Noisy Channel Model

• Language is generated and passed through a noisy channel.
• Resulting noisy data are received
• Goal is to recover the original data from the noisy data.
• Same model can be used in diverse areas of language processing e.g. spelling correction, morphological analysis, pronunciation modelling, machine translation.