Computational Linguistics (CL) and Natural Language Processing (NLP) are scientific disciplines studying the understanding of natural language and communication from a computational and cognitive perspective. CL/NLP researchers typically develop formal and computational models of linguistic processing by blending together methods from linguistics and computer science.
CL/NLP researchers investigate foundational topics in linguistics (e.g., cross-linguistic analyses of structure and meaning) in order to integrate the results in models of language comprehension/production and human-human/human-machine communication. A big challenge, e.g., is semantics and pragmatics processing, which requires both a better understanding of natural language semantics/pragmatics and technical tools for integrating such processing effectively into computer programs. Accordingly, in collaboration with artificial intelligence researchers, we pursue the formalization and computational implementation of the semantic ontologies and algorithms that enable both spoken language interaction and text-based information extraction in order to develop NLP applications that automatically facilitate and analyze social media interactions, web searches, and multi-agent dialogue systems. Formal and computational models are grounded psychologically through experimental methods that probe the cognitive and neural mechanisms enabling human linguistic processing, on which researchers of CL/NLP and psycho- and neurolinguistics work closely together.
CL/NLP is a very lively and challenging scientific field, due to the constant development and improvement of theoretical models and rapid technological advances. Joining the CL/NLP group means that you can get involved in further developing the technical tools implementing both neurologically and psychologically plausible models of human language processing and novel practical applications.
How can we analyze the language used in spoken or text-based communication?
Our research focus is the ubiquitous phenomenon of context dependence. We develop robust analyses of linguistic structure that can cope with degenerate input, e.g., the fact that both human dialogues and large corpora of text from the internet provide us with noisy, fragmentary, and ill-formed data. We design formal systems involving semantic/pragmatic representations that can model, e.g., the interaction of lexical and compositional meaning with discourse and background knowledge. Such representations can be integrated into automated reasoning and human-computer interaction systems, providing more naturalistic user interfaces.
How does linguistic information interact with non-linguistic experience?
Using eye tracking methods, we measure at which point in time people focus on which objects in a visually presented scene while listening to a brief discourse.
The pattern of eye movements over time reveals how auditory language processing interacts with visual input.