clean company names

[blanche]

cleaning company names by 1) removing special chars / lower casing / stripping and so on 2) removing high scoring idf terms using deterministic method such as 'ltd', 'group', 'international'

this reduces the number of unique companies from 18k to ~12k

an open issue in this topic is that some companies names should be merged e.g.

• 24 7
• 24 seven
• 247

but doing that automatically while preserving companies that should NOT be merged is kind of hard and does more harm than good e.g.

• red
• red rockhip
• red sky
• red snapper
• red teachers

[utkuozbulak]

Cool