Lab Meeting, Summer 2015, Week 4
Awe- and Aw-inspiring readings (or just stuff you think deserves a mention)
Michael Wagner's lab has been updating their train-able auto-aligner and its looking more user friendly. It comes pre-trained with lab speech as compared to the Penn aligner which comes trained on the Supreme Court corpus and can be retrained on ~100 hrs of audio.
Morgan Sonderegger's lab has a publicly available trainable https://github.com/mlml/autovot tool.
Both the auto-aligner and autovot extractor can (theoretically) be installed to slate for lab-wide use.
What we did over last week
Florian
Finished writing/editing the paper with Dan Gildea. Submitted it to Science. Had it triaged later that week with standard form letter. sigh. Now we're submitting to Proceedings of the Royal Society, Series B.
Finished editing paper with Liz Karuza et al on self-paced reading during statistical learning. The paper is about to be submitted to JEP:General and after being triaged there, we will submit to JEP:LMC ;).
Wrote a draft for a Frontiers commentary on "What the heck is salience?".
- Edited (twice) through Experiment 1 and 2 of paper with Lindsay Butler et al on animacy effects in Spanish and Yucatec.
Agreed with Joe Toscano on final steps for his paper. He discovered a bug in his code, so that the code didn't (quite) do what we had wanted to do (and said we did in the paper). After some discussion, we decided that this could be anti-conservative and that the simulations therefore need to be rerun.
Contributed to an abstract/draft for Language together with Damian Blasi, Balthasar Bickel, Dan Dediu, and Sean Roberts. We hope to send it to Greg Carlson next week.
Started using a task-organizer in a hopeless attempt to get more organized. So far, so happy.
Figured out how we will pay McGill for the lab retreat. The total cost was about $3100 for food and accommodation. Michael, Meghan, and Morgan said that they will cover $800 of the food and $1000 of the accommodation since we had to travel. Chigusa and I will split the rest.
Talked with Tal Linzen about his various job offers. He defended last week (I skyped in) -- an interesting thesis with studied on phonological learning, and a cluster of work on uncertainty in morphological and syntactic processing. It seems he will go to Paris to work with Emmanuell Dupoux, who does really interesting behavioral, computational, and imaging work on language processing. Tal will be working on distributional semantics. We will, however, likely write an NSF grant together this summer, with the goal to continue to collaborate.
- I might put out an ad for another post-doc for the phonetic adaptation grant. I probably would wait until September, but if you know someone we should recruit, let me know.
Emily Morgan will be in town this week. We'll continue to talk about her NRSA. Zach, if you want to talk binomials, email her? She's been interviewed by Gina Kuperberg and it looks like she'll go there and be co-advised by Gina and Ani Patel (with me serving as an external advisor). In parallel, we'll start working on a language-focused NRSA, with the goal to find funding for her to continue related research between here and Tufts.
Edited through two wonderfully detailed result reports by Scott Fraundorf on his need+PARTICIPLE experiments. We are now finally getting quite interpretable and interesting results. This includes a first VWP eye-tracking study on adaptation to this novel syntactic structure in speech. This is quite critical, as it --in my view-- substantially reduced the likelihood that the omission of to be is interpreted as a typo/error (in speech, such as error would be reflected in acoustic changes surrounding the error site). This might also hold the key to some of our new reading results on need+PARTICIPLE adaptation, which suggest that normal readers seem to generalize expectations from need+PARTICIPLE to other omission errors. If you're interested, remind me and I could give you an overview.
- Finished my (very small) contribution to NIH equipment grant proposal for fMRI update (PI: Dick Aslin).
Andrew
- Rewrote a function that gets called with every HIT request in the HIT balancing backend. The old version took ~600ms, the new version takes 10-15ms.
- Rewrote some other code in the HIT balancer to avoid a bug that was popping up each time a new HIT was created.
Started running round two of Scott's BeDrop experiment.
Helped Sarah figure out how to get PsychPortAudio in Matlab to use low latency audio and hopefully helped Sarah and Willemijn figure out part of why their script was crashing.
- Started cleaning up the hard drive (well, SSD) on Scott's former laptop so it can be handed off to someone else.
- In prep for adding a new trial type requested by Masha for the Alien Language Learning applet:
- Took Elizabeth Smith to the Public Market and High Falls and caught up with her.
Olga
- Worked on the tutorial. Tried to make filler more comprehensible, worked a little bit on stimuli creation page, thought about methods in general. Tried to figure out how to make a dropdown menu, with little success.
Set up a meeting with Lauren about Redcap only to have to reschedule, so relooked at RedCap and started planning on how to actually start sending out the demographic information to subjects beforehand. Hopefully training Lauren this week on how to input data and to start sorting through the consent/demographics forms again.
- Updated the wiki lab meeting schedule (or maybe that was the week before)
- Have thought about the directed reading page I need to set up on the wiki. Reading through the Tily et al paper.
Esteban
- Continued working on JIT submission for NRSA.
Created a workflow for using the ProsodyLab and autovot toolboxes for analysis.
Emailed with Kyle Gorman about some quirks with the ProsodyLab aligner and possible fixes.
- Talked with Cory Bonn about inter- vs intra-speaker variability corpus project.
- Tested for replication of bugs and latency issues in my mturk production experiment.
Dave
Came up with outline, framing, and v. rough draft of paper for NIPS on inferring listeners' prior beliefs
Cleaned up models and analysis for that paper, and re-ran some models with fixed parametrization
Prepared (dug up) STRAIGHT tutorial for lab meeting.
Updated dissertation outline with recent progress/lack thereof.
Sarah
- Finished presentation for ETAP3
- Attended, and presented at, ETAP3
- Continued debugging Gotta eyelink script
Dan
Amanda
- worked on the production task for the mammals study
- worked on my MXPrag talk (to be given on the 1st, so it'll be done by the time lab meeting happens)
- tentatively set up a meeting with Elspeth Wilson (a student of Napoleon Katsos) at MXPrag, because she emailed me about my talk and wanted to chat
- returned job's belongings, hung out with job, maybe tried to set up plans to see geertje (I'm writing this before taking off, but I assume I'll do these things by Monday)
Zach
- Did my advisory committee meeting
- Emailed a whole bunch of people for stimuli questions
- Attended sister's graduation
- Coded up some stuff for my modeling tidbit
- Coded up more stuff to organize all the Kamide replication stimuli so that they can actually be used
- Coded up stuff for the pupillometry study, so now I can do transcriptions in Pygaze
Linda
- Finished writing all the stim for syntax experiment.
- Talked to AP + Mike; edited paradigm to make eye movements less complex.
- Made all the lists for running Kamide Replication, original version.
- Looked at Australian materials; sent email to possibly obtain more materials, namely complete versions of chicken little story.
- Met with Lauren to discuss stimuli-making; discussed more via email. She's bed-ridden from dislocating her knee -- ouch!
- Started looking at code to dynamically adjust stimuli for causal stuff.
Maryam
- Attended ETAP3
- Read some papers for Compass review paper
- coordinated with RAs to get filler words separated from raw audio files, improve written list of stimuli