Boffins build AI that can detect cyber-abuse – and if you don't believe us, YOU CAN *%**#* *&**%* #** OFF

Alternatively, you can try to overpower it with your incredibly amazing sarcasm

Trolls, morons, and bots plaster toxic crap all over Twitter and other antisocial networks. Can machine learning help clean it up?

A team of computer scientists spanning the globe think so. They've built a neural network that can seemingly classify tweets into four different categories: normal, aggressor, spam, and bully – aggressor being a deliberately harmful, derogatory, or offensive tweet; and bully being a belittling or hostile message. The aim is to create a system that can filter out aggressive and bullying tweets, delete spam, and allow normal tweets through. Pretty straight forward.

The boffins admit it's difficult to draw a line between so-called cyber-aggression and cyber-bullying. And the line between normal and aggressive tweets is often blurred: after all, people enjoy ranting about things, from feminism to Brexit to tabs-versus-spaces. Having said that, the goal was to craft a system that can automatically and fairly – and by fairly, we mean consistently – draw a line between each category.

After analyzing more than two million tweets that discussed touchy topics such as Gamergate and gender pay inequality at the BBC, as well as more neutral matters like the NBA, the eggheads selected a sample containing 9,484 tweets, and hand labelled them as normal, aggressor, spam, and bully. Obviously, this means the academics' definition of what is aggressive or bullying forms the basis of the model.

About 80 per cent of these tweets were used to train the recurrent neural network, and the remaining 20 or so per cent was used to test it, according to one of the scientists: Jeremy Blackburn, an assistant computer science professor at Binghamton University in New York. We're told the code could sort the test tweets into the four categories with over 80 per cent accuracy. That is to say, 8 of 10 times, the AI would categorize a tweet as expected by the human boffins.

Their research was published in the journal ACM Transactions on the Web – here's the Arxiv version [PDF]. While it's not revolutionary, it's a good introduction text analysis – and don't forget, it is an academic study.

Cyber-bullying is 'dehumanizing'

The neural network analyses not just the content of a tweet but also the tweeter's profile, and the rate at which they tweeted. All the words are encoded as vectors, and various algorithms were used to determine the overall sentiment or emotion of the message and its sender, and whether or not any curse words were used, plus how many hashtags were included, and the number of followers someone has. All of this information is fed into the network so it can predict the category its human masters would have assigned the tweet.

It’s not easy to pinpoint which features are more indicative of online harassment, Blackburn told El Reg on Monday. “It is not straight forward to describe this because we are differentiating between several categories," he said. "For example, we saw that bullies and aggressors had less time between their tweets than spammers, but similar to spammers, bullies used fewer adjectives than normal users and aggressors.”

Aggressors and bullies were more likely to tweet multiple times and use more hashtags than spammers and normal accounts. Bullies tend to harm others by directing their messages at specific people, whereas aggressors were more likely insult groups of folks. Spammers, on the other hand, are less likely to use abusive language and tend to sell things like smut pics and videos.

“Normal users tend to discuss a variety of topics, such as political and social issues, whereas bully users seem to organize their attacks against important and sensitive issues, such as feminism, religion, and pedophiles, using aggressive and in some cases insulting language. Aggressive users express their negativity on popular topics, such as the ‘brexit’ case, ‘maga’, and the spread of the Zika virus. Spammers typically post inappropriate content in an effort to gain more followers or attract victims to malicious sites of questionable or malicious content,” the team wrote in their paper.

Blackburn said cyberbullying has only recently been taken seriously. “This type of behavior is often dehumanizing, which most rational people would consider a bad thing," he said. "From a more pragmatic point of view, the intensity of this behavior, enabled by the scale of the Web and social media, can and has led to acts of real world violence.

'I radically update my course module almost every year to keep up with the rate of change'


“I think that efforts are being made, but that much more can be done. There are difficult decisions for social media companies to consider. For example, some of the biggest offenders also have large followings on social media; simply silencing these people can have unforeseen consequences.”

The researchers hope to use their model on other platforms, such as YouTube or Facebook, though there are other challenges when using machine learning to tackle hate speech and harassment online.

In previous studies, computer scientists discovered that abusive messages bypassed classifiers if the text contained spelling mistakes. Not all rude comments are hateful, either.

“Understanding sarcasm is difficult, especially because different people and cultures express it in different ways,” noted Blackburn. "We did not examine sarcasm in particular, however our algorithm is designed to be extensible, and we intend to add more linguistic features to it in the future." ®

Send us news
Get our AI newsletter

Salesforce beats banks to top UK exec salary survey while Microsoft drops out of league

Come for the 'ohana', stay for the massive piles of cash

Salesforce has topped a survey of UK executive salaries with median earnings of around £100,000 per annum, beating investment bank Man Group and consultancy Kearney to the top spot.

The SaaSy CRM company, which makes a big deal of allegedly world-improving "stakeholder capitalism" and "ohana" (the Hawaiian concept of family), obviously feels the need to put its money where its mouth is when it comes to competing for executive talent.

The survey was carried out by pay and workplace comparison site Glassdoor, which also lists Facebook (median total compensation of £86,423), VMware (£86,625), and data analytics and ML company G-Research (£86,625).

Continue reading

39 Post Office convictions quashed after Fujitsu evidence about Horizon IT platform called into question

Japanese firm has a reckoning to face

Post Office employees were wrongly prosecuted by the company as a direct result of it covering up software bugs in its Horizon IT system, the Court of Appeal has said as it quashed 39 convictions this morning.

Those 39 convictions were obtained by the Post Office's in-house lawyers who ignored their own barristers' advice that the institution's behaviour was trampling over established prosecutorial codes intended to promote fairness and honesty.

Lord Justice Holroyde said that the one-time state monopoly had, by representing Horizon as reliable, "effectively sought to reverse the burden of proof," leading to criminal defendants having to prove their innocence instead of the Post Office showing they were guilty. Its lawyers compounded this by withholding evidence from courts and defence lawyers alike – evidence which clearly showed the Post Office and Fujitsu knew Horizon wasn't generating accurate accounting records.

Continue reading

NASSCOM shakeup: Accenture's Rehka Menon becomes first woman to chair Indian IT trade org

Also, Dell and HP miss out on expanded board positions

Indian IT trade association and advocacy group National Association of Software and Services Companies (NASSCOM) has appointed Rehka Menon of Accenture as chairperson – the first woman to take the position in the organisation's history.

Continue reading

SpaceX flings another bunch of humans into orbit in reused capsule atop reused booster

Second jaunt to the International Space Station for Endeavour

SpaceX has launched the second operational Crew Dragon mission, sending another four astronauts to the International Space Station.

The capsule itself, dubbed Endeavour, already saw action during the Demo-2 mission in 2020 and the first stage of the Falcon 9 booster is also flight-proven, having first been used to launch the Crew-1 mission last year before safely returning to Earth in the usual crowd-pleasing fashion.

The launch had been delayed for a day due to unfavourable weather conditions along the flight path; while the launch site looked good, possible recovery areas were less ideal (should an abort be required, which it was not.)

Continue reading

We've finally hit Peak Bork: Microsoft man reveals home-grown welcome back BSOD at Redmond HQ

A treat for returning workers as meeting cancelled due to roomful of bork

Bork!Bork!Bork! Microsoft is famed for eating its own dogfood and this week chowed down on a bowl of fresh bork as its consulting boss encountered what we can only assume is the company's latest attempt to deal with Meeting Culture.

Ben Rudolph, chief of staff for Microsoft Consulting tweeted the screen of baleful blue that greeted him outside a room in Building 115 on the Microsoft's Redmond Main Campus.

The screen would normally cheerfully inform passersby of meetings due to happen or the gatherings already occurring within.

Continue reading

Penguin takeover: We tried running some GUI Linux apps on Windows the official way – and nothing exploded

Microsoft's own distro lurks under the covers and applications magically appear in the Start menu

Hands on Microsoft has released the first public preview of Linux GUI applications on Windows 10 – so we wasted no time in taking it for a spin around the block.

The ability to run GUI applications on Windows Subsystem for Linux 2 is not new – enthusiasts have been able to run them via separately installed X server utilities for years – but the arrival of official support is still something of a game changer. The official support is more ambitious and better integrated than the various unofficial approaches.

The preview comes via the Windows Insider Program, by which developers and enthusiasts can get an early look at forthcoming releases. The latest downloadable build is 21354, but the version that supports WSLg is build 21364, so in our case it was a matter of installing 21354 from an ISO image, logging in with a Microsoft account signed up as an Insider, and then waiting while the later build came down from Windows Update. We also chose to run it in a Hyper-V VM. To do this, it is necessary to enable nested virtualization since WSL also uses Hyper-V. That requires a PowerShell command from the host machine.

Continue reading

BOFH: Postman BOFH's Special Delivery Service

Back to the office we trudge, dragging our hooked tails behind us

BOFH logo telephone with devil's hornsEpisode 4 WE'RE BACK in the office – and I feel like I 'm only now just starting to feel the effects of COVID.

Continue reading

Something went wrong but we won't tell you what it is. Now, would you like to take out a premium subscription?

Get with the disruptive app design, granddad: meaningless errors are cool

Something for the Weekend, Sir? An error has occurred… Of course it did – I'm in a hurry and the login is sensing my urgency. Big mistake. Let's try again, more casually. An error has occurred…

So it doesn't like my casual manner. How else could I type my credentials into the login screen to fool the remote computer into letting me view my own data? I try typing them r-e-a-l-l-y s-l-o-w-l-y. An error has occurred… I try elaborately. An error has occurred… I try viciously, nonchalantly, softly, insistently, accidentally, musically, and a variety of other adverbs. An error has occurred…

Hmm, it might not be the manner in which I am tapping the keys after all. Perhaps an error really has, well, occurred. The one thing I do know is that my login name and password are correct. It looks like I shall have to contact customer support.

Continue reading

From pregnancy tests to obscure industrial kitchen appliances, Doom really will run on almost anything

As long as the technical nous is there

If you've ever worked in a commercial kitchen, you probably know what a "bump bar" is. These plank-like computerised boards are the organisational force behind every restaurant, enabling sweat-drenched chefs to see pending orders and ensure a steady flow of food to the front of house.

Would it surprise you to learn that someone managed to coax '90s shooter Doom onto one? Of course it wouldn't.

Doom is the videogame equivalent of a glitter bomb, attaching itself to literally everything with an electrical current and a microcontroller. We've seen it played on digital cameras, iPods, calculators, even ATMs and pregnancy tests. So the idea that it would run on a low-powered computer device used exclusively by eateries maybe isn't so outlandish.

Continue reading

George Clooney of IT: Dribbling disaster and damp disk warnings scare the life out of innocent user

Practical jokes: just don't

On Call Welcome to another entry in The Register's On Call files, where we learn that the hilarious pranks of an IT joker can be enjoyed as much as millionaire actor George Clooney's "fun" leg-pulling.

"Jim" returns once more to pages of On Call with a tale of poorly targeted japery and an unfunny practical joke. Is there any other sort?

We skip once more to the 1980s and the mighty IBM XT, a fleet of which Jim was tasked with supporting. Data entry had been done by punch cards back in the day, but the PC now reigned supreme. Keypunchers banged in information via keyboard, as Jim explained: "Floppy disks holding the keypunch program were inserted to provide the OS, and then the other floppy in a dual system held the precious data, from which the company earned its crust."

Continue reading

Starlink creates risk of internet investment doom cycle, says APNIC researcher

Early users get speed, so policy-makers may stop building, but more users means slower speeds ...

Elon Musk's Starlink project has copped more criticism, this time from a researcher at APNIC, the Regional Internet Registry for the Asia-Pacific region.

APNIC's George Michaelson says the project, already under fire for offering low capacity and high prices, risks being loved to death by the wrong sort of users, and may therefore stymie much-needed investment in broadband.

“The bandwidth provided by these LEO satellites is really very good … for now. Starlink, unfortunately, runs the risk of being a victim of its own success,” wrote Michaelson.

Continue reading