Sign Language (SL) in Graphical Form : Methodology, modellisation and representations

Authors : Brigitte Garcia (coordinator)¹, Dominique Boutet², Annelies Braffort³, Patrice Dalle^4
1 UMR 7023 SFL, Paris, et Université Paris 8, Brigitte.Garcia@univ-paris8.fr
² UMR 7023 SFL, Paris et Université EVE, dboutet@noos.fr
³ LIMSI-CNRS, UPR 3521, Paris, annelies.braffort@limsi.fr
⁴ UMR 5505 IRIT-TCI, Toulouse, and Université Toulouse 3, dalle@irit.fr

Abstract

This paper presents a French national multidisciplinary Project (LS Script), whose aim is the elaboration of a graphical form for French Sign Language (LSF). It includes the presentation of three of the steps that have been achieved or undertaken up until now :

• Assessing the results of a qualitative survey, of practices and expectations of LSF signers relative to a graphical representation of their language (LSF). This investigation allows us to define the functions expected from the projected formalism, so as to determine the type of encoding to be conceived.

• Setting up the theoretical and methodological grounds we establish in order to explore the lowest structural level in LSF, so as to determine a viable minimal graphical unit for the formalism to be conceived. This exploration is grounding in Cuxac's hypothesis of a morphemic rather than phonemic organization of this lowest level.

• Developing digitized representations of signs, of signing space and of the signer's face to allow the analysis of a video corpus, aid in transcription, encoding of the formalization, writing editors and reading assistance with a signing avatar and the development of educational applications in schooling, in distance communication, in LSF, and in the Internet use.

Key-Words : Sign Language, Writing, Morpheme, Database, Avatar

The multidisciplinary project presented here¹ aims at creating a graphical form of French sign language (LSF). The key question that motivates this project is the following : for deaf citizens of literate societies, whose language is an institutionalised SL, what can fulfil cognitive, social, and pedagogical functions that the written form of language is used to fulfil ? The main specificity of our programme is its grounding in a preliminary qualitative investigation of the population concerned. At the same time, allowing written communication on the basis of a gestural language opens the way for numerous interactive applications based on the gesture.

A variety of notation systems for SL have been elaborated². However : 1) for the vast majority, these are transcription systems dedicated only to research, and not systems aiming at permitting written exchanges and besides, 2) these are systems conceived from de-contextualised lexical signs (outside discourse). In this sense, they do not take into consideration what seems to be the most striking structural specificity of SL : the multi-linear use of corporal parameters which are specialised linguistically (gaze, facial expressions, facial and corporal movements, hands) ; massive iconicity ; the relevant use of space and of space-time and the crucial modifications which spatialization induces in the form of signs³.

This paper presents three of the steps which have already been completed or undertaken :

1. The results of a qualitative survey of practices and expectations of LSF signers relative to a graphical representation of their language. Such an inventory, which aims at defining the functions expected from a graphical system, plays an essential role in determining the encoding to be conceived (notation of the form and/or notation of the morphemo-graphical type).

2. In order to determine the choices to be made in the elaboration of the structural bases for a graphical formalism, the exploration of the lowest structural level in LSF ; the crucial point here is to choose a viable minimal graphical unit for the projected formalism.

3. The refinement of computer models and digitised representations of signs, of signing space and of signer's face, allowing for :

   • the analysis of a video corpus and aid in transcription,

   • the encoding of the new formalism (Unicode) and assistance in its editing (writing editors),

   • assistance in reading (through a virtual animation of a signing humanoid),

   • the development of educational applications in schooling, in distance communication, in LSF, and in Internet use.

1. Which written forms are adequate to French Sign Language (LSF) ? The Results of a qualitative survey of LSF signers

There are two main motivations for our choice of basing our enterprise on a qualitative survey. The first of these is moral and political : how can we imagine providing a language with a written form without, from the outset, involving its community of speakers ? Our other motivation is theoretical. Any system of notation is based not only on the integration of hypotheses concerning what structures language, but also on formal and implicit choices that depend on the potential uses and potential users of the system : it is essential to define these.

Beyond a preliminary investigation into the notion that the Deaf we encountered have of writing in general, we tried to ascertain their relationship to their actual uses and practices of graphical representation, and to evaluate with them their expectations in this matter. Our investigation was meant to be open, prospective and not only limited to graphics in a strict sense ; we include indeed an analysis of the role of video, and new visual technologies functioning as written forms, and explore the possible interest of providing LSF with a specific written form.

1.1. Methodology

In order to carry out a thorough investigation on graphical representations, uses and practices, individual, semi-guided interviews were necessary. We carried out 31 two-hour interviews. Aiming to better understand the actual diversity in terms of geographical location, profession, age and personal history, we have nevertheless chosen to prefer the milieu of teaching - the teaching of, and with LSF, where the subject is crucial. In a society where teaching remains mainly structured by written forms, how can we imagine generalising the teaching of, and with LSF without asking what will function as a written form for this language ? In regards to teaching in schools, our study being prospective, we have also given greater place to bilingual structures of education (these structures represent only 1% of the educative structures for the deaf children today). The choices made imply that this survey does not claim to take into account the situation of all the deaf in France.

The persons we encountered are adult deaf signers, aged 24 to 59. With the exception of one person who was schooled from primary school onward in bilingual structures, the interviewees were in specialised schools and /or in integration, and therefore had an "oralist type" of schooling. This reflects the educational situation of the deaf in France. The table below shows the distribution of the persons encountered, according to their profession ; they were natives of various French cities : Paris (12), Toulouse (10), Bayonne (2), Poitiers (1), Bordeaux (1), Champ sur Marne (1), Grenoble (1) et Suresnes (1).

We set the number of interviewers at 5, all hearing and familiar with the Deaf community. The reason for these interviewers being hearing, and not deaf, results from the fact that up to now no deaf person is both trained in the problems under study and in conducting such interviews. We adopted the framework of a three-person team, interviewer, interviewee and interpreter. The exchange was recorded both on audiotape and on videotape. The interviews were transcribed into written French in their entirety. We will show here only some of the main results of the survey, to the extent that they directly determined the orientation our research was given.

1.2. Practices and Needs

Our first major observation was that alongside written French, there are numerous specific graphical practices, i.e., which aim at a notation of LSF, independently of any degree of mastery of written French. These are practices maintained for oneself, or between deaf persons. The most frequent case is, even if one masters written French, a specific use of this form that the interviewees designated as equivalent either to LSF-French or to written LSF : the ordered alignment of French words, according to them, in the syntax of LSF. The weaker the mastery of French, the more this LSF-French is mixed with drawings, up to the point where it is essentially composed of drawings and other graphical symbols. This process can lead to the perfecting of completely original systems, from which all "word-signs" are absent. However, we will henceforth insist on situations where the deaf we encountered have recourse to these specific practices, even when they have mastered written French - since, as they say, it does not satisfy their needs.

It is firstly in situations where one has recourse to a graphical support to aid in the construction of a reflection, especially when this has to lead to a production in SL (conferences, appointments). One then either utilises only written French but limits his/herself to a list of "keywords" or "key phrases", presented as simple "reminders" referring more to ideas conceived in LSF than to French. Therefore, most often, and especially for those who have little or no mastery of written French, one has recourse to the continuum evoked above which ranges from "LSF-French" to mere idiosyncratic graphical symbolisation. In any case, however, as the concept functions in LSF, this situation is presented as one of the most unsatisfactory or most frustrating -while the use of the written word as support for a cognitive elaboration leads back to one of the key functions of any written form of language.

Another situation that becomes a problem is that of a support for an oral presentation in LSF - the equivalent of written notes for a hearing lecturer. The particular difficulty here is that even if one feels at ease in written French, its utilisation as a support perturbs the fluency of production in LSF and gives rise to a production of 'signed' French. Here again, many solutions can be found, but are always described as less than satisfying. The two problems that we have just evoked crop up, according to the teachers we encountered, in LSF teaching in a school setting, when the pupils have to prepare (in class or at home) then sign a production in LSF in front of the camera for exercises of auto-correction or for evaluation (the composition of hearing pupils).

We will evoke two other cases which pose a problem. The first is note-taking in a course, a conference or a meeting in LSF ; all recourse to French is often said to be inadequate and worse if one wants to conserve a specific formulation in LSF. There is the case, moreover, in which one creates a literary or artistic production in LSF, and where one wants recourse to a graphical expression which will surely allow it to be memorised and which will allow it to be re-worked in detail -and for which video is not sufficiently flexible. This is a case that could have given rise to the most systematic graphical inventions, even if their use is usually limited to a small number of persons. A similar situation, and one just as rich in idiosyncratic graphical practices, is encountered by actors and directors for the graphical adaptation to LSF of a text in written French, but also of their personal interpretation, whether they want to memorise it or to re-work it.

According to the interviewees, however, a part of the functions that writing fulfils specifically for the language of which it is the written form are, for SL, already - or could be - fulfilled by video and the new visual technologies -and indeed, most of them consider it as "the written form of LSF"- : preservation and archiving, pedagogical support for teaching (in) LSF (feed-back for a production, support and record of lessons, support for evaluations), support normalising the language (discourse referents, video dictionaries), and as a means of dissemination. For a few, the rapidity of technological progress allows them to predict that certain present limitations of video - difficulty of use, especially for editing and selection of data for a document - will soon be resolved.

However, as the majority of interviewees have said, certain limits will not be resolved. The first is linked to access to data banks (search engines). The predicted potential in the medium term regarding digital image recognition (movement and form) are a long way from being the economical equivalent of those representing the expression of query via a specific graphical representation of the language data sought. More fundamentally however, for the majority of persons encountered, video presents intrinsic limitations, which prevent it from specifically playing the role of support for the elaboration of a reflection, and for which as we have seen, written French, even when mastered, is unsatisfactory. This is due firstly to the maintaining of the visible physical presence of the 'speaker' : video, by this very fact, remains bound to face to face communication ; it forbids, above all, the distancing that the written form authorises, a fortiori when it's a matter of one's own image. It is secondly due to the fact that, in production as in "play" - because video as it plays does not allow for the apprehension of what is being recorded and of what has already been recorded - a simultaneity which is the very basis for the potential of "graphical reasoning".

1.3. Implications

It is above all in teaching that difficulties linked to the absence of a specific written form for LSF are to be found in great number. We mentioned above the problems encountered in the teaching of LSF to deaf children, regarding the conception, then the production in front of the camera, of a "composition" in LSF. There are also the difficulties evoked by teachers of written French in LSF, who are tied, in their view, to an imbalance between a basically oral language (without its own written form), SL, and French language, which disposes of two registers. They note thus a double difficulty : first, when they want to work starting from a creation by the children done in LSF (the classic stage of the search for ideas) and to go towards written French, the necessity of passing by a sort of graphical formalisation of the content in the SL produced, which is "slapped together" ad hoc and which does not satisfy them. Then there is, for the study (in LSF, the language of instruction) of a French text, the feeling that they have, by sticking to the translation of the content into LSF, of missing what defines a written text as such.

These various points confirm the interest in elaborating a system that allows for written production in LSF. They make it possible, moreover, to specify the foreseeable options according to the desired functions. Our survey bears out the fact that, for the most expected functions, especially those of support for cognitive elaboration, for a search for ideas, for note taking, for the formulation of a query in a database, one may conceive a type of morphemo-graphic or at least mixed notation (morpho-phonographical).

2. Database of standard signs with morpho-phonetic structuring in French sign language (LSF)

2.1. Problem

As we said above, a preliminary necessity for this project of graphical formalisation is to determine the minimal graphemic units and to choose the nature of encoding (Garcia & Boutet, 2003, 2006). We proceed on the basis of Cuxac's theoretical hypothesis (Cuxac, 2000, 2004) which favours a morphemic rather than a phonemic organisation at the lowest level in LSF, and allows a morphemic composition of both standard signs (STS) and structures said to be of high iconicity (SGI) -"productive signs"-, these including "Personal Transfers" (role-play), "Transfers of size and of form", and "Transfers of situation". The systematic exploration of this hypothesis in the LSF lexicon currently on record (Girod et al., 1997) will make it possible to establish low-level morphemic components, to assess their productivity and to set out some rules for their compositional potential. Finally, it allows for testing of the validity of a strictly parametrical approach and moreover for evaluating the merits of a graphical formalisation which is at least partially founded on a notation of the signified.

The implications of this project are of two different orders :

1. It will serve to find an alternative to current notation systems for SL. These systems encode the mere visual aspect of sub-lexical parametric units (configuration, orientation, location, movement, ± contact, ± facial expression) - that is to say the graphical principle they use is based on that which structures alphabetical systems invented for vocal languages. Beyond the acknowledged limits of present systems for the notation of discursive phenomena in SL, the interest of identifying a low-level morphemic economy is :

   1. to take an account of the structural specificity of SL, especially of HIS and their permanent interaction in discourse with the STS. In our theoretical framework STS and HIS use a common set of morphemes : thus we can take the inventory of graphical units that are transversal to both types of structures and also graphically take stock of numerous discursive phenomena of co-referentiality and of the differentiated utilisation of space (metrical egocentric space, allocentric and topological spaces) ;

   2. to provide an answer to the recurrent problem of the cumbersomeness of existing notation (more than 600 graphic symbols for Sign Writing⁴) : a graphical encoding that gives direct access to meaning (morphemo-graphic notation), at least for part of the notation, inducing a lower cognitive cost.

2. It will corroborate Cuxac's hypothesis, through a formalisation of the low-level morphemic stratification. As Cuxac's model is grounded in the assumption of a morpho-phonetic structure (form-meaning constants) anchored in the perceptivo-practical (anthropological structuring), this formalisation should, in due time, furnish a solid basis of comparison with other SLs and allow an evaluation of the part to be attributed to cultural specificity.

2.2. Methodology

A database was created from 4000 STS currently listed, and checked by native signers. The analysis of this corpus focused on configuration and location parameters, for which the morphemic values were inventoried, sign by sign, and on the movement relative to the principal components of each configuration. Regarding the configuration, the approach was carried out using configurations classified by Girod (1997), and by distinguishing, on the basis of the inventory of configurations already established by Cuxac for HIS (Cuxac, 2000, pp. 97-130), the configurations attested both in HIS and in STS, and those that exist only in STS. Due to the lack of any pre-existing classification of types of location, the analysis for this parameter follows an ad hoc logic. The direction of movement was described according to salient characteristics of the configuration (planes, axis, points) and can be re-contextualized for each configuration. Generally speaking, for this parameter and for the orientation parameter, the description is made relatively to a position and/or to a segment, including the other hand, thus offering the possibility of moving the frame of reference so as to better understand the different utilisations of space.

The main methodological difficulty is due to the validation of morphemic values retained, and to their labelling. There were three types of validation of those units of meaning : 1) the internal economy of morphemic networks that were identified ; 2) the productivity of these form-meaning constants (for one meaning reposing on several parameters, preponderant parametric form is retained for graphical encoding) and their capacity to explain neologisms ; and 3) the feed-back of native signers.

2.3. First Outcome

For the parameters under study, the analysis confirms the existence of an optimised economy of morphemic values (2 to 4 per parametric unit) and the coincidence of the most productive of them with those values attested in the HIS. At the same time, the study allowed us to identify, for a single parametric unit attested in both types of structures, which morphemic values were specific to STS. Our analysis also brings out a stratified organisation : around a large morphemic value we find an organisation in families, the signs of each family being linked together by the exploitation of a particular semantic specialisation of this morphemic value. In regards to location, the entry by morphemic value furnishes criteria of classification and points up interdependencies with the choice of configurations (which is particularly the case of locations on the body). A graphic economy can therefore be achieved at this level. Moreover, the analysis brings out a model for a semantic segmentation of the body surface and of the signing space. It displays numerous concordances with semantically predetermined areas used in the spatialization of certain syntactic relations - hypotheses, time, choice of loci - (Engberg-Pedersen, 1993), shedding new light on the relations between lexical and grammatical levels in SL. The relative direction of movement reveals a great economy of means in the structure of STS, and provides a strong link between the configuration and types of movement. This leads to a very restricted ensemble of items which can be used for notation.

In this way, morphemic structuring has been found to be largely confirmed. The existence of several values for a single configuration requires one to consider either an infra-parametric mode of composition (Boutet, 2005), or a crystallised deposit on a particular parametric value of a previous composition (diachrony), as seen in Bonnal (2005). For these two options, merely analysing the parametric level cannot exhaust the resources of elaboration of meaning in LSF. Inter-parametric links, especially between configuration and movement parameters, argue in favour of a consideration of an articular level of production that would be more phonetic than phonemic. However, a certain economy of graphemic encoding is henceforth possible.

3. Sign language processing : models, representations, tools for video analysis, for signing avatars and for communication

This section is related to the computer science aspect of the project. Our aim is to study the basis needed for the design of educational tools. These tools aim at facilitating the writing and the reading of utterances in LSF.

As said earlier, the notations will be based on visual aspects and on aspects of meaning. This implies bottlenecks for computing modelling at several levels : at the lower for modelling the constitutive parameters of the gestural units ; and at the higher, for modelling utterance representations of any kind, including those of high iconicity described by Cuxac (2000, 2004).

These models must include all kinds of linguistic phenomena needed for writing or reading, including manual and non manual representation, spatio-temporal rules, and cognitive knowledge. In this section, we present two studies related to the utterance level : the first one is related to the modelling of spatial relations in LSF, and the second is related to the modelling of the construction of the narration scene.

3.1. Modelling of spatial relations

Sign languages allow us to show in space what is said, as for example spatialized semantic relations. These relations can be of different kinds. They can be static - as in "the cat is in the car" - or cinematic, when we can distinguish initial and final situations - as in "the cat jumps into the car". When the action is controlled by one of the protagonists of the situation, like in "I put the cat in the car", the relationship is dynamic. Depending on the type of the relationship, more or less movement will be present in the hand, arm and body of the signer. For example, in the case of a dynamic situation, the signer will add a body movement to the realisation of the predicate sequence of signs, to express the involvement of one of the protagonists.

The modelling of such kinds of properties is based on cognitive grammars, which are based on abstract schemes anchored in mental representations, and build from the experiences of perception and action on the world.

Our modelling is based on the LSF model proposed by Fanch Lejeune (2002, 2001). The aspects we are taking into consideration in the model are :

• the preferential order of the gestural units that will be represented as grammar rules,

• the selection of two locations in the signing space in order to show the relationship,

• the use of gaze to instantiate locations or to refer to these locations,

• the use of proforms, which give a point of view on the entity or on the relation.

The formalism is based on a formal notation using typed operators and operands (Braffort, 2005). Each scheme is associated to an operating sequence that allows us to generate a representative utterance.

For example, to represent a geographical relationship between two locative entities, such as in the sentence shown in Figure X1, where University of Paris 8 is situated in relation to Paris, we manipulate the following scheme :

<POINT(university) REP (IN(PROJ(FR(DET(LOC(Paris)))))))>

where ‘university' is located by means of a pointing gesture (POINT operator) directed on the location in space related to the entity ‘Paris' (REP relator), considered as a location (LOC operator), which is determined by its frontier (DET operator), which is the base of an oriented projection (PROJ operator) defining a new location, and the university is located in this new location (IN operator).

Figure Example 1 : LSF sentence "The Paris 8 university is situated in relation to Paris"
1 : [PARIS]_LSF, 2 : Frontier of Paris, 3 : [UNIVERSITY]_LSF, 4 : spatial relation

This modelling is used to automatically generate a representative sentence performed by a signing avatar (see Fig. 2).

Figure Example 2 : Elsi, the LIMSI signing avatar

The computing system, not detailed here, is composed of several knowledge databases, about entities, geographic relations, and proforms. A representation of the signing space allows the various locations to be stored and used for proform locations and gaze direction. More details can be found in Braffort (2005a, 2005b).

This study is a first step toward a more general modelling that will include other kind of representations, and a more sophisticated representation of the signing space, as the one proposed in the next section.

3.2. Modelling of the construction of the narration scene

The second study is related to the construction of the narration scene and the sequence of gestures produced to build the scene. The narration scene contains all the elements to which the signer and his addressee make references. It is the framework for structuring discourse. Its modelling enables an image analysis system to anticipate and interpret visual events in linguistic terms and produce a syntactico-semantic description of the discourse. Conversely, a graphic editor interpreting this representation should allow us to generate the structural framework of an utterance produced by a signing avatar.

The proposed model is composed of two parts : the narration scene construction and the gesture sequence used to update the scene.

The modeling of the narration scene is represented by a UML class diagram (Fig. 3) which shows the various concepts and their relations. The signing space is an oriented 3d volume around the signer. It is divided into sites, which contain referents. This volume is used to locate the entities.

Our model is an object hierarchy with a mechanism that insures consistency of the construction when a new entity is added. The representation allows all entities that may be further referenced in the discourse to be stored.

Figure Example 3 : UML class diagram of the semantic representation of the signing space

There is a set of sites ; each site may contain entities. The entities have several functions : temporal, actions and spatial location. There are different kinds of entities, according to their potential functions :

• date can be temporal relationship,

• action can be an agent or temporal location,

• animate can be implied in action or used to locate others agents,

• location can only be used as locative.

Entities can be used in shift role (person) or not (object). These kinds of entities are only defined by their potential function and not on a lexical categorization base.

The video (Fig. 4) shows an example, extracted from a LSF corpus of requests about a movie theatre program, with its signs to words translation. On the right, we have a symbolic 3d representation that is built with an interactive tool that implements our model. We can see how entities are located in the signing space : Two locations (the city of Toulouse and the movie theatre), an "object" (the movie), a date, a person and an action (directional verbs), which links the person to the notion of movie. An internal XML representation keeps track of this construction.

For automatic SL processing, we need a second model, a formal representation of sign language grammar that links the narration scene construction to the sequence of gestures used to add a new element in the scene.

This model represents our knowledge about the gestures that are used to create an entity of a given type at a given place in the narration scene, and about the aspects of the components that perform theses gestures, in a reduced context.

We use the description logic as formalism to represent this knowledge in the computer.

The main concept is the notion of ACT (an ACtion Transforming the narration scene), a hierarchy that describes the creation of an entity in a given location by means of a gesture sequence associated with the corresponding time interval.

ACTS contains a gesture sequence for each kind of ACT and for each part of the sequence there are different ways to evoke the entity.

The gesture concept hierarchy consists in a description of the signer's body components used to evoke the entities.

Finally, time intervals are used for gesture synchronization by means of rules that define their order in a sequence, based on the interval algebra.

Thus, if we want to automate the video analysis to help writing sign language, we can use an image analysis system to verify the state of the component (head, hand...). If the predicted state is found, the creation of the entity is validated and is added in the narration scene description with the correct type.

These two studies are our first attempts at modeling high-level knowledge to be integrated into future tools dedicated to writing and reading graphical forms of LSF.

The proposed models should be compatible with both shape-based notation and morphemographic or morpho-phonographic notations.

The next steps in these studies include three levels of integration :

• integration of the two proposed models,

• integration with the lexical representation presented in section 3,

• connection with the future graphical form.

Figure Example 4 : construction of the signing space during the realization of a sentence.

Bibliography

Bonnal, F. (2005). Sémiogenèse de la langue des signes française : étude critique des signes attestés sur support papier depuis le XVIIIe siècle et nouvelles perspectives de dictionnaires . Toulouse : Université Toulouse Le Mirail.

Boutet, D. (2005). Pour une iconicité corporelle. Proceedings from TALS 2005 : Traitement automatique des langues des signes (TALN workshop), 6-10th june 2005, Retrieved from http://tals.limsi.fr/actes/s1.pdf

Braffort, A., & Lejeune, F. (in press). Spatialised semantic relations in French Sign Language : toward a computational modelling. In S. Gibet, (Ed.), Selected revised papers from the International Gesture Workshop : Gesture in Human-Computer Interaction and Simulation. LNAI, Springer.

Braffort, A., Bossard, X., Segouat, J., Bolot, L., & Lejeune, F., (2005). Modélisation des relations spatiales en LSF. Proceedings from TALS 2005 : Traitement automatique des langues des signes (TALN workshop) 6-10th june 2005. Retrieved from http://tals.limsi.fr/actes/s12.pdf

Cuxac, C. (2000). La langue des signes française (LSF). Les voies de l'iconicité. Faits de Langues 15-16. Paris : Ophrys.

Cuxac, C. (2004). ‘Phonétique' de la LSF : une formalisation problématique. In A-M. Berthonneau & G. Dal. (Eds). (2003, September). Linguistique de la LSF : recherches actuelles. Proceedings from de Villeneuve d'Ascq Symposium. Silexicales (pp. 93-113). Lille : Université de Lille 3.

Engberg-Pedersen, E. (1993). Space in Danish Sign Language. Hamburg : SIGNUM-Verlag.

Dalle, P., & Lenseigne, B. (2005, May). Vision-based sign language processing using a predictive approach and linguistic knowledge. IAPR-MVA : International Association for Pattern Recognition Conference on Machine Vision Applications (pp.510-51). Japan : Tsukuba Science City.

Dalle, P., & Lenseigne, B. (2005). Modélisation de l'espace discursive pour l'analyse de la langue des signes. Proceedings from TALS 2005 : Traitement automatique des langues des signes (TALN workshop), 6-10th june 2005. Retrieved from http://tals.limsi.fr/actes/s7.pdf.

Garcia, B., & Boutet, D. (2003). Vers une formalisation graphique de la Langue des signes française (LSF) : éléments d'un programme de recherché. In F.Bertin & C. Cuxac, (Eds.), Langue des signes française (LSF), enjeux culturels et pédagogiques, La Nouvelle Revue de l'AIS, 23, 49-62. Suresnes : Editions du Cnefei.

Garcia, B., & Boutet, D. (2006). Finalités et enjeux linguistiques d'une formalisation graphique de la langue des signes française (LSF). In R. Sabria (Ed.), Glottopol 7. Les langues des signes (LS) : recherches sociolinguistiques et linguistiques. Retrieved 2006, http://www.univ-rouen.fr/dyalang/glottopol/numero_7.html

Girod, M., et al. (1997). La langue des signes, dictionnaire bilingue LSF/français [Sign language, LSF/French bilingual dictionary]. (Vol. 2-3). Paris : IVT Éditions.

Lejeune, F., & Braffort, A. (2002). Study on semantic representations of French sign language sentences. In I. Wachsmuth & T. Sowa (Eds.), Gesture and Sign language based human-computer interaction. LNAI : Lecture Notes in Artificial Intelligence, 2298 , Berlin, Springer Pub.

Lejeune, F., Braffort, A., & Desclés, J. (2001, June). Analyse sémantico-cognitive de quelques structures en langue des signes française. In C. Cavé, I. Gaïtella & S. Santi (Eds.), Conference proceedings from Orage 2001, Aix en Provence : Oralité gestualité, interactions et comportements multimodaux dans la communication (pp. 334-339). Paris : L'Harmattan.

Lenseigne, B., & Dalle, P. (2006). Using signing space as a representation for sign language processing. In S. Gibet, N. Courty & J.F. Kamp (Eds), Gesture in Human-Computer Interaction and Simulation, LNAI : Lecture Notes in Artificial Intelligence, 3881. Berlin, LNAI Springer.

Miller, C. (1994). A note on notation. Sign Post : vol. 7/3. Durham : University of Durham.

Miller, C. (2001). Some reflections on the need for a common sign notation. In

R. Wilbur (Ed.), Sign Language & Linguistics, vol. 4 : 1/2, 11-28. Amsterdam / Philadelphia : John Benjamins Publishing Company.

Pizzuto, E., & Pietrandrea, P. (2001). The notation of signed texts : open questions and indications for further research. Sign Language and Linguistics, (Special Issue on Sign Transcription and Database Storage of Sign Information), 4 : 1/2, 29-43, John Benjamins Publishing Company, Amsterdam/Philadelphia.

Sutton, V. (1999). Lessons in SignWriting. Textbook & workbook (2^nd ed.). La Jolla : Deaf Action Commitee for Sign Writing.

 Notes

1 The work presented was carried out in two successive frameworks : a project subsidized by the DGLFLF, Ministry of Culture and Communication (2003-2005) ; and project RIAM, LS Script (2005-2007, Minister delegated to research). They associate researchers in humanities and social sciences (linguistics, psycholinguistics and sociology), researchers in computer science (IRIT-TCI Toulouse 3 and LIMSI Orsay), the IRIS institute of Toulouse and the WebSourd company of Toulouse, provider of services for the deaf and new technologies.

2 For an inventory, see Miller (1994, 2001) and for a critique Pizzuto & Pietrandrea (2001)

3 See Boutet & Garcia (2003)

4 Sign Writing System (Sutton, 1999)