Aristotle University Of Thessaloniki (AUTH)
Πειραματικές διερευνήσεις στον τονισμό της ελληνικής
Abstract
dc:descriptionThis study investigates the nominal stress in Greek, where stress is marked lexically. The main goal is to investigate experimentally which are the most productive stress patterns on native speakers’ grammar, as they are reflected on their decisions on stress assignment of unknown words. Consequently, we explore the parameters of morphological information at the right edge of the word and lexical frequency as potential cues for stress position. In addition, we investigate differences between children and adult speakers’ stress grammars. For this purpose, we designed and conducted the following experimental tasks: a) a production task on 7-8 year old native speakers of Greek and on 18-23 year old adults and b) a perception task on 18-23 year old adults. For the purpose of the experiment, we constructed two-syllable and three-syllable pseudonouns with real class markers (-os, -as, -is, -a, -i, -o, -j, -ma), relied on innovative research tools (corpora, questionnaires). The main issue was the construction of pseudowords according to the Greek phonotactics, by taking into consideration the degree of familiarity with real words. The statistics revealed a strong preference for stress on penultimate, but not for all the experimental conditions. Pseudowords of specific morphological endings, highly motivated from the lexical frequencies, are strongly connected with specific stress patterns (p.e. stress on antepenultimate is the tendency for three-syllable nouns in -os & -o). However, the frequency effects are salient only on adults’ stress preferences. On the contrary, children overgeneralize penultimate stress across all morphological conditions. As a result, we assume that morphological information and lexicon tendencies are highly activated in adult’s stress grammar. On the other hand, young speakers’ grammar exhibit strongly effects of early grammars’ biases (p.e. The Emergence of The Unmarked). Finally, an extremely innovative part of this dissertation approaches the stress grammar as a machine learning problem. More specifically, we trained two algorithms with different object orientation (Maximum Entropy Model & Random Forest) with the experimental results, in order to make predictions about the stress patterns of unknown words.
Degree
thesis:*- Grantor dc:publisher
- Aristotle University Of Thessaloniki (AUTH)
- Year dc:date
- 2018
Author and committee
dc:creator, dc:contributor.*- Authors dc:creator
-
- Apostolouda, Vasiliki
- Αποστολούδα, Βασιλική
Subjects
dc:subject × 16Rights
- Language dc:language
- gre
Identifiers
dc:identifier.*- Identifier
- 10.12681/eadd/44406