Speech to text is based on sound patterns and word prediction. On 
Android, I get good results over all, but I find that personal (non-work 
related) dictation is the most error prone, probably because it's more 
unpredictable to the recognition engine. I started to notice it when I 
was dictating a list of poems into a note taking app. Even though I 
spoke clearly, the results were way off. Then I thought about the word 
sequences I was dictating and realized there was no way anyone, human or 
machine, could predict what I'd say next. On Android, recognition 
improves over time if you check an item that says something about 
personalizing your recognition and if you're consistent about correcting 
errors. I'm assuming iOS works in a similar way.


    VICUG-L is the Visually Impaired Computer User Group List.
Archived on the World Wide Web at
    http://listserv.icors.org/archives/vicug-l.html
    Signoff: [log in to unmask]
    Subscribe: [log in to unmask]