Issue of Monday, September 15th, 2008
Friday, Sept. 19, 3:30 PM
University of Chicago
“Learning and Linguistic Typology”
In this talk, I present results on the learnability of phonological grammars for two constraint-based models, Harmonic Grammar (HG; Legendre, Miyata, and Smolensky 1990) and Optimality Theory (OT; Prince and Smolensky 1993). I first establish that grammars in these models are learnable from reasonably sized samples of data and then present a learning algorithm for OT that is guaranteed to make no more than k log2 k mistakes when learning grammars with k constraints, which is only logarithmically worse than the best possible learning algorithm for OT.
The proposed learning algorithm calculates the number of rankings that are consistent with a set of data. This makes possible a simple and effective Bayesian heuristic for learning – all else equal, choose can- didates that are preferred by the highest number of rankings consistent with previous observations. This general strategy can be applied to HG or to any parameterized model of grammar, and it associates with each language generated by the theory an abstract quantity measuring the fraction of the parameter space corresponding to grammars that generate that language.
The p-volume seems to encode ‘restrictiveness’ in a way similar to Tesar and Prince’s (1999) r-measure. Preliminary investigations indicate that p-volume is significantly correlated with typological frequency (cf. Bane and Riggle 2008). This fact is neatly explained if language learners use a strategy that is some- times called a Gibbs leaner wherein they keep track of the region of the parameter space consistent with previous observations but make guesses according to a single hypothesis grammar randomly selected from that region. Upon making an error the Gibbs leaner updates the parameter region and randomly selects a new hypothesis grammar from that region. Following this strategy, learners will be predisposed towards grammars with large p-volume in cases where the hypotheses are underdetermined by the data. Moreover, priors other than the ‘flat’ distribution over rankings can be included to implement models of ranking bias.
One of the primary assets of this strategy is that it allows linguistic theory to be informed by the relative frequencies of patterns in linguistic typologies rather than only by the boolean distinction of whether or not a pattern is attested. Though some of the frequency asymmetries surely come from non-linguistic historical accidents, a model of learning that is able to account for some of the frequency variance is clearly of interest and makes a range of predictions that can be tested in experimental settings.
The LF Reading Group will meet this semester on Wednesdays from 3:00-4:30, in room 26-310.
Our first meeting will be this coming Wednesday, September 17th. Patrick Grosz will lead an informal discussion on Christopher Potts’ “The Logic of Conventional Implicatures”.
Phonology Circle will move to Wednesday at 5PM this term: this works for a majority of those who sent in responses.
With appropriate time adjustments, from Monday to Wednesday, the schedule this term looks like this:
- October 1st: Seokhan Kang
- October 8th: Anthi Revithiadou
- October 15: Olga Vaisman
- October 22: Stefano Versace
- November 19: Jonah Katz
- December 3rd: Jen Michaels.
Titles and abstracts will be announced closer to the time of each talk. If you want to schedule a talk, talk to Donca.
Jillian Mills successfully defended her master’s thesis, ‘Non-specific objects in the pseudopassive’, on September 3. The thesis investigates the properties of nominal phrases when they appear in the direct object position of English pseudopassive sentences. It is argued that due to the particular syntax of pseudopassives, these nominals are restricted to being D(eterminer)-less NPs, which in turn correlates with their non-specific interpretation, as well as other peculiarities of their syntax and semantics. The committee was Michel DeGraff, Kai von Fintel, and Norvin Richards.
Hyesun Cho is speaking at TIE (Tone and Intonation in Europe) 3, held September 15-17 at the University of Lisbon. Hyesun’s paper is entitled ‘Effects of speech rate on segmental anchoring: A model of F0 timing as a function of slope and alignment targets’.
Jessica Coon is spending the fall semester in Chiapas, Mexico, continuing her work on Chol (Mayan) with support from an NSF Dissertation Improvement Grant. Jessica and her husband are living in San Cristóbal de las Casas, where she is a visiting scholar at CIESAS-Sureste, a Mexican language and culture research institution. CIESAS offers a master’s degree in Mesoamerian Linguistics, which is designed to train native speakers to document and analyze their languages. Currently the program includes speakers of Chol, Tzeltal, Tzotzil, Zapotec, Zoque, and Nahuatl. Jessica will be helping out in the morphosyntax class during the week, and will spend most weekends in a Chol-speaking village, collecting data. She sends saludos to the department, and looks forward to seeing everyone in the spring.
The date for the second annual of the Northeast Computational Phonology workshop has been set:
- Date: Sat. November 15th
- Location: Yale (New Haven, CT)
Like last year, it will feature a limited number of talks in an informal and friendly setting, designed to keep up to date on computationally-related work going on in the region.
Please talk to Adam if:
You are interesting in attending— and if so, whether you would be able to drive, or if you’ll be looking for a ride. (It is, of course, also reachable by train or bus, but it’s quicker by car, so hopefully we may be able to coordinate rides)
You might have something you are interested in presenting.