next previous contents
Next: Evaluation of Translators' Aids Up: Methods Previous: Composition and structuring of

Instantiations of the interpretational scheme

For instantiating interpretational schemes, collections of spontaneous and non-proofread natural language texts are a prerequisite. Note that this process is another way of looking at how our writer model is to be constructed.

To give a rough idea, the following collections are required:

  1. A collection of texts from native speakers, possibly subdivided by:
  2. A collection of texts from non-native speakers, possibly subdivided by:
For a broad and general purpose instantiation of 1 to be applied to a grammar checker for native speakers, we would need a balanced corpus of texts to generate the rates for the frequency of phenomena. This is to be used for setting the weights and constructing the rating scales of the interpretational scheme for the measures. Along the same lines, we would get a similar instantiation for 2, but for non-native speaker grammar checkers for a language.

There is another function of such text collections; in the process of constructing the test suites they feed into these by providing examples and tokens of erroneous inputs to different categories. In fact, they may lead to definition of categories of error types.