![](bg2.png)
the hospital, numbers of statistics of numerical data
amounts to over 1500. Since the authors has been
working on supporting the self-assessment reports of
Kyushu University, we recognize that the problem of
large numbers of statistics is crucial. Those statistics
in the context of university evaluations are acquired
by sorting and cross-tabulating tables of numerical
data. Persons in charge of assessment in each school
have to study such skills as cross-tabulating and SQL
language. Though these persons are not full-time as-
sessment staff, they must learn those skills in addtion
to jobs of self-assessment.
The purpose of self-assessment is not creating re-
ports, but reflecting activities of the university. To do
so, it is useful to visualize statistics of numerical data
in time series. Fundamental visualizations of numeri-
cal data are line or bar graphs, their combination, pie
chart, radar chart and so on. It is difficult to choose
generally which type of visualization is suitable for
each statistic.
In order to manage statistics and their visualiza-
tion, we propose a data warehouse with sharing skills
on statistics and visualization using mashup technol-
ogy. Data warehouse has become popular in insti-
tutional research. In our approach, mashup tech-
nology is introduced in data warehouse. By facil-
ity of mashup technology, persons in charge of self-
assessment can become familiar with statistics and
analyse from data warehouse.
The rest of paper consists as follows. In section
2, we will introduce related work and argue an issue
of data warehouse that enterprise data warehouse is
not suitable for institutional research. In section 3,
we propose a system sharing programs for statistics
and their visualization. In section 4, we conclude our
approach.
2 RELATED WORK
Enterprise data warehouse has two styles, Inmon’s
style (Inmon, 1999) and Kimball’s style (Kimball and
Ross, 2002). While Inmon’s style is a collection of
data tables in 3rd normal form, called data warehouse
and user interface of analysing tools, called data mart,
data warehouse of Kimball’s style forms a dimen-
sional data model from which data mart extracts nec-
essary data.
Considering problems mentioned in the introduc-
tion, common issue of both style data warehouse is
that users are not supposed to develop analysing tools
and visualizing skills. Staff in institutional research
should be analysts as well as they are respected to
develop data analysing programs. A feature of our
approach is that every user can contribute to devel-
opment of data warehouse tools without expertise on
programming.
Interaction between users and data warehouse
would be useful. Ferr´andez and Peral (2010) pro-
posed such interaction that data warehouse offers
question to users. Our approach focus on interaction
between users in order to share their tools and skills.
Mashup is a new concept to combined web ser-
vices and web databases with existing technologies
such as XML and HTTP communications. Swashup
DSL(Maximilien et al., 2007) and WMSL(Sabbouh
et al., 2007) are domain specific language (DSL) for
mashup, which do not assume programming skills
like JavaScript, CGI and so on. Integrated user in-
terfaces of mashup are proposed, for instance, Ya-
hoo! Pipes, Mashup Feeds (Tatemura et al., 2007)
and Damia (Altinel et al., 2007).
From viewpoint of simple and easy mashup pro-
gramming and visualizing mashup data, we proposed
and implemented a mashup programming environ-
ment (Mori et al., 2006) with GUI (Mori et al., 2007)
which combines web databases. In our study Web
databases are supposed to have Web API which of-
fers structured data with HTTP requests. The system
has four features. Firstly, every source of structured
data, including search engines and local CSV files
1
,
is defined an abstract I/O machine in the system, we
call them resource component. Secondly we general-
ize UNIX-like pipeline and filters on the system. In
the architecture of our system, UNIX-like filter oper-
ations like sorting and creating histogram are imple-
mented. Graph generation programs are also imple-
mented as kinds of filter operation as well. Thirdly,
the system provides a GUI programming environment
which helps users to create mashup programs intu-
itively and easily. Finally the system is served as a
multiuser content management system in which users
share mashup programs.
3 THE PROPOSED SYSTEM
Our proposing system consists of three parts.
Overview is shown in figure 6.
The first part is a data warehouse for institutional
research with Web API. The mashup programming
environment can include the data warehouse as a re-
source component. The second part is a GUI mashup
programming environment for visualization and its
execution environment. The GUI mashup program-
ming environment serves fundamental data opera-
1
User can upload CSV files as resource components
CSEDU 2011 - 3rd International Conference on Computer Supported Education
324