LSA Annual Meeting Dispatch, pt. 1
It's late and I still haven't finished my handout for my talk, but I wanted to jot a few notes, albiet breif, about today's LSA meeting.
First, if someone else doesn't take it up next year, I'd love to set up an on-line discussion group with a thread for each talk. The time limits are so short, the discussion periods too brief, and for some, the distances too long to allow the talks to reach their full potential of prompting exicting discussion by the country's best and brightest. Of course, "mingling" is the old-school way of achieving dialogue on the different topics, and can be augmented by email. But this risks exclusion of the shy and those that aren't in on the email chain. In our brave new era, a threaded discussion extending days or weeks beyond the conference, allowing for folks to reference their data that they didn't have at their tongue-tips at the meeting, seems like an idea whose time is... on... its... way.
Second, the talks were, for the most part, really great. This is my first LSA, and I had low expectations since the acceptance process is not intended to be rigorous (I think) and I was pleasantly surprised by the tight logic, ellegant experiments and interesting ideas in many of the presentations.
Jeri Jaeger found an interesting correlation between speech-error directionality and information structure. Essentially, the anticipatory nature of many speech errors in English, German and Dutch is not due to some universal property of speech production, but rather includes other factors such as the way new information is prosodically or syntactically marked.
Jason Riggle tried to address the critical question of how constraint rankings can be learned for more than one grammar at a time. Laying out the hypothetical grammars as sets and looking at the statistical distribution of the results, including factoring out the impossible grammars seems to provide a possible solution with one major caveat: both input and output forms are needed by the learner. It seems that addressing this caveat in the framework Riggle suggests results in an intractable factorial problem - I'd love to see it attempted.
Laura Dilley had the most significant talk I saw which challenged the idea of using the syntagmatic formal representation of tone we currently have (H, L, etc) and suggested, instead, that we adopt a theory based on the interstitial changes: the change of tone between segments. She was forced to speak fast because she tried to pack so much into the 30 minutes and unfortunately didn't get to address a lot of the question that a hypothesis like this, which to me seems so drastic, will obviously raise. Her formalism basically has higher/lower/same markings. I'm not sure how this would account for a language like Kuki-Thaadow, where the L-H transition results in almost a downstepped H as compared to the H-L, or if it would work for 3, 4 or 5 tone systems. Contour tones were left out, and there wasn't much data discussed, but the idea fit well with Peter Ladefoged's talk. He argued against a universal set of features and suggested a representation beyond the phoneme. This was the most controversial talk of the night, I imagine purposefully so, but it's late...
More tomorrow.
First, if someone else doesn't take it up next year, I'd love to set up an on-line discussion group with a thread for each talk. The time limits are so short, the discussion periods too brief, and for some, the distances too long to allow the talks to reach their full potential of prompting exicting discussion by the country's best and brightest. Of course, "mingling" is the old-school way of achieving dialogue on the different topics, and can be augmented by email. But this risks exclusion of the shy and those that aren't in on the email chain. In our brave new era, a threaded discussion extending days or weeks beyond the conference, allowing for folks to reference their data that they didn't have at their tongue-tips at the meeting, seems like an idea whose time is... on... its... way.
Second, the talks were, for the most part, really great. This is my first LSA, and I had low expectations since the acceptance process is not intended to be rigorous (I think) and I was pleasantly surprised by the tight logic, ellegant experiments and interesting ideas in many of the presentations.
Jeri Jaeger found an interesting correlation between speech-error directionality and information structure. Essentially, the anticipatory nature of many speech errors in English, German and Dutch is not due to some universal property of speech production, but rather includes other factors such as the way new information is prosodically or syntactically marked.
Jason Riggle tried to address the critical question of how constraint rankings can be learned for more than one grammar at a time. Laying out the hypothetical grammars as sets and looking at the statistical distribution of the results, including factoring out the impossible grammars seems to provide a possible solution with one major caveat: both input and output forms are needed by the learner. It seems that addressing this caveat in the framework Riggle suggests results in an intractable factorial problem - I'd love to see it attempted.
Laura Dilley had the most significant talk I saw which challenged the idea of using the syntagmatic formal representation of tone we currently have (H, L, etc) and suggested, instead, that we adopt a theory based on the interstitial changes: the change of tone between segments. She was forced to speak fast because she tried to pack so much into the 30 minutes and unfortunately didn't get to address a lot of the question that a hypothesis like this, which to me seems so drastic, will obviously raise. Her formalism basically has higher/lower/same markings. I'm not sure how this would account for a language like Kuki-Thaadow, where the L-H transition results in almost a downstepped H as compared to the H-L, or if it would work for 3, 4 or 5 tone systems. Contour tones were left out, and there wasn't much data discussed, but the idea fit well with Peter Ladefoged's talk. He argued against a universal set of features and suggested a representation beyond the phoneme. This was the most controversial talk of the night, I imagine purposefully so, but it's late...
More tomorrow.
0 Comments:
Post a Comment
<< Home