phonoloblog | all things phonology | camba.ucsd.edu/phonoloblog

Workshop on Computational Modelling of Sound Pattern Acquisition

When and where: University of Alberta, Edmonton, February 13-14, 2010.  Robert Kirchner and Anne-Michelle Tessier, organizers

Theme: Major advances have been made in recent years towards explicit  modelling of phonological acquisition, including increasingly  sophisticated OT learning algorithms, as well as application of general machine learning techniques (e.g. expectation maximization and maximum entropy learning). At the same time, evidence of token and type frequency sensitivity in the propagation of both categorical and gradient patterns in speech has spurred growing interest in exemplar-based models of acquisition and processing.  This workshop aims to bring together these two strands of research, promoting dialogue between those pursuing symbolic and subsymbolic approaches to acquisition of the sound patterns of spoken language. We invite oral and poster presentations from phonologists, phoneticians, psycholinguists, computational linguists, and speech scientists on this general theme.  Though relevant analytic, programmatic, or experimental presentations are also welcome, priority will be given to abstracts reflecting original computational modelling results for some aspect of phonological/phonetic acquisition.

Invited speakers will include: Adam Albright (MIT), Michael Becker (Harvard), Andries Coetzee (Michigan), Robert Daland (UCLA), Bruce Hayes (UCLA), Jeff Mielke (Ottawa), Ben Munson (Minnesota), James Myers (CCU, Taiwan), Janet Pierrehumbert (Northwestern), Alan Yu (Chicago).  Titles to be announced.

Funding and registration fee: The organizers anticipate sufficient funding to cover travel and hotel costs of all presenters whose abstracts are accepted, above and beyond the invited speakers.  A registration fee of $70 ($50 students) will be charged to cover the cost of coffee break refreshments.  Late registration (after Jan. 1, 2010) is $100 ($75 students).  Registrants are encouraged to order tickets for a Saturday evening banquet, at an additional cost of $35.  All prices are in Canadian dollars.

Submission: Abstracts for oral or poster presentations should be no longer than one page (US letter or A4, 11 pt, 1 inch margins) with a second page for references, data and/or figures. Abstracts should be emailed as a PDF attachment to

phonmod@ualberta.ca, deadline: midnight (Mountain Time), November 20, 2009.

Unless the submitter indicates otherwise, the organizers will consider each abstract’s suitability for oral or poster presentation. Authors should include the title, name(s), and affiliation(s) in the body of the email.

See http://www.ualberta.ca/~kirchner/cmspa.htm for more information.

[ Via the UCLA Linguistics Department Newsletter. ]

At the time of his death, the late Professor Peter Ladefoged was engaged in an NSF-supported project to digitize and post online many recordings from the Phonetics Lab’s archives. In 2006, Professor Russ Schuh stepped in to see the project to completion. The UCLA Phonetics Archive, now on line, mostly comprises field recordings by Ladefoged and others, but also includes some recordings made for undergraduate term papers. Over the course of the project, many recent UCLA Linguistics undergraduates worked to digitize the audio recordings and accompanying wordlists. These are not teaching materials (not like http://www.phonetics.ucla.edu/), but rather raw unedited recordings, which are primarily intended for use by researchers (though they are also great fun to browse). Nonetheless, the Archive provided excellent materials for class assignments in acoustic analysis for Linguistics 104 [at UCLA] in Fall 2007 and 2008. In November 2008, near the end of the project, [the LA-area public] radio station KPCC ran a story about the Archive.

You can find a file listing the errors I’ve found in my phonology text, previously announced in this forum several months ago. The file is on my Wayne State website:

http://clas.wayne.edu/faculty/nathan

I’m sad to report that Nick Clements passed away in Chatham, Massachusetts, on August 30, just over a month shy of his 69th birthday (according to his Wikipedia entry). Beth Hume wrote an obituary for him on LINGUIST List, which I reproduce in full further below. Beth co-organized a symposium on tones and features in honor of Nick in June; the speaker list was a veritable who’s who of phonology, of which Nick was also of course a prominent member. He will be missed.

(more…)

Greetings, all two readers of phonoloblog. It’s been a long, long while, and for that I apologize. I’ve been pretty busy with various things, like this, for instance. And while I was busy, I was caught with my pants down: my claim to the phonoloblog.org domain name expired, and before I could even become aware of it, this douchebag snatched it. If you count yourself among my friends, you will bombard this person with obscene messages until s/he relents. Or, you will convince me that I can live with phonoloblog.com, phonoloblog.net, or even just good ol’ camba.ucsd.edu/phonoloblog — which is, after all, free as in beer.

Apologies (perhaps too late) to those of you who may have had your bookmarks and subscriptions set to phonoloblog.org. Please use camba.ucsd.edu/phonoloblog from now on, no matter what, and you should be fine.

Seminar: Approaches to word accent (word stress)

Organized by Rob Goedemans, Jeroen van de Weijer and Marc van Oostendorp (Leiden University)

April 2, 2009; Leiden University, Lipsius Building (http://www.visitors.leiden.edu/lipsius.jsp), room 235c

14.00 - 16:00 Harry van der Hulst (University of Connecticut): A new theory of word accentual structures (abstract below)
16:00 - Comments by Marc van Oostendorp and Jeroen van de Weijer, followed by discussion

Participation in this seminar is free for all. If possible, please announce your intention to come with Marc.van.Oostendorp@Meertens.KNAW.nl

A New Theory of Word Accentual Structures
Harry van der Hulst
University of Connecticut

The key insight of standard metrical theory (Liberman and Prince 1977, Vergnaud and Halle 1978, Hayes 1980, Halle and Vergnaud 1987, Idsardi 1990) is that syllables (or perhaps subsyllabic constituents such as skeletal positions, rhymes or moras) of words are organized into a layer of foot structure, each foot having a head. Primary accent is then derived by organizing the feet into a word structure in which one foot is the head. The head of the head foot, being a head at both levels, expresses primary accent. In this view, rhythmic accents are assigned first, while primary accent is regarded as the promotion of one of these rhythmic accents. In this seminar, I defend a different
formal theory of word accent. The theory is non-metrical in that the account of primary accent location is not based on iterative foot structure. The theory separates the representation of primary and rhythmic accents, the idea being that the latter are accounted for with reference to the primary accent location. This means that rhythmic structure is either assigned later (in a derivational sense),
or governed by constraints that are subordinate to the constraints that govern primary accent (as is possible in the approach presented in Prince and Smolensky 1993). The present approach has been called ‘a primary-accent first theory’ (see van der Hulst 1984, 1990, 1992, 1996, 1997, 1999, 2000a, 2002, 2009, van der Hulst and Kooij 1994, van der Hulst and Lahiri 1988 for earlier statements; see web page below for these and other references). I will demonstrate the workings of the theory using a variety of examples from bounded and unbounded (weight-sensitive and insensitive systems) taken from the StressTyp database developed by Rob Goedemans and Van der Hulst (http://stresstyp-test.leidenuniv.nl/).

The Department of Linguistics at the University of Alberta is inviting applications for a PhD student research assistantship position, beginning September 2009, on exemplar-based approaches to phonology.

A fundamental question of linguistic theory is how spoken language is encoded in the mental lexicon. The standard view, that phonological patterns are learned over symbolic representations, has increasingly come into question, as failing to provide a natural account of frequency effects and gradient sound change. An alternative view is that words and phrases are stored as clusters of exemplars, including detailed, individuated memories of the speech signal. Speech processing, under this view, involves massive comparison of exemplars, with phonological patterns (as well as low-level patterns of gradient phonetic variation) emerging as abstractions over the raw data. 

(more…)

Faithful readers: the e-mail barrage about phonoloblog being down the past few days has been overwhelming. I can’t possibly reply to it all, so I’m just going to apologize here to each and every one of you for depriving you of your phonolofix. I can’t promise it will never happen again, but I can assure you that I have no intention of allowing it to happen again…

OK, I’m kidding. No, not about that last bit — it’s true that I have no such intention — but about the e-mail barrage. Exactly two people wrote (evidence for binarity?), and neither one of them was Ed (which was surprising, because he has apparently not had anything better to do). I guess the remaining mass of readers relies on the RSS feed and don’t visit the blog directly, so they didn’t even notice we were down. (Should we post more often? Nah, this isn’t Language Log, just all things phonology.)

Anyway, the deal is that the security of our server (camba.ucsd.edu) was somehow compromised by comment spam and the like. This came to the attention of the systems security folks at UCSD, who brought it to the attention of the Linguistics computing staff — big thanks and props to Ezra van Everbroeck and even more to Marc Silver for doing what it took to bring things back online quickly and safely. If you notice anything that doesn’t seem to have survived the change-over intact, please let me know and I’ll take care of it if I can (or I’ll very politely ask Marc and Ezra for help if I can’t).

[ See the obituary on LINGUIST List. ]

For most phonologists, the process of Final Devoicing, which we can observe in languages such as German, Dutch, Yiddish, Russian, Polish, Catalan and Turkish, did not deserve a lot of attention. One would write a rule of approximately the shape [-son] → [-voice] / __ #/$, and declare the issue resolved.

However, recent years have seen a revived interest in phenomena surrounding devoicing, for a variety of reasons. One of them are developments in the formalism, like that of OT. For one thing, it appears much easier to view devoicing as a rule than as the result of a constraint. There is no consensus yet as to what the constraint should be in OT (e.g. a general constraint against voicing *Voiced, dominated by a faithfulness constraint for onsets, a conjunction of NoCoda with *Voiced, a positional markedness constraint, etc.) and further, Final Devoicing is one of the most famous cases of the so-called Too-Many-Solutions Problem: why would the relevant constraint always be satisfied by deletion of the voicing feature?

Further, lots of empirical work has come out which does not fit very easily with classical views of phonology (including most of OT). First, we find final devoicing both in languages in which the relevant contrast is indeed [voice] (such as Catalan), but also in languages in which it rather involves [spread glottis] (like German), which raises the question what these phenomena have in common from a phonological point of view. Secondly, there is a large body of work showing that final devoicing in many cases is not neutralizing completely, but that there are phonetic traces of voicing in the acoustic signal, and that listeners to some extent can detect these traces at least in experimental circumstances. Thirdly, it turns out that whether or not a given stem is subject to final devoicing is to a large extent predictable given lexical statistics.

Finally, it has become clear over the years that devoicing interacts with many other phonological processes in (varieties of) European languages, such as voicing assimilation, but also lexical tone. It has been claimed as well that certain dialects of French, for instance, have developed interesting phonological phenomena as a result of contact with West-Germanic final devoicing systems.

What is the place of devoicing and other voicing phenomena in phonological theory? Which phenomena need to be accounted for by our theory? Which phenomena CAN be understood by it? This will be the topic of a workshop at the Meertens Instituut in Amsterdam on September 11, 2008, and the University of Leiden on September 12, 2008. The workshop will end in a very big party. Participation (including the party) is free for all readers of Phonoloblog. Invited speakers will be Harry van der Hulst (University of Connecticut) and Ben Hermans (Meertens Instituut).

Please submit an abstract (2 pages max; does not need to be anonymous; pdf file) to Marc.van.Oostendorp@Meertens.KNAW.nl. Deadline: June 28.