This is a Sarawak 6C3D Malay conversation data for the purpose of speech technology research. At the moment, this is an experimental data and currently used for investigating speaker diarization. The data was collected by Faculty of Computer Science and Information Technology, Universiti Malaysia Sarawak.
The data consists of 38 conversations that have been transcribed using Transcriber (see TextGrid folder), where each file contains two speakers. Each conversation was recorded by different individuals using microphones from mobile devices or laptops thus, different file formats were collected from the data collectors. All data was then standardized to mono, 16000Khz, wav format.
We provide files:
- wav
- rttm
- textgrid
This data was experimented for a speaker diarizarion task, where it was used for evaluating our speaker diarization models. Our work was presented at the recent IALP 2023 in Singapore.
Cite our work when this data is used:
@INPROCEEDINGS{
10337314,
author={Rahim, Mohd Zulhafiz and Juan, Sarah Samson and Mohamad, Fitri Suraya},
booktitle={2023 International Conference on Asian Language Processing (IALP)},
title={Improving Speaker Diarization for Low-Resourced Sarawak Malay Language Conversational Speech Corpus},
year={2023},
pages={228-233},
keywords={Training;Oral communication;Data models;Usability;Speech processing;Testing;Speaker diarization;x-vectors;clustering;low-resource;auto-labeling;pseudo-labeling;unsupervised},
doi={10.1109/IALP61005.2023.10337314}}
For further details:
Sarah Samson Juan sjsflora@unimas.my
Mohd Zulhafiz bin Rahim mzhafiz1999@gmail.com