[ Skip to the content ]

Institute of Formal and Applied Linguistics Wiki


[ Back to the navigation ]

Differences

This shows you the differences between two versions of the page.

Link to this comparison view

Next revision
Previous revision
courses:rg:2011-report-baby-steps [2011/10/26 09:49]
kruza vytvořeno
courses:rg:2011-report-baby-steps [2011/10/26 14:49] (current)
kruza
Line 1: Line 1:
-====== Report from RG on 2011-10-24 ======+====== From Baby Steps to Leapfrog: How “Less is More” in Unsupervised Dependency Parsing ======
 //Talk by Martin Majliš\\ //Talk by Martin Majliš\\
 Report by Oldřich Krůza//\\ Report by Oldřich Krůza//\\
  
 +==== Introduction ====
  
 On Monday, October 24th 2011, we heard a talk about a paper by Valentin On Monday, October 24th 2011, we heard a talk about a paper by Valentin
Line 10: Line 10:
 unsupervised parsing, and reports a success in a rate of percents, which unsupervised parsing, and reports a success in a rate of percents, which
 certainly makes it a paper worth notice. certainly makes it a paper worth notice.
 +
 +==== Notes ====
  
 Most of the attendants apparently understood the talk and the paper well, and a Most of the attendants apparently understood the talk and the paper well, and a
Line 39: Line 41:
 progressed from short sentences to longer, and identified the threshold, where progressed from short sentences to longer, and identified the threshold, where
 it's best to start ignoring any more training data, at sentences of length 15. it's best to start ignoring any more training data, at sentences of length 15.
-However, it seems they used evaluation on gold data to find out this threshold.+However, we were not 100% clear how they computed this constant.
 If the model was to be fully unsupervised, it remains a question, how to setup If the model was to be fully unsupervised, it remains a question, how to setup
 this threshold, because it cannot be safely assumed that it would be the same this threshold, because it cannot be safely assumed that it would be the same
Line 48: Line 50:
 face to face with words like "unbridled" or "jettison", which I personally had face to face with words like "unbridled" or "jettison", which I personally had
 never seen before. never seen before.
 +
 +==== Conclusion ====
  
 All in all, it was a paper worth reading, well presented, and thoroughly All in all, it was a paper worth reading, well presented, and thoroughly
 discussed, bringing useful general ideas as well as interesting details. discussed, bringing useful general ideas as well as interesting details.

[ Back to the navigation ] [ Back to the content ]