Many attorneys like to dictate letters or messages using their Android device. I use voice to text quite often to compose text messages, and when combined with Google Now’s voice commands, your phone becomes a handy personal assistant. Of course, I’ve talked about making your own dictation system with Google Drive and your Android Device. But Christian Williams wonders the following:
I wonder what punctuation can be used, if there is a command for new line or paragraph breaks.
Google is really improving Android dictation capabilities, and while you’re not going to experience Dragon-like dictation, I think you’ll be pleasantly surprised with the results.
STT types what you tell it to
Android’s speech to text engine is quite literal. That means, when you say, “new paragraph,” the STT engine translates and types, “new paragraph.”
This “stupidity” happens for essentially two reasons:
- STT isn’t really designed for dictation (at this time); and
- Google’s interested in conversational context for its STT engine.
In the grand world of speech to text, dictation — at least as attorney’s think of dictation — takes a very distant place. The reality is that most Android consumers will never use their Android devices with the intent of dictating long, legalese-filled, memos and diatribes. The true purpose, at least in the STT world according to Google, is to translate and send short bursts of text (think text messages or Twitter updates — 140 characters), rather than lengthier discourses.
When STT accomplishes that goal, 95% or more of users are happy.
The conversation is the important factor
Those happy users, sending short bursts of messages, is what Google’s trying to capture. Google’s search engine is all about conversational elements, as opposed to choppy keyword searches.
If you remember conversational, it’s easier to get engaged with STT.
What commands work?
Ultimately, figuring out what commands work, and when, will drive you bat crazy. Here’s a list of usable commands to make STT work a little better.
Here are some of the commands that are guaranteed to work:
- Period = period (.)
- Comma = comma (,)
- Question mark = question mark (?)
- Exclamation or exclamation point = exclamation point (!)
- Apostrophe = apostrophe (‘)
- Enter or new line = moves to a new line
- New paragraph = new paragraph
- Tab key = tab
- Colon = colon (:)
- Dash = dash (-)
- Ellipsis or dot dot dot = ellipsis (…)
- Ampersand = ampersand (&)
- Asterisk = asterisk (*)
- At sign = at sign (@)
- Backslash = backslash (\)
- Forward slash = forward slash (/)
- Open bracket = open bracket ([)
- Close bracket = closed bracket (])
- Open parenthesis = open parenthesis (()
- Close parenthesis = close parenthesis ())
I’m noticing there are a lot of comments regarding the performance of periods, spaces, and parenthesis.
Admittedly, there are some problems with the text recognition and performance. But remember that Google’s STT program is not a true dictation program. That means you aren’t going to be able to train STT to perform complex punctuation commands. First and foremost, as I said above, STT is conversational versus true dictation. Thus, you eliminate complex function commands for less complex commands.
For example, Jack Calhoun, wants to be able to add three periods with spaces between them — . . . — to his dictated sentences. This is a useful texting or commenting technique, especially when you’re trying to show a pause in time in thoughts or conversation. Unfortunately, that type of editing is beyond Google’s capability right now. When you try, you’re going to see something like, “. space. Space.”, as opposed to the “. . .” he’s wanting. Again, it’s a limitation of the program, and partially beyond the design purpose.
Now, as far as the issue with creating open or closed quotation marks, this is just plain stupid. Google should be able to add quotes.
Making it all work together
I’ve found that the best way to get dictation working is to have a continual conversation with your device. This means that if you pause for long periods of time, usually 2 or more seconds, STT won’t connect your command as a command. This is especially true for commands like “new paragraph” and “new line” or “enter.”
I find that if you pause only slightly, STT will spell out “enter,” rather than execute the command. It’s a really finicky function, but if you do it right dictation works well.
I also find that the accuracy of the dictation depends a lot on your location. Obviously, quiet areas in your office or home will make the speech engine more accurate. If you’re in your car or another busy location, I find the engine isn’t great, but speech to text still works relatively well.
But it’s not Dragon Naturally Speaking
If you’re dead-set on having the Dragon experience, Android’s STT will disappoint. But, if you recognize Android’s dictation for what it is, then you’ll probably be satisfied with what you’re getting. For me, Google’s dictation engine works fine when I’m sending short texts or dictating small briefs. It also helps to proofread your dictation before sending it off, since grammar and spelling errors definitely lower professionalism.
If you know of more commands that work, let me know so we can help everyone.
This post was originally published November 18, 2013.