We sat down with Mary Kenny, Product @ Temi, to discuss transcription services.
“What is Temi?”
Mary: Temi’s an automated transcription product–we convert audio to text, to make it searchable & readable.
The key value of Temi vs. other transcription services is its fast turnaround time and cheap price.
Temi is one of the cheapest providers of transcription with a turnaround time in mere minutes.
“Who uses Temi?”
Mary: People like journalists, filmmakers, podcasters, religious speakers, students & researchers.
People who record interviews “on-the-go” and face time-sensitive deadlines.
You might record an interview and by the time you get back to your desk you’ve got a transcript in your inbox.
In film-making, documentarians create scripts from interviews by stitching them together.
“How do you measure the quality of a transcript?”
Mary: Tough question! It’s an ongoing area of study.
The most commonly used metric is what’s called “word error rate.”
It’s a formula that takes the number of substitutions/insertions/deletions required to correct the transcript divided by the number of words in the transcript.
Different companies have discrepancies in how to benchmark their performance vs. each other.
For example, Google has publicized a word error rate of 18%–however this advertised rate doesn’t share the clarity and quality of the audio files they were testing.
As a customer, you can determine quality by how well a service identifies different speakers or identifies sentence punctuation.
These are examples of transcript quality that word-error-rate doesn’t cover!
Temi highlights parts of the transcript that are more likely to have transcription errors–something that other tools won’t surface.
“Any pro-tips about how to maximize the value of Temi as a user?”
We provide an interactive web-based editor–which is much better than editing in Microsoft Word or Google Docs, since we align your text with timestamps in your audio file you can use our “find” feature, making it super easy to compare your transcript with the source audio file.
Besides using our editor, you can improve the quality of your transcripts by improving the audio quality of your recordings.
This can be done by:
- moving speakers close to the microphone
- avoid cross-talk (people talking simultaneously in your recording)
- speak slowly, loudly, and clearly
- invest in better audio-capturing devices (microphones)
- minimize background noise
“Any hints at what’s on the horizon for the transcription business at large?”
Mary: Applications for speech recognition is a $60B market.
One area of high growth in voice-to-text services is text chat in a business context. This includes call centers, marketing automation, customer support
I’m personally excited about this huge area of automated transcription!
Interested in Jagaloon Studios?
We are a content production studio specializing in short-form video content for businesses.