Remind yourself daily, that every time you start the day, you're starting anew...

 

theyellowbrickroad:

psyducked:

theyellowbrickroad:

imagine being in a business meeting 20 years from now and your boss is pulling up posts you made on tumblr as a teenager and having a discussion about them with you and your colleagues

image

im ddeleting the internet

matthen:

Before producing a transcription of what you said, speech recognition software often represents its results as a ‘confusion network’. An example is shown above, where differently weighted words are listed at each stage.  The most likely transcription is the one with the thickest lines, ‘Hi my name is Matt’, but the other options are still possible.  The animation shows the top 5 transcriptions being selected.  In Dialogue Systems (which is where I am doing research), the computer system must try to extract meaning from the speech recognition output.  A traditional approach is to pick the top few transcriptions, and attempt to decode each of them in turn, before combining the results.  I have been working on a method which considers individual arches and paths in the confusion network, and directly decodes from this much richer source of information. This seems to work very well! [more] [code]

This is awesome, I’ve been wondering lately why speech to text isn’t more developed and looking at this, might give some inspiration for others because I don’t fully understand what I could do …

matthen:

Before producing a transcription of what you said, speech recognition software often represents its results as a ‘confusion network’. An example is shown above, where differently weighted words are listed at each stage.  The most likely transcription is the one with the thickest lines, ‘Hi my name is Matt’, but the other options are still possible.  The animation shows the top 5 transcriptions being selected.  In Dialogue Systems (which is where I am doing research), the computer system must try to extract meaning from the speech recognition output.  A traditional approach is to pick the top few transcriptions, and attempt to decode each of them in turn, before combining the results.  I have been working on a method which considers individual arches and paths in the confusion network, and directly decodes from this much richer source of information. This seems to work very well! [more] [code]

This is awesome, I’ve been wondering lately why speech to text isn’t more developed and looking at this, might give some inspiration for others because I don’t fully understand what I could do …

ummok:

Floating public swimming pool on the Hudson River, New York.
Most interestingly, the funds for this project were raised via a crowdfunding campaign on KickStarter, which took just six days to raise the $41,000 needed for the construction.
It feels like 2012 could be a blue ribbon year for crowdfunding and also for socially-organised urban developments. Let’s hope we see more of these projects as the year unfolds.

I never paid attention to the amazing implications the Kickstarter idea might have… public places that actually matter to people is an incredible idea! It will never remove taxes…or could it?

ummok:

Floating public swimming pool on the Hudson River, New York.

Most interestingly, the funds for this project were raised via a crowdfunding campaign on KickStarter, which took just six days to raise the $41,000 needed for the construction.

It feels like 2012 could be a blue ribbon year for crowdfunding and also for socially-organised urban developments. Let’s hope we see more of these projects as the year unfolds.

I never paid attention to the amazing implications the Kickstarter idea might have… public places that actually matter to people is an incredible idea! It will never remove taxes…or could it?

(Source: grabyourpack)