|View single post by Sparkz|
|Posted: Sep 24th, 2014 01:52 AM||
|Yes, they are spelled on the screen as you hear them. Words that are synonymous with common, physical things that one could use as a name will alter between capitalization or none. Like the difference between a spirit describing a "river" or naming themselves, "River". I think that it simply uses the text -> speech engine present in all Android O/S's to speak all textual input. I say this because it can regenerate words that are not in a common person's speaking lexicon. I've had it produce rather technical words, too - it will reproduce some things that sound made up, later to be clarified with further, singular word responses usually received very quickly back to back. I have had it string several words quickly into coherent if not poetic, short sentences as well. In the logs it produces, the words will be recorded to memory in fractions of a second between them. (It time/date stamps all received/decoded words for review, e-mail, texting)
So the software app seems flexible in that common parlance is immediately displayed, spelled correctly. Unique names too will come though, with capitalization. The average working dictionary of most text -> speech engines is only about 650 words or so, and as I stated before, very technical and sometimes things that sound made up come through yet with logical placement of consonants and vowels. The speech engine just muddles through it. At times, very complex concepts can be broken down, and words come back to back. As in say, describing a phenomenon:
"cord" as an example, can come in rapidly.
Most times it will find just one word to drop the idea of intent to describe a situation.
"How did you die?" could be asked. A reply I got once was "drowned". And that was that. No further communication. I guess they did not want to think or talk about it any more.
Other can get chatty and descriptive, downright imaginative too as they try to describe sensations, sights, colors, etc.
Perhaps in the marked instances where there was textual gibberish on other devices, it was because the spirit was just crunching up and using the sound snippets/allophones directly. I understand that items like the Ovilus use a speak stream/jet chip which is just a bare bones allophone sound clip set. A separate "dictionary" is stored in another discrete device. Having used such systems back in my 8-bit computing days, one had to use a "display" word, while coaxing the right pronunciation out of the allophone chip set with a hidden spelling. Some of the same techniques worked behind the scenes with those old Speak and Spell toys. A displayed text string would be say, "aspirin". But to get it to speak the word properly, a hidden allophone text string would be sent to the speech synth module spelled "asp-a-rin". If not sent in chunks like that, the module would produce speech that said, "AspAReen" and you're conjuring the speech patterns of Ralph Cramden. ;> *LOL*
So yes, it can vary capitalization for common nouns or proper.
It can and does seem to "hunt" as complex words are assembled/spelled and then spoken with clarity through Android's TTS engine.
Complex, multi-part concepts will be made up of rapidly sought and spoken words back to back and will be logged in memory as being within fractions of a second between one to the next word.
-Rich (Oh man, sorry for stepping on your forum name, Jeff!)