engleski

Annonymous Letters Examination: Where Forensic Linguistics Meets Handwriting Examination

After attending this presentation, attendees will have a better understanding of the contents and/or the way threatening letters are formulated while sent to the ruling party or an individual in power by their opponent. This presentation will impact the forensic science community by introducing a unique study in which written letters were analyzed regarding the scope of the threatening letter, the contents, sentence complexity and repetition of certain words and formulations. The aim of this paper is to examine the possibilities of profiling the author of the threatening letters by analyzing the utilized linguistic formulations and to cluster groups of anonymous texts under some potential common identity. The assumption is that there is mostly one group of people who send politically driven threatening letters. Special attention should be given to the letters that are in some way “signed”, which can then refer to the fact that the author is the same, but this does not necessarily have to be true. The importance of this research is that no one has dealt with this topic seriously in Croatia yet. In the first phase testing and analysis of the handwriting of a certain number of anonymous letters have been conducted in order to determine whether or not there is evidence that two pieces of handwriting have a common authorship (that is to say “Is there any evidence that two pieces of handwriting are written by the same person or not?”). Preliminary analysis was performed to check whether the collected anonymous letters meet the professional standards of handwriting analysis, regarding their quality and quantity, as is the standard practice under ISO 17025. Next, a detailed examination of general and individual handwriting characteristics was performed in all anonymous letters individually. A stereomicroscope and video spectral comparator were used to compare handwriting samples, as well as for capture of images and for measurement [1]. Even though the scope of our research does not necessarily include the graphic analysis spectrum, the described preliminary analysis was necessary to perform the following forensic linguistic comparison of the achieved results. For the purpose of further research, the hand-written letters have been transferred into a digital format, partially for the purpose of performing a computer text analysis [2, 3, 4]. The initial result achieved by identifying the author based on the handwriting characteristics has been verified using the method of forensic linguistics. After it was established that the author of all analyzed letters is the same, semi-automatic linguistic research of the letters was conducted in order to compare the results [5, 6]. Based on the vocabulary used, word order in sentences, sentence complexity, used nouns, adjectives and verbs, the intention was to determine whether the common text author could be established through such partially automatic linguistic analysis [7]. Computational analysis has shown that an average number of sentences in the letters is 13.538, with the shortest letter containing 9 sentences and the longest 25. When examining all the letters, an average number of tokens per sentence is 11.392. One letter contains in average only 4.364 number of tokens per sentence, while the letter with the longest sentences contains in average 19.280 tokens. The data was compared to two general corpora of Croatian language: the Croatian web corpus hrWac and the Croatian Language Corpus (CLC). hrWaC (version 2.2) contains 1.3 billion tokens compiled of documents collected on the hr domain. In this corpus, a sentence in average contains 17.952 tokens [8, 9]. CLC is a corpus containing documents written in standard language and contains 100 million tokens. An average number of tokens per sentence in this corpus is 18.074.The computational analysis has shown that the subject sentences are shorter than the ones in hrWac and CLC by approximately 6 to 7 words. Finally, it should be emphasized that the present research is a case study. In order to draw a universal conclusion, the research should be conducted on a larger number of letters. An interesting aspect of the conducted research is the analysis of hand-written letters and the comparison with the ones obtained through a linguistic and computer-helped analysis of individual words, sentences, or a text.

Forensic Linguistics, Handwriting Examination, Computational Analysis

nije evidentirano