STANDARDS FOR COMMUNICATION AND E-LEARNING IN

VIRTUAL WORLDS

The Multilingual-assisted Chat Interface

Samuel Cruz-Lara, Tarik Osswald, Jordan Guinaud, Nadia Bellalem and Lotﬁ Bellalem

LORIA / INRIA Nancy - Grand Est, 615 Rue du Jardin Botanique, 54600 Villers-l

es-Nancy, France

Keywords:

Multilinguality, MLIF, Chat interface, Web services, Communication, e-Learning, Automatic translation, Vir-

tual worlds.

Abstract:

Many of today’s applications embed textual chat interfaces or work with multilingual textual information.

The Multilingual Information Framework (MLIF) [ISO DIS 24616] is being designed in order to fulﬁll the

multilingual needs of today’s applications. Within our research activity for the MLIF standard, we developed

the Multilingual-Assisted Chat Interface, which intends to help people communicate in virtual worlds with

others who do not speak the same language and to offer new possibilities for learning foreign languages.

By developing this application, we also wanted to show the advantages of using web services for externalizing

computation: we used the same web service for two virtual worlds: Second Life and Solipsis.

In this paper, we ﬁrst propose a short analysis of social interactions and language learning in virtual

worlds. Then, we describe in a technical way the features, architecture and development indications for the

Multilingual-Assisted Chat Interface.

1 INTRODUCTION

Today, talking to people via a textual chat interface

has become very usual. Many web applications have

an embedded chat interface, with a varying array of

features, so that the users can communicate from

within these applications. A chat interface is also easy

to implement: unlike voice or video, it needs neither

additional devices nor additional signal analysis al-

gorithms. Therefore it is not surprising if brand new

technologies such as virtual worlds also embed a tex-

tual chat interface.

But all applications have their own peculiarities,

and their chats also serve various requirements. A

distinctive feature of virtual worlds is that people are

more likely to converse with other people who can-

not speak their native language. In such cases, the

need for some sort of assistance in facilitating inter-

language communication becomes obvious. A chat

interface with multilingual features can meet these

requirements.(Cruz-Lara et al., 2009). Moreover,

such an interface can be turned into an advantage for

people who want to improve their foreign language

skills in virtual life situations.

In this paper, we ﬁrst want to present some con-

siderations about social interactions and language

learning in virtual worlds. Then, we will de-

scribe chat interfaces in general and more speciﬁ-

cally the Multilingual-Assisted Chat Interface, which

we have developed within the ITEA2 Metaverse1

(www.metaverse1.org) Project [ITEA2 07016] in or-

der to give a ﬁrst answer to the question “how can

we ease communication in virtual worlds and turn the

multilinguality issue into an advantage?”

2 SOCIAL INTERACTIONS

The social interaction in virtual worlds constitutes a

particularly rich ﬁeld of research. Indeed, the ob-

jective to be reached is to offer simple and effective

means of communication, which approach the nat-

ural communication. In this context, efforts should

be made as much in improving the technical aspects

as in taking into account the socio-cognitive aspects.

Thus, it will improve the realism of the virtual envi-

ronments and increase the quality of the exchanges

between the avatars. Therefore, gestures, speech, di-

rection of gaze, postures or facial expressions have an

entire role in the construction of the social links be-

tween individuals.

The question concerning the inﬂuence of virtual

Cruz-Lara S., Osswald T., Guinaud J., Bellalem N. and Bellalem L. (2010).

STANDARDS FOR COMMUNICATION AND E-LEARNING IN VIRTUALWORLDS - The Multilingual-assisted Chat Interface.

In Proceedings of the 12th International Conference on Enterprise Information Systems - Human-Computer Interaction, pages 45-52

DOI: 10.5220/0002894800450052

 SciTePress

environments on the social behaviour of users, con-

stitutes a particularly interesting topic for the re-

searchers in sociology. In fact, they highlighted a phe-

nomenon of disinhibition and facilitation which leads

to a greater sociability (Suler, 2004)(Coleman, 2007).

As social human beings, we adjust our behaviour with

the social norms in face to face communication. We

know, thanks to our education and our culture, what is

socially acceptable or not. Communicating by inter-

posed computers strongly decreases this adjustment

because we cannot observe in real-time the effects of

our words and of our writings.

In order to propose credible and powerful commu-

nication between avatars, the characteristics of the hu-

man communication must be taken into account: lan-

guage with explicit or implicit references to the ob-

jects of the environment, gestures, postures, facial ex-

pressions. These elements of communication are all

the more important since the avatar is immersed in

a three-dimensional world populated by objects, per-

sonages and places more or less characteristic which

sometimes echoes the real world.

The models suggested by El Jed (El Jed et al.,

2006) try to take into account intentional communica-

tion as well as non-intentional communication in the

interpretation of the acts of communication between

avatars. In this context, the favoured mode of com-

munication is natural language combined with deic-

tic gestures. The difﬁculty, in this case, consists in

using markers like vocal intensity, voice intonation,

or indexical or deictic references (“I”, “here”, “over

there”) associated with the designation gesture to de-

termine the relevant interpretation of the exchanges.

The direction of the gaze can also be exploited in

order to focus the visual attention of the interlocu-

tors towards a speciﬁc place in the shared environ-

ment. The facial expressions are essential during the

exchanges and constitute the ﬁrst channel to commu-

nicate emotions. They can express mood, approval or

disapproval, but also the whole panel of the human

emotions (fear, joy, etc.). All these manifestations of

the human communication would be very useful in

the domain of education, specially for the develop-

ment of e-learning techniques used for the realisation

of virtual campuses (De Lucia et al., 2009).

3 LANGUAGE LEARNING

Language learning in virtual worlds is a new ﬁeld of

research which is still open to innovations. How can

we create technological advances in order to create

an optimal psycholinguistic environment for language

learning? What makes new proposals innovative and

helpful? In order to answer these questions we are

currently developing some empirical support.

We believe that one potential source of guidance

may be offered by some methodological principles

of Task-Based Language Teaching (TBLT) applied

to distance foreign language teaching (Doughty and

Long, 2003). It should be noted that the TBLT as-

pect of our work is currently under development.

The general idea is allowing teachers to create TBLT

units via the Moodle learning environment system

(http://www.moodle.org), and then to use SLoodle

(http://www.sloodle.org). SLoodle is an open source

project which integrates Moodle components in Sec-

ond Life.

Our approach (i.e., the Multilingual-Assisted Chat

Interface) must be considered as an innovative form of

Computer-Assisted Language Learning (CALL).

4 THE MULTILINGUAL

ASSISTED CHAT INTERFACE

The Multilingual-Assisted Chat Interface is a tool of-

fering new functionalities to the chat users in Virtual

worlds. It is directly embedded in some virtual worlds

viewers (to date: Second Life and Solipsis).

In order to implement those functionalities (de-

scribed in Section 4.2), we modiﬁed the source code

of the viewers (see Section 4.5).

The Multilingual-Assisted Chat Interface mainly

relies on the Multilingual Information Framework

(Cruz-Lara et al., nd), which is a standard being cur-

rently developed by the International Organization for

Standardization (ISO).

4.1 Generalities about Chat Interfaces

The term “chat” refers to a real-time written dialogue,

using a computer. The chat presents many similarities

with the oral dialogue. In fact, it surely is the written

means of communication which is the closest to it.

Indeed, it is closer than other tools, such as forums

and email.

Most of the time (for example in Second Life), the

chat is volatile. In fact, the contents of a chat session,

like an oral conversation, is not intended to be avail-

able to the public. Furthermore, when a user logs in,

he totally ignores what has been said before his ar-

rival. Similarly he cannot know what will be said after

his disconnection. Moreover, the chat users introduce

into their writing some elements which are speciﬁc to

oral communication.

The chat is usually based on a client-server archi-

tecture, meaning that users do not communicate di-

ICEIS 2010 - 12th International Conference on Enterprise Information Systems

rectly with each other, but through a single server. All

the chat users connected to one server do not necessar-

ily communicate together. In general, a server gives

access to several channels of discussion (also known

as rooms), which are completely partitioned. The chat

users only see what is happening in the channel they

are connected to, and they can only send messages to

this very one. On most of the public servers, the cre-

ation of channels is completely free, and any user can

open some.

In general, the client chat interface is split into

three areas:

• the ongoing conversation;

• the message editing and sending zone;

• the listing of the connected users.

A chat session can be considered as a set of mes-

sages, ordered sequentially and produced by various

authors, humans or robots.

4.2 Functionalities

In this section, we are going to describe the three

main functionalities which we implemented in the

Multilingual-Assisted Chat Interface.

4.2.1 Grammatical Analysis and Word

Colouring

This is the ﬁrst feature that we implemented. It con-

sists in coloring some speciﬁc words in a sentence

written in the chat interface, in order to show the

grammatical structure of the sentence more clearly.

The settings can be modiﬁed in a speciﬁc menu: the

user can choose which grammatical categories they

want to highlight, and the corresponding color. They

can also choose which sentences they want to analyze

(other people’s sentences, objects’ sentences, all the

sentences, etc.).

The grammatical part-of-speech analysis is per-

formed on a remote web server, using the method de-

scribed in Section 4.4. The data structure used for tag-

ging the grammatical category of each word is MLIF.

4.2.2 Providing Word Translations, Synonyms

and Deﬁnitions

An important feature is the ability to simply click on

a word in the chat interface in order to get deﬁni-

tions, synonyms and/or translations of this word. In

the current implementation, the deﬁnitions and syn-

onyms are retrieved from WordNet, and the transla-

tions from Google Translate. It is important to note

that other external corpora could also be used.

This feature could be needed in two main situations:

• When people are reading text in a language they

do not know very well, they may need some help

(e.g. deﬁnitions, synonyms, translations) in order

to understand even very common topics.

• When the discussion is about a rather precise sub-

ject, where several technical terms are often used,

people may especially need additional deﬁnitions

even if they are native speakers.

The obvious advantage of this feature is that it is

able to aggregate information from several web ser-

vices in just one click, which is much more con-

venient than looking for the information in a web

browser by oneself. Figure 1 shows the menu, the

chat interface (with the verbs in a different colour)

and the action performed when clicking on a word.

Every MLIF data structure generated after click-

ing on a word can be stored in a database. Doing so

provides the user with a feedback on the words that

he did not understand that well. Writing down the

unknown vocabulary is something very usual when

learning a new language and this is exactly what we

have implemented here.

4.2.3 Automatic Translation

When people do not know a language very well, they

will be interested in having an automatic translation

of every sentence. In this case, it is no longer an e-

learning functionality but a way to make communica-

tion easier between people in the virtual worlds.

For example, user A uses the Multilingual-

Assisted Chat Interface and wants to have all the mes-

sages displayed in French. The Multilingual-Assisted

Chat Interface will analyse the incoming messages

and translate them if they are not written in French

(using Google Translate). Then, the messages sent

by user A will be translated with respect to the lan-

guage of the latest received message in the conversa-

tion. Figure 2 shows a typical situation involving two

avatars who neither understand nor speak the same

language.

Another important point is that the source lan-

guage is automatically detected by Google Translate.

Thus, the user only needs to enable the automatic

translation functionality to be able to chat with any

other avatar in the virtual world. Also, as the source

language is stored in the MLIF data, it is possible to

link every discussion to a pair of languages and as a

consequence, it is possible to carry out several multi-

lingual conversations at the same time.

In addition to that, both messages are stored as

MLIF data. Thus, the user can click on one translated

STANDARDS FOR COMMUNICATION AND E-LEARNING IN VIRTUALWORLDS - The Multilingual-assisted Chat

Interface

Figure 1: Colouring words in the chat interface and displaying information about one word (Solipsis).

Figure 2: Automatic translation between a Japanese and a French avatar.

ICEIS 2010 - 12th International Conference on Enterprise Information Systems

sentence and see the original sentence if they want

some insight into the original language or if the trans-

lation does not seem very accurate.

4.3 General Architecture

Figure 3 shows the general high-level architecture of

the main components of the Multilingual Assisted

Chat Interface.

Three colors are used in this scheme. Each one

represents a certain category of components:

orange (on the left): components belonging to the

virtual world (especially the viewer);

blue (in the middle): the web service components

that we developed, mainly business components;

green (on the right): external web services and cor-

pora, and data storage.

The circled numbers represent the chronological

order of the interactions between the components

when a message is sent or when a word is clicked.

The corresponding explanations are written below:

1. Every message sent by a user is ﬁrst sent to the vir-

tual world server. When the client (i.e. the viewer)

of the person we are writing to receives a message,

this is forwarded to a Message Manager, (i.e. the

component set dealing with the chat messages).

We needed to modify these components both in

Snowglobe (a viewer for Second Life, see Sec-

tion 4.5.1) and in Solipsis.

2. Before displaying the message on the

Chat Interface, the Message Manager sends

an HTTP request to the web service (the

Grammatical Analyser) in order to obtain the

MLIF data representing the sentence and its

several grammatical components.

3. The Grammatical Analyser connects to external

Grammatical Corpora in order to get the gram-

matical part-of-speech tag for each word of the

message.

4. When the Message Manager receives the MLIF

data structure representing the original message

with a grammatical labeling (as an HTTP re-

sponse), the MLIF data is parsed and turned into

a format enabling the coloring and hyper linking

of each word. The coloring code depends on the

settings of the user interface (see Section 4.2.1).

5. An action is performed: while the addressee reads

the message, they click on a word that they do

not understand (in the Chat Interface) in order to

retrieve synonyms, deﬁnitions and translations for

this word.

6. After clicking on the word, the user is redirected

to a Web Interface, which will display all the de-

sired information.

7. Loading the web page involves calling the

Word Request Manager so that it retrieves infor-

mation from external web services.

8. First, the Word Request Manager retrieves some

deﬁnitions and synonyms from WordNet.

9. Second, the Word Request Manager connects to

Google Translate in order to retrieve translations.

10. Once the Word Request Manager has received all

the desired information, all of it is written into an

MLIF data structure. This MLIF data is stored in

a database, from which it can be retrieved later if

required (see Section 4.2.2).

11. The MLIF data is then sent to an

MLIF-to-HTML Parser, which is going to

transform the MLIF data into user-friendly

HTML code.

12. The HTML code thus obtained is ﬁnally displayed

by the Web Interface user interface, which makes

it easy for the user to read.

It is important to note that most data exchanges are

made using MLIF, since this data format exactly ﬁts

our requirements. This is very important to facilitate

the development within the web service (as we only

use one data format) and for future applications (as it

enhances interoperability).

4.4 Web Services for Virtual Worlds

The Multilingual-Assisted Chat Interface is based on

a web service. While the virtual world client viewer is

in charge of displaying information, the web service

deals with data processing.

The text is sent by the viewer (Second Life, Solip-

sis, etc.) to the web service. The latter gathers the in-

formation requested by the users and puts it together

into a MLIF data structure. Finally, the generated

MLIF data is returned to the viewer, which turns it

into a displayable format.

The web service has a grammatical tagger tool

(Brill, 1992). It is composed of two Python-generated

dictionaries: a default tag dictionary and a rule dictio-

nary. The ﬁrst one contributes to the matching of each

word with its most likely category, the second one to

the correction of errors by checking the context.

Interoperability is one of the most interesting fea-

tures in our web service: we can use our web service

to build new tools on any platform (other metaverse

platforms, web services, applications, etc.). Note

that it is used both for the Multilingual-Assisted Chat

STANDARDS FOR COMMUNICATION AND E-LEARNING IN VIRTUALWORLDS - The Multilingual-assisted Chat

Interface

Figure 3: The Multilingual-Assisted Chat Interface ﬂowchart.

Interface in Second life and in Solipsis. External-

izing tools on PHP servers makes them platform-

independent and saves a lot of time when applications

must be adapted to different platforms.

Many web services exist on the web and propose

speciﬁc ways to process information. Most of the

time, there are separate tools for translating, for ﬁnd-

ing synonyms or for getting the deﬁnition of a word

(Google Translate, WordNet, ConceptNet, etc.). Our

web service uses all those tools in order to allow

the user to centralize all this information in only one

MLIF data structure.

4.5 Virtual Worlds Programming

The multilingual-assisted chat interface is now avail-

able in two virtual worlds: Second Life, developed by

Linden Lab, and Solipsis, developed by Orange Labs

and ArtefactO. This section describes our work in a

technical way and is therefore directed to people who

have an advanced knowledge of programming.

4.5.1 Snowglobe

Snowglobe is an alternative viewer for Second Life

that can be downloaded on the ofﬁcial web site at

http://snowglobeproject.org. This project aims at

gathering Linden Lab and open source developers in

order to create an innovative viewer.

The developers who want to implement new

features can get the source code on the Second Life

wiki (http://wiki.secondlife.com/). This page

contains the source code, artwork and open

source libraries for all operating systems. All

the instructions concerning compilation and pri-

vate linked libraries are detailed on the wiki

(http://wiki.secondlife.com/wiki/Snowglobe) in the

“Get source and compile” section.

We are now going to explain how to modify the

chat interface and what we have modiﬁed in order to

build the Multilingual-Assisted Chat Interface.

Modifying the Chat Interface. A user can send a

message with two kinds of tools: the public chat inter-

face and the instant messaging interface (i.e. private

messages to one speciﬁc user).

• The Public Chat Interface. The text written into

the chat interface can be modiﬁed in two ways.

The ﬁrst one consists in modifying the text before

sending it to the server and the second one con-

sists in modifying the text upon its reception by

the viewer.

For the ﬁrst case, the code is located in the

ﬁles named llchatbar.cpp, llchatbar.h and llchat.h.

The LLChatBar class is directly linked to the

LLPanel class (deﬁned in the ﬁles llpanel.cpp and

llpanel.h) which is in charge of displaying the

menu elements, among the chat elements. In par-

allel, the text is sent to the server from the func-

tion named send chat from viewer. This func-

tion has three parameters. The string parameter

utf8 out text is the text that the user has written.

The type parameter (of type EChatType) indicates

the range of the message. The range will depend

on whether the user wants to whisper or to shout.

It will be different too if the message concerns all

the members of a region, the owner of an object

or just a single user. The last parameter is the

channel. There are public channels and private

channels. Generally, we use private channels to

communicate with objects.

ICEIS 2010 - 12th International Conference on Enterprise Information Systems

The code for the second case is mainly deﬁned

in the LLTextEditor class (see lltexteditor.cpp and

lltexteditor.h). Once the text has been written

by the user, received by the LLPanel class and

sent to the server by the LLChatBar class, the

LLTextEditor class deals with the next step: it pre-

pares the text for the class which is in charge of

the display.

LLTextEditor has three important functions:

appendColoredText, appendStyledText and

appendText. The appendColoredText func-

tion deﬁnes the style of the text and calls the

appendStyledText function. The latter applies

the style to the text and ﬁnds out the HTML

addresses to replace them by clickable links. If

a link is detected, a new style is deﬁned and the

appendText function is called. When the code has

completed the analysis of the HTML links, the

appendText function is called a second time with

the previous style. Each style change requires the

appendText function to be called. In other words,

appendText can consider only one style at the

same time.

• The Instant Message Interface. There is an-

other way to communicate, which allows the

user to have private communications with other

users. This means of communication is called

instant message and is deﬁned in another ﬁle:

llimpanel.cpp. The main function is called

addHistoryLine. It can edit history and call func-

tions of the LLTextEditor class in order to write

into the instant message panel.

The Multilingual-Assisted Chat Interface. The

functions we built for the Multilingual-Assisted Chat

Interface are mainly deﬁned in the LLTextEditor class.

When the viewer receives text, the buildMLIF

function converts it into a MLIF data structure

thanks to the web service, which provides the

grammatical category for each word. Then,

three functions (parseMLIFAndReturnCategories,

parseMLIFAndReturnLinks and

parseMLIFAndReturnWords) are called to parse

the MLIF data structure and they return respectively

each word, its grammatical category and the link to

its information.

In the following, a “grammatically tagged link”

deﬁnes a clickable HTTP link which embeds the

grammatical category information of the word. A

grammatically tagged link allows the viewer to rec-

ognize a word category and to apply it to the corre-

sponding style and URL. It is composed of a word

and its grammatical tag (tag://word). For example,

if you write the verb “be” and wish to colour it, the

viewer will create a “verb://be” grammatically tagged

link. We chose to represent the links like this for im-

plementation reasons. In fact, this was the clearest

way to enable coloration and clickability of words in

Snowglobe, since it is the same structure as the usual

“http://website” links. The main format is MLIF; the

“tag://word” is only a bridge between MLIF and the

chat interface in Snowglobe.

The addTags function matches a word with its

grammatical category in order to create a grammati-

cally tagged link and to apply the colour chosen by the

user. We have deﬁned as many grammatically tagged

link types as there are grammatical categories.

When user A wants an instant translation in their

language of what user B says, the function buildMLIF

requests the web service to return a MLIF data struc-

ture with user B

ıs sentence in its original language

and the same sentence translated in user A’s language.

Both original and translated sentences are stored in

the MLIF data structure. Indeed, user A can consult

the original sentence if the translation is odd or if they

want to learn the other language. The MLIF data is

parsed by theparseMLIFAndReturnTranslatedWords

function, which returns translated words.

All those functions are not called at the

same time. A menu proposes several options

(see 4.2). Each option is saved in an XML ﬁle

(panel preferences linguistic chat.xml), which is

managed by the functions deﬁned in the ﬁles

llprefslingchat.cpp, llﬂoaterpreference.cpp and

llstartup.cpp. The appendStyledLinguisticText

function in the LLTextEditor class can verify

if an option is activated before calling the as-

sociated function. For instance, if the variable

linguisticChatActivateInstantTranslation is ac-

tivated, the viewer will use the function which

allows an instant translation of a sentence to be

displayed(parseMlifAndReturnTranslatedWords).

4.5.2 Solipsis

Solipsis is a French virtual world (still under develop-

ment), which is notable for two features. First, it plans

to implement a decentralized Metaverse platform that

will use a peer-to-peer protocol. Second, it inte-

grates a web browser directly in the three-dimensional

space. All the objects created can embed a web

browser (as a texture applied on a surface), which

is compatible with many web technologies (Flash,

Javascript, etc.).

The Multilingual-Assisted Chat Interface on

Solipsis is roughly similar to the Snowglobe version.

The general development is typically the same but the

structure is different.

STANDARDS FOR COMMUNICATION AND E-LEARNING IN VIRTUALWORLDS - The Multilingual-assisted Chat

Interface

The graphical chat interface is deﬁned in an HTML

ﬁle (uichat.html) and built with HTML and Javascript.

All the classes and functions we have described for

the Snowglobe version are written in C++ in the ﬁles

GUI Chat.cpp and GUI Chat.h.

The HTML ﬁle contains a form linked to a

Javascript function (sendMessage). This calls a C++

function of GUI chat.cpp (addText). This function

decides, with respect to the user’s preference of

whether to use the linguistic functionalities or not, to

convert the text to MLIF data or not. In other words,

it calls one of the two sending methods declared in the

HTML ﬁle:

1. sendTextMessage, which writes text into the chat

interface without data processing;

2. sendMLIFMessage, which converts MLIF data

to human-readable text (with the function

convertMlifToText), and matches the tags with

both corresponding colours and links (with the

function colorWordByTagAndMatchLink).

The menu is implemented in the HTML ﬁle

menuLinguisticChat.html and the actions that the user

can execute with the in-world interface menu are de-

clared with showMenu C++ function. When a user

clicks on a word, a new in-world web browser appears

thanks to the showBrowser C++ function.

5 CONCLUSIONS

First, we would like to highlight again that all the data

exchanges are made using MLIF, because we intend

to enhance interoperability through standardization.

We believe standardization to be a key issue in the

development and dissemination of new technologies.

Up to now, the grammatical coloration of the

words in our Multilingual-Assisted Chat Interface in-

terface is only available in English. In fact, such an

analysis has to be based on linguistic corpora and the

one we use only contains English terms. However, we

plan to adapt our technology to other analysers such

as Treetagger.

In the same way, the translations rely on the

Google Translate web application. As its quality of

translation is improving every day (thanks to its “sug-

gest” functionality), we can expect better translations

in the future.

The other important point which should be noted

is that the web service which we developed has been

used for both Second Life and Solipsis. Externalizing

applications (especially service applications) is very

useful in order to minimize the development time and

to be able to use more adapted data structures.

But the Multilingual-Assisted Chat Interface is

still a prototype. In the future, we intend to improve

the reliability of this interface. We will distribute

the modiﬁed Snowglobe viewer to a large number of

users for testing and collect feedbacks. Adding new

features to the chat interface in order to facilitate so-

cial interactions and e-learning is a process which still

requires time, research and new technologies.

REFERENCES

Brill, E. (1992). A simple rule-based part of speech tag-

ger. In Proceedings of the third conference on Applied

natural language processing, pages 152–155. Associ-

ation for Computational Linguistics.

Coleman, D. (2007). Flame First, Think Later: New Clues

to E-Mail Misbehavior. New York Times.

Cruz-Lara, S., Bellalem, N., Bellalem, L., and Osswald,

T. (2009). Immersive 3D Environments and Mul-

tilinguality: Some Non-Intrusive and Dynamic e-

learning-oriented Scenarios based on Textual Infor-

mation. Journal of Virtual Worlds Research, 2(3).

Cruz-Lara, S., Romary, L., Ducret, J., Bellalem, L., Guin-

aud, J., and Osswald, T. (n.d.). MLIF: the MultiLin-

gual Information Framework. http://mlif.loria.fr/.

De Lucia, A., Francese, R., Passero, I., and Tortora, G.

(2009). Development and evaluation of a virtual cam-

pus on Second Life: The case of SecondDMI. Com-

puters & Education, 52:220–233.

Doughty, C. J. and Long, M. H. (2003). Optimal Pshy-

cholinguistic Environments for Distance Foreign Lan-

guage Learning. Language Learning & Technology,

7(3):50–80.

El Jed, M., Pallamin, N., and Pavard, B. (2006). Vers

des communications situ

ees en univers virtuel. In

National Conference on: Coop

eration, Innovation et

Technologie.

Suler, J. (2004). The Online Disinhibition Effect. Cy-

berPsychology and Behavior, 7:321–326.

ICEIS 2010 - 12th International Conference on Enterprise Information Systems