ETD

Archivio digitale delle tesi discusse presso l'Università di Pisa

Tesi etd-10212019-154642


Tipo di tesi
Tesi di laurea magistrale
Autore
PEDINOTTI, PAOLO
Indirizzo email
pedinotti.paolo@gmail.com
URN
etd-10212019-154642
Titolo
A distributional account of metonymy
Dipartimento
FILOLOGIA, LETTERATURA E LINGUISTICA
Corso di studi
LINGUISTICA E TRADUZIONE
Relatori
relatore Prof. Lenci, Alessandro
correlatore Prof.ssa Marotta, Giovanna
Parole chiave
  • sentence processing
  • metonymy
  • event knowledge
  • distributional semantics
Data inizio appello
18/11/2019
Consultabilità
Completa
Riassunto
The aim of this work is to propose a distributional model for predicting the interpretation of words in metonymic contexts. The model is based on the assumption that speakers rely on knowledge about typical events (McRae and Matsuki 2009) to figure out the additional meaning. This is in line with psycholinguistic evidence showing that extralinguistic information is available early during processing (Hagoort and Van Berkum 2007, Elman 2014) and thereby "constrains" the process of sentence interpretation (McRae and Matsuki 2013). More precisely, I argue that, along with thematic fit, the relation with the concept expressed by the metonymic word is a key factor in determining the best interpretation. A dataset with 50 metonymic sentences and their literal controls has been created to evaluate the ability of the model to deal with both conventional metonymies (like Producer-for-Product metonymy, e.g. "The editor paraphrased Hemingway") and "contextual" metonymies (known as "Reference Transfers", e.g., "The ham sandwich has been waiting for over an hour"). The hypothesis is confirmed by results showing that the quality of paraphrases is correlated to the ranking produced by the model. However, I argue that the performance of the model could potentially be improved by representing more information useful for the interpretation (constraints), such as discourse and utterance context.
File