Text synthesizer

Text synthesizer converts thought into text though.

Interface:

Output: text.

Simple implementation of text synthesizer

Sort all active concepts by their “strength” attribute value (optionally by desirability attribute value).
Select first active concept from the sorted list of active concepts.
Find the most relevant text concept for the selected active concept.
Convert the found text concept into piece of text using word dictionary or phrase dictionary.
Add the piece of text into text result.
Select next active concept.
Repeat starting from step 3 until all active concepts will be processed.

Advanced implementation of text synthesizer

Advanced text synthesizer uses the same basic principles as simple text synthesizer. The following advanced features could improve quality of synthesized text.

Feature 1. Pre-sort active concepts by softcoded routines

Strong AI should try to create its own softcoded routines for sorting active concepts. Eventually experimenting will find out what softcoded routines are more efficient for text synthesizing.

Text synthesizing experiment scenario:

Try to “sort text” by a softcoded routine.
Get text output.
Convert text output back into collection of active concepts. Use text parser for that.
Compare original set of active concepts with the final collection of active concepts.
(Similar round-trip test may be applied as well: “initial text thought” -> “thought” -> “resulting text thought”. Ideally “initial text thought” and “resulting text thought” should be the same).
Evaluate text results by external expert (another intelligence system --- could be a human). Check if external expert understand the text results. Does the expert like the text results?
Hint: Internet chatters can be used as external experts. They are free and large in number.
Update ratings of softcoded routines based on self-evaluation (c) and expert evaluation (d).
Hint: Use Event correlation analyzer to evaluate experts’ feedback.

Feature 2. Order synthesized text to look naturally

After set of text concepts is prepared to final text rendering, the set is ordered in accordance with data from text pair.