Psycholinguistics and Computational Linguistics
Many faculty members of the University at Buffalo Linguistics department use a variety of empirical, quantitative, and experimental methods to investigate the representation and processing of human languages. Here is a list of areas in which research is being conducted by both faculty members and graduate students:
- Computational Psycholinguistics
- Corpus Linguistics
- Experimental Semantic Typology
- Human Sentence Processing
- Sentence Production
Prof. Bohnemeyer pursues experimental research on the relationship between language and thought as an application of his work on semantic typology. Linguistic categorization varies across languages – if only within the bounds of constraints imposed by cognition. Semantic typology seeks to isolate universals of the language-cognition interface and determine what properties of linguistic representations are specific to particular languages and cultures. To the extent that conceptual categories are learned, linguistic categories may serve as powerful “bootstraps” for the enculturating individual learning to tune into culture-specific conceptualizations. The Linguistic Relativity Hypothesis (LRH) – that linguistic representations may constrain and shape non-linguistic cognitive representations – has long been debated in the cognitive sciences. Bohnemeyer has conducted tests of the LRH in the domains of temporal and spatial semantics within what is sometimes called the Neo-Whorfian program: identify two or more populations whose native languages differ in the constraints they impose on linguistic representations of particular states of affairs; perform experiments to assess internal cognitive representations of the states of affairs in these populations; if linguistic performance proves to be a predictor of cognitive performance, then look for additional evidence to make the case that the correlation in fact reflects causation from the properties of external representations to those of internal representations.
Prof. Candelas de la Ossa is interested in methodological triangulation involving corpus linguistics approaches, such as corpus-based discourse analysis, and corpus approaches to variation analysis.
Prof. Pate uses methods from machine learning to explore how children learn language from data. How does lexical stress compare to syllable weight as a cue to word boundaries? Are suprasegmental cues, such as word duration, helpful for distinguishing good syntax trees from bad ones? How should long-distance dependencies be traded off against local dependencies when finding syntactic derivations of observed sentences? Probabilistic models provide a mathematically well-founded framework for formulating and quantitatively comparing different answers to these kinds of questions.
Prof. Koenig’s research focuses on the use of lexical information (particularly semantic information) in on-line sentence processing. In collaboration with Gail Mauner, he has explored the various ways in which argument structure is used to integrate constituents that co-occur with verbs, trying to distinguish the role of semantic argumenthood, world knowledge about situations, co-occurrence frequency, and morphosyntactic “activeness.”