- Q: Why the Audio Transcription Center cannot use Voice Recognition Software?
- A: Because Voice Recognition Software is not yet capable of producing to our strict standards.
- Q: What strict standards?
- A: Let us count the ways:
- VRS has difficulty in recognizing, simultaneously or not, two or more voices. Of course, two or more voices are intrinsic to oral histories.
- VRS has difficulty with accents.
- VRS has difficulty in dealing with less than broadcast quality sound.
- VRS has difficulty with overlapping dialogue, idioms, collaquialisms, and especially ambient sound.
- VRS – Formatting? Fuggedaboutit!
- VRS developer IMB reached a 94.5% accuracy milestone which they are very proud of in its evaluation by “using the SWITCHBOARD corpus, a collection of telephone conversations that’s been used for decades.” “SWITCHBOARD is not the industry standard for measuring human purity, however, which makes breakthroughs harder to achieve.”
- Finally, an important factor of VRS accuracy is the need for “training” the software to recognize the speech patterns and idiosyncrasies of the speakers. Imagine asking your narrators train the software that will be transcribing the session before each of your interviews. Oy!
- A: Let us count the ways:
BUT IT AIN’T ALL BAD
There are many projects where a very rough transcript is used as a quick reference source, and an actual verbatim transcript isn’t even required. In those cases, perfect transcripts are not needed, and VRS fits the bill… As well as lowers your initial budget.
In summary, if you don’t need a near-perfect transcript, VRS is a wonderful tool at a reduced cost. If you’re looking for an accurate transcript that is also 100% guaranteed, then the only option is to call your transcription vendor of choice. You might want to try us. Call us at (617) 423-2151, or click on the GET A QUOTE link in red.