We investigate the notions of continuity and interaction in linguistic models. There is now a quite rich tradition of work based on the hypothesis that natural language is not a discrete model. Instead, continuous models consider word sense, grammar rules and categories as continuous notions: some words are hard to categorize and some rules do not fully apply in certain contexts. Word meaning in context often corresponds to different meanings according to the dictionary. New models have been proposed that provide a more accurate view of these phenomena. Recent advances in natural language processing also support this idea by providing very rich models based on a multi-level representation of the context of use of linguistic items. We describe how these models provide an interesting convergence between linguistic and computational models.

/pdf/continuity-in-the-interactions-between-linguistic-units-2it6iqbakh.pdf

Problèmes de linguistique générale

Continuity in the interactions between linguistic units

Language diversity is a subject which has long been side-stepped in scholarly work on language, the verbal faculty. At best, any discussions have been purposely restricted to considerations of superficial syntactic variation. In recent years, however, cognitive science has shown renewed interest in this subject, in connection with the diversity of the meaning representations constructed in individual languages. The diversity of the ways in which languages represent reality is a core issue for linguistics, demanding a theoretical basis for linking variation and invariance. Analogous situations exist in other scientific disciplines, e.g., biology where the phenotype/genotype articulation is crucial. The issue is equally important to cognitive science : the overall view of how language, thought, and cognition are related intimately depends on how variation and invariance are apprehended. It should by now be clear that the search for interlinguistic invariance and the search for invariants underlying semantic variation within a single language (both synchronically and diachronically) are equally reliant on an understanding of gradualness. They therefore require identical treatment at the level of theory and modelization, particularly in the form of tools which can go beyond binary oppositions and all-or-nothing conceptualizations, whence the usefulness of topologically dynamic types of modelization. Allowing for gradualness means recognizing that all languages are by nature such that the relationship between form and meaning is subject to constant readjustments and that there is no ultimate state of equilibrium. Some states are nevertheless more stable than others. These states can be modelized by approximation in a static, discrete way using classic algebraic tools. The stability of apparently fixed patterns can, however, only be explained on the basis of an underlying mobility.

https://halshs.archives-ouvertes.fr/halshs-00067929/document

Diversity in linguistic representations : a challenge for cognition

Le concept d’ « Indo-Europeens » est fortement charge affectivement : depuis un siecle, ou un peu plus, il seduit ; ou il scandalise. L'attitude qu'il determine aupres d'un chercheur est fonction, naturellement, des realia — mais pas uniquement. La formation du chercheur, sa sensibilite aux differents elements du probleme, et ses gouts, ses affects, ses fantasmes, investissent le concept de contenus variables. Parce qu'ils sont les « ancetres » des Grecs, des Latins, des Celtes, des Germains — nos peres, biologiques ou culturels ; parce que leur etude, comparative, permet une formidable plongee intellectuelle dans une prehistoire qui, autrement, faute de textes, serait reduite aux realia archeologiques, anonymes, lacunaires, frustrants ; parce que l'abondance des textes qu'ont laisses leurs successeurs historiques est une matiere inepuisable pour le travail historique — les Indo-Europeens sont bons a penser.

Penser — et mal penser — les Indo-Européens

A la fin du XIXe siecle, lorsque la France entre pleinement dans une culture ecrite, le roman a tendance sentimentale fait l’objet d’une immense diffusion, liee a l’industrialisation culturelle. Ces textes romanesques qu’on baptisera a posteriori sous l’etiquette generique de « roman de la victime » constituent alors une mode litteraire aupres des femmes lectrices fraichement alphabetisees et scolarisees. Ce travail vise donc a etudier le roman de la victime en termes de reception et de lecture. L’analyse investigue trois domaines : les contextes empiriques dans lesquels les oeuvres du roman de la victime ont ete diffusees, consommees et recues ; les strategies auctoriales et textuelles de guidage de la lecture ; le travail editorial qui participe a l’encadrement de la lecture. Cette approche multiscalaire nous parait la seule a meme de reconstruire l’histoire de la lecture du roman de la victime, rendue difficile par la rarete des traces ecrites valant pour temoignage direct.

Poétique du support et captation romanesque : la « fabrique » de son lecteur par le roman de la victime de 1874 à 1914 :

This thesis is a study of syntactic constraints on the distribution and interpretation of the Russian reflexive pronouns sebja and svoj. These are usually, but, importantly, not always subject-oriented and in complementary distribution with pronominals, and in certain conditions they trigger some hitherto poorly studied interpretive effects. Russian also demonstrates considerable versatility in binding of NP-internal reflexives, including possessives. An explanatory account of the patterns in question, which canonical Binding Theory of Chomsky (1981) has been unable to provide, is offered. The proposed account assumes no features or principles specific to the reflexives. They are taken to realize bundles of derivationally valued interpretable phi-feature that enter into feature sharing dependencies with the antecedents, which encodes binding. In Chapter 2 constraints on the Agree operation are discussed and refined. It is concluded that they must apply to every feature and each direction of valuation independently. An unvalued feature instance can only receive its value from the closest matching relation, but itself can value multiple feature instances. Thus, Agree is supposed to be subject to relativized rather than absolute locality. Unlike in other Agree-based approaches to syntactic anaphora, in this thesis it is assumed that different phi-features (specifically person and number) can be shared separately, and either of them can support an anaphoric dependency. This is made possible by defective phi-feature probes, most importantly a number probe consistently positioned immediately above every complete argument structure. Anaphoric dependencies based on different phi-features have different interpretive properties: person sharing encodes awareness effects, while number sharing yields a distributive interpretation. This also has implications for the possibility of inanimate antecedents in various configurations. Possibly because the Russian reflexives can only be partially valued and thus cannot be spelled out as fully specified personal pronouns, they never realize derivationally acquired phi-feature values morphologically. It is well known that across languages long-distance reflexives tend to be more strictly subject-oriented than local ones. In Russian this holds even if locality is understood in the relative sense.Except PRO, only antecedents that trigger overt agreement can bind reflexives across closer non-nominative potential antecedents. The latter can only enter into number dependencies and don't prevent a person dependency across them, which explains this pattern. The complementary distribution of reflexives and pronominals as an effect of Agree holds for every alternating derivation separately, so if the derivations don't converge on the same antecedent, their superposition may result in apparent non-complementarity. This presents a challenge to global competition approaches to pronominal distribution. Where pronominals are not ruled out by the effects of the Agree operation in Russian, it becomes possible to detect that they can only be bound in configurations that are not too local. Among other issues considered in this thesis are certain differences in binding of dependents of process and result nominals, which correlate with their NP-internal movement possibilities, as well as binding into higher circumstantial PPs and comparative adverbs, which appear accessible only to number dependencies.

The Grammar of Binding : A study with reference to Russian

Communication using symbols (such as words) structured with grammar

Problèmes de linguistique générale

Citations

Continuity in the interactions between linguistic units

Diversity in linguistic representations : a challenge for cognition

Penser — et mal penser — les Indo-Européens

Poétique du support et captation romanesque : la « fabrique » de son lecteur par le roman de la victime de 1874 à 1914 :

The Grammar of Binding : A study with reference to Russian

Related Papers (5)

The Minimalist Program

Cours de linguistique générale

How to do things with words

Derivation by phase

Lectures on Government and Binding