Example System II: FLUENT

The FLUENT-program mainly consisted of a microworld, in which the student could act and the "tutor", i.e. a language generation module would respond to these actions. In a second type of exercise the tutor could command the student to do things in the microworld and learner was expected to act accordingly. The system therefore resembled the ALLP-System with the "Poltergeist"-scenario in a certain sense. Most of this text was taken from Hamburger, Henry in: Appelo, L. ; de Jong, F. (Ed.), 1994 and Hamburger, Henry in: Holland, V.M. ; Kaplan, J.D. et al. (Ed.), 1995. The microworld consisted of a kitchen scene with a sink, a stove, and several cabinets as can be seen below. The aim of the system FLUENT-1 was to provide a connection between graphical action (in a microworld) and language output of the system. This was done in order to immerse the learner into a scenario, in which only the target language was visible and available.

The following figure is a scan from Hamburger, Henry in: Appelo, L. ; de Jong, F. (Ed.), 1994. It shows a graphical hand (above the sink) which could be moved by the learner to do various things in the kitchen such as filling a cup or putting a pot on the stove.

The NLP-component of FLUENT-II is the same as the one used in the ALLP-prject. According to Hamburger, Henry in: Appelo, L. ; de Jong, F. (Ed.), 1994, it was specifically adapted to be used in the FLUENT-concept. Therefore not only a language understanding module but also a generation module was included.

One important aspect is the recognition of different views on a single action or scene in a computer-program. The picking up of the pot in order to move it can be viewed from very different perspectives, which are verbalized differently. The modelling of these views allows for a great variety of verbalizations of the same facts. Additionally Hamburger claims, that this achieves a tight integration, but not interdependence, of graphics and language. The actual verbalization by the system depends on the state of the microworld in combination with the learner-model.

It seems however, that this system has not been evaluated in a true language learning setting. Therefore it is difficult to say whether the advantages of the system actually hold.