My primary research interests are in modeling semantic and discourse influences in sentence processing.
To that end, my dissertation juxtaposes predictions made by semantic predictability with those made by pragmatic inferences about verbs and how their participant roles are realized.
The dissertation is part of a larger research program to explore the organization of our mental lexicon.
By understanding properties of the connections between words and concepts (and not just properties of words and concepts), we can develop a clearer understanding of the mental lexicon as a dynamic structure.
Part of that dynamic structure which I have had the opportunity to explore in other projects is the mapping function from surface forms to underlying representations.
What are the implications of these different mapping functions and how specific are they to a given task, individual, or language?
- Topic Segmentation for Multi-Party Chat
- Initially, this project focused on using semantic features to reconstruct conversational threads in chat logs. After receiving a Small Business Innovation Research (SBIR) grant through Janya, Inc., work was extended to make a multi-genre discourse segmenter that worked with both traditional, continguous-topic discourse segmentation and less traditional, interwoven-topic discourse segmentation (e.g., chat).
- Folopa Grammar Description (pdp, ppo)
- For a whole host of reasons, I am trying to build a standard grammar description using translations of Bible verses.
I gave a twenty minute talk (condensed) on the many peculiar facets to work under these conditions with a progress report on the language features I have so far uncovered. Folopa has popped up in a few papers (solo and in harmony).
These other sources provide a little boost here and there but are sufficiently underspecified to usually leave me only with further investigation ideas.
Here lie the projects that time requires I not focus on for a spell. It doesn't take much to get me interested again.
- Contextual Vocabulary Acquisition
- The indirect benefit of this project is learning how to adapt computer models to classroom programs. The direct benefit is improving our understanding of the ex- and implicit knowledge required to understand a text and a word.
- Textual Entailment (rte.php)
- My main interest in readability is rethinking measures of complexity from a stronger linguistic and psychological standpoint. Also, I would like to see the classic indices extended to account for more language families. Lastly, I am interested in L2 curricula which include ranked readings selected from a wider corpus.
- L2 Education
- I would like to develop curricula for difficulty-ranked readings (mentioned above) to challenge learners and foster organic vocabulary growth. The reciprocal of that project is honing word choice for non-native speakers. I've laid out my initial thoughts in something I call "Ontologically Speaking." Another (as yet unpursued) venue is computer modeling of students' L2 knowledge to predict gaps and help tailor lesson plans. A simple example is documenting the implications and entailments of conjunctions in the source and target languages. A third branch of some interesting L2 research points towards text-based gaming. In the spring of 2008, I failed to get funding for my EduMUDE project.
- At my undergrad Grinnell, I explored Gricean Pragmatics as an explanatory tool. Now I am more interested in creating productive and predictive theories of humor in language.
- Fallacy and Language Abuse
links are not actively being updated)