Abstract: A look at "Tutoring Systems" reveals the main areas of applying CL: analysing learner input. But there are other areas as well.
In traditional CALL-systems there is a certain range of possible exercises as mentioned above.
These are restricted in the sense, that the pattern of solving the exercises is always very similar.
In most cases the learner either selects an item from a list or the context in the exercise is limited in order to allow for only one sensible answer.
This is the major disadvantage since the didactics call for free production of language to achieve communicative competence.
Forcing the learner to use his personal knowledge of language has a much bigger impact on the learning process than choosing from a set of answers.
Even though most exercises are intended to train the use of language in communicative situations, the exercises with more or less predefined answers actually used are integrated in so called drill-and-kill-exercises about grammatical or lexical phenomena.
Some systems also provide a kind of guidance through the material of the system.
In traditional programs there is a static system which is only able to recognize solved or unsolved exercises in order to provide guidance.
This guidance is also part of the tutorial component of an intelligent program.
The "tutor" suggests a certain way of working on the units and exercises, so that at the end the learner has covered all material following the given path.
Usually this path is build following some pedagogical goals.
These two areas - exercise design and student modelling - are the main parts for improvement with the help of methods of CL.
Exercises:
So far only pattern-matching with prestored examples is implemented in commercial programs and the content is usually static.
Some kind of linguistic analysis can improve the feedback to the exercises.
Nowadays this is possibly restricted to a orthographic or morpho-syntactic analysis.
Further analyses in the semantic field would not yield good enough results for broad usage.
The area of error analysis is further explored in the chapter
» Error-Analysis.
Guidance:
If some kind of learner-modelling is implemented the program can guide the learner according to his personal needs.
Not only the performance in working on exercises can be measured but also the behaviour using the system like the frequency of using help-pages etc.
In order to develop an intelligent tutoring system certain program-modules have to be realized as can be seen in the figure » ITS-Modules.
Most of these are necessary to provide adequate tutoring.
More on this is explained in the chapters » Intelligent Tutoring Systems and » Learner Modelling.
Based of this the help-system could also be improved.
According to the performance of the learner, the system can provide tailored help-messages either in reference to the current exercise or to the handling of the program.