Flow chart - Algorithm - Speech Recognition

i don't know the algorithm to solve Speech Recognition + AI conversion + Speech Engine English problems.
i think the following chart is the one, maybe.

voice recognition -----> convert to text -------> AI chat box ( such as Eliza algorithm) ----> Reply text --> read reply text , and we have voice output.

Right? But, with this chart, time to process can be a problem, because there're many steps.

Hi

I don't know how other people does it, but this is how I did for the BRAINS software.

Voice input > convert to text > search for keyboard > AI Chat from database > out text/speech.

Dear Tomato, Yes the flow chart is correct for most of the sound enabled chatbots out there. The best speech recognition engine is Nuance's "Dragon Naturally speaking" ver 10. or IBM's "Via Voice" product. With the recomended sound card and most any PC less than two years old, the processing lag is only 2 - 3 seconds at worst. The chatbot AI adds another 3 seconds or so. Text to speech does not add much at all to the overall lag. A good PC with 4gigs ram and 2.5+ gigaherts will do it all in less than 4 seconds. Time to process used to be a big problems for real time chat agents 5 years ago. Modern processors have eliminated the problem for STT and TTS conversions. The AI itself can still be a problem depending on how 'deeply it thinks'.
check my other post for you.

Hi Captquazar,

Have you ever tried Dragon Naturally speaking" ver 10.....
How well it recognize your voice without training...

Yes. Dragon Ver. 10. is dramatically better than version 8. You can get over 90% accuracy with NO training at all. Speed is excellent also. Conservatively 2 or 3 seconds with the recommended sound card. Of course training helps accuracy and speed. Check the independent reviews the past 4 months. Of note Apple Inc. integrated Dragon ver9 into the Mac OS 10 versions for native speech support. Nuance also provides many voices for developers for the TTS. Japanese too. Check the developer's forum.

I've used Dragon Naturally speaking version 10 a good bit, and it did require a lot of training for it to recognize my voice correctly. It's the best speech recognition software on the market by far, but it's not perfect. It still gives some errors even after training it.

I noticed if I talk casually it has a harder time accurately recognizing what I'm trying to say. Although, if I speak properly like somebody reading the news, it has a much easier time understanding me and is more accurate that way.

The only thing I find impressive about aiko is the B.R.A.I.N.S software the way it knits all those process dynamically together.
On the other side knitting software together to perform some function,
issss not likely the goal I think I would rather design it myself with help of good open source tools if needed,
That way improvements can take a radical new approach.
The other complex thing that come from that is the systems footprint which can be reduced.
Using software like ''DRAGON'' would only increase system resource you would use something that really is a closed system on a system.
At least thats what I think. The way I see it if aiko where to be build like that and one of these software application would crash she be screaming ''Blue Screen of Death , ERROR ERRORRORorororrrrrrr''. A.i is not about assembling functionality from different software tools together those functionality should be learned by the creativity of the versatile neural network and learning programs and within that hides the clue to real A.i when its level of maturity has improved it can get rid of those software functions or falls back on them only if needed. If a.i where that simple company's with money would already cannibal's that market. Project like aiko could reap that benefit because it starts with an idea from someone's garage and could improve overtime. (my god I got all serious again 'haaa )

I'm sure that some of us on the forum find his own interesting way of building his or hers own system.
And that make me curious why is there no post about your own progress using the chatterbots as an
experiment. Why is the monkey not out yet.

Dear WulfCry. An AI like you describe does not exist yet. (except wetware) An AI that can cope with a wide variety of problems has confounded all research attempts to come up with an all purpose algorithm. Even neural nets have to be trained to solve specific problems with a limited domain. However lots of AI tech exists that solve specific problems in limited (that is tightly defined) domains. Grouping these AI agents together is how the DARPA off road challenge was solved. Likewise Aiko will of necessity need several AI processes running simutaneously in real time to obtain 'situational awareness'. For example vision processing and body motion / balance for eye hand coordination and walking obstacle avoidance. That is why I favor a multi processor design; so you can write the software modules without worrying about process interference and lag time issues.

I guess the logic of my question generated a different kind of outcome in answer.
Again it was not about an all purpouse a.i but the little steps that some on this forum
where experimenting with to emulate what aiko can do like connecting an chatterbot with
speech software.

And about what you said about "A.i" is not completely true its possible to build an automonous
computer that reacts through the software tools connected as an A.i assembly. I myself am aware
of the history of A.i . And blog about the possible steps and build to form a versatile system making
that possible. Little steps and realistic view of what can be done is key as well the knowlegde to
prevent the same mistake or view world wide around A.i. ( Even I where a pessimisti and reluctant to the idea
But with an openmind you can do a lot)

Do not forget even A.i can be build in a garage. Also people forget the history why a computer should
showcase some thinking. I'm talking about Von neuman , Marvin minsky , J.C.R licklider , Alan kay its a long list
of people that where involved with DARPA now an administrative agency. Even the history of darpa and how
it where setup brings up the quest of computer that could perform autonomous. (notice that i let out ''think'').

My project is called wulfcry IDORU and blog about it for almost two years and have trouble establishing my claim.
It will take time to adapt to reading my writing style and personality mix into it thats why I say ''Keep an open mind''.
But I am confident with hard work and getting to the job the possibility for an autonomous machine based on the
recent development that broaden the A.i field a little more now to quote Barrack Obama "Yes we can " it is possible.

Last but not least we can use the new features in cpu , and software models such as parallel computing , functionel programming
web services , cloud computing and more to build a model A.i with this all and especially for the clever minds out there its an
endavour.

That is why I like the AIKO project M.r Le Trung took the bull by the horns and build his dream , vision and realised it.

So my question again for those who where posting there mix of the Eliza chatbot with speech engine what did you do and
does it work.

I can make a list of books about the subject which I read the past years and recently if some of you want to know whats in the event horizon.