A Lesk-inspired Unsupervised Algorithm for Lexical Choice from WordNet Synsets - Archive ouverte HAL Accéder directement au contenu
Communication Dans Un Congrès Année : 2014

A Lesk-inspired Unsupervised Algorithm for Lexical Choice from WordNet Synsets

Résumé

The generation of text from abstract meaning representations involves, among other tasks, the production of lexical items for the concepts to realize. Using WordNet as a foundational ontology, we exploit its internal network structure to predict the best lemmas for a given synset without the need for annotated data. Experiments based on re-generation and automatic evaluation show that our novel algorithm is more effective than a straightforward frequency-based approach
La generazione di testo a partire da rappresentazioni astratte comporta, tra l'altro, la produzione di materiale lessi-cale per i concetti da generare. Usando WordNet come ontologia fondazionale, ne sfruttiamo la struttura interna per indi-viduare il lemmapì u adatto per un dato synset, senza ricorrere a dati annotati. Es-perimenti basati su ri-generazione e valu-tazione automatica mostrano che il nostro algoritmò epì u efficace di un approccio diretto basato sulle frequenze.
Fichier principal
Vignette du fichier
Basile2014.pdf (205.66 Ko) Télécharger le fichier
Origine : Fichiers éditeurs autorisés sur une archive ouverte
Loading...

Dates et versions

hal-01228924 , version 1 (19-11-2015)

Identifiants

Citer

Valerio Basile. A Lesk-inspired Unsupervised Algorithm for Lexical Choice from WordNet Synsets. First Italian Conference on Computational Linguistics, 2014, Pisa, Italy. ⟨10.12871/CLICIT2014110⟩. ⟨hal-01228924⟩
97 Consultations
100 Téléchargements

Altmetric

Partager

Gmail Facebook X LinkedIn More