2012 | OriginalPaper | Buchkapitel
Unsupervised Grammar Inference Using the Minimum Description Length Principle
verfasst von : Upendra Sapkota, Barrett R. Bryant, Alan Sprague
Erschienen in: Machine Learning and Data Mining in Pattern Recognition
Verlag: Springer Berlin Heidelberg
Aktivieren Sie unsere intelligente Suche, um passende Fachinhalte oder Patente zu finden.
Wählen Sie Textabschnitte aus um mit Künstlicher Intelligenz passenden Patente zu finden. powered by
Markieren Sie Textabschnitte, um KI-gestützt weitere passende Inhalte zu finden. powered by
Context Free Grammars (CFGs) are widely used in programming language descriptions, natural language processing, compilers, and other areas of software engineering where there is a need for describing the syntactic structures of programs. Grammar inference (GI) is the induction of CFGs from sample programs and is a challenging problem. We describe an unsupervised GI approach which uses simplicity as the criterion for directing the inference process and beam search for moving from a complex to a simpler grammar. We use several operators to modify a grammar and use the Minimum Description Length (MDL) Principle to favor simple and compact grammars. The effectiveness of this approach is shown by a case study of a domain specific language. The experimental results show that an accurate grammar can be inferred in a reasonable amount of time.