You cannot post messages because only members can post, and you are not currently a member.
Description:
Discussions about the Dragonfly speech recognition toolkit. Dragonfly is an open source Python package providing access to various speech recognition engines.
|
|
|
Ideas on how to map spoken name to actual, in code?
|
| |
A few ways to dictate a name in code: * Use voice commands as a virtual keyboard, spelling out individual chars and symbols. This can be time consuming for a just single name, especially when components of the name are not full words that DNS recognises. Not a preferred way to do it. * Say first character in IDE and rely on popups. I'm not sure this always... more »
|
|
|
Initial result: coding by voice 30% faster than by keyboard
|
| |
[Reposting here in case it's lost within earlier thread] In case of interest to anyone, I compared the speed of writing code by voice and keyboard. Both were used to create a copy of a selected benchmark module from the Dragonfly command modules repository: Command-module for moving and controlling **windows**... more »
|
|
|
Extension to Dragonfly: Modes
|
| |
This extension, a very early implementation of which it must be said, adds support to Dragonfly for "modes", which are effectively grammars, but they support the idea of sub-grammars in some way, and to do so modes are detached from DNS grammars. Being detached from DNS grammars, in order to receive recognition... more »
|
|
|
How to turn off voice navigation in apps? (..so Dragonfly grammar can work)
|
| |
When I'm dictating commands into (code) editors for processing in a Dragonfly grammar, the editor appears to intercept certain words spoken (tree items or menu commands) and uses them for navigation within the app. Is there any way to turn this off? Other than dictation mode which then stops word flow to Dragonfly grammar.... more »
|
|
|
Number element slow to decode (5 seconds to recognize number 120)
|
| |
It takes 4.9 seconds on my computer to decode "one hundred and twenty". Code used is included below. The Number element is a great one to have, allowing natural number dictation, but this kind of delay would be noticeable during dictation. My impression is that it may be due to a large parse tree being generated... more »
|
|
|
Good result dictating code identifiers with Dragonfly
|
| |
Just to report on a small but good result obtained using Dragonfly for the abovementioned task, in case of interest to anyone using or attempting to use Dragonfly for coding. It has been possible to dictate the following identifer using 9 words and with any spoken timing of those words (single or multiple phrases):... more »
|
|
|
"reject", "other" in results_callback(words)?
|
| |
Upon words being spoken, in the results_callback method in GrammarWrapper, the spoken words are usually received. But there are two tests on these words, for matching to 1) "other" and 2) "reject" (plain strings, not tuples). But this never seems to occur. DNS never seems to send either of those strings (in DNS v10 and 11). Including when "scratch that" is said.... more »
|
|
|
Code to visualise Dragonfly parse trees in GUI
|
| |
In the interests of making it easier for people to understand the workings of Dragonfly, the following code can be used to open a window showing either node or element trees in a collapsible Tree control. It requires TKinterTreectrl at [link]. Disclaimer: the code is as rough as guts.... more »
|
|
|
Dictation in Dragonfly cannot be "played back" in DNS
|
| |
I just discovered the above, that speech matched to a Dictation element cannot be played back in DNS. This is the case both in V10 and 11 of DNS. This could be of slight drawback when it comes to error correction, as it's not possible to hear what was said to know whether or not pronunciation was... more »
|
|
|
Grammar for writing source code?
|
| |
Hi folks, If there are any code bits floating around that might assist with dictating source code, would appreciate any pointers. Thanks Mark
|
|
|
