Improved Fully Unsupervised Parsing with Zoomed Learning.

  • Reichart R
  • Rappoport A
  • 24


    Mendeley users who have this article in their library.
  • 4


    Citations of this article.


We introduce a novel training algorithm\r
for unsupervised grammar induction, called\r
Zoomed Learning. Given a training set T and\r
a test set S, the goal of our algorithm is to\r
identify subset pairs Ti, Si of T and S such\r
that when the unsupervised parser is trained\r
on a training subset Ti its results on its paired\r
test subset Si are better than when it is trained\r
on the entire training set T . A successful application\r
of zoomed learning improves overall\r
performance on the full test set S.\r
We study our algorithm’s effect on the leading\r
algorithm for the task of fully unsupervised\r
parsing (Seginer, 2007) in three different English\r
domains, WSJ, BROWN and GENIA, and\r
show that it improves the parser F-score by up\r
to 4.47%.

Author-supplied keywords

  • Computational, Information-Theoretic Learning with
  • Natural Language Processing

Get free article suggestions today

Mendeley saves you time finding and organizing research

Sign up here
Already have an account ?Sign in

Find this document

  • PUI: 362643544
  • SGR: 80053228417
  • SCOPUS: 2-s2.0-80053228417
  • ISBN: 1932432868


  • Roi Reichart

  • Ari Rappoport

Cite this document

Choose a citation style from the tabs below

Save time finding and organizing research with Mendeley

Sign up for free