[ Skip to the content ]

Institute of Formal and Applied Linguistics Wiki

• ÚFAL Wiki Home
• ÚFAL Home
• Sitemap
• RSS
• ÚFAL calendar
• Recent changes
• Login
• Hledat:

[ Back to the navigation ]

Differences

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

View differences: Side by SideInline Go

Link to this comparison view

Both sides previous revision Previous revision 2012/03/22 18:18 zeman Parsing.2012/03/22 17:28 zeman Sample.2012/03/22 17:06 zeman Inside.2012/03/22 16:47 zeman ICON 2010 Telugu data size.2012/03/22 11:46 zeman ICON 2009 Telugu data size.2012/03/22 11:34 zeman Training data size (both sentences and words) was identical in ICON 2009 and 2010.2012/03/22 11:30 zeman Size.2012/03/22 11:14 zeman vytvořeno Go		2012/03/22 18:18 zeman Parsing.2012/03/22 17:28 zeman Sample.2012/03/22 17:06 zeman Inside.2012/03/22 16:47 zeman ICON 2010 Telugu data size.2012/03/22 11:46 zeman ICON 2009 Telugu data size.2012/03/22 11:34 zeman Training data size (both sentences and words) was identical in ICON 2009 and 2010.2012/03/22 11:30 zeman Size.2012/03/22 11:14 zeman vytvořeno Go Next revision Both sides next revision
user:zeman:treebanks:te [2012/03/22 11:46] zeman ICON 2009 Telugu data size.		user:zeman:treebanks:te [2012/03/22 16:47] zeman ICON 2010 Telugu data size.
Line 51:		Line 51:
	| TOTAL       |  1756 |  6752 | 3.85 |		| TOTAL       |  1756 |  6752 | 3.85 |
-	The data distributed for ICON 2010 was slightly smaller, maybe it had been cleaned up? Note that the number of training words, 7602, is identical to the number published for ICON 2009. I cannot verify it because I only see chunks, not words in the CoNLL data format.	+	As for ICON 2010, the data description in [[http://ltrc.iiit.ac.in/nlptools2010/files/documents/toolscontest10-workshoppaper-final.pdf|(Husain et al., 2010)]] does not match the data that we downloaded during the contest. They indicate the number of words, while we can count the number of nodes, i.e. chunks. Anyway, the number of training sentences should match and it does not. Also note that they give the same number of words as they gave for ICON 2009 in [[http://ltrc.iiit.ac.in/nlptools2009/CR/intro-husain.pdf|(Husain et al., 2009)]]. In any case, the training data shrank during the year (clean up?) In the following table, we give both the published and real numbers of sentences, the published number of words and the counted number of chunks (nodes).
-	^ Part ^ Sentences ^ Chunks ^ Ratio ^ Words ^ Ratio ^	+	^ Part ^ Sentences ^ Chunks ^ Ratio ^ PSentences ^ Words ^ Ratio ^
-	| Training | 1400 | ? | ? | 7602 | 5.43 |	+	| Training    |  1300 |  5125 |  3.94 |  1400 |  7602 |  5.43 |
-	| Development | 150 | ? | ? | 839 | 5.59 |	+	| Development |   150 |   597 |  3.98 |   150 |   839 |  5.59 |
-	| Test | 150 | ? | ? | 836 | 5.57 |	+	| Test        |   150 |   599 |  3.99 |   150 |   836 |  5.57 |
-	| TOTAL | 1700 | ? | ? | 9277 | 5.46 |	+	| TOTAL       |  1600 |  6321 |  3.95 |  1700 |  9277 |  5.46 |
-		+
-	We drew our training and test data from the ICON 2010 datasets but we have fewer sentences – why?	+
-		+
-	^ Part ^ Sentences ^ Chunks ^ Ratio ^	+
-	| Training |  1300 |  5125 | 3.94 |	+
-	| Test     |   150 |   597 | 3.98 |	+
-	| TOTAL    |  1450 |  5722 | 3.95 |	+
	==== Inside ====		==== Inside ====

---

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