A spam filter targeting on Chinese spammers!

chinese_spam_filter.rb

class ChineseSpamFilter
    class << self
      def spam?(job)
        tencent_email?(job.email) &&
          (
            in_cjk?(job.title, 0.2) ||
            in_cjk?(job.company, 0.2)
          )
      end

      def tencent_email?(email)
        email.end_with?('@qq.com')
      end

      # Check whether text is in Chinese, Japanese or Korean
      def in_cjk?(text, threashold)
        chars = text.unpack('U*')

        cjk_char_count = chars.count { |char| cjk_char?(char) }

        puts "count: #{cjk_char_count}, ratio: #{(cjk_char_count.to_f / chars.length.to_f) * 100}"

        (cjk_char_count.to_f / chars.length.to_f) >= threashold
      end

      def cjk_char?(char) # Detect char in CJK char by check code block
        char.between?(0x4E00, 0x9FFF) || # Main blocks
          char.between?(0x3400, 0x4DBF) || # Extended Block A
          char.between?(0x20000, 0x2A6DF) || # Extended Block B
          char.between?(0x2A700, 0x2B73F) || # Extended Block C
          char.between?(0x2B740, 0x2B81F) || # Extended Block D
          char.between?(0x2B820, 0x2CEAF) || # Extended Block E
          char.between?(0x2CEB0, 0x2EBEF) || # Extended Block F
          char.between?(0xF900, 0xFAFF) # Compatibility Block
      end
    end
  end