Building LVCSR System for Transcription of Spontaneously Pronounced Russian Testimonies in the MALACH Project: Initial Steps and First Results

The MALACH project [1] uses the world's largest digital archives of video oral histories collected by the Survivors of the Shoah Visual History Foundation (VHF) and attempts to access such archives by advancing the state-of-the-art in Automated Speech Recognition (ASR) and Information Retrieval (IR). This paper discusses the initial steps and the first results in building large vocabulary continuous speech recognition (LVCSR) system for transcription of Russian witnesses. Russian as the third language processed in the MALACH project (after English [2] and Czech [3]) brought new problems especially in the phonetic area. Although the most of the Russian testimonies were provided by native Russian survivors we have encountered many different accents in their speech caused by a territory where the survivors are living.