Developing a Computer-Aided Pangasinense Language Learning System

Abstract

Ideally, instruction is best done one on one. However, due to the scarcity of public school teachers, this ideal remains just that, only an ideal. This ideal, however, can be realized by using a computer-assisted language learning system. One such language that this system can be applied to is the Pangasinense - one of the top ten languages of the Philippines. Using this system, any Filipino can now learn Pangasinense. Creating this involves developing the speech corpus for the Pangasinense language and designing a reading miscue detector (RMD) that employs hidden markov models (HMM) and artificial neural network (ANN). The RMD uses the reference verification (RV) method that compares the input speech to the reference speech found in the Pangasinense speech corpus. The collection of the speech corpus involved 10 native Pangasinense speakers who each recorded a total of 21 phrases and 309 words that were considered as common conversational phrases or words for Pangasinense. The system was initially tested by 10 native Pangasinense speakers, who also speak Filipino, and their scores were set as the reference scores. The system was then put to test by conducting a six-week pilot study participated by 10 Filipino speakers. The system’s effectiveness was then evaluated through the progress trends of all learners’ scores for each module. All learners’ progress curves showed to have a positive slope. In addition, the system’s efficiency was determined by its false alarm rate (FAR), misdetection rate (MdR), and accuracy. The system was able to get a FAR of 26.67% and 30%, MdR of 30.0% and 6.67%, and accuracy of 71.66% and 81.67%, for males and females group, respectively.