Huffman encoding is a technique for data compression where encoded patterns (often words for example) that occur frequently get the smallest representation (are compressed more than infrequent words).
In the context of Perl, things that are used a lot are highly compressed:
default variable: _
pattern match on default variable: //
read from standard input: <>
read everything (slurp) from standard input into default array: @_ = <>
and so on. As a result of this people often talk about Perl looking like "line noise".