Peter Zörnig

Unified representation of Zipf distributions

Abstract Certain discrete probability distributions, used independently from each other in linguistics and other sciences, can be considered as special cases of the distribution based on the Lerch zeta function. We will list the probability functions for some of the most important cases. Moments and estimators are derived for the general Lerch distribution.

Classification of Serbian texts based on lexical characteristics and multivariate statistical analysis

Abstract We study lexical properties of different Serbian text types (poems, rock songs, different kinds of spoken language, prose, scientific and journalistic texts). We investigate characteristics and text parameters based on the frequency of word forms (relative frequencies, repeat rate, h-point and related indices). Relations between parameters are studied in order to identify the principal characteristics. We also apply techniques of multivariate analysis (cluster analysis, MDS) to classify the text types adequately. One additional objective of the present paper is to illustrate the functioning of these techniques in detail by explicit calculations of all programming steps.

A Sequential Activity Measure for Texts and Speeches

Abstract We introduce a new indicator for activity in written or spoken language, defined by the number of verbs appearing before the adjectives. The concept is applied to Italian presidential speeches.

On Consensus Strings in Text Analysis

Abstract The present paper is a pilot study introducing a concept of computational biology – the closest string problem (CSP) – into linguistics. In quantitative studies, texts can be represented as sequences, and for any given set of sequences one can solve a CSP. The solution is called a consensus string (CS), which is a sequence that is „as close as possible to all the given strings“. In this article we solve some CSP’s resulting from the study of a Hungarian poem and illustrate possible interpretations.