In Two Years a Bot Will Win a Pulitzer Prize

At least that’s what Kristian Hammond of the company Narrative Science predicted in 2012. Is it possible? After looking through the data and assessing some of the technological advances in the emerging field of news writing bots, I can’t completely discount this claim. Already, Narrative Science technology is used to write corporate reports for Forbes, short sports stories on local events, and short articles for specialist publications. In fact, the algorithm behind the bot can write a news story every 30 seconds or so, and it can do it on its own without any human intervention.

Basically, the algorithm looks at, for example, corporate data (numbers), analyzes the statistics, and turns the information into an article using a standard syntactic format that is common to most company reports. They may not be the most creative reports, but they get the information across. It is even possible that the bot may find some relationships between the numbers that humans have failed to notice. That may be where the possibility of a Pulitzer Prize comes in.

That may be fine if only corporate numbers are concerned, but what else can these bots do? Well, some are programmed to read twitter feeds on various topics and reach certain conclusions which are then presented in article form. They can be programmed to search the internet and social media for trends and produce accounts based on what they find. They can do some research and produce articles for small, niche publications that no one else is writing for. However, most such articles are short with formats that lack much variety. This is because humans have written templates for these bots in different subject areas. That said, Narrative Science does offer its customers the ability to choose the tone or style that they feel most comfortable with. It’s just a matter of tweaking these templates.

When it comes to searching for individual topics to write about, the bot must be programmed to recognize certain vocabulary and phrases connected to particular topics. One bot that the company designed to write restaurant reviews, learned to recognize key phrases from customer comments such as “great service”, delicious meals”, “below average” and so on. It later would produce articles with titles such as, “The Best Italian Restaurant in Atlanta”. One would have to wonder, however, if the bot could correctly determine the intent of more linguistically challenging comments. (“Great service and a delicious meal were beyond the expertise of this establishment.”) To rid the stories of non-human errors and to add more verbal spice, human editors still need to be employed. In my opinion, these programs would work better if they had some self-learning, neural network, loop built into them, where they could compare their own output with the final human output and adjust themselves accordingly.

But bots are writing more than restaurant reviews. Some write for Wikipedia. One such bot has helped Sverker Johansson become the most prolific contributor to that service. In fact, he is responsible for 8.5% of Wikipedia’s content. That’s a lot of influence. Had he generated it without the help of his self-developed Lsjbot, it would have taken many human lifetimes. This is because Johansson produces about 10,000 articles a day and had totalled over 2.7 million as of July. True, most of his entries are short stubs about obcure insects or small towns, but they are all bot-written. His bot scours trusted internet sources for information on which to write articles.

Bots are really nothing new to Wikipedia. Vandals used to wreak havoc with entries by replacing, let’s say, the name of a famous president with that of a cartoon character. In order to stop such vandalism, Wikipedia uses a vandal detection bot called ClueBot. It also uses bots to correct spelling and grammar and add new information from databases. In fact, a recent study found that about half of all Wikipedia edits are done by bots.

Journalists and other content writers have already seen their jobs threatened by new media. Do these news bots pose an even greater threat? They could. Hammond predicts that within the next 15 years, bots will write 90% of all news. On top of that, programs now exist that take original content from news sites and blogs and spin it to make it look like it was created by the person who took it. Basically, they change the vocabulary a bit. Some of these are free, web-based programs like Spinbot. Many such tools are available online but most have Spinbot architecture. Some that don’t are so bad they are laughable. To demonstrate this, I will show what happens when I rewrite the last few sentences.

Spinbot version:

A lot of people such apparatuses are accessible online yet most have Spinbot structural engineering. Some that don’t are so terrible they are absurd. To demonstrate this, I will indicate what happens when I change the last two sentences.

Plagirisma.net

A to each of relatives such apparatuses are at hand online yet most have Spinbot structural engineering. Varied prowl don’t are so terrible they are absurd. To altercate this, I spine suggest what happens when I change the last two sentences.

Other software programs are said to be more sophisticated. One of these is Content Professor. This company basically asks the question: Why be stupid enough to use so much time researching and writing your own articles when you can take someone else’s article for free? Of course, they don’t use these words, but this is their message. This program includes a plagiarism checker to get around those pesky plagiarism laws. As they say, the program gives you, “Everything you need to rapidly Create an Unlimited Number of UNIQUE, Quality Articles That Are Ready To Flood Your Sites With Traffic, Backlinks & Sales!” And when you’re done you can distribute ‘your’ article to around 400 sites with Articlebot. Actually, this program takes care of all the article generation and distribution for you. They also hope to get a boost in search results. “The formula is a simple one – the more articles you write, and the more directories you submit them to, the more money you make”. That’s a good formula as long as you have real writers out there somewhere producing real content. Articlebot and similar programs are generally held in contempt by webmasters and bloggers, with some justification. They have also been accused of containing spyware which sends a copy of every submitted article back to the company. In any event, it’s too bad those of us who write original content can’t get a commission from these guys.

That brings us to Phillip M. Parker. Parker has developed a program that writes technical books. He’s written a lot of them; over 800,000 of them, to be exact. You can buy any of them on Amazon. He has an unlimited number of books that are available only when someone orders one. That’s because his program can compile a book on any topic in a few hours or less. You request the book. The book becomes available. True, the book may not make the most compelling reading material, but if you want some information in book form on a somewhat obscure topic, this program may be for you. In addition, the program can do much the same with collecting photos or videos.

But, at least it will be a while before computers do any creative writing, like writing novels and poetry, right? Not really. They’ve already done it. Not only that, but scientists have used algorithms to predict with 84% accuracy which books will be a commercial success. The algorithm analyzed the vocabulary, style, and plots of 800 books of multiple genres to attain such a success rate. Poetry generators fall a little short and end up like something you would create with Mad Libs. Nonetheless, look for them to improve as better algorithms are created. Where does this leave writers? Well, almost all of these programs need to have their content pass through human editors. However, this being the case, I can’t help but wonder who is working for whom?

About Steve Mierzejewski

Marketing consultant for InZero Systems, developer of the next generation in hardware-separated security, WorkPlay Technology. I've worked in Poland, Japan, Korea, China, and Afghanistan. I'm a writer, technical editor, and an educator. I also do some work as a test developer for Michigan State University.
This entry was posted in Uncategorized and tagged , , , , , , . Bookmark the permalink.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s