First, I think on the whole this is so well-written and clear. The final paragraph is especially good.
Here's a few things I noticed: "The eye-movement data come from a collection of 144 German sentences read by native speakers known as the Potsdam Sentence Corpus"
I read this as the native speakers are the PSC ;). I think a better phrasing would be: "The eye-movement data come from the Potsdam Sentence Corpus, a collection of 144 German sentences read by native speakers"
I'm kinda wondering about the phrasing "The work described in this talk"...you could probably just shorten it to "the work" unless there's a specific reason for "in this talk".
Do you mean "log-probability of a word in a probabilistic language model"?
I'm missing the word syntax in this abstract. As I understand it, this work also shows that syntax processing has a stronger impact on eye movement control than previously thought. Especially as the PSC is a collection of pretty harmless sentences (no far-fetched ambiguities). Why don't you make this point explicit.
The last column on the right (TRT) is cut in the graphs!
#1. the reason I am so pedantically saying "this TALK" is because I want the reader to get the message that we are demanding a talk. Repeating this in the abstract seems as good a technique as any.
#2. Titus, your point is well-taken. I shied away from the relevance-of-syntax angle because that is more of a significant result in the eye-movement research community than at CUNY. The CUNY-goers already believe in syntax and that it is relevant for eye-movements. they just have never seen it all bolted together before in an end-to-end computational model.
I updated the version linked here to not-cut-off the diagrams, to fix "union of" and to include the word syntax at least once at the end...bring on more comments!
5 comments:
First, I think on the whole this is so well-written and clear. The final paragraph is especially good.
Here's a few things I noticed:
"The eye-movement data come from a collection of 144 German sentences read by native speakers known as the Potsdam Sentence Corpus"
I read this as the native speakers are the PSC ;). I think a better phrasing would be:
"The eye-movement data come from the Potsdam Sentence Corpus, a collection of 144 German sentences read by native speakers"
I'm kinda wondering about the phrasing "The work described in this talk"...you could probably just shorten it to "the work" unless there's a specific reason for "in this talk".
ditto, gets at the main point pretty well, detailed, clear.
possible typo, "trained on union of two corpora", near end 1st paragraph.
on the 2nd read, this is very contentful and clear. 1st and 3rd paragraphs are sharp and interesting.
Do you mean "log-probability of a word in a probabilistic language model"?
I'm missing the word syntax in this abstract. As I understand it, this work also shows that syntax processing has a stronger impact on eye movement control than previously thought. Especially as the PSC is a collection of pretty harmless sentences (no far-fetched ambiguities). Why don't you make this point explicit.
The last column on the right (TRT) is cut in the graphs!
Thanks for these comments.
#1. the reason I am so pedantically saying "this TALK" is because I want the reader to get the message that we are demanding a talk. Repeating this in the abstract seems as good a technique as any.
#2. Titus, your point is well-taken. I shied away from the relevance-of-syntax angle because that is more of a significant result in the eye-movement research community than at CUNY. The CUNY-goers already believe in syntax and that it is relevant for eye-movements. they just have never seen it all bolted together before in an end-to-end computational model.
I updated the version linked here to not-cut-off the diagrams, to fix "union of" and to include the word syntax at least once at the end...bring on more comments!
http://comp.lin.msu.edu/cuny08/hale.pdf
Post a Comment