[ Skip to the content ]

Institute of Formal and Applied Linguistics Wiki

[ Back to the navigation ]

You are here: start » user » zeman » interset » drivers

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
Next revision Both sides next revision
user:zeman:interset:drivers [2008/03/31 22:14]
zeman de::conll 		user:zeman:interset:drivers [2008/04/03 14:27]
zeman Restructuralization.
Line 2: Line 2:
  
 This is an overview of existing tag set drivers. Tag-set or language specific issues are described here. This is an overview of existing tag set drivers. Tag-set or language specific issues are described here.
 +
 +===== Arabic (ar) =====
 +
 +The Arabic CoNLL tags are derived from the tags of the Prague Arabic Dependency Treebank.
 +
 +Created in 2006-2007.
 +Total work time: 13 hours
 +
 +===== Bulgarian (bg) =====
 +
 +The Bulgarian CoNLL tags are derived from the tags of BulTreeBank. Speciality: sophisticated system of pronouns includes interrogative adverbs and numerals.
 +
 +Created in 2007.
 +Total work time: 35 hours
 +
 +The main reasons why the implementation took so long:
 +  * Necessity to re-work the system of main word classes, especially pronouns.
 +  * Necessity to separate morphological and lexical definiteness (there are indefinite pronouns morphologically definite, and vice versa).
 +  * Necessity to separate morphological and lexical aspect (aorist vs. imperfect tense; there are perfective verbs that can occur in imperfect tense).
 +  * Driver tester required that encode(decode(x))=x. However, the CoNLL incarnation of the tags was inconsistent, in the order and form in which it presented features.
  
 ===== Chinese (zh) ===== ===== Chinese (zh) =====
Line 38: Line 58:
  
 More than half of the time was consumed during testing for tuning tags containing the Sem feature. More than half of the time was consumed during testing for tuning tags containing the Sem feature.
 +
 +===== Danish (da) =====
 +
 +Tags of the Danish Dependency Treebank converted to CoNLL format. 144 tags with complex documentation in Danish.
 +
 +Total work time: about 7 hours
 +
 +===== English (en) =====
 +
 +==== Penn Treebank Tagset ====
 +
 +Penn Treebank (45 atomic tags). Detailed classification of punctuation.
 +
 +Total work time: about 3 hours
 +
 +==== CoNLL Tagset (derived from Penn tags) ====
 +
 +The driver is just an envelope around the ''en::penn'' driver.
 +
 +Total work time: 48 minutes
  
 ===== German (de) ===== ===== German (de) =====
Line 59: Line 99:
 Total work time: 10 min Total work time: 10 min
  
-===== Time needed for tag set conversion =====+===== Swedish (sv) =====
  
-Poznamenávám si, kolik času mi zabral který ovladač, abych to mohl publikovat. Srovnání potřebného času s časem potřebným na obyčejný převod je zajímavé, i když vím, že ve skutečnosti ušetřím až při opakovaném využití ovladače.+==== Mamba and CoNLL ====
  
-Ruský treebank (nejen značky, ale vůbec převod formátu): +Mamba tagset of Talbanken05. 48 tags, no morphosyntactic categories but detailed classification of auxiliary and modal verbs and punctuation. CoNLL driver is just an envelope around Mamba.
-12:36+
  
-Arabské značky (Otovy i Buckwalterovy, ještě bez Intersetu, 22.3.2006): +Total work time: about 3 hours
-4:45+1+1:40 = 7:25+
  
-Dánské značky DDT/Parole (144 značek s košatým popisem) +==== Tags of Hajič's Swedish tagger ====
-asi 7 hodin+
  
-Švédské značky Mamba (48 značek) +Based on PAROLE Swedish tagset but some characters different (@ => W), and filled by dashes to uniform length of 9 characters (although i-th position does not always encode the same feature).
-asi 3 hodiny+
  
-Penn Treebank (36 značek) +No reliable statistics of work time; estimated 8 hours
-asi 3 hodiny, ale tady jsem to ještě neměřil, takže to je jen hrubý zpětný odhad+
  
-Hajičovy švédské značky +===== Time needed for tag set conversion =====
-0:32 - tady zjevně chybí úplná statistika+
  
-Arabské značky CoNLL +Some records about targeted tagset conversion for given tagset pairs, done in early 2006:
-4:33+5:19+3:16 = 13:08+
  
-Bulharské značky CoNLL +Ruský treebank (nejen značky, ale vůbec převod formátu): 
-0:20+1:00+0:26+5:44+2:00+6:15+1:20+0:46+1:26+2:30+0:48+12:44 = 35:19 +12:36
-(ale u bulharštiny jsem se dost natrápil s jevy, které do té doby nebyly v intersetu podchycené)+
  
-Anglické značky CoNLL +Arabské značky (Otovy i Buckwalterovy, ještě bez Intersetu, 22.3.2006): 
-0:48 - možná tady chybí statistika, ale možná taky ne, protože stačilo upravit existující ovladač Penn Treebanku, ne? +4:45+1+1:40 = 7:25
- +
-Žádné z výše uvedených převodů (tedy vše napsané před říjnem 2007) ještě neměly k dispozici chytré funkce pro nahrazování nepovolených hodnot.+
  

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