). How was the activity? A visualization
support for a case of location‐based learning design. British Journal of Educational Technology, 46(2),
317-329.
Nowke, C., Schmidt, M., van Albada, S. J., Eppler, J. M., Bakker, R., Diesmann, M., Hentschel, B., & Kuhlen,
T. (2013). VisNEST—Interactive analysis of neural activity data. In Proc. of IEEE Symposium
on Biological Data Visualization (pp. 65-72). IEEE.
Ogata, H., Li, M., Hou, B., Uosaki, N., El-Bishouty, M.M., & Yano, Y. (2011). SCROLL: Supporting to share
and reuse ubiquitous learning log in the context of language learning. Research and Practice in
Technology Enhanced Learning, 6(2), 69-82.
Roberts, J. C. (2007). State of the art: Coordinated and multiple views in exploratory visualization. In
Coordinated and Multiple Views in Exploratory Visualization, (pp. 61-71). New York, IEEE.
Shneiderman, B. (1996). The eyes have it: A task by data type taxonomy for information visualizations. In
Visual Languages, 1996. In Proceedings of IEEE Symposium on Visual Languages (pp. 336-343). New
York, IEEE press.
Spence, R. (2001). Information visualization (Vol. 1). New York: Addison-Wesley.
Verbert, K., Duval, E., Klerkx, J., Govaerts, S., & Santos, J. L. (2013). Learning analytics dashboard
applications. American Behavioral Scientist,.
Weyers, B., Terboven, C., Schmidl, D., Herber, J., Kuhlen, T. W., Müller, M. S., & Hentschel, B. (2014).
Visualization of memory access behavior on hierarchical NUMA architectures. In Proceedings of the
First Workshop on Visual Performance Analysis (pp. 42–49). New York: IEEE.
69