Irwan Sumarsono, Radina Anggun Nurisma, Salim Nabhan, Aliv Faizal Muhammad, Halimatus Sa'dyah
Through student speeches and other means, English teachers have already been able to assess their students' speaking skills firsthand. A huge number of pupils' speaking skills must be evaluated, which takes time. Thus, the need for an automated methodology to gauge English-speaking proficiency. It is good to examine English speaking with a computer using speech recognition technologies. However, to accurately assess the students' speaking performance automatically using computer requires a large dataset of students' various responses that are marked as appropriate responses. These responses are further marked as correct or expected responses, which later be used for automatic scoring generation and other various useful features. In this research we develop a dataset by collecting, analysing, categorising, and weighting various answer possibilities collected from students' responses to specific questions. The dataset, from 20 participants, was examined through Confusion Matrix, and it yielded accuracy score above 80%. However, this yielding result needs further testing with higher number of participants to get stronger accuracy. This dataset is expected to power the automatic English speaking testing app in assessing students' speaking performance. Eventually, the app is expected to be helpful for English teachers to efficiently score the speaking performance of huge numbers of students. © 2022 IEEE.
Politeknik Elektronika Negeri Surabaya, Indonesia; Universitas Pgri Adi Buana Surabaya, Indonesia
Research at a Glance
Register to unlock
Topics & SDG Alignment
Register to unlock
Collaboration
Register to unlock
Author Profile (Selected)
Register to unlock
References Overview
Register to unlock
Journal & Source
Register to unlock
Metadata & Integrity
Register to unlock