Grammar checkers are not reliable enough yet to be used in CALL programs. Their rate of overflagging is still much too high for learners of a foreign language. CALL programs, however, could profit greatly from the incorporation of a grammar checker if this could provide the option of responding more intelligently to students’ input. Several strategies are proposed to make grammar checkers more acceptable for CALL developers, among them adapting the lexicon and the parser in the grammar checker.

This paper presents IDAZKIDE, a prototype of an intelligent language learning environment (ICALL) for learners of Basque. The philosophy of the system is to make different Natural Language Processing tools simultaneously accessible to students to help them (mainly at the morphological level) to write in Basque, as well as to give advice, taking into account some characteristics of the student gathered in a student model.

A diagnostic component for natural language utterances has been devised which allows one to integrate (possibly inconsistent) evidence from a wide range of knowledge sources into a unique decision procedure. It is based on a procedure for the disambiguation of dependency structures using graded constraints. Different description levels like syntax, semantics, and others are treated by separate structural representations, which are disambiguated simultaneously. Due to the complete symmetry of the system architecture the approach allows for the successful handling of a wide variety or student errors.

In this paper, we present the Web-based CALL environment (or WALL) which is currently being experimented with at the University of Paris 13 in the Computer Science Department of the Institut Universitaire de Technologie. Our environment is being developed to teach computer science (CS) English to CS French-speaking students, and will be extended to other languages for specific purposes such as, for example, English or French for banking, law, economics or medicine, where on-line resources are available.

English, and more precisely CS English is, for our students, a necessary tool, and not an object of study. The learning activities must therefore stimulate the students' interest and reflection about language phenomena. Our pedagogical objective, relying on research acquisition (Wokusch 1997) consists in linking various texts together with other documents, such as different types of dictionaries or other types of texts, so that knowledge can be acquired using various appropriate contexts.

Language teachers are not supposed to be experts in fields such as computer sciences or economics. We aim at helping them to make use of the authentic documents that are related to the subject area in which they teach English. As shown in Foucou and Kübler (1998) the wide range of resources available on the Web can be processed to obtain corpora, i.e. teaching material. Our Web-based environment therefore provides teachers with a series of tools which enable them to access information about the selected specialist subject, select appropriate specialised texts, produce various types of learning activities and evaluate students' progress.

Commonly used textbooks Tor specialised English offer a wide range of learning activities, but they are based on documents that very quickly become obsolete, and that are sometimes widely modified. Moreover, they are not adaptable to the various levels of language of the students. From the students' point of view, working on obsolete texts that are either too easy or too difficult can quickly become demotivating, not to say boring.

In the next section, we present the general architecture of the teaching/learning environment; the method of accessing and using it, for teachers as well as for students, is then described. The following section deals with the actual production of exercises and their limits. We conclude and present some possible research directions.

This paper makes a case for a style of CALL in which learners engage in written ‘conversations’ with a computer, implemented by means of natural language processing techniques. These conversations involve using language as a tool for achieving goals through active communication rather than as the subject matter for language exercises. The paper goes on to discuss the design and implementation of prototype software developed to illustrate this approach, and reports on pilot observations of the software in use. However, many issues in the design and use of this approach remain to be clarified.

This paper presents the prototype of CALLex, a program for learning lexical functions, created by a project funded by INTAS, a European organisation promoting cooperation between the European Union and the states of the former Soviet Union, developed by the Laboratory of Computational Linguistics at the Russian Academy of Sciences, Moscow, supported by the universities of Klagenfurt, Austria and Heidelberg, Germany. The goal of the program is to facilitate language learning through several linguistic games in order to improve the lexical command of the language studied. The CALLex games access a lexico-semantic database consisting of two dictionaries, Russian and German, each containing roughly 1000 lexemes. The lexical functions cover a wide variety of lexical relationships, which can be roughly divided into three major groups: (1) collocations, which are syntagmatic relationships, such as ‘do x, nave x, or being in the state of x’, (trade) = conduct (trade) or (anger) = feel (anger), (2) substitutions, i.e. paradigmatic relationships, e.g. ‘a lexeme whose meaning is opposite to x’, (appear) = disappear, (courage) = cowardice, and (3) other prototypical relationships, ‘head of what is denoted by x’, (university) = rector, (tribe) = chief. While studying the combinatorial capabilities of a word and its most ‘idiomatic’ collocations, the student can get a feel for semantic fields and obtain structured access to the vocabulary and its syntactic expression in the foreign language. The strict separation of the CALL program and the underlying database facilitates the expansion of the linguistic resources on the one hand and the adaptation or the linguistic games to new didactic approaches on the other hand. This paper highlights the function of the database in the background of the program and the treatment of illrormed student input. Although some adjustments were made during the course of the project, a more flexible approach seems necessary. Here we envisage a component separate from, but interacting with the database, allowing for a more robust treatment of ill-formed input.