Second International Conference on Spoken Language Processing (ICSLP'92)

Banff, Alberta, Canada
October 13-16, 1992

Two Different Methodologies for Evaluating the Comprehension of Synthetic Passages

Cristina Delogu (1), S. Conte (2), A. Paoloni (1), C. Sementina (1)

(1) Fondazione Ugo Bordoni, Roma, Italy
(2) UniversitÓ di Roma, La Sapienza, Italy

In the present work we focus our attention on the type of understanding which is caused by listening to passages produced by vocal synthesizers, using two different methodologies: multiple choice test (Test 1) and click monitoring task (Test 2). Two texts taken from a published Italian reading comprehension test and pronounced by a human speaker and by the Olivetti PC Vox synthesizer were used. In Test 1, 150 subjects after listen to a passage had to answer ten multiple choice questions. In Test 2, other 50 subjects during the listening of a passage had to recognize some computer generated clicks distributed in the passage. The results of Test 1 show no statistically significant differences in performance between subjects who listened to natural passages and those who listened synthetic passages. The results of Test 2 show that click monitoring responses were faster for clicks in natural speech compared to those in synthetic speech.

Full Paper

Bibliographic reference.  Delogu, Cristina / Conte, S. / Paoloni, A. / Sementina, C. (1992): "Two different methodologies for evaluating the comprehension of synthetic passages", In ICSLP-1992, 1231-1234.