Dragonfly - a Python library which extends NatLink

Hello Speech Computing readers,

I would like to introduce Dragonfly, a Python library which
extends NatLink. It provides a language object model which
takes care of all the nitty-gritty details required to
interface with NatLink and Dragon NaturallySpeaking. Dragonfly
allows Python programmers to use high-level pythonic objects
to specify grammars and rules, and to process recognition
results.

Dragonfly is hosted here: http://code.google.com/p/dragonfly/

Features of the Dragonfly library:

• High-level language object model; offers dynamic and
  flexible control of command grammars.
• Integrated recognition parsing; gives easy and language-
  independent access to recognized data.
• Extensible activity contexts; allows precise control of
  when grammars and rules are active.
• Built-in action framework; simplifies the modular
  definition of actions, such as keyboard events, to be
  executed when a command is recognized.

Instead of telling you how Dragonfly works, I'd like you to
see for yourself. Below I've included a code snippet from an
example Python module demonstrating Dragonfly usage. Anybody
who's ever written a NatLink macro will see how Dragonfly can
make life easier.

spec   = "(I ate <food> <time> | <time> I ate <food>) [and thought it was <opinion>]"
time   = {
          "(two days ago | day before yesterday)":  2,
          "yesterday":                              1,
          "today":                                  0,
         }
food   = {
          "(a Granny Smith | an) apple":  "fruit",
          "an orange":                    "fruit",
          "a hamburger":                  "meat",
          "a [juicy] steak":              "meat",
         }
extras = [
          Choice("time", time),
          Choice("food", food),
          Dictation("opinion"),
         ]

def _process_recognition(self, node, extras):
    days_ago  = extras["time"]
    foodgroup = extras["food"]
    day_word = (days_ago == 1 and "day" or "days")

    print "You ate %s %d %s ago." % (foodgroup, days_ago, day_word)
    if "opinion" in extras:
        print "You thought it was %s." % (extras["opinion"])

I say:          "I ate an orange yesterday"
Module prints:  "You ate fruit 1 day ago."

I say:          "day before yesterday I ate a hamburger"
Module prints:  "You ate meat 2 days ago."

I say:          "I ate an apple today and thought ..."
                "... it was Cap delicious exclamation-mark"
Module prints:  "You ate fruit 0 days ago."
Module prints:  "You thought it was Delicious!."

Complete code for this module is available here:
(browsable repository)

Q: who is the target audience of Dragonfly?

A: people interested in speech recognition and Python.
Users of Dragonfly generally have at least a little
experience with programming in Python. Although, you
don't have to be very skilled at all to use it,
because the library is fairly intuitive and easy-to-use.

Q: why not just use NatLink's GrammarBase class?

A: because that requires a lot of unnecessary work. It
involves defining a grammar in the form of one big static
string, and then manually extracting relevant
information from the raw sequence of literal words
recognized.

A significant disadvantage of using NatLink directly is
that it makes it very difficult to write multi-language
command grammars, because literal words are used both in
the grammar definition and in the recognition parsing
logic.

Q: why not just use an existing framework, such as vocola?

A: because all existing frameworks I could get my hands on
use their own scripting language. This simplifies
defining new commands, but limits their power.

Dragonfly is a Python library. Developers can therefore
do everything that Python can do when a command is
recognized: keyboard & mouse events, Win32 API calls, COM
interfacing, and everything else the developer's
imagination can come up with.

I'm putting Dragonfly out into the world in the hopes that
other people will benefit from it. I use it every day. It
has enriched my speech-recognition experience. It allows
me to rapidly develop new application modules to control
all kinds of software. I believe that other people, if
they take the time to look at Dragonfly, will also see its
potential and will be able to increase their productivity
by using it.

As I pointed out above, the Dragonfly project is
hosted here: http://code.google.com/p/dragonfly/

Have fun and let me know what you think of Dragonfly,
-- Christo.