NLP-based scripting for CALL activities 
Antoniadis G., Echinard S., Kraif O., Lebarbé T., Loiseau M., Ponton C. 
LIDILEM, Stendhal University  
Grenoble, France, F-38025 
{Antoniadis; echinard; kraif; lebarbe; loiseau; ponton}@u-grenoble3.fr 
Abstract 
This article focuses on the development of 
Natural Language Processing (NLP) tools for 
Computer Assisted Language Learning 
(CALL). After identifying the inherent 
limitations of NLP-free tools, we describe the 
general framework of Mirto, an NLP-based 
authoring platform under construction in our 
laboratory, and organized into four distinct 
layers: functions, scripts, activities and 
scenarios. Through several examples, we 
explain how Mirto's architecture allows to 
implement state-of-the-art NLP functions, 
integrate them into easily handled scripts in 
order to create, without computing skills, 
didactic activities that could be recorded in 
more complex sequences or scenarios. 
1 CALL:  Conjugating NLP and language 
didactics 
It is generally reckoned that computer science 
can prove itself to be a great aid in language 
learning, when in fact, most often computer 
scientists and didactics experts do not agree on the 
notion of “language”. For the former, it 
corresponds to a sequence of codes, while for the 
latter it is a system of forms and concepts. 
This divergence can easily be explained, when 
considering the fact that computer science, by 
definition, can only consider and process the form 
of the language independently of any 
interpretation, while, for language didactics, the 
form only exists through its properties and the 
concepts it is supposed to represent. 
The consequences of these diverging approaches 
are “visible” in the great majority of language 
learning software. Many an imperfection of the 
latter’s stem from the divergence mentioned above. 
Most language learning software are thought and 
implemented as computer products, only able to 
take into account a language form deprived of all 
semantics, or with extremely poor semantics. 
Caricaturely, rules as basic as that of the 
interpretation of the space remain ignored, which 
leads to unfortunate learning situations. For 
instance, if the learner answers “la   casa” 
(sequence containing two spaces), his or her 
answer will not be accepted for the expected 
answer was “la casa” (sequence with one space). 
The pedagogical consequences of this poor “space 
processing” are obvious; the software teaches that 
the sequence of two spaces is not part of the 
language, and also, that all word preceded or 
followed by a space has nothing in common with 
the same word without the space! This down-to-
earth example of the “spacebar syndrome” 
characterizes, in our opinion, the deficiencies of 
today’s language learning software. 
As (Chanier, 1998) and (Brun & al., 2002) point 
it out, and as (Antoniadis & Ponton, 2002) and 
(Antoniadis, 2004) have shown it, only the use of 
NLP methods and techniques allows to consider 
and process language as a system of forms and 
concepts. Considering them might lead to answers 
for two of the issues of existent CALL software. 
The first concerns the rigidity of software: the 
data (instructions, examples, expected answers…) 
is to be predefined and, a few exceptions aside, can 
neither be modified nor enriched. Answer handling 
processes are intimately connected to this data. 
They are thus unable to consider new entries, 
unless they were explicitly anticipated. 
The second problem concerns the inability of 
CALL software to adapt the course to the learners. 
Two types of courses are generally proposed. The 
first, the more classic, offers a predefined linear 
activity sequence. Whatever his (or her) answers 
and expectations, the learner will do (and do over) 
the same activities, using the same data. The 
second type of course offered is a “free” 
progression within a scenarized environment. It is 
the case of exploration software in which the 
learner is given a mission in a given environment 
(virtual reality). The dialogue, grammar or other 
activities are predefined, but will be performed in 
an order which will depend on the learner’s 
mission completion process. This latter type of 
course, despite allowing a wider field of action for 
the learner (order of the mission, choice of 
activities…) does not offer real personalization or 
adaptation of the activities to the learner. Indeed, 
the course of action is independent of his or her 
answers for each stage, out of the incapacity of 
evaluating them. Last, we should bring to the 
reader’s attention that if the order in which the 
learner is confronted to the activities can vary 
according to his (or her) mission, the content of 
each activity remains invariable and will remain 
the same, whenever included in the course. 
The last problem, which partly derives from the 
first two, characterizes current CALL software. As 
didactic products, this software should, a priori, be 
solely designed according to didactic solutions, 
expressed without constraints using pedagogical 
concepts. Now, current learning software are in 
fact computer products which require their users 
(language teachers, with little or no computing 
knowledge) to manipulate concepts and notions, 
which, a priori, do not belong to their language 
learning set of problems. Thus, instead of 
expressing pedagogic answers thanks to tools of 
their own discipline, they are forced to look for 
computerized solutions, which connect as much as 
possible with their own models or pedagogic aims. 
They might even have to give up on some 
pedagogical solutions, for they are unable to 
express them in a computer understandable way or 
because computer science is not able to handle 
them. To our knowledge, language didactics is 
presently able to imagine open pedagogic scenarios 
with exercises adapted according to each learner, 
examples changing when repeating the same 
activity within a given session, appropriate texts 
chosen to illustrate pedagogical contexts and, open 
and variable learning situations… Computer 
science is (and will be) unable to take into 
consideration these aspects with its own set of 
problems. Resorting to other knowledge 
(linguistics and language didactics) and to their 
modeling is essential. The use of NLP tools can 
constitute a way to resort to linguistic knowledge; 
the collaborative work of language didactics and 
NLP experts ought to provide answers concerning 
language didactics knowledge. 
The problems that we have just presented 
explain, in our opinion, about the nature of 
language learning software so far. They were 
thought and implemented as computing problems 
and products which only use the aspects of 
language didactics that computer science is able to 
consider. The pedagogical solutions are often 
altered or truncated so that they can be computed. 
This approach), and also most of CALL software 
deficiencies, stem from computer science’s narrow 
view of language (a simple sequence of codes. 
Our approach towards the development of 
language learning software is radically different 
from those mentioned above. We consider that 
language learning software is above all a didactic 
product, a program which provides a didactic 
solution to a problem of language didactics, 
without altering, neither the solution nor, a fortiori, 
the problem. The design of such software requires 
that we should be able to adapt the possibilities of 
computer science to the implementation of 
pedagogical solution previously determined. In this 
approach, considering language properties, which 
are invariably present in every pedagogic solution 
concerning languages, is a must-have. Considering 
NLP methods, techniques and products only are 
capable of satisfying this condition, then a 
language learning software should be defined as 
the adaptation of NLP possibilities to the 
predefined didactic aims of language learning. In 
our opinion, such an approach is the only way to 
offer to language didactics experts not only tools 
that would not narrow the scope of treatment of 
their set of problems, but also tools with 
pedagogical added-value, capable of widening the 
set of problems of their discipline. 
The use of NLP in the design of CALL software is 
not a new idea; systems like ELEONORE (Renié, 
1995), ALEXIA (Chanier & Selva, 2000), or the 
EXILLS platform (Brun & al., 2002) resort to NLP 
methods and use NLP resources. Nevertheless, 
such examples remain marginal and concern non 
commercial products. Paradoxically, CALL and 
NLP, two fields centered on language, still seem to 
be ignoring each other. Most of the time, not using 
NLP is justified through the added cost resulting 
from its use. But more than the often-invoked extra 
cost, it is the lack of NLP culture, which should be 
held responsible for its absence. 
In the line of the systems mentioned above, the 
Mirto platform (Antoniadis & Ponton, 2004) 
(Forestier, 2002) is aiming at providing a global 
answer to the problems of CALL software, through 
an NLP approach on the one hand and on the other 
hand a collaborative work with didactics experts. 
More than a finished product, Mirto seeks to be a 
tool for the creation of didactic solutions for 
language learning. We present in the rest of the 
paper the aspects of the system, which describe our 
approach and its implementation. 
2 Mirto description 
The Mirto project is determinedly 
pluridisciplinary, and aims at giving an NLP 
toolbox to language teachers in order to design 
scenarios in their own pedagogical set of problems. 
The main goal of Mirto is to propose to the 
language teacher the possibility of designing 
pedagogical scenarios while fully taking advantage 
of NLP technologies in a user-friendly manner. 
Thus, those scenarios will be open (dynamical text 
database), will allow an individualized adaptation 
according to the learner (automated generation of 
exercises, qualitative evaluation of the answers…) 
and should allow new possibilities (work on long 
texts, automated production of aids or exercises, 
design of non-linear scenarios …). The approach 
of Mirto is determinedly user-oriented since it is 
meant for language teachers who, a priori have 
little or no skill in computing nor in NLP. The 
technical nature of NLP has to be transparent to the 
language teacher and only the didactic aspects are 
to be visible and available to him. 
In that way, four hierarchical levels (function, 
script, activity and scenario), associated with the 
text database, structure Mirto as it is illustrated on 
fig.1. 
 
2.1 Function level 
The functions (1 to 5 in fig.1) represent the 
Mirto lower level objects. They correspond to a 
basic NLP process such as tokenization (text 
splitting in forms) or language identification. 
Considering its technical nature and its 
independence from a didactic application, this 
level is not visible for any final users of Mirto (i.e. 
teachers and learners).  
2.2 Script level 
This level corresponds to the application of NLP 
functions to language didactics. A script (S1 to S3 
in fig. 1) is a series of functions with a didactic 
purpose. So, this level needs both NLP and 
didactical competences and its design will be the 
result of an interdisciplinary work. For instance, 
the automated design of a gap-filling exercise is 
considered as a script because it connects the 
functions of language identification, tokenization, 
morphological analysis and gap creation depending 
on parameters chosen by the user. 
2.3 Activity level 
This level with the next one (scenario level) is 
the didactic core of Mirto. An activity (A1 to A4 in 
fig. 1) corresponds to the didactic contextualization 
of a script (previous level). Its goal is to associate a 
script with a text from the corpus database, some 
instructions, possible aids and an optional 
evaluation system. In order to create a gap-filling 
exercise, one only has to choose to apply the script 
of the previous example to a text while specifying 
the gaps criteria (for instance, hiding the preterit 
verbs and replacing them by their infinitive form), 
associating an instruction as “Fill in the blank with 
the preterit form” and specifying the evaluation 
form of the activity. 
2.4 Scenario level 
This level allows the teachers to define the 
sequence of activities in order to answer to their 
pedagogical objectives throughout the learner 
progression. This expected progression is not the 
same for each learner. Effectively, each of them 
will have a personal learning process linked to 
different factors. Mirto is dealing with that reality 
while proposing non-linear scenario creation. The 
path through the scenario depends on the 
individual process of each learner (learning course, 
evaluation…). That course is stored in a learners’ 
tracing database. For instance, according to his 
progress in a given scenario, a learner can be 
redirected to remediation activities, or retry an 
activity on another text or simply advance in the 
scenario. 
2.5 Levels and users 
There are three kinds of users in Mirto: NLP 
specialists, language specialists (didactic experts, 
linguists and teachers) and students. The following 
table shows the intervention level of each user of 
Mirto.  
 
Level Use User 
Function Conception NLP specialist 
Script Conception NLP specialist + 
Language specialist 
Activity Conception Language teacher 
Conception Language teacher Scenario 
Playing Student 
Tab.1 – The intervention level of each user 
This article deals more precisely with the 
NLP/CALL meeting, which takes place in the 
« script » level. However, before exposing the set 
of problems of script designing, it is necessary to 
stress on the activity level, which uses that script 
level first. 
3 Activity design 
An activity is the implementation of a precise 
minimal pedagogical aim (for instance, having a 
work on a grammatical notion, revising 
conjugations, writing a paragraph, etc.). Activities 
are designed by language teachers through a 
specific interface: the activity editor. The activity 
editor (cf. Fig.2) is an authoring system. It allows 
to manipulate and format pedagogical objects such 
as texts (or text corpora), scripts and instructions. 
In order to illustrate the steps of activities 
design, let us give the example of a teacher who 
wants to create an activity for the systematic 
revision of the preterit, using a gap filling exercise. 
The design work is then broken up into five 
steps (cf. Fig.2). The first consists in selecting a 
script in the toolbox, which allows him to generate 
a gap filling exercise. The second is the definition 
of a didactic context for the script application. This 
script setting operation allows the teacher to select 
elements from a text base and determine the 
elements (criteria on the form, the category or/and 
morpho-syntactical features). These first two steps 
produce the desired gap-filling exercise, which will 
be integrated into the activity. Before the effective 
production of the activity, three steps remain: 
writing the instructions, precising the aids, which 
will be given to the learner, and finally specifying 
the evaluation criteria. 
 
 
4 CALL/NLP scripts 
The script level represents the computing side of 
the didactic tools available in the Mirto 
environment. Scripts are integrated modules that 
implement one or several NLP standard resources 
and processes such as tagging, stemming, 
lemmatizing, parsing, dictionaries, etc. The 
standardization of these functions is an important 
aspect, because Mirto does not aim at developing 
new NLP techniques, but only at giving a 
framework to take advantage of the existing state 
of the art: Mirto is a car running with a NLP 
engine, and the engine may be changed, as a 
simple interchangeable part, if a new engine allows 
to get better performance. 
Thus, scripts are the core of Mirto's architecture: 
their design should allow to transform the engine 
kinetic energy into movement and direction on the 
road of didactic activities, without requiring that 
the driver to have mechanic skills. 
4.1 Parameters 
As any computing module, a script will be 
directed by a set of parameters. These parameters 
shall not be accessible to the end-user directly, but 
through a control panel. This control panel shall be 
relevant from the didactic point of view; that is 
why the controls may be transcribed into a set of 
parameters. Let us take the example of the gap-
filling exercise generator. By the mean of a simple 
form, the user may define: 
a) which the units are to be removed from the 
text. Any linguistic feature should be used for this 
definition: lemma (e.g. to drive), part-of-speech 
(ex. verb), morphosyntactic description (ex. past 
tense), or even meaning (e.g. "car" semantic field - 
this functionality has not been implemented yet). 
b) what information has to be given in the gap : 
none, the lemma, the morphosyntactic features, a 
synonym, a definition (not implemented yet) etc. 
c) if the removed words should appear or not as 
an ordered list in the text header. 
d) if the learner's answer should initiate a feed-
back process immediately after it was entered. 
On the user interface, the controls have to be: 
- simple: two many features could discourage the 
user  
- declarative: the user is not supposed to handle a 
tough formal language, so the control definition 
has to be intuitive and immediately 
understandable. 
- user-friendly: the interface must allow to pick 
out the important information. For instance, a first 
form may present the standard settings for a 
control, and a second optional form may give 
access to advanced settings of the generator. 
It is clear that the definition of linguistic features 
in a) involves a simple transcription process in 
order to determine the script parameters: the 
tagged and lemmatized texts handled by the 
generator use specific codes for morphosyntactic 
description. Declarative features as "Verbo, Prima 
coniugazione, Indicativo, Presente, Prima persona, 
Singolare" will be transcribed into a parameter set: 
"base=er$", ctag="verb", msd="IndP SG P1". 
Even if this transcription process appears to be 
unavoidable, the script design must render the 
Selection 
criterion 
Script 
type 
Example of 
activity 
Expected answer Involved NLP 
functions 
Semantic lexical 
spotting 
Spot every word 
related to the "car" 
topic 
Spotting of "drive", 
"taxi", "engine", "road", 
etc. 
morphosyntactic tagging, 
lemmatization, semantic 
net interrogation 
Semantic  lexical 
question 
Give an Italian 
translation for "to 
drive" 
Entering of "guidare" morphosyntactic tagging, 
lemmatization, bilingual 
dictionary interrogation 
Morpho-
syntactic 
gap-
filling 
Replace every 
infinitive verb in the 
gaps, using the 
appropriate tense 
Replacement of "to wait" 
by "have been waiting"... 
morphosyntactic tagging, 
lemmatization 
Morpho-
syntactic 
lexical 
question 
What would be the 
contrary of the 
adverb "lentement"? 
Entering of "rapidement" morphosyntactic tagging, 
lemmatization, semantic 
net interrogation 
Morphologic
al 
lexical 
spotting 
Spot every word 
derived from the 
verb "traduire" 
Spotting of "traducteur", 
"traduction", 
"retraduite", etc. 
morphosyntactic tagging, 
lemmatization, stemming 
Morphologic
al 
gap-
filling 
Fill every gap by a 
word of the 
"traduire" verb 
family 
Entering of "traducteur", 
"traduction", 
"retraduite", etc. 
morphosyntactic tagging, 
lemmatization, stemming 
Tab.2 - Example of scripting for activity generation 
parameters as close as possible to the user's 
control. 
4.2 Incremental approach 
It is impossible to determine from scratch what 
the exact form of a script must be. There are two 
reasons for this uncertainty:  
- NLP functions are multifaceted, they may 
require complex sets of parameters to give an 
expected result, and the form of their input and 
output may have many different forms. 
- the application field of NLP for a didactic use 
has been so far unexplored. New activities, new 
pedagogical habits, and new teachings are likely to 
emerge from these new technologies.  
We strongly claim that only the pedagogical 
practice can pave the way. 
Thus, designing the script, one may offer 
complex functionalities without real interest. Other 
scripts may appear to be very useful in some 
applications for which they were not initially 
designed. What we propose is to combine both top-
down and bottom-up approaches: the proposed 
tools may offer wide possibilities, among which 
the pedagogical practice may select a few 
interesting features. Conversely, the practice may 
give rise to new needs that the technology will try 
to meet. 
As suggested by (Kraif, 2003), to initiate the 
incremental process of script designing, we have 
chosen existing activities that may take advantage 
of simple improvements from NLP techniques. For 
these activities, we have tried to define scripts with 
a major modularity, i.e. scripts that may be 
reusable in different contexts and for a large 
spectrum of didactic applications. At last, another 
important criterion was given by the performances 
and limitations of the implemented functions: 
when a NLP task yields a 20% error rate, the 
results may not be valid for every kind of activity: 
erroneous information may be very confusing for a 
learner. 
4.3 Examples of scripting 
Most of the following examples are not 
implemented in the Mirto platform yet: but they 
are all realistic, given the current NLP state of the 
art, and may be added to Mirto in the short term. 
The scripts fall into three categories 
4.3.1 Activity generators 
Given an input text, NLP techniques allow to 
select lexical units and expressions that bear 
specific lexical, idiomatic, grammatical or 
semantic features. This ability makes it possible to 
create a wide range of activities using generators 
for gap-filling, lexical spotting (i.e. identification 
of specific units of the text) or lexical questions 
(i.e. questions about units occurring in the text, 
concerning synonyms, contraries, translation 
equivalents, etc.). Table 2 shows various examples 
of generated activities. 
 
Other scripts can be used upstream for the input 
text constitution: for instance a concordancer 
allows to extract from a corpus every unit (and the 
surrounding context) that satisfies the former 
selection criteria. 
Such a concordance script, integrated with an 
appropriate interface, may give rise to a full 
activity, in order to allow the learner to search by 
him/herself examples (in context) that may help 
him/her solve a problem. A bilingual concordance 
script, involving an NLP aligning function, may 
also be very useful for this kind of text mining. 
 
Similar activity generators may work without 
any input text, applying the selection criteria on a 
dictionary, and using, if necessary, a random draw 
to select a single unit:  
- Conjugator: e.g. "Conjugate the expression 
tomber en panne sèche to : subjonctif imparfait, 
première personne du singulier" 
- Lexical question: e. g. "Give a synonym for the 
word phare."  
- Morphological question: e.g. "Give a noun 
derived from the verb conduire". 
 
Another interesting application of NLP 
technique for activity generation is to implement a 
kind of "chat-bot", following the classical model of 
Elisa, able to simulate a conversation with a virtual 
interlocutor on a given subject. 
4.3.2 Comprehension aid 
For any kind of activity (reading a text, doing an 
exercise, etc.), it is possible to propose interactive 
aids for the learner. Most of the NLP tools 
available on the Exills platform belong to this 
category: at any time the learner can ask questions 
to a robot, that gives access to dictionary 
definitions (after a context sensitive 
disambiguation) and to a conjugator, or allows to 
find the correct form of a wrongly spelt word.  
 
Such aids can be either generic (like dictionary 
or concordance consultation, grammar lessons, 
conjugator, phonetiser) or context dependent (a 
click on a word can give access to its 
morphological tags, lemma, syntactic function, 
definition and/or translation). As an example, we 
have implemented a contextual aid that 
automatically links to specific grammar lessons 
according to the morphosyntactic features of the 
clicked word: when an Italian verb is at the 
"passato remoto" tense, a hyperlink is 
automatically pasted in the contextual popup, 
giving access to the corresponding grammar lesson 
(see fig. 3). 
 
 
Fig.3 - Example of generated contextual aid 
For the teacher, the handling of these scripts 
corresponds to specific settings of the final activity 
interface.  
 
4.3.3 Automated evaluation 
The learner production, in the framework of an 
activity, may have very various forms: clicks on 
check box, words, sentences or even texts.  
The evaluation of sentences and texts is a tough 
problem: NLP techniques cannot really give 
reliable information about features that require a 
human interpretation (meaning, style, etc.). Even 
for the simplest task of error detection, the existing 
models are both silent and noisy at the same time: 
some errors are not detected, and correct 
expressions are wrongly pointed out as errors. 
On the opposite, the evaluation of a multiple 
choice questionnaire is a trivial problem that does 
not need the expensive implementation of NLP 
tools. 
For now, we think that the most realistic and 
promising application concerns the evaluation of 
simple lexical productions. We are currently 
studying a three levels protocol for the evaluation 
of a given answer with respect to the expected 
correct answer. If the given answer is different, 
three cases are considered:  
1- Spelling error: if the entered chain does not 
exist in an inflected form dictionary, one can 
assume that it bears a spelling error. If the chain is 
very close to the correct answer, a message can be 
displayed, warning about the spelling error. Else, a 
list of resembling existing words can be proposed 
to the learner, asking him to make a choice. 
2- Morphosyntactic level: at this stage, the 
answer is integrated in the linguistic context of the 
activity (for instance, the sentence where the gap 
was done, in a gap-filling exercise), in order to 
compute a morphosyntactic analysis with tagging 
and lemmatization. If the lemma is the same than 
the lemma of the correct answer, a warning can be 
displayed about the difference in the 
morphosyntactic features (e.g. "wrong tense", 
"wrong number", etc.).  
3- Semantic level: in the case of a different 
lemma, a semantic wordnet is searched in order to 
check whether a close semantic link (synonymy, 
hyperonymy, hyponymy, meronymy, antonymy) 
can be found between the given answer and the 
expected one. Then, a warning can be displayed 
such as "be more precise", "not exactly", etc. 
 
In the global architecture, such a script could be 
useful for the evaluation of various activities: gap-
filling, lexical questions, etc. According to the 
specific context and aim of a given activity, the 
feed-back to the learner may be very different. For 
instance, if a gap-filling exercise is designed to test 
the ability to conjugate verbs in a given tense, the 
fact that the lemma of the learner's answer is 
different is not very important, provided that the 
verbal flexion is correct. 
Therefore, in the design of such an evaluation 
script, it is important to separate the comparison 
and the feed-back. We plan to implement too 
scripts: 
 
- the comparison script that takes as an input: the 
linguistic context, the expected answer, the given 
answer; and returns a difference code such as: 
0: no difference 
1.1: spelling error on the expected answer 
1.2: spelling error on another answer (with a list 
of close words) 
2.1: different lemma 
2.2: different morphosyntactic features 
2.2.1, 2.2.2,...: different number, different 
gender, etc. 
3.1,3.2,...: synonym, hyperonym, etc. 
 
- the feed-back script that takes as an input the 
difference code, and returns a message, such as : 
"yes, but the spelling is wrong", "be more precise". 
Even if one can propose standard messages for 
each difference code, the teacher should obviously 
be able to edit an adapted message set depending 
on the didactic context of a given activity. 
 
5 Current functionalities of Mirto and 
perspectives 
The development of Mirto started about a year 
ago. A total of three years should be necessary to 
complete the first prototype. A handling period is 
to be foreseen in order to allow teachers to master 
the use of the product. We plan to integrate Mirto 
to the Stendhal Intranet for experimentation.  
So far, the development of Mirto mainly 
concerned the script creation module. Completing 
this module allowed the integration of various NLP 
(and non-NLP) software. Other software, 
especially NLP-based, ought to be integrated. The 
choice of the number and nature of integrated 
software can only be done through a process of 
exchange involving both language teachers and 
NLP experts. We consider that, the software 
integrated so far allowed the creation of enough 
scripts for an experimental use of Mirto. 
In order to perform tests and validate the global 
approach, a first version of the activity and 
scenario editor has been implemented. It allows the 
creation of almost every type of activity (excluding 
the evaluation) and the design of linear scenarios 
that will not trace the learner training history. 
The definition of the approach underneath Mirto, 
along with making use of it, originated various 
research works, which are currently being carried 
out. Apart from the implementation of the 
prototype of the system, our efforts particularly 
concern the following aspects:  
- the pedagogical annotation and indexation of 
texts towards the creation of a corpus to be used by 
language teachers (Loiseau, 2003) 
- the automatic analysis and pedagogical analysis 
of the learners’ answers using NLP based tools and 
techniques.  
- scripting and interfacing for activity generation 
The first results of these works should find their 
application in Mirto. 
At the crossroads of three branches – language 
didactics, NLP and computer science – Mirto 
raises new problems, not only in each of these 
branches (advances in NLP for instance) but also 
problems for which no solution can be reached 
unless the branches (and their specialists) work in 
quasi osmosis. One can mention among others the 
examples of the automatic definition of the 
appropriate response for the learners’ answers, the 
modeling and implementation of computer 
functions manipulating language didactics 
concepts (so as to provide language teachers with 
no specific computational skills with tools they can 
handle), the definition and pedagogical 
exploitation of the trace of the learners’ activity or 
the modeling of non-linear scenarios… Unless we 
find answers to these problems, CALL will have to 
settle for the creation of pedagogical-added-value-
less-products. 

References  
G. Antoniadis and C. Ponton. 2004. Mirto : un 
système au service de l'enseignement des 
langues, UNTELE'2004, Compiègne, France. 
G. Antoniadis and C. Ponton. 2002. Le TAL : une 
nouvelle voie pour l'apprentissage des langues. 
UNTELE'2002, Compiègne, France. 
C. Brun, T. Parmentier, A. Sandor and F. Segond., 
"Les outils de TAL au service de la e-formation 
en langues", Multilinguisme et traitement de 
l’information. F. Segond, ed., pages 223-250, 
Hermès Science Publications, Paris, France 
T. Chanier. 1998. Relations entre le TAL et l'ALAO 
ou l'ALAO un "simple" domaine d'application du 
TAL ?. International conference on natural 
language processing and industrial application 
(NLP+IA'98), Moncton, Canada. 
T. Chanier and T. Selva. 2000. "Génération 
automatique d’activités lexicales dans le système 
ALEXIA". Sciences et Techniques Educatives 
7(2):385-412, Hermès, Paris, France 
G. Forestier. 2002. Plates-Formes pour 
l’enseignement des langues : le cas de Mirto. 
Mémoire pour l’examen probatoire en 
Informatique, CNAM, Grenoble, France. 
O. Kraif. 2003. "Propositions pour l’intégration 
d’outils TAL aux dispositifs informatisés 
d’apprentissage des langues, Intercompréhension 
en langues romanes". LIDIL, n° 28, C. Degache 
ed., pages 153-165, Grenoble, France. 
M. Loiseau. 2003. Vers la création d’une base de  
données de ressources textuelles indexée 
pédagogiquement pour l’enseignement des 
langues. Mémoire de DEA en Sciences du 
Langage, Stendhal University, Grenoble, France. 
D. Renié. 1995. Modélisation informatique de 
l’acquisition des interrogatives directes en 
français langue seconde dans leur dimension 
pragmatique, proposition d’un environnement 
offrant un apprentissage collaboratif : 
ELEONORE. PhD Thesis, Clermont II 
University, Clermont-Ferrand, France.
