Use this URL to cite or link to this record in EThOS: https://ethos.bl.uk/OrderDetails.do?uin=uk.bl.ethos.684079
Title: The genitive ratio and its applications
Author: Glover, Kevin
ISNI:       0000 0004 5919 9544
Awarding Body: University of Essex
Current Institution: University of Essex
Date of Award: 2016
Availability of Full Text:
Access from EThOS:
Full text unavailable from EThOS. Please try the link below.
Access from Institution:
Abstract:
The genitive ratio (GR) is a novel method of classifying nouns as animate, concrete or abstract. English has two genitive (possessive) constructions: possessive-s (the boy's head) and possessive-of (the head of the boy). There is compelling evidence that preference for possessive-s is strongly influenced by the possessor's animacy. A corpus analysis that counts each genitive construction in three conditions (definite, indefinite and no article) confirms that occurrences of possessive-s decline as the animacy hierarchy progresses from animate through concrete to abstract. A computer program (Animyser) is developed to obtain results-counts from phrase-searches of Wikipedia that provide multiple genitive ratios for any target noun. Key ratios are identified and algorithms developed, with specific applications achieving classification accuracies of over 80%. The algorithms, based on logistic regression, produce a score of relative animacy that can be applied to individual nouns or to texts. The genitive ratio is a tool with potential applications in any research domain where the relative animacy of language might be significant. Three such applications exemplify that. Combining GR analysis with other factors might enhance established co-reference (anaphora) resolution algorithms. In sentences formed from pairings of animate with concrete or abstract nouns, the animate noun is usually salient, more likely to be the grammatical subject or thematic agent, and to co-refer with a succeeding pronoun or noun-phrase. Two experiments, online sentence production and corpus-based, demonstrate that the GR algorithm reliably predicts the salient noun. Replication of the online experiment in Italian suggests that the GR might be applied to other languages by using English as a 'bridge'. In a mental health context, studies have indicated that Alzheimer's patients' language becomes progressively more concrete; depressed patients' language more abstract. Analysis of sample texts suggests that the GR might monitor the prognosis of both illnesses, facilitating timely clinical interventions.
Supervisor: Not available Sponsor: Not available
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral
EThOS ID: uk.bl.ethos.684079  DOI: Not available
Keywords: QA75 Electronic computers. Computer science
Share: