Not only police officers work in the Federal Criminal Police Office, but also German specialists. The team found the alleged sender of the “NSU 2.0” letters and the bomber from the Dortmund stadium: through text analysis.
Wiesbaden – On Sabine Ehrhardt’s desk there are books that very few would expect from the Federal Criminal Police Office: the collected works of the Duden publishing house.
The BKA employee works with language, she and her team analyze texts in order to identify the author and thus to help solve criminal cases. This subject is called forensic linguistics, and the term author recognition has become common in German.
Ehrhardt heads the language and audio department at the Forensic Institute of the BKA. “Text can play a role in every crime,” says Ehrhardt. The classic is the extortion letter, but letters play a subordinate role today. “Written texts have become more diverse,” says the doctor of linguistics. Typewriter and cut-out letters are rare.
It is a fight against the flood of texts
That’s why there are SMS, emails, chats, tweets, posts. Since then, the investigators have been fighting against a flood of text – but that has advantages, as the specialist explains: “The language is more spontaneous.” When you were still writing letters, you avoided errors, read them in corrections before sending them, and thought about the design. Electronic typing takes place and whoops: send. “The more you write, the less you can make an effort and the more individual there is in the language.”
To find out who wrote a text or whether several texts come from the same author, the employees look at the “linguistic characteristics” of the text: What mistakes does the author make? Are there words that come from a dialect or another native language? Sometimes the vocabulary allows conclusions to be drawn about age, a job or a hobby. “This allows you to come very close to the author’s attitude,” says Ehrhardt, who has been with the BKA for 17 years.
For example, anyone who uses the word “nest dirtier” has the idea that there is a protected area that has to be kept clean and that is endangered by external influences. Anyone who says “unbeliever” is more likely not to come from a secular Christian environment.
Incidentally, emojis are also considered language. “They are a conscious decision by the sender,” explains the linguist. So they belong to the “character of a text”, similar to the layout. In the age of electronic communication, the visual design of a text only plays a subordinate role. At the end of the analysis there is usually an expert opinion that helps with the investigation or is used in court.
“We must not develop any form of zeal for hunting”
When the language forensic experts at the BKA are working on a case, they often do not even know what it is about in detail. “We don’t need any case information,” says Ehrhardt, “that is actually rather a hindrance”. As experts, language analysts should be as neutral and impartial as possible. “We mustn’t develop any form of zeal for hunting.”
The latest great success: the language department was able to help identify the alleged author of the threatening letters with the sender “NSU 2.0”. The man was arrested in Berlin in May and will soon be on trial in Frankfurt. He is said to have sent a total of 116 threatening letters between August 2018 and March 2021, including to the lawyer Seda Basay Yildiz, who represented relatives of murder victims of the terrorist cell as a co-prosecutor at the NSU trial.
The fact that the man could be identified was initially due to an officer from the State Criminal Police Office. It monitors right-wing blogs and forums and came across a user “whose posts in the form and style of the utterances showed similarities with the threatening letters of the so-called” NSU 2.0 “, as the State Criminal Police Office of Hesse (LKA) reported after the arrest.
Suspicious user in the chess forum
In a chess forum, the man not only used the same user name, but also the same profile picture and, according to the LKA, posted “verbatim insults in the chat function”. Linguistic reviews confirmed the suspicion that it was the same person. The investigators finally got their name and address through the operator of the chess platform.
“Text can play a role in every crime,” explains Ehrhardt. Her team has already tracked down a murderer or kidnapper who took his victim’s cell phone and used it to send text messages to cover up the crime. Because the messages sounded “wrong” to the recipients, the suspicion arose that someone else was typing here – and thus giving themselves away.
In 2011, a stranger threatened with an explosives attack on Borussia Dortmund’s football stadium. The explosion was thwarted – a whistleblower had warned the authorities of the danger. The linguists scrutinized the letter and compared it with the national fact sheet collection. The collection was created in the 1980s and currently contains around 6,000 texts. It turned out that the whistleblower himself was the one who had set the explosives.
Sometimes it’s about purchase negotiations on the Darknet, where criminals get drugs, weapons or child porn. Language analysis can help to consolidate cases: the buyer can use different nicknames or addresses, if his language gives him away, it stands to reason that they are dealing with the same person. The language forensic experts are not allowed to make a legal decision, the team only provides a probability estimate.
At the BKA, clients of such analyzes are exclusively the police, courts and public prosecutors. The team is not allowed to work for the defense or private individuals – there are private providers for this. One of them is the private institute for forensic text analysis in Munich. Most of the orders come from companies, as CEO Leo Martin says. Often it is about slander, false claims, defamation, “for example, when a manager is to be dismantled”.
“The leading authority in Europe”
Private individuals also commission the institute, for example to have the authenticity of a will checked – or the police inquire if the case is below the threshold at which the BKA acts. There is no sign of competition: the language team at the BKA is “the leading authority in Europe,” says Leo Martin. “They don’t take anything away from us,” says BKA employee Ehrhardt, and in any case there are many more inquiries than their team can handle.
A prominent case of the Munich language profiler took place in Mainz: During the mayor’s election campaign in 2019, the media received an anonymous letter intended to discredit the incumbent. The then director of the university medicine should also be discredited by similar letters, reports Martin. A comparative speech analysis revealed that the same sender was behind both cases.
Language can be extremely tell-tale
“Show me what you write and I’ll tell you who you are,” says Leo Martin’s colleague Patrick Rottler. He too thinks language is extremely treacherous. “Misrepresentation only works in one direction: the perpetrator cannot pretend to be any smarter than he is,” says the communication scientist. Martin does not expect artificial intelligence to do the work for the language profilers: “Language is too individual for that.” Supporting systems already exist today, for example they would generate lists that help to recognize significant patterns. “But the main work will remain with the human expert for a long time.”
At the Ruhr-Universität Bochum there is a research focus “digital forensic linguistics” in the German Institute. The scientists develop methods based on artificial intelligence to work out linguistic characteristics. “We apply this basic linguistic research to uncover hate speech and disinformation in social media,” says Junior Professor Tatjana Scheffler.
Her team mainly does basic research “on the variability of language”, for example on the question of how style changes when you change medium. A research project on IT in author recognition is running at the institute with the BKA. Despite the demand, forensic work does not exactly open up a huge opportunity for unemployed German scholars: the number of positions is manageable. There is no training path or subject. dpa
#text #reveals #perpetrator