In this paper, a new self-training method for domain adaptation is illustrated, where the selection of reliable parses is carried out by an unsupervised linguistically-driven algorithm, ULISSE. The method has been tested on biomedical texts with results showing a significant improvement with respect to considered baselines, which demonstrates its ability to capture both reliability of parses and domain-specificity of linguistic constructions.
Unsupervised Linguistically-Driven Reliable Dependency Parses Detection and Self-Training for Adaptation to the Biomedical Domain
Felice Dell'Orletta;Giulia Venturi;Simonetta Montemagni
2013
Abstract
In this paper, a new self-training method for domain adaptation is illustrated, where the selection of reliable parses is carried out by an unsupervised linguistically-driven algorithm, ULISSE. The method has been tested on biomedical texts with results showing a significant improvement with respect to considered baselines, which demonstrates its ability to capture both reliability of parses and domain-specificity of linguistic constructions.File in questo prodotto:
Non ci sono file associati a questo prodotto.
I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
