The following is a list of the research projects I’m currently working on; where possible, the list includes links to a relevant manuscript, and/or code to reproduce the results reported in that manuscript. I welcome feedback on any or all parts of the projects described below!
Ambiguity in Language
Ambiguity pervades all levels of human language. Individual wordforms often have multiple, unrelated meanings (e.g., bank), a phenomenon known as homophony. And longer expressions, such as It’s getting cold in here, can serve both as declarative statements about the world and as indirect requests for the hearer to take action.
The prevalence of ambiguity in language raises several questions:
How do comprehenders resolve ambiguity?
What sources of information do comprehenders draw on to enrich the meaning of ambiguous or under-specified utterances? And how do individuals vary in their likelihood of sampling and deploying particular disambiguating cues?
In our work, we’ve asked about the role of prosody, constructional cues, and cognitive resources like mentalizing, in indirect request comprehension. Moving forward, we hope to ask how these disparate information sources are integrated as language comprehension unfolds in real-time.
Relevant code and papers:
Trott, S., Reed, S., Ferreira, V., & Bergen, B. (2019) Prosodic cues signal the intent of potential indirect requests. Proceedings of the 41st Annual Meeting of the Cognitive Science Society. [Preprint link][Data and code for analysis]
Trott, S., & Bergen, B. (2017). A Theoretical Model of Indirect Request Comprehension. In Proceedings of the AAAI Fall Symposium Series on Artificial Intelligence for Human-Robot Interaction (AI-HRI). [Link]
Do speakers attempt to reduce the ambiguity of their utterances?
One solution to the apparent prevalence of ambiguity would be for speakers to design their utterances to be less ambiguous. This could mean using linguistic or extra-linguistic cues (like prosody and co-speech gesture) to provide additional inferential information, tailoring a referring expression as a function of what’s mutually known across interlocutors, and more. To what extent do speakers engage these design processes at all, and how much of this design process is engaged automatically vs. strategically?
Why is language ambiguous?
Why does a system ostensibly evolved for communication contain properties such as ambiguity?
Relevant code and papers:
(Non-)Arbitrariness in Language
The morpheme is generally considered to be the basic unit of meaning in a language, while phonemes are thought to be arbitrary.
However, there is evidence for sub-morphemic systematicity in form-meaning pairings, such as phonaesthemes (e.g. the onset gl–). Which linguistic and non-linguistic factors affect the evolution of systematicity in a language, and how does the presence of non-arbitrariness promote (or hinder) learning and memory?
Relevant papers and resources:
Trott, T., Semenuks, A., Bergen, B. (2019). Sub-morphemic form-meaning systematicity: the impact of onset phones on word concreteness. Proceedings of the 41st Annual Meeting of the Cognitive Science Society, Poster presentation. [Code]
Natural Language Understanding
My previous work at the International Computer Science Institute involved building a modular framework for natural language understanding––producing action from language.
Trott, S., Eppe, M., & Feldman, J. (2016). Recognizing intention from natural language: clarification dialog and construction grammar. In Workshop on Communicating Intentions in Human-Robot Interaction. [Link] [Code]