Replacing Sports Bloggers With an Algorithm

tesmar tips a report up at TechCrunch that begins "Here come the robo sports journalists. While people in the media biz worry about content mills like Demand Media and Associated Content spitting out endless SEO-targeted articles written by low-paid Internet writers, at least those articles are still written by humans. We may no longer need the humans, at least for data-driven stories. A startup in North Carolina, StatSheet, today is launching a remarkable network of 345 sports sites, one dedicated to each Division 1 college basketball team in the US. For instance, there is a site for the Michigan State Spartans, North Carolina Tar Heels, and Ohio Buckeyes. Every story on each site was written by a robot, or to put it more precisely, by StatSheet's content algorithms. 'The posts are completely auto-generated,' says founder Robbie Allen. 'The only human involvement is with creating the algorithms that generate the posts.'"

Comments

Close, but still not pratical

[kevinNCSU]

I tried reading the first article on the Tar Heels, and as much as I hate reading anything about the Tar Heels the sentences just don't flow together. It's disjointed and mentally uncomfortable to read. I can't imagine anyone using it as an actual replacement for even semi well-written content.

Re:Close, but still not pratical

[John Hasler]

I can't imagine anyone using it as an actual replacement for even semi well-written content.

They aren't. They're using it as a replacement for the output of sportswriters.

Re:Close, but still not pratical

[WrongSizeGlass]

Hard to read? Disjointed? Mentally uncomfortable? Sounds like it could fit right in here on /. ;-)

Re:Close, but still not pratical

[St.Creed]

Hard to read? Disjointed? Mentally uncomfortable? Sounds like it could fit right in here on /. ;-)

A clever attempt, RoboWrongSizeGlass, but not clever enough! Trying to point the finger at humans while sneaking in another templated contribution! Haha! Your plans will never work! :P

Re:Close, but still not pratical

[mysidia]

It doesn't have to fool humans (unfortunately). It just has to fool Googlebot

Re:Close, but still not pratical

[St.Creed]

I can't imagine anyone using it as an actual replacement for even semi well-written content.

It's a bit uncomfortable to read in spots, but way above the quality of most blogs and nothing you can actually point out as an error. So if they manage not to swamp the sites in ads, and provide good statistics as well, I can't see why they couldn't get a rather large take of the advertising action - large in proportion to the manhours invested in writing articles, that is. It could be very lucrative and if it is - well, say goodbye to a lot of reporters who aren't the primetime writers but just pad out the papers: they're going to be automated away or at best, write the templates for the robowriters.

Re:Close, but still not pratical

[ThreeGigs]

Lucrative is right. They seem to have garnered 1.3 million dollars in funding for their efforts.

Re:Close, but still not pratical

[JustinOpinion]

You joke, but I think this pretty much nails it. There's a lot of content out there that is just a bunch of numbers wrapped up in some formulaic sentences. The results of sports games is an obvious example. Analyses of political campaigns might also be amenable. Perhaps even presenting the results from surveys or scientific studies.

The important thing here is that this isn't replacing deep, insightful thoughts and analysis, which still has to be done by a human. If you want a reasoned opinion that pulls together the statistics, external factors (e.g. a player's mind-set or personal life), and adds in some humor, then you're going to want a skilled human doing the writing. But if your interest is more along the lines of "Who won, by how much, and what were the main things that led to them winning (e.g. was it strong offense or good defense)?" then auto-generated content is fine. In fact, as with all aspects of automation, the point is to free up humans from doing the boring, silly tasks, so that they can concentrate on the more important tasks.

After reading some of the auto-generated articles (Michigan State Spartans, North Carolina Tar Heels, and Ohio Buckeyes) I must say I'm quite impressed with how good the content is. Obviously it won't be winning any prizes, but I can't say that it's any worse than human-generated summaries of matches. It goes through the details, throwing in some contextual commentary (e.g. "the underdogs") obviously based on a nice database of stats. What's even better is that the articles also present some of the stats themselves, allowing the reader to skip the writeup and focus on the numbers/graphs if they prefer.

So, frankly, I see this as a good thing. It's a waste of human talent (even mechanical-turk caliber talent) to write a bunch of formulaic summaries when a computer can clearly do a decent job. This lets the humans focus on tasks that are more difficult to automated.

Re:Close, but still not pratical

[houghi]

Perhaps even presenting the results from surveys or scientific studies.

Or interviews with actors about their latest movie, where they are telling how great it was working with X and Y. Or about any PR stuff going out into the world.

If they are just formula's around numbers, just give us the numbers. No need for all the fluff around it.

It says more about the lousy writing people are used to then the quality of a new Lisa script.

Re:Close, but still not pratical

[Tacvek]

The raw numbers are useful, but many people would like to read a quick summary of the highlights of the statistics, rather than having to read through them themselves.

Somebody who is not well acquainted with the specific stats may have trouble telling what is unusual, or combining them together to reach a conclusion. Even those familiar with the statistics would often find it quicker to read the computer generated summary than trying to skim the numbers to determine if they are worth spend more time on.

But I would agree that providing the actual numbers is important, so the summaries can be verified if desired, etc.

Re:Close, but still not pratical

[Tom]

If they are just formula's around numbers, just give us the numbers. No need for all the fluff around it.

You're a geek. So am I, but here's a secret I learnt: Lots of people are afraid of numbers. Much in the same way they are afraid of punks - they don't really think they will harm them, but they prefer to have them accompanied by words/police.

Re:Close, but still not pratical

[Anonymous Coward]

here's a secret I learnt: Lots of people are afraid of numbers.

Generally speaking, you're right. That said, I'm an engineer, and I've never met a geek more into numbers than the sports fans I've met. They might not understand the implications of the stats they're spewing out, but that doesn't mean they don't have them memorized.

The site is going after the right demographics. Sports fans are hungry for numbers like these.

This post....

[nitehawk214]

This post was written by a robot.

Re:This post....

[gandhi_2]

What makes you say this post was written by a robot?

Re:This post....

[bunratty]

Please tell me more about this post was written by a robot.

Re:This post....

[WrongSizeGlass]

Please tell me more about this post was written by a robot.

If I am correct, you are having trouble believing about this post was written by a robot? I do apologize for the inconvenience. Is there anything else I can do for you today? Thank you for calling Dell technical support.

Re:This post....

[Anonymous Coward]

Do you enjoy your inability to avoid making puns?

Not bad...

[Jazz-Masta]

I've read a couple articles and they are no worse than the SEO-targeted content written by freelancers odesk for $2/hr (and english as a second or third language).

Seems as though the "algorithm" is quite elaborate - taking into account odds of winning as well. Lines such as "The [team] was not supposed to win this game, but made it happen" and combined player statistics "Coming off a poorly put together team last year, this year, the [team] looks to have greater talent."

It reminds me of how someone in Junior high would write. Impressive. Similar to MIT's paper generator: http://web.mit.edu/newsoffice/2005/paper.html

PHP + MySQL + Mad Libs for Sports.

Help reading

[tocs]

Now I just need to find a robot to read all these sports blogs to free up time for things I want to do.

fans

[emkyooess]

Now we need a sports fan algorithm to rid ourselves of all these needless sports fans in the world and replace them with something more worth the resources.

Re:fans

[John Hasler]

That wouldn't get rid of them. Better to keep them occupied with a cheap diversion (which also keeps the athletes busy). Do you really want the jocks and their fans wandering around looking for something to do?

xkcd

[Jazzbunny]

Mission. Fucking. Accomplished.

Obligatory Simpsons

[Anonymous Coward]

This is the DJ 3000. It plays CDs automatically, and it has three distinct varieties of inane chatter:
- Hey hey -- how about that weather out there?
- Woah, that was the caller from hell.
- Well, hot dog -- we have a weiner.

- Those clowns in congress did it again -- what a bunch of clowns.
How does it keep up with the news like that?

Impressed

[Andy Smith]

I read the first article on the first linked site and I was impressed. I wouldn't have known it was generated by a computer. Even knowing that it was computer-generated, I'd still be happy with the quality for this kind of reporting. Very good.

Human involvement

[Chemisor]

I am going to guess that there will not be any humans involved in reading the output either.

tiny issue

[sribe]

Can you copyright the output of an algorithm? Seriously, copyright requires a creative element...

Re:tiny issue

[maxwell demon]

Tell that to a phone book or other assemblage of facts.

I tried, but the phone book wouldn't listen to me.

Re:tiny issue

[slashqwerty]

Tell that to a phone book or other assemblage of facts.

Perhaps you're confused about the outcome of Feist Publications, Inc., v. Rural Telephone Service Co. Phone books and other collections of facts may not be copyrighted because they lack creativity. Hence the question:

Can you copyright the output of an algorithm? Seriously, copyright requires a creative element...

While sribe focuses on the creative element one must also ask who the copyright would go to. The constitution grants congress the power

To promote the Progress of Science and useful Arts, by securing for limited Times to Authors and Inventors the exclusive Right to their respective Writings and Discoveries;

To my understanding this has always been interpreted to mean authors have rights to their writings and inventors have rights to their discoveries.

There's a T-shirt for that...

[MachDelta]

Why am I suddenly reminded of this t-shirt? :)

Including the core cliché of sports reporting

[PapayaSF]

As I suspected it would, the first sentence includes the word "momentum."

Slashdot -- proudly Luddite

[1u3hr]

At least we know that Slashdot isn't generated by robots. A robot wouldn't make the idiotic mistakes that the current human (for want of a better word) editors do. E.g. "one dedicated to each Division 1 college basketball tam in the US." Robots don't suffer from dyslexia, and aren't too lazy to use a spell check.

I like it

[Anonymous Coward]

Now sports editors have something to show novice reporters. "If you can't give me something a whole lot better than this, you're fired".

It's a reminder that standards for every knowledge-based profession are going up every year, driven by the combination of the Internet, globalization, and Moore's Law. And this is just the start of it for journalism.

Emotionless Facts

[denshao2]

Part of good sports writing is that it evokes emotions. I read some samples and it's devoid of feeling. It is also completely unable to recount similar events in the past. In fact, no actual events are mentioned beyond statistical data. I want to know about fights during a game or the nearly perfect game that got spoiled.

Re:Sentences don't lead into one another

[pipatron]

You mean like BBC News? I just clicked on the first UK article I found to give an example: http://www.bbc.co.uk/news/uk-england-nottinghamshire-11751079

Verb Selection

[nuckfuts]

I've always been amused by how sports reporters vary the verb used to describe a win. They can't just keep saying "Team A beat Team B" over and over, so they mix it up, based on how wide the score was. For a win with a small margin, they might say "Detroit edged Ottawa", or "The Rangers slid past the Ducks". For a large margin, perhaps "The Coyotes pummeled the Blues". I give extra credit if the verb matches the subject, as in "So-and-so doused the Flames".

I think it would be a lot of fun to write a program for this.

Re:Sports Journalism Turing Test ...

[John Hasler]

"The RoboSportReporter is broken again. It looks and smells like someone poured a beer into him."

They were just trying to make him more realistic.