This week’s Phonology Circle features a double-header, with practice talks by Igor Yanovich and Youngah Do.
Time: Monday 9/27, 5pm, 32-D831
Speaker: Igor Yanovich (MIT)
Title: Making OT learning less algorithmic
(Practice talk for the upcoming meeting of NECPhon, Oct 9, UMass)
Investigations into OT boundaries of expressivity and learning in OT have been mostly empirical, with focus on devising algorithms that work. Yet the line of research trying to formally analyze OT properties and then devise algorithms that work better has also been prominent, including the foundational ERC theory of Prince 2002, which served as the basis for effective algorithm of FRed discovered by Brasoveanu & Prince 2005 that transforms tableaux into a much more comprehensible equivalent form, or Riggle 2008, who used the measure of number of rankings compatible with a set of data, r-volume, in a useful heuristic for his learning algorithm.
Following that line of research, I present an analysis of the structure of the domain of OT rankings, which provides the theoretical framework for computing the full solution to the OT Ranking problem and allows to make the OT learning theory less dependent on algorithms, using robust analysis in terms of static constraints instead. The framework provides means to describe the range of an algorithm non-procedurally, allowing a meaningful comparison between different algorithms; and provides the benchmark of the full set of rankings compatible with a given tableau, against which smaller sets returned by learning algorithms trying to mimick human learning can be compared, allowing a more analytical approach to the OT Subset problem.
Speaker: Youngah Do (MIT)
Title: When focal cues are conflicting: Focus perception in Korean
(Practice talk for the 20th Japanese/Korean Linguistics Conference at the University of Oxford)
This study explores perceptual cues to focus in Seoul Korean. While the production of focus marking has been widely studied, little attention has been given to focus perception in Korean. I tested focus perception by presenting listeners with spoken sentence frames (e.g. “John ate apples, not ___”) and asking them to fill in the blank: e.g., “not Mary” (if the subject was perceived to be focused) or “not bananas” (if the object was perceived as focused). In order to check whether listeners use both prosodic and structural cues, the first phrase was manipulated in order to vary the pitch range of subject and object, as well as the word order of the sentence. The results show that Korean listeners use (a) prosodic cues as well as (b) word order cues, and (c) they are biased to perceive object over subject as focus. Logistic regression analysis shows that the weight of the word order cue is highest, followed by prosodic cue and object bias. As expected, uncertainty is observed when cues are conflicting. Using the results of the logistic regression, I consider to what extent the observed uncertainty can be explained by the relative cue weights.
Upcoming talks:
Oct 4: Kirill Shklovsky (MIT)
Oct 18: Ari Goldberg (Tufts)
Oct 25: Youngah Do (MIT)
Nov 8: Michael Kenstowicz (MIT)
Nov 29: RUMMIT Practice talks
You can view the current, up-to-date version of the schedule here (click ‘agenda’ to see the schedule as a list), or subscribe via iCal here.