NLP serving the cause of language learning 
Frédérique Segond 
Xerox Research Centre Europe 
6, chemin de Mauperthuis 
F-38240 Meylan - FRANCE 
segond@xrce.xerox.com  
Thibault Parmentier 
Xerox Research Centre Europe 
6, chemin de Mauperthuis 
F-38240 Meylan - FRANCE 
parmentier@xrce.xerox.com 
 
 
Abstract 
E-learning paves the way to a new type of 
course, more student centred, granulized, on 
demand, and highly interactive. Natural 
Language Processing (NLP) technologies 
associated with other multimedia technologies 
can help to address the major issues raised by 
this new type of courses: interaction, 
personalization and reliable information 
access. This paper presents Exills, a true e-
learning solution which integrates natural 
language processing tools and virtual reality
1
. 
Exills is unique in that,, unlike most of the 
language learning systems, it focuses on 
improving learners’ performance rather than 
learners’ competence.  
Introduction 
This paper is not a theoretical paper. It 
describes a true e-learning system that concentrates 
on improving users’ performance in a foreign 
language and integrates natural language 
processing technologies. 
 
The system, Exills (www.exills.com) represents an 
innovative way of integrating Natural Language 
technologies and multimedia technologies in order 
to provide a visionary e-learning tool for learning 
foreign languages.  
Exills is the result of a close collaboration 
between language teachers, specialists in virtual 
reality, linguists and computer scientists. While 
retaining “traditional content” such as language 
exercises, grammar worksheets and speech acts, 
Exills takes advantage of new technologies for 
interaction (asynchronous and synchronous), game 
aspect and course scenarization.  
Exills has been conceived with the constant 
concern for pedagogical quality and for 
differentiating itself from both CDROMs and face 
to face language courses. We have been careful to 
                                                      
1
 The software we use for Virtual reality comes from 
Blaxxun (http://www.blaxxun.com/en/site/index.html). 
renounce using technologies, even innovative ones, 
if they did not serve the cause of a pedagogical e-
learning system. The focus has been to improve 
learner’s performance rather than to insist on 
improving learner’s competence
2
. Indeed, 
experimenting with different available language 
learning solutions focusing on competence (via 
vocabulary exercises and grammar) led us to the 
conclusion that these types of activities do not 
insure that learners are then able to interact in a 
foreign language. While they might acquire the 
conceptual knowledge of the language they are not 
necessarily able to use it in practical contexts. In 
the case of Exills, learners are put into practical 
situations. The situations are represented via 
different scenarios. In most cases the scenario 
corresponds to everyday work situations. 
Exills focuses on the interaction aspect of 
language learning driven by the use of the Internet, 
in particular, on interaction and information access.  
Interaction within Exills includes language 
understanding through reading and listening. It 
also improves language production in writing. As 
we will see in the conclusion adding speech 
technologies to Exills will enable users to also 
improve on language oral production.  
In the demonstration scenarios that are 
accessible online, targeted users are professionals 
with an intermediate level in English or French. 
These scenarios put users in typical work situations 
such as introducing themselves, reading emails, 
searching for and understanding information, 
writing and making a presentation, ordering their 
meal, interacting with their colleagues and 
customers. Part of the scenario is dedicated to the 
discovery of cultural aspects.  
                                                      
2
 Competence and performance as defined by 
Chomsky (Chomsky, 1965): Competence is 
the speaker-hearer's knowledge of his 
language. Performance is the speaker-
hearer's actual use of his language in 
concrete situations  
 
In what follows, after presenting our view of e-
learning, we describe Exills and its different 
technical components. We then conclude by 
suggesting other linguistic technologies that could 
serve the cause of language e-learning. 
How e-learning affects our way of learning 
Learning through the Internet, or e-learning, paves 
the way for a new type of course which, most of 
the time, complements the face-to-face course 
(blended learning), but can also be substituted for 
the face-to-face courses. While traditional and 
Computer-Based training is targeted to a group of 
students, e-learning is often referred to as 
“my-learning” as it is highly personalized, 
granulized, on demand and highly interactive. The 
characteristics of e-learning are that it provides 
learners with the possibility to learn exactly what 
they need, from anywhere and at any time 
E-learning completely changes both the 
teaching and the learning processes. Teachers 
become tutors that support students’ work. 
Students take more responsibility in the learning 
process as they know what they need to learn to be 
efficient. This is especially true in a life long 
training context. They become actors in their 
learning process by choosing their course topics, 
elaborating their learning paths, exchanging with 
other students as well as with tutors. 
E-learning is definitively a new way of learning 
that can take advantage of new Internet 
technologies.  
However, despite the explosion of e-learning 
platforms also called LMS for learning 
management system (see thot.cursus.edu) there are 
very few systems which take real advantage of the 
functionalities the Internet offers such as: 
interaction, personalization and huge information 
access (for both students and tutors). Indeed, most 
of the current e-learning solutions are LMS that are 
former Knowledge Management Systems 
customized for learning. Because they have not 
firstly been designed for a learning purpose they 
often miss the pedagogical dimension. Other 
solutions are a simple web link to already existing 
CD-ROM content. This solution usually offers a 
simple type of interaction such as emails with the 
tutors and, sometimes, forums with other students. 
Within the e-learning domain, language 
learning is one of the most successful market 
segments as it comes second only to IT and 
marketing. This is probably due to the fact that for 
many years it has been possible to learn a language 
using electronic media (CD-ROM). Medium to 
large publishers have been successful in 
developing language training multimedia content 
of good quality. Still, the majority of so-called e-
language learning solutions that are on the market 
today are CD-ROM content transferred onto the 
web rather that real online solutions.  
The reason is probably that most of the 
publishers in this area are specialists in producing 
good quality pedagogical content but are not 
specialists of the Internet, nor do they have the 
technological resources necessary to transfer their 
content onto the Web. Offering on-line content on 
the web is very different from offering content on a 
CD-ROM. These media organize content in a 
completely different way and this has a strong 
impact on the basic pedagogical principles. This 
shows up at different levels. For instance, studies 
have shown that users will spend at the most 15 
minutes on an Internet learning session versus 45 
minutes on a CD-ROM. learning. Besides, there is 
the issue of how to organize learning material so 
that navigation among short learning modules is 
possible without affecting the learning pedagogical 
process. Finally, record keeping, tracking, tracing 
and flexibility are all essential functions for 
making an on line course successful as well as for 
keeping user motivation high. These are just a few 
of the issues that need to be addressed when 
creating high quality content on the Internet for 
language learning. 
Exills is a true e-learning solution, that on the 
technological part, integrates virtual reality with 
linguistic and collaborative technologies as well as 
with a smart search engine. 
 
Exills’ general features  
In very general terms, Exills could be described as 
a hybrid between an authoring tool and a delivery 
platform particularly well suited to building and 
distributing language learning applications. 
Exills user interface provides interactive access 
to the most up-to-date Internet technologies. 
The user interface is composed of: 
• a 3D scene to put the user in a cognitive 
context in line with the topic of the course;  
• a chat frame to allow students, robots and 
tutors to interact via text (written 
production); 
• a frame to provide learners with any 
information they might need as well as with 
activities to do. 
• a toolbar to provide learners with easy 
access to specific functionalities such as: a 
linguistic toolbox (contextual dictionary 
look up, phonetic spell checker, conjugation 
engine), a notepad to take notes, a history of 
their work to review it at any time, as well 
as shortcuts, through maps, to navigate 
inside 3D scenes. They can also get help 
from tutors, access their profile and a list of 
personalized exercises, grammar 
worksheets, and speech acts. 
Each course is associated with a scenario. These 
scenarios are usually associated with a mission to 
accomplish. This mission integrates learning 
activities, game aspects and competition between 
students. In order to achieve the mission, learners 
need to interact with each other, with the tutor as 
well as with the virtual environment. Thanks to the 
virtual reality this interaction takes place like in 
video games. Learners play a video game in a 
foreign language.  
In its current incarnation Exills has the 
following underlying features. 
 
Learner 
Linguistic Services 
Collaborative and 
Community 
Management Tools 
Smart Search Engine 
 
Figure 1: Technologies involved 
Virtual reality 
While virtual reality presents real advantages 
for learning, this technology is not yet strongly 
used for this purpose.  
In Exills, virtual reality is used both for interaction 
and for personalization. Personalization, because 
when they are in virtual reality, learners can 
choose a representation of themselves, also known 
as an avatar, and can move around the virtual 
places like in the real world. For instance, learners 
can meet with other avatars/learners, interact with 
them and access the Internet.  
The virtual reality induces a type of 
interaction that completely differs from email, 
forum or face to face. Part of this difference is due 
to the fact that people hide themselves behind an 
avatar but also because of the game aspect 
permitted by the virtual reality.  
First experiments showed that in virtual 
reality, as in games, people behave completely 
differently and tend to interact much more. For 
instance, we have noticed that while in a face to 
face or in a video course, most of the time, only the 
most confident students speak up; in a virtual 
environment, everyone dares more to express 
himself/herself. This definitively helps the tutor in 
evaluating students’ ability.  
To ease and push learners’ interaction, it is 
important to provide them with an interaction 
topic. This is where the notion of scenario comes 
in and where, again, virtual reality turns out to be 
an important technology. The scenario is there to 
help students to interact among themselves but also 
to immerse them directly into “prototypical 
situations”. For instance, if students need to learn a 
language for work reasons, a scenario will be built 
around their everyday life at work, including the 
physical work place and this induces reaction and 
discussion correlated to the 3D scenes. Because it 
provides learners with a reason to interact, like in 
role playing games, the scenario makes interaction 
more effective than in traditional chat and forums.  
  
 
 
Figure 2: Set the student in a cognitive context 
using the 3D. 
What is more natural than clicking on a 
computer screen to access emails, than clicking on 
a CD to hear music, than visiting a library to read 
books or than being invited to chat with the 
receptionist when arriving at the reception desk? 
Because it mimics the real world, students can 
enter quickly into the virtual reality environment. 
Students usually just need a few hours’ training to 
easily manipulate the mouse in order to move from 
one virtual place to another. 
Linguistic technologies 
Exills adds other technologies to enhanced virtual 
reality in such a way that the resulting learning 
system makes real use of the potential of the 
Internet. 
In its current incarnation, Exills offers learners 
contextual bilingual dictionaries, morphological 
analyzers and syntactic taggers, a language 
identifier and phonetic spell checkers. These 
technologies are there to help students express 
themselves or understand a foreign language. 
Linguistic technologies associated with virtual 
reality retain the personalized aspect of individual 
courses while being immersed in the collective 
course. 
Within the chat, Exills offers a comprehension 
help service which uses the context to first retrieve 
the most appropriate translation of a word 
(syntactic and semantic disambiguation). As a 
result, the students’ comprehension is sped up and 
interaction becomes smoother. 
The same contextual dictionary look up could 
be used for providing the student with a list of 
sentences where the word they don’t understand 
appears with the same meaning. In that case, the 
pedagogical process is different: instead of 
providing the student directly with the translation, 
the system provides him/her with other contextual 
uses of the same word  
 
 
Figure 3: Understanding assistance. 
If we now consider the linguistic analysis of 
student production (chat/forum contribution, 
exercise, writing, etc.), linguistic technologies 
provide students with customizable services which 
allow them to parse and tag their own production 
in order to check its correctness. Here again the 
goal is not to provide them with the correct answer, 
this is left to the tutor, but rather to help them 
detect their own mistakes.  
 
 
Figure 4: Through linguistic services, students can 
analyze and correct their written production. 
Along the same lines as “learning by example”, 
linguistic tools can be used to highlight specific 
language constructions such as functional relations, 
terminology or idiomatic expressions. 
Derivational morphology can be used to discover 
the words of a same family and improve students’ 
vocabulary. 
A simple morphological analyzer gives students 
access to verb conjugations or noun declensions. 
 
Figure 5: Access to the conjugation of a verb 
Exills also includes a language identifier to 
force students to interact only in the language they 
are learning. This integration relieves the tutor of 
checking on each student, and pushes students to 
express themselves more in that language.  
The language identifier is integrated into the 
chat. This module identifies the language used by a 
student and if different from the one the student is 
supposed to learn, the system blocks any type of 
interaction. Students then receive a message saying 
that they are not using the appropriate language.  
The language identifier can also be plugged into 
mail or forum services between students or even 
between students and tutor. 
 
 
Figure 6: Automatic control of the language used 
in the chat. 
A phonetic spell checker is also available to 
students to help them to learn the correct spelling 
of words. 
Linguistic technologies can be used to reinforce 
the personalization aspect of e-learning. Indeed, 
linguistic technologies open the way to learning on 
any text that is of interest for the course and for 
each individual learner. For instance, professionals 
can work on their everyday technical 
documentation and can learn their professional 
terminology in a foreign language. In particular, 
linguistic technologies can automatically extract 
technical terminology for them. 
 
Smart search engines 
Exills uses smart search engines in order to 
strengthen the personalization and interactive 
aspects. Indeed, having a smart search engine that 
can take into account students’ needs and interest 
through a detailed user profile and can retrieve 
personalized content from the web clearly 
improves the personalization aspect and course 
efficiency. Tutors and students can use the Internet 
as an infinite source of information to enrich 
course content. This is where collaborative 
technologies play a crucial role. Information on the 
Internet varies in quality, and it is especially 
important in a course context to be able to separate 
the wheat from the chaff. 
 
Smart search engines strengthen the 
personalization aspect of e-learning. They can 
retrieve documents that match the user’s profile. 
This user profile encompasses student-specific 
difficulties in the courses as well as the student’s 
personal interest, be it for work reasons or in life in 
general. An example of this is a student who is 
interested in American politics and has difficulties 
with the use of idiomatic expressions in French. A 
smart search engine uses morpho-syntactic 
analysis of the retrieved text as well as an ontology 
to provide the student with texts about American 
politics that contain idioms. Using smart meta-
search engines permits a search on the Intranet as 
well as on student emails. This is especially useful 
in a context of professional language learning 
courses. In that case, the search can be performed 
on technical texts from the company. 
 
WWW 
DB 
Local Files 
Mails 
Smart Search 
Engine 
Requests 
Results 
Aggregated Results 
Request 
Community 
Management Tool 
Interesting  
results 
Community 
databases 
Request 
  Results 
 
Figure 7: Correlated use of a content aggregation 
engine and of a community manager tools. 
Collaborative technologies  
Within Exills, collaborative technologies are used 
to filter information found on the web by 
commenting and interacting on it. At the end of the 
interaction process, the tutor will decide which 
documents will be retained to enrich the content of 
the course.  
Using collaborative technologies impacts the two 
kinds of users in several ways. 
• First, the students. Using this tool and 
contributing to its content allows the student 
community to have access to a very 
qualified and annotated text database. 
• Second, the tutors. Tutors can use the 
community repository in two different ways: 
first, to get specific and adequate content in 
order to generate new exercises or activities; 
second, to evaluate the level and the 
understanding of the students by analyzing 
their comments on the different documents 
posted. 
Infinite number of possible scenarios 
The Exills scenario that is currently available on 
the Internet relates to professionals working for a 
printing company. It is just one of many possible 
scenarios. In fact, Exills permits the construction 
of an infinite number of scenarios. This possibility 
also points out the high pedagogical adaptability of 
Exills which leaves to the tutor the possibility to 
decide about scenario, content and activities. In the 
next sections we give some examples of other 
possible scenarios that could use Exills as an 
underlying framework. 
Schools 
In an academic context, it is possible to organize 
virtual travel for students, similar to language 
exchanges that already exist. In that, case pupils 
from two different countries connect at the same 
time and organize a virtual tour of their school, 
their district, their city, they describe their cultural 
habits. In this type of experiment one class plays 
the role of the tutor and corrects the language of 
the other class.  
This kind of experimentation is known as 
‘Tandem’ in the pedagogical domain. According to 
several studies, this way of learning languages is 
more efficient as pupils are more motivated.  
 
Life long training 
One important pedagogical consequence of the 
notion of scenario is that it allows users to learn 
both a work process and a foreign language. 
Another scenario, we are currently working on, 
is targeted to people working at hotels reception 
desk. This scenario is developed in the framework 
of an eContent European project, Thetis 
(http://www.thetis-project.org/), which groups 
together a language publisher (Q group
3
), a 
training company specialized in tourism (Grupo 
                                                      
3
 http://www.qgroupplc.com/ 
GDT
4
) and a research center (Xerox Research 
Centre Europe
5
). 
 
To summarize, while using linguistic technologies, 
students can process online texts on the fly and can 
work on their own documents. This strengthens the 
personalization aspect of the solution. Going 
forward, using Exills in a professional context 
allows students to work on the documents of their 
own company. 
Evaluation 
Exills has been tested in different contexts. We 
tested the current scenario (professional working 
for a printing company) in language schools as 
well as in companies. In all cases we have spent 
two hours beforehand with the tutors to train them 
with the concept as well as with virtual reality. 
Between the training of the tutor and the real test 
we allowed one week so that tutors can get familiar 
with the environment. We attended the testing just 
to help in case of technical problems. We 
distributed a questionnaire at the end so that users 
can immediately provide us with feedback. The 
feedback was globally highly positive. Users like 
the game aspect, although sometime they would 
prefer a non professional scenario. Because of this 
game dimension they tend to use the linguistic 
technologies more than usual and in particular the 
contextual bilingual dictionary look-up. This 
demonstrates that the wish to interact combined 
with the impossibility of speaking any other 
languages, because of the language identifier, push 
them to use linguistic technologies to progress. 
Teachers like the language identifier as it 
relieves them of checking on the language students 
use. They report on the fact that correcting on the 
fly is sometimes difficult because the chat is so 
quick between participants. In any case they do 
have the possibility of correcting any student 
production afterwards as everything is stored on 
the server.  
The other context where we tested Exills is 
within classrooms. In that case users were 
teenagers from age 13 to 14. This type of 
population is obviously very comfortable with 
virtual reality and the game aspect. Again they 
enjoy very much the fact that they learn a language 
while playing a game, and again they made 
extensive use of the linguistic technologies to be 
able to write sentences. School teachers were 
impressed by the fact that their students stayed 
attentive for the entire course. This should 
                                                      
4
 http://www.grupogdt.com/superior.html 
5
 http://www.xrce.xerox.com/ 
probably be balanced by the novelty aspect of the 
system. 
One thing we have not yet been able to test, but 
which would be very interesting to evaluate, is the 
difference in users performance in language after 
having used Exills. 
We are now working with a research team of 
specialists in pedagogical issue to define a 
methodology that could measure this performance. 
Conclusion 
To summarize, Exills integrates virtual reality, 
linguistic, collaborative and smart search 
technologies into a true e-learning solution that can 
be used both synchronously and asynchronously. 
Thanks to this technology integration Exills 
allows: 
• Students to benefit from the interest’s 
potential for interaction, , to have a 
controlled and efficient access to the 
information on the Internet, to learn by 
themselves at their own rhythm. 
• Tutors to get support in supervising their 
students. 
• Learning companies to offer courses based 
on a new concept (virtual reality and 
linguistic technologies) instead of just 
putting classical lessons on the web. 
• End-user companies to have a course 
completely customized to their needs, their 
working environment and their terminology. 
Exills is a true e-learning solution created to take 
advantage of the web. By taking advantage of the 
interaction, Exills escapes the drawbacks of many 
e-learning solutions such as: the strenuousness of 
reading notices, the impression for students of 
being left to their own devices, or the temptation to 
give up the course. This is made possible because: 
• Virtual reality induces a cognitive context 
for the students and pushes them to interact. 
• Smart search technologies allow one get 
information, texts and documents from the 
Internet or specialized databases related to a 
specific subject matching the language level 
of the student. 
• Collaborative technologies allow the sharing 
of information gathered on the Internet 
between students and the provision of 
adequate content to the tutor to create new 
exercises or lessons almost on the fly. 
• Linguistic tools provide autonomy to 
students by showing them concepts, give 
them assistance to understand word 
meanings or different senses within a 
particular context by presenting various 
examples, to provide feedback for their 
production (in the chat as well as during 
exercises or even free activities). 
Clearly Exills would benefit from adding new 
technologies and in particular speech technologies. 
With speech technologies, students would be able 
to interact via voice, to record themselves and to 
compare their own oral production to a model.  
Finally, the linguistic technologies 
presented in this paper could also be used to 
provide more functionality. For instance, 
morphology and tagging could be used in order to 
automatically create online filling-in-the-blank 
type of exercise from any text found on the 
Internet: The process is quite similar to the 
exercises generator framework but the 
functionalities provided to the student are fewer: 
given a text and an exercise pattern, the exercise is 
generated just for her/him. This exercise is then 
corrected automatically using pattern matching, 
morphology or spellchecker. 
Part of these ideas, especially the speech 
components
6
, will now be implemented in the 
framework of Thetis, the on-going European 
project mentioned before. 

References 
Chomsky N., . Aspects of the Theory of Syntax. 
Cambridge: The MIT Press, 1965. 
Baudit, Apprendre à deux. Études psychosociales de 
situations dyadiques, PUF, 1998. Paris. 
Bayer V., Farah J., Apprentissage des langues en 
Tandem sur Internet, ELA, n°113, pp.73-78, 1999. 
Bergholz A., Experiences with the knowledge Pump, 
ISIC 2002: Information Needs, Seeking & Use, 
September 2002, Lisbon. 
Brun C., Parmentier T., Sandor A., Segond F., Les 
outils de TAL au service de la e-formation in 
Multilinguisme et le traitement de l'information, 
Frédérique Segond Ed., Hermès 2002, pp. 223-251  
Chidlovskii B., Glance N., Grasso A., Collaborative Re-
ranking of Search Results, Proc. AAAI-2000 
Workshop on AI for Web Search, July 2000, Austin. 
Kashny M., Les usages des Technologies d'Information 
et de Communication par des enseignants dans un 
dispositif de formation tutorée en langues vivantes 
étrangères. Une approche ergonomique. Thèse 
Université Pierre Mendès France, 2001, Grenoble. 
Kindley, R., Scenario-Based E-Learning: A Step 
Beyond Traditional E-Learning, 
http://www.learningcircuits.com/2002/may2002/kind
ley.html 
Leacok C., Chodorow M., Miller G., “Using corpus 
statistics and WordNet relations for sense 
identification, Computational Linguistics, 24(1):147-
-166, Mars 1998.  
Paulsen, J. (2001-section 4). "Authentic Online Target 
Language Reference Resources." New Era Trends 
and  Technologies in Foreign Language Learning: An 
Annotated Bibliography - Interactive Multimedia 
Electronic Journal of  Computer-Enhanced 
Learning, April 2001, Wake Forest University. 
Puren, C. Histoire des méthodologies de l'enseignement 
des langues, , Clé International, Nathan, 1988, Paris. 
