Первый круглый стол по практикам и стандартам судебного автороведческого анализа (обзор 3)

Abstract

The article provides an overview of the reports of the 1st Roundtable on Practices and Standards in Forensic Authorship Analysis held by the International Association of Forensic Linguistics and the Centre for Digital Humanities at the University of Manchester on May 15, 2019. It includes the reports by Erica Gold, Lecturer on forensic speech at the University of Huddersfield “Probability coefficient in the Forensic Speech Science: Current Situation” and Stefan Evert, Professor of the University of Erlangen-Nuremberg “Statistical Significance in the Literary Authorship Attribution”. At the end of each report, there are questions to the speaker and answers to them. The first (fifth) report examines a situation where a sample of the speech of a known speaker from among the suspects and samples of the speech of an unknown speaker are compared. Usually two methods are combined: auditory (perceptual) and acoustic phonetic analysis; automatic (computer) speech recognition. It is proposed to use a probability coefficient when attributing the text, which expresses the degree of correspondence between the speech samples under consideration. The disadvantages of computer speech analysis include the arbitrary choice of reference groups, the lack of statistical information, and the need to take into account different aspects of texts (stylistic, communication channel, recording quality, etc.). The next report is devoted to stylometry and is based on the analysis of works of art. Relevant parameters for such studies include: sentence length, word length, frequency class; vocabulary richness; syntactic complexity; spelling; choice of synonyms. Different methods of statistical calculations including the Burrows s delta are considered. The potential limitations of these methods are formulated (the influence of genre on style; applicability to small-volume texts; resistance to noise , for example, errors in automatic text recognition).