This paper presents an ensemble system for dependency parsing: three parsers are separately trained and combined by means of a majority vote. The three parsers are (1) the MATE parser [http://code.google.com/p/mate-tools/], (2) the DeSR parser [http://sites.google.com/site/desrparser/], and (3) the MALT parser [http://maltparser.org/]. The MATE, that was never used before on Italian language, drastically outperforms the other parsers in the SPLeT shared task. Nonetheless, a simple voting combination further improves its performances
Simple parser combination
MAZZEI, Alessandro;BOSCO, CRISTINA
2012-01-01
Abstract
This paper presents an ensemble system for dependency parsing: three parsers are separately trained and combined by means of a majority vote. The three parsers are (1) the MATE parser [http://code.google.com/p/mate-tools/], (2) the DeSR parser [http://sites.google.com/site/desrparser/], and (3) the MALT parser [http://maltparser.org/]. The MATE, that was never used before on Italian language, drastically outperforms the other parsers in the SPLeT shared task. Nonetheless, a simple voting combination further improves its performancesFile in questo prodotto:
File | Dimensione | Formato | |
---|---|---|---|
MazzeiBoscoSPLET2012.pdf
Accesso aperto
Tipo di file:
POSTPRINT (VERSIONE FINALE DELL’AUTORE)
Dimensione
186.21 kB
Formato
Adobe PDF
|
186.21 kB | Adobe PDF | Visualizza/Apri |
I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.