![](bg3.png)
includes a number of ready setups for chemical
experiments as well as a virtual lab for open
exploration. The JFLAP environment (Rodger,
2013) allows students to create, analyze and test
finite-state machines — the devices that constitute
the basis of computer science.
We consider such systems as great examples of
well-grounded uses of computer technology in
education. Virtual labs provide safe and controlled
environments in which students can test their ideas,
and in this sense they can be likened to flight
simulation software, used to train pilots: the students
perform predefined training routines, but also can
experience the outcome of any arbitrary maneuver.
Furthermore, virtual labs contribute to the modeling
of the problem domain in the learner’s mind, and
thus are consistent with constructivist views on
educational process.
It is interesting to note that from the
technological point of view, virtual labs are not
necessarily complex systems. The possibility of
open experimentation outweighs many technical
limitations and constraints.
Unfortunately, environments for open
experiments are barely provided by the existing
CALL systems. This perhaps can be attributed to the
unclarity of the notion of an “experiment” in
language learning. It is evident, however, that a large
portion of active language learning is related to the
process of combining words and phrases into
meaningful sentences, and the analysis of the
subsequent feedback. We learn a language both by
comprehending other people’s speech and writing,
and by creating our own phrases that are to be tested
for admissibility by our interlocutors.
Within such a concept of experiments, even a
feature-rich electronic dictionary can be a powerful
experimental tool in the hands of an avid learner.
Indeed, with full-text search it is possible to check
actual word use, test the correctness of certain word
combinations, the compatibility of certain prefixes
with certain stems, etc.
The ways in which students could do
“experiments with the language” are still to be
identified. Here we can only quickly introduce our
own work-in-progress system that is intended to help
language learners master basic grammatical rules.
5 TOWARDS WORDBRICKS
One of the most basic aims of language learning is
to train the ability to formulate grammatically
correct sentences with known words. Unfortunately,
traditional exercises lack active feedback
mechanisms: learners are unable to “play” with
language constructions to find out which word
combinations are admissible and which are not. The
best (and maybe the only) way to train active writing
skills is to write (essays, letters…), and to get the
writings checked by the instructor. Some intelligent
CALL systems, such as Robo-Sensei (Nagata,
2009), can assess students’ writings by using natural
language processing technologies, but the success of
these instruments is limited.
We suggest that active skills of sentence
composition can be improved by forming a
consistent model of language in the learner’s mind.
Metaphorically speaking, the difference between a
“consistent model” and a set of declarative grammar
rules in this context is the same as the difference
between a Lego construction kit and a lengthy
manual describing which Lego bricks can be
connected and in which ways. A child does not need
manuals to play Lego: the rules of brick linkage can
be easily inferred from brick shapes and with some
trial-and-error process. Unfortunately, there is no
such way to easily check whether it is correct to
combine certain words in a sentence.
The idea of modeling syntactic rules with shaped
bricks was implemented in the educational
programming environment Scratch (Resnick et al.,
2009). In Scratch, individual syntactic elements of a
computer program are represented with shaped
bricks that have to be combined to constitute a
program (Figure 1a). While Scratch code may have
logical errors, syntactically it is always correct, since
it is impossible to combine mismatching bricks.
Scratch’s graphical editor is not just a simpler
way to write computer programs, helpful for the
beginners. It can be treated as a construal (Gooding,
1990) that forms a model of a programming
language in the learner’s mind, though this aspect is
not explicitly emphasized in Scratch.
In our research, we are working towards
implementation of a similar scheme for natural
language sentences. Undoubtedly, natural language
grammar is much more complex and less formal
than the syntax of any programming language.
However, for the purposes of novice language
learners, it is reasonable to teach restricted grammar
(as it happens in traditional language teaching),
which is technologically feasible.
Even in the case of Scratch, the design of brick
linkage principles is not trivial. One important
problem is to make sure that the links between the
bricks reflect actual structure of the corresponding
computer program. For example, a loop control
CSEDU2014-6thInternationalConferenceonComputerSupportedEducation
406