DEVELOPER ROLES IN THE SPEECH CORPUS FOR PRONUNCIATION VARIATION OF RUSSIAN SPEAKERS

Abstract

he article describes the user roles in the speech corpus for studying pronunciation variability of native speakers in the Russian language. The need for systems of speech recognition of speakers with dialects and disabilities is stated. It is emphasized the need to study the pronunciation variability of pronunciation by different speakers, taking into account regional and individual speech characteristics. Subsequent creating a speech corpus as the basis for regional and individual speech recognition is discussed. The speech corpus being developed contains recordings of the same text fragments by different speakers. The system of audio speech markup for researching the pronunciation variability of native Russian speakers is described. The roles of administrator, moderator, marker and analyst are provided for working with the corpus. Each user rights in the speech corpus are described: the administrator is the role which has all possible rights in the system. The marker is the user whose main task is to mark up the audio recordings. Analyst is the user who can assess and process the data in the speech corpus. The necessity of the moderator’s role in controlling the quality of markup is proved by the fact that it’s mostly students who mark up the audio recordings. The information in the developed speech corpora is supposed to be useful for phonetic studies in linguistics and a database for oral speech recognition.