Exploiting Dialect Identification in Automatic Dialectal Text Normalization