Abstrakt
Language, as a method of communication, is a robust system that is able to overcome various noisy channels (in the Shannonian sense). The contemporary quantitative linguistic research takes into account the theory of information, understanding that various language features function as a redundancy, which is necessary to get over the noise.
Thus using an appropriate model of typical noisy channels is necessary to model both language and communication. This paper provides empirical testing of some existing models and proposes several new ones.
The paper consists of four parts: 1. The first part explores some well-known models (Shannon, 1948; Gilbert, 1960) of the noisy channel and introduces a few new ones (1st author 2015). 2.
The models are compared with some typical real noise which interferes with communication between people and their parameters are fitted to the empirical data that were obtained 25 during building a spoken corpus (2nd author 2013). 3. The ability of a recipient to filter out various kinds of noise is predicted according to the theory of information. 4.
The theoretical predictions are compared with perception tests.