ISCA Archive ICSLP 1992
ISCA Archive ICSLP 1992

Usage of words and sentence structures in spontaneous versus text material

Noriko Umeda, Karen Wallace, Josephine Horna

This paper reports striking differences in the usage of words and sentence structures between spontaneously spoken material and written texts in English. The text material consists of twenty texts, containing an average of 700 words each, which are easy to read aloud and use minimal technical vocabulary. The spontaneous material consists of the transcribed conversations of eleven native speakers of English (six males, five females), each of whom participated in two hours of conversation. Using the AT&T UNIX system (Writer's Workbench) parser, all sentences were analyzed in terms of sentence type, verb types, content/function ratio, grammatical classes of words, average word length, part of speech of the sentence beginning, and overall complexity. Material from different texts and speakers was analyzed separately. All speakers show similar patterns with regard to these analysis categories. Different texts, on the other hand, show considerable variation. As a group, texts show higher complexity than the spontaneous speech. The results imply that although speakers' personalities and topics of conversation vary, they use easier words and simpler constructions than carefully written texts.