Wednesday 4 April 2012

Introducing the Evening Post Article Factory!

When introducing our automated bingo card generator, we mentioned that an ultimate goal was to generate whole articles. This actually turns out easier than expected. Rather than use stochastic Markov Chain modelling to predict the follow-on words from an initially n-tuple of corpus-extracted words, we can just use the sentences extracted from the evening post articles and pick some in a random order, initially making sure not to repeat any. This works so well that we cannot even believe it ourselves:

"There are more motorists than cyclists, it makes perfect sense to just ban them. Dark clothing, no lights and certainly no sense but, incredibly, he shouted at me as if I was at fault for not wearing my night vision goggles. As soon as I moved here I started having issues with cyclists, she told the Evening Post. We all aspire to live in attractive, liveable cities where it is safe to get around by foot and by bike, and so we need to encourage people to take up these modes of transport. There were also a couple of pedestrian crossings where cyclists just didn't stop. How about spending some money on the poor down-trodden motorist. But a bit of tolerance might help matters a little. I don't think there are any particular blackspots in the city for riding on the pavement - the whole city is a blackspot. Motorists don't pay enough attention towards cyclists and they seem to forget that we are very vulnerable. If Bristol is the first cycling city then there needs to be more education out there. "

Even though there's been no attempt at finding some coherence between individual sentences, it turns out not to matter. Look at this one. You could get this as a letter in the evening post without even bothering to do any proof-reading:
"Do all this and we can all be friends. Cyclists going on the pavement is really quite dangerous. ). Cyclists shouldn't expect respect from motorists if they can't even be bothered to wait for the lights to go green before launching themselves forward. I have yet to hear of a cyclist that rides according to the law. It's illegal to ride without a helmet and high visibility clothes. The lad was without lights or reflective clothing, and clearly also a responsible adult to look after him. Not so much hug a hoodie but cuddle a cyclist perhaps. And what would have ensued had he collided with me. Whatever happened to the old 'Be safe, be seen' advertising campaign. "
And again.
'Cyclists should observe the highway code just as motorists do. I have yet to hear of a cyclist that rides according to the law. If a driver hit him off his bike they would no doubt get charged with driving with undue care and attention. Concerned residents made their points to MPs from across the region at a meeting of the Royal National Institute of Blind People (RNIB). Cyclists need to realise that the road is for motorists and cyclists and the pavement is for pedestrians. I realise not all cyclists behave like this, just the vast majority, especially the Lycra ones who think they are as wide as cars. There desperately needs to be something shown on TV perhaps to illustrate these dangers. Bristol cyclists still running red lights. I personally think a Cyclist free week would be far more beneficial. We are working with the police to try and overcome this problem. '
This is transformational. We can use the past works of the Evening Post article and letter writers to actually render themselves obsolete. It takes 17 milliseconds to generate one of these articles on a four year old laptop, which means that five million articles can be pushed out every single day. The evening post web site could be set up to generate one every time a visitor hit "reload" on the anti-cycling page. Then, once the comment generator was hooked up to generate comments from the collected comment articles, we could go on to simulate an entire Evening Post comment argument. Finally, the same code could be hooked up the Ashton Value stadium dataset and the whole town green/football stadium battle could be fought to the death between two competing servers. Like skynet only more provincial.

To close, then, with a fifteen sentence article:
"That is very dangerous because you expect to be able to cross safely without watching out. 20 billion pounds of damage to business is caused by traffic jams and cyclists are the main cause. The old Bristol to Bath railway track, Leigh Woods, the Avon Walkway and Ashton Court estate have all been badly scarred by cycling louts as they trash pedestrian paths making them fit only for cyclists, or gouge cycle tracks through the woodland for their hobby. Sometimes this has been where small children are present or when people are walking their dogs. Bodywork repair costs a lot these days. Every single one I see cuts through red lights like a knife through butter. They give me the impression that it is only them on the track and seem blissfully unaware of the fact that there are other 'freelance' cyclists and pedestrians who have just as much right as they have. Mr Drummond said the police seemed to turn a blind eye to cyclists on the pavements, too. It's illegal to ride without a helmet and high visibility clothes. Then, as I reached my destination another cyclist ignored the rules of the roundabout and pulled out straight in front of me. If something isn't done soon, I am afraid there is going to be a terrible, terrible accident. I do think that I will get hit at some point, especially if the number of cyclists in the city keeps increasing. We won't be coming again. I have nothing against cyclists and if I could see properly I would be one, but I would like to think I would be a considerate cyclist. "
Bristol will never be the same again