ANOPHONE: An Annotation Tool for Phonemes and L2 Annotation Systems for Czech

Abstract

Abstract The goal of this text is the presentation of the ANOPHONE annotation system, which allows for the management and annotation of speech data to develop a tool for the automatic transcription of speech of non-native speakers of Czech. This system is currently designed for annotations on the segmental level of recordings of non-native speakers of Czech, with the aim to train automatic speech recognition (ASR) models used in this tool. After an introductory section that discusses the use of technology in pronunciation teaching and mentions some of the e-learning applications for teaching the pronunciation of second languages (L2), we address both general and more specific aspects of speech data annotation to train ASR models and mention attributive and synthetic segmental systems of speech data annotation for Czech as L2. We also briefly introduce the annotation system of non-native speakers of Czech called BV1, which is used for testing the ANOPHONE tool. The main part of this text focuses on presenting the annotation tool itself, while the conclusion describes the experience of testing the speech data annotation tool using BV1 annotation system for Czech as L2.