Assisting in natural language abilities.

[thesystemera]

Hi all..   First and foremost.  Fantastic little app with a shite load of potential here.  I'm currently running the system through my iPhone 3GS (as I could not find a good array mic including the kinect) using a little app called pocketaudio.  

I'm currently working on my VoxCommando Treeeditor to introduce more fluent language abilities and am sitting here wondering why this is not introduced into the software more intuitively.  I'm finding myself repeating a lot of the xml code when I could be using strings or a natural language branch to apply to all sub branches...  

For instance, I call my system Brian as he is using the fantastic Ivona Text to Speech engine, he sounds like an English butler.  

I have got certain functions working fluently with almost every language variable in place and it is amazing, I just can't for the life of me be excited about applying these changes to all the commands.  Especially if lets say I wanted to change my computers name from Brian to Sam.  

Here is an example of a simple tree.  


Here the first five lines and the last line could apply to almost any command.    Yet if I wanted to change what could be a global features with lets say another greeting such as 'howdy' (which I would never actually do, I'm from New Zealand) you get my drift.  

Could a global tree be introduced for words for voxcommander to ignore with implemented rules of syntax obviously.  for instance 'hey, hi, hello' would always come before 'can, could, would' and this list would not apply directly after a command as to avoid confusion in dictation of titles for films such as 'analyse THAT'

I do stress the order of these lists is important as to not allow for complete jibberish of variables, a weighted priorty system would be the best where I could set more weight to some words that others but this is a whole new story.  I would love it if we could address natural language in to VoxCommander and then you vere well would have a true Siri killer...  Especially considering the nature of open-source add-ons..

Here is another exampe, this one works so well.. I've tried to think of variables out if it, but there are few....



Once again, the phrase 'please' could be removed three times...  So what do you guys think...

Thanks


 

[Kalle]

Here is an example of a simple tree...

Here the first five lines and the last line could apply to almost any command.    Yet if I wanted to change what could be a global features with lets say another greeting such as 'howdy' (which I would never actually do, I'm from New Zealand) you get my drift....

So what do you guys think...

Hi thesystemera, nice work. If you would perfect it a bit, you can set for each command a random answer.
In the command it self you can set as last a TTS.speak action and use as parameter "yes master | enjoy it | or what ever"
If you use the "vertical dash" between the text, VC speak it random.
You can use also a "logic block", but that is an other story  

Kalle

[thesystemera]

I have all ready addressed that, not so much with each command, but I generally find myself saying Thanks after Brian has done a command.  He does say 'playing genre jazz' and I then say Thanks and then he says 'your welcome'...  I have about five different custom responses. 

My biggest problem right now is my network, having trouble with both Headset and pocketaudio as an application to communicate via my iPhone and the skype method does not work for me... Am going to have to save up for a good array mic I think... 

[Kalle]

I have all ready addressed that, not so much with each command, but I generally find myself saying Thanks after Brian has done a command.  He does say 'playing genre jazz' and I then say Thanks and then he says 'your welcome'...  I have about five different custom responses. 

My biggest problem right now is my network, having trouble with both Headset and pocketaudio as an application to communicate via my iPhone and the skype method does not work for me... Am going to have to save up for a good array mic I think... 

If you have a chance to buy a amulet-remote www.amuletdevices.com that is the best solution for all. If you have a VC license you can save 40$ for the amulet-remote.

Kalle

[thesystemera]

Yeah nah... Brought the unlimited update version almost straight away after viewing this software.. First piece of software I have ever brought at that amount.  Poor audio engineer student and all LOL...

Will look at the amulet remote... PS: How do I get my account to gold status LOL...

[Kalle]

Yeah nah... Brought the unlimited update version almost straight away after viewing this software.. First piece of software I have ever brought at that amount.  Poor audio engineer student and all LOL...

Will look at the amulet remote... PS: How do I get my account to gold status LOL...

haha, gold status - you mean supporter - take a look here: http://voxcommando.com/forum/index.php?topic=734.msg5678#msg5678

Kalle

[jitterjames]

Hi all..   First and foremost.  Fantastic little app with a shite load of potential here.  I'm currently running the system through my iPhone 3GS (as I could not find a good array mic including the kinect) using a little app called pocketaudio. 

I'm currently working on my VoxCommando Treeeditor to introduce more fluent language abilities and am sitting here wondering why this is not introduced into the software more intuitively.  I'm finding myself repeating a lot of the xml code when I could be using strings or a natural language branch to apply to all sub branches... 

For instance, I call my system Brian as he is using the fantastic Ivona Text to Speech engine, he sounds like an English butler. 

I have got certain functions working fluently with almost every language variable in place and it is amazing, I just can't for the life of me be excited about applying these changes to all the commands.  Especially if lets say I wanted to change my computers name from Brian to Sam. 

I will give this some thought...

[thesystemera]

Thanks James.. If you want some assistance I would be happy to help.  I have some experience in project management as I'm currently working on another project, similar in ways...

Thanks again...

[thesystemera]

I noticed the "Payload from XML"...  I think using this rather than "Payload Dictation" would be better as then you only introduce a certain number of variables.  What I noticed though is "Payload from XML" only allows one response from all rows...  I had made an XML like this instead of manually entering it into every command.

Value 1: Hey, Hi, Hello
Value 2: Brian
Value 3: can, could, would
Value 4: you, we, I

How I imagine this should work is that the order in this list should be strictly respected so not to allow for to many nonsensical variables.  

For instance this would not allow for "you can Brian Hey" but would allow for "Hey can you" or "Brian could we" or just "can we"

Another aspect to consider is that the placement of this base language xml should not dictate command placement, as actual command placement could be any where amongst these optional variables.  However likelyhoods could be determined.  For instance you are likely to be prompting a command after the phrase "that, the, a"

the following example could cover almost all variables any request command, obviously not included the command it's self. Any variables that are not included could be then just added as you currently do using the 'optional phrase' command...

Worth noting, the system once pairing up two words or more could abolish them.  For instance once "Hey [Insert Name Here]" or "could we" has been confirmed the system has no need to return back to prior entries.  It's like following language branches.    

Hey, Hi, Hello
[Insert Name Here]
please
can, could, would
you, we, I, I'm, I am
want, wanting, need, needing
us to, you to, to
please
[Command Likely]
us, me
that, the, a
[Command Likely]
called, named
[Payload Likely]
please
[Insert Name Here]


I have noted that such words as 'Please' and the 'computers name' are harder to dictate placement as there are many more variables.  Maybe for this reason these two variables could be in an additional language file with the paramaters suggesting placement would be anywhere, wildcard optional.  

I hope this all makes sense LOL...