Peter Norvig's essay critiques Noam Chomsky's dismissal of statistical methods in linguistics, defending them with historical evidence and practical successes in computational linguistics. Norvig argues that probabilistic models, trained on data, outperform Chomsky's innate grammar theory in explaining and processing language.
Norvig's essay responds to Chomsky's criticisms by demonstrating the effectiveness of statistical models in language tasks like speech recognition and translation. He refutes specific points, such as the 'colorless green ideas' example, by showing statistical models can handle grammaticality probabilistically. The piece highlights the empirical success of data-driven approaches over rigid innate structures, concluding that statistical methods better capture language's complexity.
Comments
Please log in or register to join the discussion