I have some recordings on an Olympus WS-700M from a few years ago on my computer, and it would save a ton of time if anyone could suggest a good transcription program. I'm looking to convert them into Word files. Thanks!

Nuance dragon software home has a large percentage of the market.

You can get an older version on eBay for cheap.

Retail new version is around $200.

Be aware that there will be many unusual and sometimes funny mistakes. I have the fun of reading medical records that have been transcribed via Dragon. Human beings are much better at this task. Having a human transcribe is expensive, I have been paying the cost for many years.

Hello!

I paid a service to transcribe all the interviews that were conducted as part of my dissertation. While I do not recall the name of the service, I do recall that it was not too costly.

Jim

Do you have an Apple device?

Go to the Notes app on your phone or tablet and press the microphone and see how it does before buying a program. It picks up and transcribes regular voice pretty well. If it works decently you can email the note to yourself, open it on your PC and copy and paste it into Word.

When I was contracting for a R&D Govt. Agency back in the 90's, we did an extensive study of voice -> text software. Some of findings:

1) there was little difference between available commercial products in terms of accuracy. The open source projects were about as good. Academic researchers in the subject were generally pessimistic about ever achieving acceptable results, saying they would need at least 100x more processing power than was available at the time, even in so-called super-computers.

2) no software-only product came close to even a mediocre professional transcriber.

3) no software-only product was deemed to be "good enough" for production use. For purposes of the study, "good enough" was one or two mistakes per page.

4) accuracy depended on the speaker more than any other factor. Speakers can be generally classified as "sheep" or "goats" in their speaking mannerisms, transcribing a sheep isn't too tough, a goat not so much. No matter how hard they try, a natural goat cannot imitate a sheep's vocal mannerisms. Note that varying regional accents made much less difference, however background noise interfered with accuracy considerably.

5) software that acceptably transcribed an individual speaking in a quiet room failed miserably at transcribing a meeting, a telephone call, or other multi-speaker situation.

6) there was no measurable difference between live transcription and a transcription of a recording of the same oration. Little accuracy was lost due to lower fidelity of the recording, some of those we tested were very bad/muddy. Ironically, software was often better at transcribing very poor quality recordings than humans.

7) The language used by the speaker made little difference, French, Spanish, Russian, etc. had all about the same accuracy levels. However, the software for transcribing Chinese, Indonesian, and Vietnamese performed much more poorly.

7) Accuracy levels were generally misrepresented by vendors of commercial software.

8) Accuracy had to be much greater than available software could promise, note that even a 96% accuracy leaves several mistakes per page.

9) running a spelling and grammar checker over the transcription made a significant difference, but still wasn't good enough.

10) the best results were achieved with an experienced sheep monitoring and correcting their own transcription in real time. This with the aid of an on-the-fly spell checker. Results were better doing this than with the best professional native language transcribers. Productivity went way down, however, due to speakers backing up and re-doing language with which they were unhappy.

11) Fronting a professional transcriber with a software transcription didn't make much difference in the quality of the results. In some cases, this made things considerably worse. However, we felt that this approach held much promise, with changes to the software, and training/experience of the operators. I recall that we recommended additional research in this approach, and the Agency rejected our proposal.

I was left with the impression that accurate automatic voice->text was not something that I was going to see in my lifetime. It does not appear that things have changed much in the past 20 years.