While walking my husky after work yesterday, I Jott’ed myself:
Another great work out today on the electrical, you had over 3 kilometers and over 550 calories burned in 32 minutes. Nice work and then some good wait listing …
Most human readers would automatically parse this Jott as:
Another great workout today on the elliptical, you had over 3 kilometers and over 550 calories burned in 32 minutes. Nice work and then some good weight lifting …
Even though I don’t know a lot about Jott’s transcription engine, I’ll share my perspective on the identified differences:
- “work out” vs. “workout” and “wait” vs. “weight” – These are subtle differences. Differences that can only be resolved with an understanding of context. In other words, a human reader knows that I was attempting to capture some data on my lunch-time exercise routine, and re-parses the Jott with contextually correct words. In order to correct such subtle ‘errors of transcription’, Jott will need to develop semantic filters. Filters that can take context into account.
- “electrical” vs. “elliptical” and “listing” vs. “lifting” – These are glaring differences. I know, from past experience, that Jott has words like “elliptical” and “lifting” in its ‘dictionary’. Therefore, I regard these as errors originating from Jott’s inability to ‘hear’ what I’m saying. And although a context-based filter may also help here, I feel I must share some of the responsibility for not clearly articulating my Jott.
What does all of this mean?
Meaning, indeed, is the root of it all!
What this means is that some future version of Jott will need to do a better job of capturing meaning. What I had intended. The context in which I framed my Jott.
What this means is that in the longer term, a few major releases of Jott down the road, Jott will need to become as interested in the Semantic Web as companies like Google are today.
And as we’re experiencing with search engines like Google, this’ll take some effort and some time!
Pingback: Jott Meets the Semantic Web
Ha-ha. I think this belongs on the Misjott page (http://misjott.com).
What makes your case doubly ridiculous is that Jott’s “transcription engine,” from what I’ve been able to gather, is humans at a call center in India. Quality of the recording aside, you’d think they would be able to judge the “context” of words and their meanings far better than the latest and greatest speech recognition software.
Well, I did write: “What this means is that in the longer term, a few major releases of Jott down the road, Jott will need to become as interested in the Semantic Web as companies like Google are today.” 😉