Use this URL to cite or link to this record in EThOS:
Title: Generating coherent references to multiple entities
Author: Gatt, Albert
ISNI:       0000 0001 3492 3653
Awarding Body: University of Aberdeen
Current Institution: University of Aberdeen
Date of Award: 2007
Availability of Full Text:
Access from EThOS:
Despite intensive research on the Generation of Referring Expressions (GRE), many GRE algorithms either lack empirical backing, or are motivated by concerns which arguably shift their focus away from the crucial problem, which is to generate natural descriptions, much as a person would generate them in a comparable situation. This problem becomes more acute in the case of plural reference. This thesis focuses on the generation of plurals, with particular attention to the semantic heart of the problem, that is, content determination. The first part presents an in-depth theoretical and empirical evaluation of the state of the art of GRE, and makes three main contributions.  The main contributions are:  a) The construction of a semantically transparent corpus of singular and plural descriptions; b) An empirical investigation into reference by human authors in this corpus; and c) An evaluation study on various existing algorithms, the first to explicitly address plurals. The second part of the thesis focuses directly on plurals.  It begins with a test of the similarity hypothesis on corpus data, leading to the development of a new algorithm which addresses the issues of similarity and conceptual categorisation.  This work is generalised to pluralities in discourse, starting from the hypothesis that pluralities should be conceptually coherent, that is, should conceptualise entities from the same perspective.  The investigation of this hypothesis, through five psycholinguistic experiments leads to an integrated framework for content determination in GRE.  Among the contributions of this second part of the thesis are: a) The use of an experimental psycholinguistic methodology to test hypothesis that are relevant to generation; b) The development of a novel approach to content determination that seeks to satisfy conceptual coherence through the use of corpus-derived similarity metrics.
Supervisor: Not available Sponsor: Not available
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral
EThOS ID:  DOI: Not available