Merge pull request sebastianruder#108 from stompchicken/master

Add section for unsupervised dependency parsing

_data/dependency_parsing.yaml

@@ -112,3 +112,40 @@ Reference:
   UAS:         95.66
   LAS:         94.03
   comment:     Stanford conversion **v3.5.0**
+
+Unsupervised_Penn_Treebank:
+-
+  model:       Iterative reranking
+  authors:     Le & Zuidema
+  year:        2015
+  UAS:         66.2
+  paper:       Unsupervised Dependency Parsing - Let’s Use Supervised Parsers
+  url:         http://www.aclweb.org/anthology/N15-1067
+-
+  model:       Combined System
+  authors:     Spitkovsky et al
+  year:        2013
+  UAS:         64.4
+  paper:       Breaking Out of Local Optima with Count Transforms and Model Recombination - A Study in Grammar Induction
+  url:         http://www.aclweb.org/anthology/D13-1204
+-
+  model:       Tree Substitution Grammar DMV
+  authors:     Blunsom & Cohn
+  year:        2010
+  UAS:         55.7
+  paper:       Unsupervised Induction of Tree Substitution Grammars for Dependency Parsing
+  url:         http://www.aclweb.org/anthology/D10-1117
+-
+  model:       Shared Logistic Normal DMV
+  authors:     Cohen & Smith
+  year:        2009
+  UAS:         41.4
+  paper:       Shared Logistic Normal Distributions for Soft Parameter Tying in Unsupervised Grammar Induction
+  url:         http://www.aclweb.org/anthology/N09-1009
+-
+  model:       DMV
+  authors:     Klein & Manning
+  year:        2004
+  UAS:         35.9
+  paper:       Corpus-Based Induction of Syntactic Structure - Models of Dependency and Constituency
+  url:         http://www.aclweb.org/anthology/P04-1061

dependency_parsing.md

@@ -37,4 +37,18 @@ The following results are just for references:
   results=site.data.dependency_parsing.Reference
   scores='POS,UAS,comment' %}
 
+# Unsupervised dependency parsing
+
+Unsupervised dependency parsing is the task of inferring the dependency parse of sentences without any labeled training data.
+
+## Penn Treebank
+
+As with supervised parsing, models are evaluated against the Penn Treebank. The most common evaluation setup is to use
+gold POS-tags as input and to evaluate systems using the unlabeled attachment score (also called 'directed dependency
+accuracy').
+
+{% include table.html
+  results=site.data.dependency_parsing.Unsupervised_Penn_Treebank
+  scores='UAS' %}
+
 [Go back to the README](README.md)