Current Projects

TXT – Language as a Trace

Project Details

Funding

FFG Kiras

Grant Number

926222

Duration

2026/01/01-2027/12/31

Contact

Florian Kleber

Persons

Robert Sablatnig
Florian Kleber
Rafael Sterzinger

Development of an AI-based tool for analyzing the need for forensic linguistic expertise

Forensic linguistics is an emerging discipline that is used when documents of any kind become the subject of an investigation and thusconcerns criminal offenses such as stalking, blackmail, hate postings and defamation. In addition, anonymous tips, letters of confessionand manifestos can also be the focus of analyses. If, as is often the case in such investigations, the only lead to the perpetrator is alinguistic one, forensic linguistics can determine characteristics of unknown authors, compare them to other texts of possible suspectsand create language profiles for further investigation. As these analyses have to be carried out manually by experts, their current use islimited. It is therefore particularly important to know in which situations experts should be consulted – and this is precisely where this project comes in.


The “TXT – Language as a Trace” project aims to create an AI-supported analysis tool that examines texts to determine whether in-depthanalyses by experts are possible and worthwhile. In addition, a preliminary assessment will be made whether a similar linguistic tracealready exists in the database, which could then be subjected to further analysis by experts. As a data basis for this project, thehandwritten CV collection of the Document & Handwriting Investigation Unit in the Department of Forensic Science of the CriminalIntelligence Service Austria (BK) in the Federal Ministry of the Interior (BMI) will be digitized and supplemented by further incriminatedtexts. Methodologically, this goal is to be achieved through the use of machine learning techniques, whereby an AI-based tool can make atime-efficient statement in these two areas whether an analysis is possible and whether comparison texts should be analyzed in greaterdepth.

Partners

  • Universität Graz, Institut für Strafrecht, Strafprozessrecht und Kriminologie
  • Bundesministerium Inneres (BMI), Herrengasse 7, 1010 Wien
  • WareTect IT Solutions GmbH
  • Universität Graz, Business Analytics and Data Science-Center (BANDAS-Center)
  • Universität Graz, Institut für Anglistik

The project is funded by the security research program KIRAS/K-PASS.