Reinventing Entropy – Compression and Intelligence Part 1 (3b1B) [video]
youtube.comI've tried to make my own text compressor and here was the main idea.
One side is a table of all the most probable phrases up to 12, the other side is a unique identifier.
Then, then you simply string together the unique identifiers, take the bigint of the overall message, then utilize base62 encoding which shrinks the message even further since we are dealing with numbers.
It worked pretty well, some messages would compress very well, but on average, it compressed the original text by 50%.
For example, here is me trying to compress that message.
From 400>237 characters!