The Brainy Read for the Business Strategists: Voice Activation

The Brainy Read: Voice Activation

Click on this to toggle to the Heart of the Matter

pillar.gif (3924 bytes)

Not needing dictionary definition, but word order, to define semantic meaning
Having universal syntactic characteristics for adapting to all language
Differentiating among the multiple meanings of words having more than one meaning
Changing word-identification to fixed word-order, not semantic attributes
Programming English in binary code
Formalizing language with mathematical rules
Making natural-language computer commands possible
Tying automatically to voice-synthesis technology
Providing algorithmic structure emanating from natural language structure and not a forced construct
Serving as the "math" of language
Accessing infinite mobility within finite set of context formulae

theExactWord^Ž seeks partners in this venture to integrate its baseline development with state-of-the-art technologies in voice recognition and activation fields.

Integration Goal: Recognize Each Word in a Sentence in Natural-language Context

A frustrating weakness inherent in computer technology has increased as data-bases and voice technology have gained sophistication: users need the ability to spontaneously interact with computers beyond scripts or limited vocabularies which force prompted answers. Context, in other words, needs to have equal importance with word definitions. theExact Word^Ž has isolated such context patterns in English, making voice-recognition programs more like natural speaking.

Historical Problems

theExact Word's^Ž technology has solved some thorny issues which affect voice products. The following examples illustrate those apparently unresolvable issues.

Examples:

Homonyms: I would like two eggs with coffee.

Some current voice input products for "speaking to a screen" rely on a pop-up window for some words prompting the user to decide among the options. An easy example illustrates:

two

too

theExact Word's^Ž technology solves that issue by examining the sentence's context without needing word definition. Each one of the forms above can only occur in exclusive positions in any sentence. It is almost impossible for them to occupy the same position in a sentence as each other; such a case would be a possible but unlikely occasion. theExact Word's^Ž program knows which is which.

Another type of example:

A cute girl drew an acute angle.

"A cute" and "acute" sound alike in rapid speech. Again, their position in context gives them "exclusive rights" to their own position in the sentence for their distinct meanings. It is impossible for them to occupy the same sentence position while having multiple meanings in a given position.

Further:

We will wind the clock.

The wind blows.

Clock my time as I run.

theExact Word's^Ž prototype identifies the differences in multiple meanings for identical forms. Without the context placement, the user cannot know which meaning of "wind" or "clock" applies. In the case of "wind," the pronunciation changes as the context changes; in the case of "clock," however, the meaning change does not indicate a change in pronunciation to reflect the change in meaning. Context alone solves these problems.

The three types of examples demonstrate context problems unique to English and very troublesome problems for computational linguistics. Meaning results from placement and cannot be ignored or "worked around" for natural-language applications.

For Practical Application

In contrast, theExact Word's^Ž technology has identified the trigger mechanisms which allow information to re-sort and respond by context, driven by the user. This technology gives the user a natural flow in speech, thereby eliminating scripted answers altogether in favor of natural-language patterns, or augmenting and integrating with existing technologies. For example, solving the problems that these examples represent could make the keyboard obsolete where voice activation and recognition commands would better serve. Or, databases do not need to restrict users to "allowable text" word choices.

For the Business Strategist

theExact Word's^Ž natural-language product provides an automatic baseline context and more natural conversation between a speaker and a machine.

Back to Voice Recognition Intro