[ 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

Both sides previous revision Previous revision
Next revision
Previous revision
Last revision Both sides next revision
user:zeman:treebanks:hu [2011/12/13 13:20]
zeman Sample.
user:zeman:treebanks:hu [2011/12/13 13:42]
zeman Parsing results.
Line 54: Line 54:
 ==== Inside ==== ==== Inside ====
  
-Both versions (CoNLL 2007 and BDT-IIare in the CoNLL 2006/2007 format.+The original Szeged Treebank is a phrase-based treebank and it is distributed in XML-based, TEI-compliant format. The CoNLL 2007 version is dependency-based (i.e. the head of each phrase was identified), distributed in the CoNLL 2006/2007 format.
  
-The syntactic guidelines (structure and labels) are described in Spanish in this [[http://ixa.si.ehu.es/Ixa/Argitalpenak/Barne_txostenak/1068549887/publikoak/guia.pdf|technical report]]. See Appendix 3 for some lists of tags+Morphological annotation includes lemmas. Morphosyntactic tags were probably disambiguated manually. The tagset used in SzTB seems to be same or similar to [[http://nl.ijs.si/ME/V4/msd/html/msd-hu.html|Multext-East]]. In the CoNLL version, tags were decomposed into CPOS columnPOS column and the list of feature-value pairs in the FEAT column.
- +
-Multi-word expressions have been collapsed into one tokenusing underscore as the joining character (e.g. Espainia_Poliziak, iduri_zait).+
  
 ==== Sample ==== ==== Sample ====
Line 136: Line 134:
 ==== Parsing ==== ==== Parsing ====
  
-BDT is a mildly nonprojective treebank. 1925 of the 151,604 tokens of combined BDT-II training and test sets are attached nonprojectively (1.27%).+SzTB is a mildly nonprojective treebank. 4032 of the 139,143 tokens of the CoNLL 2007 version are attached nonprojectively (2.9%).
  
-The results of the CoNLL 2007 shared task are [[http://nextens.uvt.nl/depparse-wiki/AllScores|available online]]. They have been published in [[http://aclweb.org/anthology-new/D/D07/D07-1096.pdf|(Nivre et al., 2007)]]. The evaluation procedure was changed to include punctuation tokens. These are the best results for Greek:+The results of the CoNLL 2007 shared task are [[http://nextens.uvt.nl/depparse-wiki/AllScores|available online]]. They have been published in [[http://aclweb.org/anthology-new/D/D07/D07-1096.pdf|(Nivre et al., 2007)]]. The evaluation procedure was changed to include punctuation tokens. These are the best results for Hungarian:
  
 ^ Parser (Authors) ^ LAS ^ UAS ^ ^ Parser (Authors) ^ LAS ^ UAS ^
-| Malt (Nilsson et al.) | 76.94 82.84 | +| Malt (Nilsson et al.) | 80.27 83.55 
-| Titov et al. | 75.49 | 81.93 +| Sagae | 79.53 83.51 
-| Sagae | 74.64 81.19 | +| Nakagawa | 76.74 | 82.49 
-| Carreras | 75.75 | 81.11 +Titov et al. | 77.94 82.18 |
-| Nakagawa | 72.56 | 81.04 | +
-| Malt (J. Hall et al.) | 74.99 80.61 +
-Johansson et al. | 75.08 80.43 |+
  
 The two Malt parser results of 2007 (single malt and blended) are described in [[http://aclweb.org/anthology-new/D/D07/D07-1097.pdf|(Hall et al., 2007)]] and the details about the parser configuration are described [[http://w3.msi.vxu.se/users/jha/conll07/|here]]. The two Malt parser results of 2007 (single malt and blended) are described in [[http://aclweb.org/anthology-new/D/D07/D07-1097.pdf|(Hall et al., 2007)]] and the details about the parser configuration are described [[http://w3.msi.vxu.se/users/jha/conll07/|here]].
  
-Parsing results on BDT-II have been published in Kepa Bengoetxea, Koldo Gojenola: [[http://aclweb.org/anthology-new/W/W10/W10-1404.pdf|Application of Different Techniques to Dependency Parsing of Basque]]. In: Proceedings of the First Workshop on Statistical Parsing of Morphologically Rich Languages (SPMRL 2010), NAACL Workshop, Los Angeles, California, USA, 2010. They report only Labeled Attachment Score (LAS) and their best system achieved LAS = 78.98%. 

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