Dutch verb spelling is a rather complicated system of rival orthographic pre-scriptions. The spelling of Dutch verbs not only depends on their pronuncia-tion, but also on syntax and semantics. Pupils are often incapable of learning the system, even after many hours of verb spelling instruction. Investigations lead to the insight that a rule-oriented algorithmic method is more effective than traditional methods. But pupils still make a lot of errors. In this study we investigate how computers can improve verb-spelling performance of pupils. A first way could be the use of a word processor with an built-in spelling checker. Spelling checkers find errors in typed text, but existing spelling checkers cannot correct Dutch verbs automatically. Artificially Intelligent spelling checkers will do (but will be expensive for instructional use). A second way of improving spelling performance is the use of tutorial programs. We discuss algorithm-based instructional programs that introduce and train the verb spelling system and its underlying concepts. As a third way to improve spelling performance we introduce the instructional program DT-DUIVELTJE (DT-DEMON). DT-DEMON can be characterized as a half-automatic spelling checker for Dutch verbs. In most cases it can find and correct verb spellings automatically. In the remaining cases DT-DEMON asks the user one or two questions, in order to complete the analysis. The core of the program is a simulation of a human expert in verb spelling. An expert is supposed to find correct orthographic prescriptions as a result of very economic decisions, i.e. based on an analysis of lower-order language regularities. For instance, when the spelling of a verb can be found by analyzing letter sequences, a syntactic analysis can be omitted. In this way an expert writes far more verbs correctly as a result of automatic pattern recognition. Although the internal inference mechanism of DT-DEMON is one of pat-tern recognition, the dialogue with the user is rule-oriented. The program also knows the rule-oriented algorithm for verb spelling. A final remark about the way DT-DEMON internally processes, concerns the number of analyses performed. The program does not just stop searching when it finds a solution, but continues until all solutions are found. The number of successful redundant analyses is an indication of the degree of difficulty of the spelling problem. A lot of solutions means: easy; just one solution means: a less simple problem; not even one complete, automatically found solution means: pay attention, this is a difficult problem. DT-DEMON is suitable for instructional purposes once a pupil has learned the rules that govern verb spelling. The program works like an online spelling checker. When there is a difficult spelling problem, and the program cannot find a complete solution itself, it will ask the student for help. By doing so, it prevents errors due to premature automatization of spelling skills. The use of DT-DEMON is not restricted to precooked exercises, it can handle free language productions.

The compatibility of didactic resources and the linguistic logic of spelling in Dutch verb spellingDespite years of dedicated education, a significant number of Dutch pupils leave primary school each year without mastering verb spelling. At this point, the spelling system appears to be a wolf in sheep’s clothing. The system underlying Dutch verb spelling is logical, but it violates the basic rules of Dutch spelling and leads to homophonic forms that have to be spelled differently. The effect of frequency and context increases the uncertainty on how to spell these verb forms. The latest research, by now already about thirty years old, indicated that verb spelling is learned best by whole-class teaching and by means of an algorithm. In this article we discuss the available didactic resources and the problems that students have to overcome when learning Dutch verb spelling. It provides us with a tentative answer as to whether didactic resources and the logic of Dutch verb spelling are compatible.

No abstract

Software systems convert between graphemes and phonemes using lexicon-based, rule-based or data-driven techniques. SHOTGUN combines these techniques in a hybrid system which converts between graphemes and phonemes bi-directionally, adds linguistic and educational information about the relationships between graphemes and phonemes and provides estimates about the likelihood that the generated output is correct. We describe the components from which SHOTGUN is built and determine its accuracy by running tests on two data sources, the BasisSpellingBank and CELEX, comparing the results to Nunn’s (1998) rule-based conversion system. SHOTGUN converts phonemes to graphemes and vice versa with precision of 81% and 86% when tested on the BasisSpellingBank, and 80% and 81% when tested on CELEX. SHOTGUN proves to be a powerful new conversion tool.