Computing and Informatics

The paper focuses on the process of selecting representative sample documents written in a natural language that can be used as the basis for automatic selection or classification of textual documents. A method of selecting the examples from a larger set of candidate examples, called automatic biased sample selection, is compared to random and manual selection. The methods are evaluated by experiments carried out with real world data consisting of customer reviews, with different document representations and similarity measures. The presented approach, that provided satisfactory results, faces problems related to processing user created content and huge computational complexity and can be used as an alternative to manual selection and evaluation of textual samples.