Computational linguistics studies the computational processes involved in human language comprehension, production and acquisition. (We hypothesize that these processes have an important computational aspect because language involves meaning-bearing symbolic representations). The field has both a scientific and an applied side. The scientific side seeks to understand what kinds of computational processes can in fact comprehend, produce or learn a human language. Here computational linguistics builds on the mathematical theory of grammars and parsing (to understand what kinds of devices could possibly solve these problems at all), as well as results from psycholinguistic and language acquisition research (since humans are the only kind of entity around today that truly understand language).
The applied side of the field builds on this scientific understanding to enable computers to do useful things involving language, such as translating a text from one language into another, producing answers to questions, and summarizing a text. The recent growth in computational linguistics is largely because of its commercial and intelligence applications in text processing and information extraction.
Over the past decade and a half, computational linguistics has been revolutionized by the use of statistical techniques. Comprehension, production and acquisition can all be viewed abstractly as specialized inference problems. For example, language comprehension can be viewed as the task of inferring the meaning of a sentence from its words or sounds, given knowledge of the language from which it came. Employing statistical inference to solve this task amounts to using distributional and probabilistic cues in this inference.
This class surveys computational models of phonology, morphology, syntax, semantics and pragmatics. While it focuses primarily on the scientific side of computational linguistics, applications of computational linguistics will be covered and discussed in class.
For more information, email the instructor Mark Johnson.