There are a number of word predictors on the market. However, they have, typically, been developed for English, which means that they are not well suited for morphologically rich languages such as Dutch, French, German and Swedish. Simply adapting the English programs to these highly inflecting languages by replacing the English dictionaries usually implies a massive reduction of keystroke saving rate. These effects are due to the simplistic language description that is used for predicting English text and that fails to predict the correct inflectional form of a word as required by inflectional languages. The English language description is, as a rule, limited to frequency data on individual words (unigrams) and sequences of words (bigrams, trigrams). Attempts have been made in research systems for Swedish and Spanish to use a more elaborate language description, including n-grams of word classes and syntactic grammars. The experiences made in these projects are taken into account in the FASTY project. They do not, however, present solutions that will ensure a keystroke saving rate of above 50% for the FASTY languages. An additional problem with most of the FASTY languages is the fact that new compounds can be created on the fly, thus making it hopeless to strive for a complete lexicon. Other methods need to be employed for coping with dynamic word formation processes. Being able to cope with compounds, even if they are new, is of great importance, since compounds are usually rather long words and failing to predict them can cause a significant drop of the keystroke saving rate. Since no word prediction system currently available is able to handle new compounds, this aspect of the FASTY system is a true innovation.
At a general level, the innovative aspects of the FASTY system are represented by
In particular, the innovative nature of the FASTY system is reflected in the following features: