• 日時：平成22年7月26日（月）午後3時より
• 会場：東京大学駒場キャンパス 18号館4階 コラボレーションルーム4
• 発表者：Darrell Larsen and Jeff Heinz
• 発表題目：A corpus study and comparative analysis of formal learning proposals of Korean sound-symbolic vowel harmony
• 要旨：Neutral vowels in vowel harmony in sound-symbolic words in Korean (Sohn 1999) raise two challenges for researchers. First, word-medially these vowels are transparent (i.e. invisible) to the harmony process. This means a correct analysis must admit long-distance dependencies. Second, word-initially neutral vowels function not as transparent vowels but as harmony-triggers (Kim-Renaud 1976, Sohn 1986, Lee 1984, Cho 1994, inter alia). In other words, a correct analysis must account for the dual nature of neutral vowels word-initially versus word-medially.
  This research examines a corpus of sound-symbolic words to determine the extent to which this data supports (1) the dual functionality of neutral vowels and (2) the claim that all high vowels are in fact neutral, which extends a hypothesis of Cho (1994). We conclude the corpus strongly supports both (1) and (2).
  Finally, we compare two different proposals for automatically detecting vowel harmony patterns in data. The conclusion here is that, while neither approach at present captures the dual functionality of the neutral vowels, strictly piecewise methods (Heinz and Rogers 2010) more accurately extract the long-distance pattern in the corpus than tier-based bigram methods (Hayes and Wilson 2008, Goldsmith and Riggle, to appear).

懇親会の様子