Speech recognition for Dutch users

Question: I am considering to start using speech recognition software, and I would appreciate the advice of other users on which type of software to buy. It appears to me that “Dragon naturally speaking” may be the best choice. My native language is Dutch, but I tend to write about 95% in English. My major worry is that the speech recognition software may not recognize my “broken english”. I would like to hear from other people (Dutch/ foreign/ obscure accents) about their experiences on this. Secondly, I am looking for information on vendors of speech recognition in the Netherlands. Preferably some place where I can try different systems and where I can get good information about the products. Please reply and help me make up my mind!

Answer: I have no knowledge of Dragon Dictate, but I’ve been using IBM VoiceType Dictation software professionally. I’m a translator for the past two years and am very pleased with it. The vendor was Cosmos Automatisering in Valkenswaard (040-2046705). I’m English, so I bought the UK version, but the program has no trouble recognising my wife’s speech –she’s from Chicago. Obviously a certain amount of “training” is necessary before it can cope with a new user. The program specifically claims to be able to deal with non-standard accents.

A number of my Dutch colleagues and students have tried it out on my machine. We discovered that the accuracy of recognition was very much dependent on the quality of the particular individual’s English pronunciation. If the accent was truly “near native” (Nijmegen graduate “anglisten”), there was no problem. If the speaker was your common-or-garden Cloggy, there were a lot of mis-recognitions, but not so many as to render the program useless. Fourth-year students did significantly better than first-years. All this applies to IBM VoiceType, but I see no reason to assume that Dragon won’t be similar. BTW, I get something like 98% correct recognition with this program, which is higher than any claims I’ve seen for continuous speech programs. The mis-recognitions usually involve unusual words – Dutch place names for example – but the next time such words occur they are recognised without difficulty.

Leave a Reply

Your email address will not be published.

Notify me of followup comments via e-mail.