Latin had served as an official language across Europe from the Roman Empire until the 19th century. As a result, vast amount of Latin language historical documents (charters, account books) survived from the Middle Ages, waiting for recovery. In the digitization process, tremendous human efforts are needed for the transliteration of textual content, as the applicability of optical character recognition techniques is often limited. In the era of Digital Humanities our aim is to accelerate the transcription by using automatic speech recognition technology. We introduce the challenges and our initial results in developing a real-time, medieval Latin language LVCSR dictation system for East-Central Europe (ECE). In this region, the pronunciation and usage of medieval Latin is considered to be roughly uniform. At this phase of the research, therefore, Latin speech data was not collected for acoustic model training but only for test purposes — from a selection of ECE countries. Our experimental results, however, suggest that ECE Latin varies significantly depending on the primary national language on both acoustic-phonetic and grammatical levels. On the other hand, unexpectedly low word error rates are obtained for several speakers whose native language is completely uncovered by the applied training data.
Cite as: Mihajlik, P., Szabó, L., Tarján, B., Balog, A., Rábai, K. (2017) First Results in Developing a Medieval Latin Language Charter Dictation System for the East-Central Europe Region. Proc. Interspeech 2017, 2058-2062, doi: 10.21437/Interspeech.2017-1558
@inproceedings{mihajlik17_interspeech,
author={Péter Mihajlik and Lili Szabó and Balázs Tarján and András Balog and Krisztina Rábai},
title={{First Results in Developing a Medieval Latin Language Charter Dictation System for the East-Central Europe Region}},
year=2017,
booktitle={Proc. Interspeech 2017},
pages={2058--2062},
doi={10.21437/Interspeech.2017-1558}
}