A language model
is not simply a list of words. A language model is a statistical database
of information analyzed for unique patterns used within a given specialty.
Since speech recognition is a statistics program by nature, the better
the language model, the better the accuracy.
A good analogy
would be a standard speech recognition program would be the equivalent
of an "unskilled" black jack player in a Las Vegas casino. A speech
program with a language model would be the equivalent of a "card counting"
black jack player. Like the skilled "card counter" the odds are in their
favor. The person using NaturallySpeaking with Smart
Vocabularies language model has the "odds" weighted toward
their specialty and thus will receive much better speech recognition.
A language model
is constructed by analyzing thousands of reports within the specialty.
EACH Voice Automated language model was created with a minimum of 15,000
reports and 12-15 million words of SPECIFIC dictation within the field.
Smart
Vocabularies uses proprietary software tools to develop the
language model independent of the speech engine through applied computational
linguistics processes. From this analysis, Smart
Vocabularies builds an Language Model statistically weighted
towards the field based on the "big ram" analysis used within the Dragon
NaturallySpeaking program.
Sounds
good, but what does it really mean to an end user?
By analyzing 15,000
reports and 12-15 million words, Smart Vocabularies
assures you that the specialty is adequately covered for 97%+ of the
words used within that field. This means less time for the end user
to train words OR teach the system CONTEXT or their specialty. 8,000-
12,000 words are added with correct orthographic (a typed version of
how the word sounds to the speech program, i.e. Tomato, would sound
like, Toe-may-toe, etc.) meaning the user does NOT need to train any
of these words. Words are also added within the correct CONTEXT of the
specialty with hundreds of examples of how the word is used within the
specialty.