Limit Theory Forums

Hmm...if you analyse the language to get the word usage frequencies and apply binary node tree, that could work. You could, for example, use 2 bits to store the word "the", because it's used so often. Yeah, that might work. But you would still need at least 2 bits for often used words and more bits for less frequently used word. So, in the end it's still a little unrealistic that you have an average of 1-2 bits per word...

JanB1 wrote: ↑

Sun Jun 18, 2017 1:07 am

So, in the end it's still a little unrealistic that you have an average of 1-2 bits per word...

Thats why

Silverware wrote:it's 1.0-1.2 bits per letter.

Cornflakes_91 wrote: ↑

Sun Jun 18, 2017 1:18 am

JanB1 wrote: ↑

Sun Jun 18, 2017 1:07 am

So, in the end it's still a little unrealistic that you have an average of 1-2 bits per word...

Thats why

Silverware wrote:it's 1.0-1.2 bits per letter.

Yeah it was 1-2 bits per word I wrote originally, but that was from misremembering it.
Just change bits to bytes there and you get the same effect, and the same speed transmissions are still easy enough to send text over.

Also, letters like e would have a higher entropy = more bits, because they are used so frequently.
Q however, would make the next letter almost always a U and reduce entropy.

Except that things that appear often (like "e") have less informational content = bits because they dont convey much data.
If exactly every third letter were an e the information it contains would be 0, because its position is predictable.

rarer letters like q and z would contain more information because they're rare and thus "surprising", telling you something you couldnt have predicted.

Q would increase the effective entropy of any u directly following it, though.

I would expect that because after an E could be a much larger selection of numbers, it would result in increased entropy....

Yes, but increased entropy means the e contains less info, and thus needs less bits

Silverware wrote: ↑

Sun Jun 18, 2017 1:03 pm

Cornflakes_91 wrote: ↑

Sun Jun 18, 2017 1:18 am

JanB1 wrote: ↑

Sun Jun 18, 2017 1:07 am

So, in the end it's still a little unrealistic that you have an average of 1-2 bits per word...

Thats why

Silverware wrote:it's 1.0-1.2 bits per letter.

Yeah it was 1-2 bits per word I wrote originally, but that was from misremembering it.
Just change bits to bytes there and you get the same effect, and the same speed transmissions are still easy enough to send text over.

Also, letters like e would have a higher entropy = more bits, because they are used so frequently.
Q however, would make the next letter almost always a U and reduce entropy.

Aaah. Sorry Corn, got confused by Silver because, as he said, he wrote "word" before, not letter.