If I understand your post correctly you intend to perform lexical analysis of the logs with aim of detecting paedophiles etc. With all due respect: you are wasting your time. No amount of positive identification of phrases can or will positively identify a person as a child molester. Writing an AI (artificial intelligence) script that could do that would be something I would put in the “impossible” basket.

1. Anyone with logs positively identifying a paedophile should contact police not you
2. If you are asking for logs it means you have no practical knowledge on the subject – psychoanalytical profiles of such people, their “grooming” techniques etc.
3. You assume each paedophile is the same (uses the same language mannerism)
4. how do you propose the script will tell apart a paedophile from people who discuss paedophilia and may use the same phrases
5. Your script can only see public conversations – I doubt very much an offender would commit his crime in an open channel
6. If it is for your own “protection” than your brain should send alarm signals long before any script would and if it doesn’t neither would your script as your AI programming would only follow your own logic
7. mIRC script is not really suitable as an AI language

Do I make it harder than it is? Perhaps I am. Write a script that will detect people who like apples and yet they never say “I like apples”. If you can do that, ignore my rumblings.