Robot Journalist writes a better story than human sports reporter | Hacker News Reader

Robot Journalist writes a better story than human sports reporter

thenextweb.com

18 pointsEwanToo28 comments

Only because of the horrid sports journalism in the US that reads like a police blotter. I follow some Euro-soccer and was amazed at the difference in quality and breadth of the writing.

Not sure about European sports reporting, but the problem I've found with both the US sports reporting and the Indian sports reporting that I've read is that journalists are allowed to -- perhaps even encouraged to -- use sports jargon without definition. If a tech reporter were to file a story containing undefined three-letter acronyms (CRM, ROI, etc.) and refer to Python without mentioned that it's a "programming language," his or her editor would mark it up and send it back (as she should).

That would be the case if writing for a more general audience... but in a publication whose audience is, say, digital marketing executives, "ROI" would be perfectly fine. Of course, depending on the editorial guidelines that publication follows, it might not be allowed because they don't like acronyms, but not because it needed a definition.

With sports writing, you can't expect definitions for common sports jargon. In F1 racing you have to assume that readers know what "pole position" is, in football (US: soccer) you assume readers understand "offside", and so on. If you were to explain every piece of sporting terminology in every article, it would drive readers crazy.

As to acronyms, while it makes it even more confusing for people who don't understand the sport, the vast majority of the readers are people who do understand it. And for those who don't, it often wouldn't make a difference. If I wasn't a baseball fan, "earned run average" wouldn't make any more sense to me than "ERA".

The biggest difference between American sports and European sports (I'm not familiar with MLS, so I'm not sure if this difference lies with the sport or where it's being played, MLS perhaps fits with European football or perhaps with sports such as baseball) is that American sports use way more terminology and statistics, not just from journalists, but talked about by fans, shown in stadiums, etc. If you watch a US baseball game, and a cricket game (not that cricket is all European, but it is non-US), you'll see that difference hugely.

And that's actually something that fascinates me - I'm English, so such a big focus on stats wasn't a part of sport for me growing up, but once I got into NFL/MLB, I really love it. But, it still seems like something that most sports fans wouldnt care about, to me. Even though I know that they clearly do.

Going off-topic, but what the hell. I grew up memorizing stats on the backs of baseball cards and playing sports video games, so I agree that knowing the numbers is as much a part of my understanding of the game as the actual rules.

What I enjoy about the English football writers I read, is they'll link sport with culture, history, and politics. An upcoming match preview can feel like a history lesson for an entire region. That's something you rarely find over here (though Hunter S. Thompson used to do a sports column for ESPN that covered all kinds of ground). I'd say most Americans get their sports news through highlight shows, but there's always been the tradition of "checking the box score" in the paper.

When soccer does hit mainstream news here, like during the World Cup, most of the conversations seem to be around "how can we fix the game?", ie. goal-line technology, etc. Maybe as a culture we like things that are more concrete, which leads to something like the current NFL, new rules every year so it takes 5 minutes for the officials to sort out a play.

I still think that the number one threat to professional journalists is the low barrier to entry for people to get into journalism these days, not robots. That said, this is really interesting and will likely play a role in generating content from structured data (like baseball stats) at some point in the relatively near future.

My take, as a professional journalist:

The number one threat to professional journalists is the fact that newspaper readership has shifted from print to online, but revenue has not.

Until news companies figure out how to make as much money per online reader as they used to make per print reader, the journalism world is going to continue to be in trouble.

I've never really understood this. The cost of keeping our website up is not even a percent of what it costs to lay out, print and distribute one dead tree issue. We don't need as much money as we once did. It just looks like a reluctance to adjust to me.

Are you including the cost of people? That is, the cost of paying people to maintain the servers, develop the backend and design the frontend?

> Are you including the cost of people? That is, the cost of paying people to maintain the servers, develop the backend and design the frontend?

They're a lot cheaper than people involved in printing paper and distributing it.

Can we put numbers on this? Even ballpark numbers. Because I'm thinking that if you have one system admin, one frontend dev, and one backend dev, that will cost at least $200,000 a year. That assumes that the cost of each person is at least $66,666 a year - that is how much they cost, not how much they are paid. I think that's a very low number, both in terms of number of people and in how much they cost.

Here's a ballpark:

http://ask.metafilter.com/73789/Cost-of-printing-national-ne...

And that's just printing, not counting the physical paper design folk, who also cost money and who actually are needed to design each individual issue (as opposed to website developers, who I maintain are not needed constantly to update a site whose primary product is not software).

Developing the backend and designing the front-end are essentially fixed costs, which amount to little more than the cost of printing a couple of paper issues of a fairly small publication.

Good sites (such as the NY Times) are continually innovating their front and backend.

I know the NYT spends a lot of money on its site, but that's just their choice. It's nice, but usually not strictly necessary for the purpose of delivering news online, and I don't think the NYT's site is really that much better than if you secretly switched its CMS to Brand X. If they want to constantly throw money into that pit and then complain about their financial woes, I guess that's their call, but I would personally prefer to focus on things with a good ROI.

While web production costs may be lower than print-edition production costs ... the cost of editors, reporters, photographers, etc. -- that is, the people who actually produce the content -- remains the same.

Actually, it may even be more, since you've also got videographers and programmers and other content-producing positions that did not exist at most newspapers before the Web ... as well as a 24/7 news cycle, rather than just a one- or two-edition daily cycle.

Your average newspaper does not need huge squads of full-time programmers, sysadmins and videographers. I know several that are transitioning from an over-engineered third-party solution to a $50/mo VPS with WordPress. I'm our only in-house programmer, with freelancers filling in where it's needed. Is it the most amazing website ever? No, but it's capable of delivering the news (24/7, no less). If you want a little less seat-of-your-pants, you'd need maybe one or two more people. The overhead of running a website is substantially less than the overhead of printing ink on paper with expensive equipment and physically driving it all around town.

Yes, you'll still need to pay for reporters, but those have always been dirt cheap. Yes, there will still be costs, but I didn't say it was free — just that the Web is not this absolute harbinger of doom the industry thinks it is. They're just not adapting well to the new environment.

Earlier submission from another publication:

http://news.ycombinator.com/item?id=2457864

Sorry missed that one, you're right yeah.

God this stuff drives me insane. The "robot journalist" wrote a good article because the person who created it PROGRAMMED IT to write a good article! There were no skills inherent to the software that eclipsed the requirement of having a creative human involved in the process. Articles like this don't mean anything.

Actually, articles like this do mean something -- at least to the journalists that the software is already beginning to make fear for their jobs.

As a full-time journalist who has also written some "robot reporter" code -- see "I wrote this article with one mouse click," http://coding.pressbin.com/60/I-wrote-this-article-with-one-... -- I know first-hand that even though it takes a creative human to code the software, once the software is written, it's "set it and forget it" ... which means you no longer need to pay a journalist to write the article, which means bye-bye journalism jobs. (Of course, it also means hello-hello journalist-programmer jobs.)

How long before we have robot programmers?

So what's wrong with the article? I don't really understand with what you are disagreeing.

I'm not sure why it doesn't mean anything, surely the simple point of the software being written is to automate the task?

Software like this can be used to amplify the productivity and creativity of 50 individuals, or indeed to replace 50 individuals each typing out their own report with 1 individual producing 1 computer program to produce the 50 reports.

Do you know what «fire and forget» means?

While the article doesn't mention it, the Narrative Science software and company is a spin-off of this project[1], developed here at Northwestern University as a joint project between the Medill School of Journalism and the AI department.

[1] http://infolab.northwestern.edu/projects/stats-monkey/

Isn't it clear already that robots are disruptive to the human kind(regarding jobs)?

"Damn robots taking our jobs!" - That's what Jesse Jackson will say 10 years from now.

If you're gonna complain about machines 'taking' humans' jobs, you're 300 years late to that party.

This is a single data point. Its best not to draw conclusions from it.

Reply on news.ycombinator.com