Automatic Disambiguation of Homographic Heterophone Pairs Containing Open and Closed Mid Vowels
Christopher Shulby, Gustavo Mendonça, Vanessa Marquiafável · 2013
Abstract. The issue of openness in Brazilian Portuguese vowels is a question not yet satisfactorily explored in the field of automatic classification of homographic heterophones (HH). Therefore, we aimed to develop and test a pilot classifier which assists in the automatic disambiguation of HH. For this purpose, a set of 226 word pairs of HH with the unique grammatical classes, distinguished by alternating mid vowels [e, E] and [o, O], was analyzed. The results showed that the rules proposed herein solve most disambiguation problems of HH word pairs containing mid vowels in the corpus analyzed and can be applied to TTS and ASR applications. The data also revealed that a predominant trend of non-verb classes exists, and, for some word pairs, that value can reach 95 % occurrence. 1.