An investigation of the pronunciation of English words in German sms texts

Abstract:

This paper reports on pronunciation variations by German speakers of English words in German SMS speech database. The text database contains SMS communication in different domains, such as on holidays, families, congratulations, schools, etc. As expected, there exist many English words in the German texts. The speech database comprises 100 German speakers, these speakers are chosen with balanced age, female-male difference and dialect region. There are 13 different promptsheets for the speakers to read, so that only 7 or 8 speakers share the same promptsheet. Due to various characteristics of the English words and different English levels of the speakers, the English words in German SMS texts are pronounced in different ways, from standard pronunciation with an English accent to very poor mispronunciation with a very strong German accent. It is thus important to find out possible pronunciation variations of each word by average German speakers. These pronunciation variants can be built into the lexicon for the speech recognition of English words with German accent. It is further interesting to make a statistical study on the realization of particular English phonemes by average German speakers, the information can be exploited to derive pronunciation rules that can be applied to new vocabulary. We have investigated the pronunciation variations of all words that can have potential English pronunciations both at word level and phoneme level, only selected examples will be presented in this paper.


Year: 2009
In session: Sprache und Didaktik
Pages: 396 to 402