Fixing the Infix: Unsupervised Discovery of Root-and-Pattern Morphology

We present an unsupervised and languageagnostic method for learning root-andpattern morphology in Semitic languages. We harness the syntactico-semantic information in distributed word representations to solve the long standing problem of root-and-pattern discovery in Semitic languages. The root-and-pattern morphological rules we learn in an unsupervised manner are validated by native speakers in Arabic, Hebrew, and Amharic. Further, the rules are used to construct an unsupervised root extractor called JZR, which we show to compare favorably with ISRI, a widely used and carefully engineered root extractor for Arabic.