Language experts to help identify internet 'trolls'

  • Published
Georgia Varley
Image caption,

A Facebook tribute page for Georgia Varley was targeted by trolls

University researchers are developing a system to help identify people who are behind offensive comments posted on the internet.

Linguistic experts at the University of Central Lancashire (UCLAN) are working with police and child protection agencies to develop an automated system to recognise language patterns.

The aim is to track people who post anonymous malicious messages.

The activity, known as "trolling", was recently raised with Facebook by Liverpool Walton MP Steve Rotherham following messages that were left on a tribute page to 16-year-old Georgia Varley, who died when she fell between a platform and a train.

Claire Hardaker, lecturer in linguistics and English language at UCLAN, said: "Everyone has a unique way of writing, of putting certain words together, which is subconscious.

"Many teenagers say they are able to identify who sent a text to them - just by the style of writing and word habits or the way the words are written.

"Someone might be pretending to be someone else, but by analysing the way they write online, we can determine a probable, age, gender - even a probable region from where they come from.

"In its simplest form, people use different words for things - for example a bread roll. Some people would say a tea cake, some people would say a barm - it is these sort of elements that help to narrow down a search."

It is proving hard for authorities to trace so-called trolls and there have only be two people in England successfully prosecuted.

Sean Duffy, from Reading, was jailed for 18 weeks after posting abusive remarks on pages set up in remembrance of girls who had died.

Meanwhile, Colm Coss, from Manchester, was jailed for 18 weeks for sending "malicious communications" after posting on a memorial page for Big Brother star Jade Goody and a tribute site to John Paul Massey - a Liverpool boy mauled to death by a dog.

Ms Hardaker added: "It is not just anonymous posts.

"Online there are all these extra problems of 'did this person genuinely author this?'

"Can you actually prove that the person sat at that computer at that moment and wrote that message?

"So, even though we know some of these things are patently wrong, just actually getting the person in question can be really difficult."

'Different posts'

Every computer connected to the internet can be traced through IP addresses, but trolls often use special software to disguise their location making it harder for authorities to find.

Ms Hardaker said she believed this was where linguistic profiling could help.

Researchers at the university in Preston will spend the next year working with Merseyside and Greater Manchester Police, as well as the Child Exploitation and Online Protection Centre (CEOP), to develop a linguistic profiling system.

Ms Hardaker said: "The analysis process is currently done by hand by a linguistic specialist manually analysing online text.

"We are looking to develop an automated computer system to analyse the different components of written text.

"If the police have a suspect paedophile, for example, we can work with them to help to identify whether different posts across different sites were actually written by the same person."

Tuesday marks Safer Internet Day, an annual event designed to encourage people to use internet technology more responsibly.

This year it coincides with the week-long BBC Share Take Care campaign, in partnership with the UK Safer Internet Centre, to raise awareness of issues surrounding safeguarding reputation online.

Around the BBC

Related internet links

The BBC is not responsible for the content of external sites.