Apprentissage automatique de la morphologie : le cas des structures racine-schème

RÉSUMÉ Cette thèse porte sur le développement de méthodes algorithmiques pour découvrir automatiquement la structure morphologique des mots d'un corpus. On considère en particulier le cas des langues s'approchant du type introflexionnel, comme l'arabe ou l'hébreu. La tradition linguistique décrit la morphologie de ces langues en termes d'unités discontinues : les racines consonantiques et les schèmes vocaliques. Ce genre de structure constitue un défi pour les systèmes actuels d'apprentissage automatique, qui opèrent généralement avec des unités continues. La stratégie adoptée ici consiste à traiter le problème comme une séquence de deux sous-problèmes. Le premier est d'ordre phonologique : il s'agit de diviser les symboles (phonèmes, lettres) du corpus en deux groupes correspondant autant que possible aux consonnes et voyelles phonétiques. Le second est de nature morphologique et repose sur les résultats du premier : il s'agit d'établir l'inventaire des racines et schèmes du corpus et de déterminer leurs règles de combinaison. On examine la portée et les limites d'une approche basée sur deux hypothèses : (i) la distinction entre consonnes et voyelles peut être inférée sur la base de leur tendance à alterner dans la chaîne parlée; (ii) les racines et les schèmes peuvent être identifiés respectivement aux séquences de consonnes et voyelles découvertes précédemment. L'algorithme proposé utilise une méthode purement distributionnelle pour partitionner les symboles du corpus. Puis il applique des principes analogiques pour identifier un ensemble de candidats sérieux au titre de racine ou de schème, et pour élargir progressivement cet ensemble. Cette extension est soumise à une procédure d'évaluation basée sur le principe de la longueur de description minimale, dans- l'esprit de LINGUISTICA (Goldsmith, 2001). L'algorithme est implémenté sous la forme d'un programme informatique nommé ARABICA, et évalué sur un corpus de noms arabes, du point de vue de sa capacité à décrire le système du pluriel. Cette étude montre que des structures linguistiques complexes peuvent être découvertes en ne faisant qu'un minimum d'hypothèses a priori sur les phénomènes considérés. Elle illustre la synergie possible entre des mécanismes d'apprentissage portant sur des niveaux de description linguistique distincts, et cherche à déterminer quand et pourquoi cette coopération échoue. Elle conclut que la tension entre l'universalité de la distinction consonnes-voyelles et la spécificité de la structuration racine-schème est cruciale pour expliquer les forces et les faiblesses d'une telle approche. ABSTRACT This dissertation is concerned with the development of algorithmic methods for the unsupervised learning of natural language morphology, using a symbolically transcribed wordlist. It focuses on the case of languages approaching the introflectional type, such as Arabic or Hebrew. The morphology of such languages is traditionally described in terms of discontinuous units: consonantal roots and vocalic patterns. Inferring this kind of structure is a challenging task for current unsupervised learning systems, which generally operate with continuous units. In this study, the problem of learning root-and-pattern morphology is divided into a phonological and a morphological subproblem. The phonological component of the analysis seeks to partition the symbols of a corpus (phonemes, letters) into two subsets that correspond well with the phonetic definition of consonants and vowels; building around this result, the morphological component attempts to establish the list of roots and patterns in the corpus, and to infer the rules that govern their combinations. We assess the extent to which this can be done on the basis of two hypotheses: (i) the distinction between consonants and vowels can be learned by observing their tendency to alternate in speech; (ii) roots and patterns can be identified as sequences of the previously discovered consonants and vowels respectively. The proposed algorithm uses a purely distributional method for partitioning symbols. Then it applies analogical principles to identify a preliminary set of reliable roots and patterns, and gradually enlarge it. This extension process is guided by an evaluation procedure based on the minimum description length principle, in line with the approach to morphological learning embodied in LINGUISTICA (Goldsmith, 2001). The algorithm is implemented as a computer program named ARABICA; it is evaluated with regard to its ability to account for the system of plural formation in a corpus of Arabic nouns. This thesis shows that complex linguistic structures can be discovered without recourse to a rich set of a priori hypotheses about the phenomena under consideration. It illustrates the possible synergy between learning mechanisms operating at distinct levels of linguistic description, and attempts to determine where and why such a cooperation fails. It concludes that the tension between the universality of the consonant-vowel distinction and the specificity of root-and-pattern structure is crucial for understanding the advantages and weaknesses of this approach.

Validity, reliability and responsiveness of the French language translation of the Western Ontario Shoulder Instability Index (WOSI).

BACKGROUND: The WOSI (Western Ontario Shoulder Instability Index) is a self-administered quality of life questionnaire designed to be used as a primary outcome measure in clinical trials on shoulder instability, as well as to measure the effect of an intervention on any particular patient. It is validated and is reliable and sensitive. As it is designed to measure subjective outcome, it is important that translation should be methodologically rigorous, as it is subject to both linguistic and cultural interpretation. OBJECTIVE: To produce a French language version of the WOSI that is culturally adapted to both European and North American French-speaking populations. MATERIALS AND METHODS: A validated protocol was used to create a French language WOSI questionnaire (WOSI-Fr) that would be culturally acceptable for both European and North American French-speaking populations. Reliability and responsiveness analyses were carried out, and the WOSI-Fr was compared to the F-QuickDASH-D/S (Disability of the Arm, Shoulder and Hand-French translation), and Walch-Duplay scores. RESULTS: A French language version of the WOSI (WOSI-Fr) was accepted by a multinational committee. The WOSI-Fr was then validated using a total of 144 native French-speaking subjects from Canada and Switzerland. Comparison of results on two WOSI-Fr questionnaires completed at a mean interval of 16 days showed that the WOSI-Fr had strong reliability, with a Pearson and interclass correlation of r=0.85 (P=0.01) and ICC=0.84 [95% CI=0.78-0.88]. Responsiveness, at a mean 378.9 days after surgical intervention, showed strong correlation with that of the F-QuickDASH-D/S, with r=0.67 (P<0.01). Moreover, a standardized response means analysis to calculate effect size for both the WOSI-Fr and the F-QuickDASH-D/S showed that the WOSI-Fr had a significantly greater ability to detect change (SRM 1.55 versus 0.87 for the WOSI-Fr and F-QuickDASH-D/S respectively, P<0.01). The WOSI-Fr showed fair correlation with the Walch-Duplay. DISCUSSION: A French-language translation of the WOSI questionnaire was created and validated for use in both Canadian and Swiss French-speaking populations. This questionnaire will facilitate outcome assessment in French-speaking settings, collaboration in multinational studies and comparison between studies performed in different countries. TYPE OF STUDY: Multicenter cohort study. LEVEL OF EVIDENCE: II.

In vitro engineering of human stratified urothelium: analysis of its morphology and function.

PURPOSE: Gastric or intestinal patches, commonly used for reconstructive cystoplasty, may induce severe metabolic complications. The use of bladder tissues reconstructed in vitro could avoid these complications. We compared cellular differentiation and permeability characteristics of human native with in vitro cultured stratified urothelium. MATERIALS AND METHODS: Human stratified urothelium was induced in vitro. Morphology was studied with light and electron microscopy and expression of key cellular proteins was assessed using immunohistochemistry. Permeability coefficients were determined by measuring water, urea, ammonia and proton fluxes across the urothelium. RESULTS: As in native urothelium the stratified urothelial construct consisted of basal membrane and basal, intermediate and superficial cell layers. The apical membrane of superficial cells formed villi and glycocalices, and tight junctions and desmosomes were developed. Immunohistochemistry showed similarities and differences in the expression of cytokeratins, integrin and cellular adhesion proteins. In the cultured urothelium cytokeratin 20 and integrin subunits alpha6 and beta4 were absent, and symplekin was expressed diffusely in all layers. Uroplakins were clearly expressed in the superficial umbrella cells of the urothelial constructs, however, they were also present in intermediate and basal cells. Symplekin and uroplakins were expressed only in the superficial cells of native bladder tissue. The urothelial constructs showed excellent viability, and functionally their permeabilities for water, urea and ammonia were no different from those measured in native human urothelium. Proton permeability was even lower in the constructs compared to that of native urothelium. CONCLUSIONS: Although the in vitro cultured human stratified urothelium did not show complete terminal differentiation of its superficial cells, it retained the same barrier characteristics against the principal urine components. These results indicate that such in vitro cultured urothelium, after being grown on a compliant degradable support or in coculture with smooth muscle cells, is suitable for reconstructive cystoplasty.

Age-dependent impairment of spine morphology and synaptic plasticity in hippocampal CA1 neurons of a presenilin 1 transgenic mouse model of Alzheimer's disease.

Presenilin 1 (PS1) mutations are responsible for a majority of early onset familial Alzheimer's disease (FAD) cases, in part by increasing the production of Abeta peptides. However, emerging evidence suggests other possible effects of PS1 on synaptic dysfunction where PS1 might contribute to the pathology independent of Abeta. We chose to study the L286V mutation, an aggressive FAD mutation which has never been analyzed at the electrophysiological and morphological levels. In addition, we analyzed for the first time the long term effects of wild-type human PS1 overexpression. We investigated the consequences of the overexpression of either wild-type human PS1 (hPS1) or the L286V mutated PS1 variant (mutPS1) on synaptic functions by analyzing synaptic plasticity and associated spine density changes from 3 to 15 months of age. We found that mutPS1 induces a transient increase observed only in 4- to 5-month-old mutPS1 animals in NMDA receptor (NMDA-R)-mediated responses and LTP compared with hPS1 mice and nontransgenic littermates. The increase in synaptic functions is concomitant with an increase in spine density. With increasing age, however, we found that the overexpression of human wild-type PS1 progressively decreased NMDA-R-mediated synaptic transmission and LTP, without neurodegeneration. These results identify for the first time a transient increase in synaptic function associated with L286V mutated PS1 variant in an age-dependent manner. In addition, they support the view that the PS1 overexpression promotes synaptic dysfunction in an Abeta-independent manner and underline the crucial role of PS1 during both normal and pathological aging.

Prognostic ability of practitioners of traditional arabic medicine: comparison with Western methods through a relative patient progress scale.

The ancient Greek medical theory based on balance or imbalance of humors disappeared in the western world, but does survive elsewhere. Is this survival related to a certain degree of health care efficiency? We explored this hypothesis through a study of classical Greco-Arab medicine in Mauritania. Modern general practitioners evaluated the safety and effectiveness of classical Arabic medicine in a Mauritanian traditional clinic, with a prognosis/follow-up method allowing the following comparisons: (i) actual patient progress (clinical outcome) compared with what the traditional 'tabib' had anticipated (= prognostic ability) and (ii) patient progress compared with what could be hoped for if the patient were treated by a modern physician in the same neighborhood. The practice appeared fairly safe and, on average, clinical outcome was similar to what could be expected with modern medicine. In some cases, patient progress was better than expected. The ability to correctly predict an individual's clinical outcome did not seem to be better along modern or Greco-Arab theories. Weekly joint meetings (modern and traditional practitioners) were spontaneously organized with a modern health centre in the neighborhood. Practitioners of a different medical system can predict patient progress. For the patient, avoiding false expectations with health care and ensuring appropriate referral may be the most important. Prognosis and outcome studies such as the one presented here may help to develop institutions where patients find support in making their choices, not only among several treatment options, but also among several medical systems.

L'enseignement et l'apprentissage de la conjugaison en FLE : comment réduire les difficultés engendrées par l'orthographe ?

French verb morphology has always been a major challenge for learners as well as teachers of French as a foreign language. Learning difficulties arise not only from the inherent complexity of the conjugation system itself, but mostly from the traditional description found in specialized books, grammars, etc. French spelling alone tends to complexify the actual oral verb morphology by more than 60%, thus hindering efficient learning. Following Dubois (1967), Csécsy (1968), Pouradier Duteil (1997), etc., I suggest an alternative approach, exclusively based on phonetic transcription, and starting with plural forms instead of singular ones (Mayer 1969). For more than 500 verbs of the 2nd and 3rd groups, this strategy allows learners to first memorize the present tense plural form e.g. /illiz/ (ils lisent, "they read") and take the stem's final consonant away to get the singular /illi/ (il lit, "he reads").