Projects

• Linguistic documentation of Tahltan (Northern Athapaskan)
• Mechanisms of consonant assimilation
• Process-based morphology in Optimality Theory
• Synchronized learning strategies in two cognitive domains: artificial neural networks and constraint-based optimization

This program of research is a collection a of projects designed to give a general understanding of the phonology (sound patterns) and morphology (word structure) of Tahltan, a critically endangered language of Northwest British Columbia. One theme in this research is the correct analysis of suprasegmentals, including stress, tone, and length, and how they referred to by word structure. The research is documented in a set of published and unpublished works and sound recordings that are archived in a way appropriate to the content of the recording. These works include conversational dialogues, recordings and transcriptions of folklore, questionnaire data, catalogs of the data in public institutions, and an annotated bibliography. A larger goal is to document the language in a grammar sketch, a set of texts, and a lexicon, and to make public records of this information so that they can assist language learning.

Mechanisms of consonant assimilation

Funding: SSHRC Standard Research Grant
Principal investigator: Alexei Kochetov (University of Toronto)
Co-investigators: John Alderete (SFU), Louis Goldstein (University of South California), and Marianne Pouplier (Edinburgh)

The project investigates consonant assimilation - a common phonological process by which a consonant becomes more similar to, or identical to, another adjacent or non-adjacent consonant. The goal is to provide an explanatory account of major types of consonant assimilation as an interaction of phonological and phonetic mechanisms: the higher-level mechanism of featural agreement and the lower-level mechanisms of spreading and repetition of articulatory gestures. Our specific objectives are (i) to investigate cross-linguistic perception of adjacent consonants - stops of different places of articulation, (ii) to examine speech errors in the production of sibilants in several languages, and their cross-linguistic perception, (iii) to conduct a thoroughly controlled typological survey of assimilation processes and to compare attested grammars of assimilation to those predicted based on our experimental results.

Process-based morphology in Optimality Theory

Funding: SSHRC Standard Research Grant
Principal investigator: John Alderete
Co-investigators: Keren Rice (University of Toronto), Peter Avery (York University)
Collaborator: Alexei Kochetov (University of Toronto)
Research assistants: Tim Choi (SFU), Angela Cooper (SFU), Andreea Kosa (SFU), Tzu-ying Vivian Lee (SFU)

This grant brings together researchers from three different Canadian universities to study the interaction between morphology and phonology. We investigate, both empirically and theoretically, one of the best types of evidence for this interaction, process-based morphology (PBM), or the use of phonological processes like metathesis or deletion to mark morphological distinctions. Optimality Theory provides a set of mechanisms for rigorous formal analysis of PBM as a consequence of one of its central tenets, namely that a grammar of a language is constituted by the interaction of well-formedness constraints. Hypothesis formulation and testing in OT is thus guided by the assumption that phonology-morphology interaction in PBM follows from the interaction of intrinsically phonological constraints with morphological ones. Empirical investigation of PBM involves primary linguistic description of feature-inserting PBM in a critically endangered Aboriginal language, Tahltan (Northern Athapaskan) through a community partnership with the Tahltan Nation, description and analysis of consonant mutations in Dholuo (Nilotic), also supported by research relationships with three African universities, and typological investigation of the range of attested phonological operations in PBM.

Synchronized learning strategies in two cognitive domains: artificial neural networks and constraint-based optimization

Funding: pending
Principal investigator: John Alderete (SFU) Collaborators: Alexei Kochetov (University of Toronto), Paul Tupper (SFU)

The objective of this research program is to develop and test concrete algorithms that are involved in learning the complexities of human language. Computational learning systems in the past have largely focused on one of two types of learning: numerical algorithms that are sensitive to pattern frequencies and abstract symbol-manipulating algorithms that work on mental data structures of language to approximate human language use. This research program assumes that there is in fact a role for both types of learning, and indeed, the synchronization of numerical and symbol-manipulating computation can address a number of long standing problems in language learnability research. Progress in testable language learning systems is of fundamental interest to numerous fields, including computer science and artificial intelligence, as well as to linguistics and cognitive science, which seek explicit computational mechanisms for characterizing what one knows when one knows a language.