Type

PhD Thesis

Authors

Jer Hayes

Subjects

Linguistics

Topics
concept combination polysemous compunds knowledge base structural model noun noun compounds evaluation study linguistics machine translating

A structural alignment model of noun-noun compound interpretation (2003)

Abstract The interpretation of noun-noun compounds is complex, yet compounds such as 'web surfer' and 'beef baron' are generated and interpreted easily by native English speakers. Concept combination is the core process in the generation and interpretation o f noun-noun compounds. Such compounds may be read literally or metaphorically suggesting that the combination process is capable of both literal and metaphoric interpretations. The motivation for this thesis is to tackle three problems which occur in concept combination. These problems are: (1) compounds are often polysemous, (2) compounds often appear to be understood by evoking a context (or world knowledge) and (3) compounds can be interpreted figuratively. We suggest that adopting structural alignment allows us to deal with each of these problems. Structural alignment is a process whereby conceptual structures are placed into correspondence and similarities are found. The structural alignment model proposed in this thesis suggests that there are six core combination types and that an interpretation of a nounnoun compound will fall into one of these combination types. Some of these combination types are figurative and some rely on finding a context. We provide an implementation of the model, the fNCA system. The INCA system is a program where a user can find interpretations for noun-noun compounds. INCA has a knowledge base and attempts to find fixed patterns in a network representation of concepts. Depending on the type of pattern found, several types of interpretation can be generated. The performance of INCA is compared with that of a number of human subjects in a brief evaluation study. The study shows that combination types proposed by our structural alignment model to offer a good coverage of the interpretations that people generate. Finally we set out proposals for developing INCA further and outline directions for future research.
Collections Ireland -> Dublin City University -> DCU Faculties and Centres = DCU Faculties and Schools: Faculty of Engineering and Computing: School of Computing
Ireland -> Dublin City University -> Subject = Humanities: Linguistics
Ireland -> Dublin City University -> Publication Type = Thesis
Ireland -> Dublin City University -> Subject = Computer Science
Ireland -> Dublin City University -> DCU Faculties and Centres = DCU Faculties and Schools
Ireland -> Dublin City University -> Status = Unpublished
Ireland -> Dublin City University -> Subject = Computer Science: Machine translating
Ireland -> Dublin City University -> Subject = Humanities
Ireland -> Dublin City University -> DCU Faculties and Centres = DCU Faculties and Schools: Faculty of Engineering and Computing
Ireland -> Dublin City University -> Thesis Type = Masters Thesis

Full list of authors on original publication

Jer Hayes

Experts in our system

1
Jer Hayes
Dublin City University
Total Publications: 24