On the use of multi-lingual approach for a cloud-based transcription system for the ‘Ilonggoish’ dialect

Abstract

The study is aimed at the development of a Transcription System for ‘Ilonggoish’ Dialect, which is a widely-spoken local language in the Philippines. It is a software that records speech in .wav file format, transcribes speech into text, and generates text file containing the transcribed text. The system has a built in speech recognition that has the capability to recognize pre-recorded speeches spoken in different languages such as English, Filipino, Hiligaynon, and Ilonggoish dialect. Integrated into the system are the recording tool for the input speech data, data storing capability in . wav format, and text storing capability in .txt format. This study presents an approach to extract features of the spoken words by using the Mel Frequency Cepstral Coefficients (MFCC) algorithm from speech signals of isolated spoken words, and Hidden Markov Model (HMM) method in presenting the recognized spoken words in text format. The system uses the Google Cloud’s database of words as the baseline for standard words. It was evaluated by linguists specializing in Filipino, English, and Hiligaynon languages, and IT experts in different fields such as the academe and industry.