The major fundamental challenge of this research is to develop a model of the contact points (or interfaces) between the language faculty and the other faculties of the cognitive system. Besides form this great intellectual venture, there is also the one of developing information technologies that integrate the cognitive interfaces model. The fundamental task is an answer to the necessity of advancing our knowledge of the language faculty and of its interaction with the other sub-systems of cognition. The technological task is an answer to the necessity of developing parsing systems that are able to handle information most effectively on an electronic platform and thus be more useful to the society at large and more specifically to the users that are affected by language disorders.
Anna Maria Di Sciullo
Département de linguistique
Université du Québec à Montréal (UQAM)
Tel: (514) 987-3000, poste 3519
Fax: (514) 987-0377
achievements | demo | advances | partnerships
Nature and impact of the major funding related to the latest grant 2004-2012
Our advances in theoretical linguistics in the context of the GTRC (2004-2012) have led to the formulation of a model of interfaces where asymmetrical relations, i.e. unidirectional, are prioritized. According to this model, the relation of proper subset characterizes the dependencies between the elements of conceptual-intentional interface representations and sensory-motor. We have shown that this relation characterizes and unifies the properties of morphological (Di Sciullo (2004, 2005a, b, f, 2007a), and syntactic chains (Di Sciullo and Isac, forthcoming). We have also shown that asymmetry persists in micro-variation (Di Sciullo and D’Alessandro, forthcoming; Hale 2007; Hale, Kissock and Reiss, 2007). In addition, we have put forth supplementary proofs to the asymmetry of categories (Dechaine and Tremblay, in preparation), structural arguments (Di Sciullo 2005a; Hill and Roberge 2006), null arguments (Cummins and Roberge 2004; Pirvulescu and Roberge 2005; Roberge 2007), modification (Androutsopoulou and Echevarria 2006; Androutsopoulou and Echevarria, in revision), and the relations of operator-variable (Di Sciullo and Landman 2007; Di Sciullo and Banksira 2007). These results will have important benefits for the formulation of a model of the language faculty.
We have brought forth experimental proofs to the hypothesis that asymmetries are perceived by the human cognition, by performing experiments on the perception of functional categories by very young children (Shi and al. 2006a, Shi 2005; Marquis and Shi, submitted), the acquisition of pronouns (Di Sciullo and Aguero, forthcoming; Roberge, Perez-Leroux and Pirvulescu 2006; Androutsopoulou, Español-Echevarria and Prévost, in revision) and on concepts (Niyogi and Berwick 2005). Our work on subjects suffering from Parkinson’s disease also supports our hypotheses (Cohen 2003; Cohen and Pourcher 2007; Frak, Cohen and Pourcher 2004). The results of our experiments indicate differences in the perception (time of reaction) of verbal structures with either external or internal prefixes (Tsapkini, Jarema and Di Sciullo 2004), between compounds with either an object or an adjunct (Di Sciullo and Tamioka, in preparation) passive and middle syntactic structures (Di Sciullo, De Almeida and al. 2007), and structures with an infinitive complement (De Almeida and Dwivedi 2006). These works have important benefits for a unified model of the human cognition that is based upon experimentally measurable results. We have also supported the hypothesis that asymmetrical relations are part of other faculties of cognition, as seen in the works of Frak and al. 2004, 2006) on the asymmetry of intentional movement. This supports our hypothesis that the properties of relations assure communication between the cognitive subsystems that give rise to, for instance, speech and movement.
We have developed prototypes for the automatic analysis of natural languages based on the recognition of interface asymmetry. These morphological and syntactic prototypes simulate human cognitive processes since they privilege asymmetrical relations (Di Sciullo and Fong 2005; Fong, forthcoming; Di Sciullo, Gabrini, Batori and Somesfalean 2006; Niyogi and Berwick 2005). We have developed morphological and syntactic prototypes by incorporating constraints that can treat derivational complexities (Di Sciullo 2005c, 2006; Fong 2005, forthcoming; Niyogi and Berwick 2005). These works have important benefits for natural language processing since they are guided by the properties of grammar rather than by statistical calculations. Also, along with our partners Delphes Technologies International, we have contributed to the development of a software program intended for information retrieval and extraction that is based on the processing of linguistic configurations. Delphes Technologies International proposes to donate this software program to UQAM; it will benefit the academic community for it will demonstrate the advantages that this search engine has over other search engines that are solely based on keywords. This software program has proven worthy for it has received numerous prizes for its quality in the precision of information retrieval and extraction.
Advances in the three areas since the midterm visit
As defined in our work schedule for 2006-2008, in the core theoretical linguistics section, we have analyzed the interface asymmetries specific to modification relations (Androutsopoulou and Espanol-Echevarria 2006, 2007; Hill 2006, 2007), binding relations (Di Sciullo 2006; Canac Marquis 2007) and operator-variable relations (Di Sciullo and Banksira 2007; Di Sciullo and Landman 2007). These works support the minimalist asymmetry hypothesis that asymmetrical relations are interpretable in an optimal manner. The works of Dechaine and Tremblay (in preparation) on categories and cognition, and the works of Tremblay (in preparation) and of Burnett and Tremblay (submitted) on aspectual relations also support this hypothesis. The works of Roberge (2007) and Hill and Roberge (submitted) on the derivation of indirect thematic objects in French, supports the hypothesis of minimalist asymmetry and the visibility of fragments hypothesis in which the interpretation is performed in a cyclical manner. The works on peripheral elements, truth values oui?/non? (Di Sciullo 2007b), attention markers eh?/hein ? (Gold and Tremblay 2007; Tremblay, in preparation) and pragmatic markers (Hill 2008) all support the dynamic modularity hypothesis in which the contact between the derived representations by the systems of performance are performed in the domain of fragments.
In the psycho and neurolinguistics section, the results obtained indicate that there are differences in reaction times in the human perception of arguments and modifiers: brain processing is longer for adjunct-verb compounds than it is for object-verb compounds (Di Sciullo and Tomioka, in preparation). These results confirm those which we have previously obtained on brain processing of verbal structures in French, including internal or external prefixes (Tsapkini and al. 2004). We have also observed a difference in brain processing for the perception of passive structures vs. middle structures (Di Sciullo, De Almeida and al. 2007): the reaction time for middle structures is longer than that for passive structures. According to our hypotheses, humans perceive linguistic expressions in terms of hierarchical structures: Op > Mod > Pred. These hierarchical structures are analysed in terms of interface asymmetry relations. Moreover, we have continued our work on the acquisition of functional categories (Di Sciullo and Aguero, in press; Canac-Marquis 2007; Roberge and Pérez-Leroux, in preparation; Gauthier, Shi & Yi 2007; Pietzak, Cohen and Snyder 2007), and on the asymmetrical relations linking language and intentional movement. Frak and al. 2007). These experimental results support our hypotheses of interface asymmetries and their processing in the brain.
In the computational linguistics section, we have developed a syntactic prototype by expanding its empirical coverage (Di Sciullo and al. 2006). We have extended the grammar for the processing of modification structures (prepositional and adjectival constituents) and operator-variable structures (relative clauses, complex questions and quantifier structures). We have also imported, in the analyser, the basic lexicon developed in previous sentences of the GTRC. This database (ABELEX) now includes a representative ensemble of the structures of formal (category) and semantic traits (argument structures, aspectual and conceptual) for both lexical items and functional categories. We have also optimized the parser for the processing of discourse by extending data structures thus, permitting the accommodation of a sequence of sentences. A web application for the current version of the parser has been developed. We have also developed a prototype for the analysis of English and French compounds based on the articulation of functional categories (Dungan, Fong & Harrison, in preparation; Di Sciullo 2005h). Finally, we have presented arguments against the probabilistic statistical approach to natural language processing (Berwick and Fong 2007.
The results since my midterm visit are manifested by major publications published by well recognized editors such as MIT Press, Oxford University Press, and John Benjamins, in the three sections of this research : i) the theory of interfaces (Di Sciullo and Hill, forthcoming a, b; Ramchand and Reiss (2007); Hale and Reiss, in preparation; Hale and Kissock 2007) ; ii) psychology and neurolinguistics (De Almeida, in preparation; Cohen and Stemmer 2007) ; iii) computational linguistics (Fong and Berwick, in preparation.
Our work has also lead to the development of an efficient and precise information retrieval and extraction software. Moreover, our work has driven us to explore the domain of biolinguistics that deals with the origin of language, evolution and linguistic variation. Two volumes, dedicated to this domain, will be published by MIT Press (Di Sciullo and Aguero, forthcoming) and Oxford University Press (Di Sciullo and Boeckx, forthcoming). These volumes will include contributions by researchers who have made their mark in this domain; such as, N. Chomsky, M. Hauser, and T. Fitch. We have also published in this domain. In addition, we expect to organize an internal network of biolinguistics where one of the centers will be located at UQAM.
The partnerships that have been established
We expect to pursue our partnership with the company Delphes Technology International and to develop a system to represent our research results by a configuration that takes into account user preferences. The system requires the creation of knowledge bases (BC), ontology and a system of reasoning. We will focus on the configuration approach to information processing by developing an inference engine that will also apply to the configuration properties of queries in the user’s research history.
We have planned a new partnership with the company Humanix which provides an interactive web platform for the development of the well-being of individuals. We want to develop a biolinguistic and a biomusicologic axis that can be integrated into this platform. This partnership will thus benefit the general public by the results of our research on the biological bases of language and the relation between language and cognition, as well as on the biological basis of music, the properties of the interface language-music, and their implications for well-being and health.
We have also planned a new partnership with the University of Lille for the optimization of reasoning in ontologies. The question that we want to consider is how to reason efficiently with multiple knowledge bases (BC) and how to improve the efficiency of the reasoning. Reason in a knowledge base (BC) is one of the most important applications of description logics (LD). The requirements for the execution of time and memory space are two significant factors that directly influence the performance of a reasoning algorithm. We want to study the existing techniques and propose a new technique to optimize the reasoning of description logics (LD) with, the aim of significantly reducing the two factors mentioned above. This new technique will be applied to accelerate reasoning in the case of a knowledge management system, or potentially a system of cognition and other subsystems. Furthermore, the incorporation of such a technique, along with previous optimization techniques in current description logics (LD) systems, can efficiently resolve insurmountable inferences. We opt for a decomposition approach of the ontology dedicated to systems of cognition in multiple sub-ontologies that always preserve the semantics and the inference services of the original ontology.
The links that have been created thus far with the partnerships, including government bodies, academics and the population at large, were long and fruitful. It is likely, given the work accomplished thus far, that the new partnerships will also be as successful.
List of publications deriving from the 2004-2012 cycle
We would like to thank the Social Sciences and Humanities Research Council of Canada for the constant support given to our research.