PC able to talk and listen

hi there

I think it is time that the computer should be able to PERFECTLY understand the human through speech and respond back to him, like in movies (iron man even flubber :D).
I used the speech recognition software in vista but it still needs loads of improvement.

So i was wondering if any of you know about any software which helps in doing this, or atleast something that can make the PC have cool conversations with you and read texts with you (i know about narrator, very robotic).

Eagerly waiting for any discussions regarding this.

Compuwhiz93

Although great improvements have been made with speech recognition, home computers currently don't have the capacity to comprehend fully spoken language. For a start off, words can have a slightly different meaning depending on how it is pronounced, and the context in which it is used. In addition, the human brain can recognise the same word when pronounced in many different ways by people with differing accents and dialogue inflections. Unless the speech recognition system in a computer has already learnt the word, and the ways that it may be pronounced, it cannot understand a word if asked to recognise it through audio input. Even if it does know the word, it might not recognise it if it comes in a different accent and inflection.

As regards the computer outputting speech, it is currently very robotic because the number of voices available is very limited. Although the pitch and speed can be varied, this usually affects all of what is being output by the same amount, unlike normal human speech which can change in pitch and speed dynamically. Whilst certain punctuation can make the pitch of the word prior to it rise or fall, the end result is still very much robotic.

Unfortunately, home computers simply don't have the memory capacity or resources in order to be able to speak and to understand speech in the same way as a human brain can. Mainframe computers fair better because they have more memory and resources available, in addition to more processing power thanks to them supporting multiple CPUs (realistic speech comprehension and reproduction is CPU-intensive).

Yeah a pity thought, looks like we still have a long way to go until we get that technology.
Anyways thanks a lot for the reply.

It is funny that humans struggle doing things that computers find easy like answering calculations and that computers struggle at things that humans can do easily like speaking

Humans tend to work best with analog data, whereas a computer is primarily a digital entity. Certain things lead themselves better towards digital interpretation, and others towards analogue. Some are at home in both worlds.

Calculations and text recognition (picking letters and/or words from a given piece of text) is something that computers find easy to do because both can be programmed digitally.

Speech is one of those things that is, by its nature, an analogue quantity. To the human brain, this is of no consequence, but to a computer it is a completely alien concept. The nearest that we can get to this is listening to someone in a foreign language. In order to understand and process the information, there has to be some form of conversion so that the brain and/or computer can understand. With languages, we typically learn this whilst we are at school, although for a good many of us much of this is lost when we grow older and we end up primarily using our home language. A computer needs to have an analog signal translated into its digital equivalent in order to process it. Here lies the main limitation. Conversion is done by means of quantitization, whereby an analog signal is sampled and assigned an equivalent digital value. The rate of sampling, and the number of quantitization steps between an input of 0% and one of 100% is of prime importance. The higher both of these values is, the more accurate the processing of the signal becomes. Unfortunately, this is its limitation, and it is why I stated in my earlier post that realistic speech comprehension and reproduction is CPU-intensive.

In addition to what I have said above, you need to remember that certain words, although spelt differently, can sound exactly the same, and it is only the context in which they are used that differentiates the spelling that we use (if writing it down) and its meaning.

For example: Where (as in place), Wear (as in clothing, or eroding), and Ware (as in items) all sound exactly the same, and a computer would have a great deal of difficulty in deciphering which particular meaning that you wish to convey.

Nuance - Dragon NaturallySpeaking Speech Recognition Software
This is better than the Windows version.

Thanks, i'll give it a try and tell you.