Requesting several message filtering methods in learn.

[Artoria2e5]

• [ ] Believe it or not, you should remove the length checking. Jieba is fast enough for tg's max msg length.
• [ ] Replace all english punctations in zh context with regexes.
• [x] Consider ignoring input with < 60% typical Chinese stuffs.（[“\u4E00-\u9FFF，\u3400-\u4DBF？\u20000-\u2A6DF！\uF900-\uFAFF…\u2F800-\u2FA1F”]）。
• [ ] Normalize the input to a specific locale varient of zh, with things like OpenCC.
  • Using the big dict file in jieba may also help with zh_Hant but it is less useful.
• [x] Delete the URLs.
  • [x] URLs are not only http(s). Use the RFC regex.
• [x] Delete common IRC forwarding headers.

Finally again, consider adding /learn MSG TIMES command.

[Artoria2e5]

@PeterCxy URLs still have some way to go.

[PeterCxy]

URLs will be exluded because of too much unrecognized characters.