tranfree issue 15 - 6 June 2000

Dictation Software

By Alex Eames

Since I first mentioned it in a previous tranfree edition a few months ago, people have been asking me for articles about dictation software. I have not yet found anyone to write an authoritative article on this subject. But I can give you the benefit of my limited experience.

I have been using IBM VoiceType 3.0 on and off for a couple of years now. I have found it to be quite useful. In fact, I like it, and I've trained it! It is good enough for me not to need to upgrade to one of the more modern packages. This is precisely why I don't know much about the newer versions.

Caveats

Before I give you specific details on performance let me mention a few points...

1. IBM VoiceType 3.0 was state of the art in 1998. It is the older type of discrete speech recognition where you have to enunciate each word separately with a tiny pause between each. Newer software versions of several manufacturers allow you to speak naturally or almost naturally.
   
   
2. I speak with a very neutral accent - almost "BBC English" - it may be harder if you have a very broad or "foreign" accent. I have not measured this.
   
   
3. I don't think there is speech recognition software available in all markets yet. If you speak a "minor" language like Polish (my wife is Polish - please don't all write in this kind of software is not available yet.

Performance

Rate and accuracy.

I have been able to achieve sustained rates of 80 words per minute at around 97% accuracy.

By my definition this means that...

• the overall time taken to get the text into the machine and correct it worked out at a rate of 80 words per minute.
• for every 100 words I read out, 3 were incorrectly recognised and had to be corrected or learned by the computer

When not to use it.

Generally I find this satisfactory, but there are texts which I do not consider using VoiceType for..

• When there is a lot of formatting it is quicker to use the keyboard. I won't even consider tables.
• Abbreviations, acronyms, symbols and formulas are a nightmare.
• When the text contains a lot of jargon which is not in the standard recognition dictionaries you will have to correct every single word that it does not know. If you are patient this can be a good investment. If it is the kind of material you do regularly it will pay off. If not, you may be better off with the keyboard.

Other Problems

Dictation software does also bring problems of its own. I have read many published books over the last couple of years which contain dictation errors. These have been missed by professional editors and proofreaders in major publishing houses!

For example to, too, two and 2 are all possible variations on the exact same sound. The software has to decide which is the most likely from the context. The bad news? Your spell-checker will not find these, so your proofing must be ever more vigilant!

Conclusion

That's my experience with older software. It definitely works! The newer software is meant to be better and allows you to speak in a more natural and continuous manner.

Which one should you buy? I can't recommend one because my knowledge is not current. Dragon, IBM and Lernout & Hauspie all have different offerings that I know about.

What I do know is that if you are willing to spend a few days training and getting used to the system, you will get increased efficiency from pretty-much any of them. This will cost you far more in time than the most expensive product costs to buy, so don't automatically buy the cheapest "to save money".

You may also be better off buying a really good...

...computer microphone, rather than settle for the one which comes bundled with the software. By good, I mean comfortable and effective. Comfortable is very important!