ILSP Feature-based multi-tiered POS Tagger
- Autoři
- Papageorgiou, Haris ; Prokopidis, Prokopis and Prokopidis, Prokopis
- Identifikátor
- http://hdl.handle.net/11372/LRT-1308
- Datum vydání
- 2014-07-30
- Typ
- toolService
- Popis
- ILSP FBT Tagger is an adaptation of the Brill tagger trained on Greek text. It uses a PAROLE compatible tagset of 584 different tags which capture the morphosyntactic particularities of the Greek language. Working on the output of a sentence detection and tokenisation tool, the tagger assigns initial tags, looking up in a lexicon created from a manually annotated corpus during training. A suffix lexicon is used for initially tagging unknown words. 799 contextual rules are then applied to improve the initial phase output.
- Nakladatel
- ILSP/R.C. "Athena"
- Klíčová slova
- POS tagger