Trouble with multiprocessing: Parallelized version slower than single-process version

Question

I\'m building a tool to search words in multiple files, dealing with big text files (>450MB).

The first version of this tool (full code here) included multiprocess