Differences between revisions 4 and 7 (spanning 3 versions)
|
Size: 764
Comment:
|
← Revision 7 as of 2013-01-29 18:26:00 ⇥
Size: 857
Comment:
|
| Deletions are marked like this. | Additions are marked like this. |
| Line 1: | Line 1: |
| ## page was renamed from TnT | |
| Line 3: | Line 4: |
| = TnT = | = NKJP model for TnT Tagger = |
| Line 5: | Line 6: |
| Tu można pobrać utworzony z ręcznie anotowanego podkorpusu milionowego NKJP model dla taggera [[http://www.coli.uni-saarland.de/~thorsten/tnt/ | TnT tagger]]. [[attachment:nkjp.zip|Model]] udostępniony jest na licencji w stylu BSD. Pobrany plik należy zdekompresować programem obsługującym format zip. | Here you can download the model for [[http://www.coli.uni-saarland.de/~thorsten/tnt/ | TnT tagger]], which was created by training the tagger with the one-million manually annotated subcorpus of the Polish National Corpus. The [[attachment:nkjp.zip|model]] is available under BSD license. The downloaded file has to be unzipped. |
| Line 7: | Line 8: |
| Uwaga: aby skorzystać z taggera, należy uzyskać jego kopię od autora. Uruchomienie taggera: | To use the TnT tagger, you need a copy from its author (see [[http://www.coli.uni-saarland.de/~thorsten/tnt/ | his webpage]]). To run the tagger, use: |
| Line 11: | Line 12: |
| Plik wejściowy musi być wcześniej podzielony na zdania (oddzielone dwoma końcami wiersza) i wyrazy zgodnie z regułami podziału stosowanymi w korpusie NKJP. Tagger uzyskuje średni wynik około 88% poprawnych znaczników; nie dokonuje lematyzacji. | The input file needs to be tokenized into sentences (separated with two end-of-line characters) and words in the way the Polish National Corpus is tokenized. (Use Morfeusz if unsure). The tagger's quality is around 88%. |
NKJP model for TnT Tagger
Here you can download the model for TnT tagger, which was created by training the tagger with the one-million manually annotated subcorpus of the Polish National Corpus. The model is available under BSD license. The downloaded file has to be unzipped.
To use the TnT tagger, you need a copy from its author (see his webpage). To run the tagger, use:
tnt nkjp <nazwa_pliku>
The input file needs to be tokenized into sentences (separated with two end-of-line characters) and words in the way the Polish National Corpus is tokenized. (Use Morfeusz if unsure). The tagger's quality is around 88%.