Machine-generated content and data storytelling will become more common and as content strategists and content creators we might as well wrap or a head around it. And use it to our advantages to create better experiences.
How does data storytelling work?
In a nutshell this model of content creation includes – and I highly over simplify the process – these steps:
- Some kind of data input in a structured format.
- A trigger point that prompts content creation off the available data.
- The content creation and public output.
A example of machine-generated data storytelling content
In the case of my daughter’s softball games, one parent enters the plays as they happen into the Game Changer app.
Anyone with access to that team can watch the live action. Within seconds of the conclusion of the game the Game Changer app with the help of Narrative Science creates a quick and often pretty decent game summary story.
Here’s a look at recent headlines from a tournament that we attended:
The first headline is kind of so-so but the other two headlines I find pretty creative. And descriptive.
Early in my career I actually worked as a sports writer and I can tell you these kind of stories would’ve taken me a lot longer to write than the seconds it’s takes the machine.
The one thing that they do miss are quotes and additional context from the players and coaches. But overall they do a fantastic job telling the game story and tell it quickly.
Here’s an example of one of the stories and I would say it’s pretty well written, minus quotes of course.
One of these games stories written by a machine is certainly better than no game story at all. And let’s be honest who is gonna write them other than the machines? Nobody. But sometimes the stories are also a little too formulaic and not well written. For example take this lead of an article written by a computer:
“Linn-Mar Freshman Lions scores more runs than Cedar Falls, takes victory on Wednesday to the tune of 5-1”
In other words, the one team won because it scored more. I mean that could be used for opening sentence for every sports story. Except for the few games that are tied.
Data storytelling with Narrative Science
Cassidy Shield, vice president of marketing at Narrative Science, joined me on this livestream to discuss data storytelling and how it works in Game Changer. He also mentioned that companies use this automatic data storytelling technique internally as well. Different people want different information and data storytelling can help pull out the right information for each person.
Some content creators can feel threatened by the evolution of this technology. But I think the way you need to look at it is how do you leverage machine-generated content in your overall strategy.
In addition to machines creating content it also can be used already in creating social media posts and in our content creation process. And of course many of us use email automation and other technology-based workflows.
Machine-generated content: Social media posts from your blog or article
I’ve run the following model with writers for years now and it appears that now machines are taking over the same process. Welcome to 2022.
Historical: Writers repurposing blog content as social media posts
The process has been to create the editorial calendar with ideas and then produce the content for the blog in the form of articles – usually 1,000 to 2,000 words.
Once those articles have been approved, writers would write social media posts for the different networks. When the writing has a variety of sentence lengths and structures often authors can just take different pieces for their social media updates.
Why this is better than posting just headlines to social media
Many people – including myself – share links on Twitter to blog posts and other website content that we think our audiences might find interesting.
That, of course, is because it works. People on social media do still click on links – though less than a decade ago.
Only pushing out headlines word for word doesn’t really add much to the social media – especially Twitter – experience. Here’s why: Twitter shows a preview of the headline anyway.
So it’s like we are just repeating ourselves a bit.
Here is one reason why that happens:
When you click the share to tweet button below articles, WordPress automatically grabs the headline and the link. It’s still the default and as far as I can tell there’s no way (at least from mobile) to set it to tweet something else. Sure, people could delete the headline, and write their own tweet, but what reader takes the time for that?
My blog posts are also auto-pushed to Twitter, Facebook and LinkedIn and that function also pulls the headline over.
I can actually change the copy before the post publishes. Even if I don’t change it the default push has been been improved and doesn’t generate the preview link:
It’s something to think about and it’s a good idea to mix up the copy.
What else might the copy say?
- Grab a soundbite from the blog post.
- Write a short teaser and then link.
- Share soundbites of content without linking.
Social media is not just a content delivery service!
It’s something to consider as you are continuing to use social media in your content marketing mix.
It’s worthwhile to use, but make it even more worthwhile to the reader by adding even more value by not just repeating headlines.
How technology can now write social media posts
It’s much easier for writers to draft social posts from the articles they just finished writing.
Some teams have social media editors write them which can take even more time. They first have to familiarize themselves with the article and may even have to run their posts by the writer for an accuracy check.
Now I ran across the technology solution that promised to automagically pull social posts from a submitted website address – (i.e. your article or blog post).
Visit variations.meetedgar.com and simply paste in the URL of your blog post, article or even webpage to get recommendations for posts.
Once I pasted the website address in Meet Edgar gave me several recommendations for social posts based on the content in the article. Here they are:
Not bad at first glance. Had a writer written these would I be approve them? Maybe, leaning more yes than no.
I’m not sure the auto posts always entice readers to click for more but they certainly could work or at the very least as a start.
Technology doesn’t always have to get us there 100% of the way either. For example, I was preparing to speak in Hamburg, Germany, at a content conference. They asked for my bio in German and while I speak German it is much easier to use technology to get most of the way there.
I grabbed my bio from LinkedIn, ran it through Google Translate, made a few updates and shipped it off. Certainly I could’ve translated it from scratch and that is what anyone would’ve done 20 years ago but why not use the technology to help me get there quicker?
This technology here presented by Meet Edgar certainly can help you get closer to where you want to be with social updates and quicker – even if you have to make some updates yourself to what has been written by the machine
Machine learning for content efficiency
Jake Goldman of 10up shared several ways on the Business Storytelling Podcast how WordPress allows content creators to use machine learning.
Tagging and classification of content
Traditionally content creators would add tags themselves and sometimes even type them in. That can lead to multiple versions of the same tag.
In a machine-learning model, the computer will read the content and send back relevant tags. It can also add it to categories.
This can be helpful with large teams, which can end up with several versions of the same tag as people type it in. Categorization can be non-standardized when each content creator picks their own categories. I have a few dozen categories on here and what should go into Workflow or Content Creation or both can be debatable.
A computer taking care of categorization of content can be very helpful here.
Automatic tagging is being used elsewhere for personalization, too. For example, when content gets distributed via Flipboard the service categorizes stories by tags which users follow.
It’s not always perfect. Sometimes a story is tagged “Cedar Rapids” which is where I live but the city wasn’t mentioned in the story.
Other times, it helps stories take off. My story on Bacardi and its ties to Puerto Rico got a ton of readership from the “Puerto Rico” tag. Flipboard correctly showed it there.
Do tags even help with SEO?
I’ve seen tags show up in Google Alerts – like when somebody tagged my name in a post. With more general keywords it might not be as important as it once was, but can help with some.
I would predict that Google is using way more text analysis but it certainly wouldn’t hurt to provide tags. By text analysis I mean they will analyze the text to determine what it should rank for.
It’s similar to meta descriptions. When I reviewed some results I noticed that Google results uses fewer suggested meta descriptions and more of text analysis already. Google then picks what it thinks the meta shown to searchers should be.
Read next: [Podcast] How important are meta descriptions and does Google even use them?
Visual Services
This is where machines tag and identify what’s in an image and even add description of what’s in an image and alt text. Example: “Mountains.”
Alt text is often overlooked and WordPress now reminds creators to add it. Using a machine to add descriptions can help content be found and takes care of an often forgotten strategy.
This will likely get more and more sophisticated. Facebook already tags people based on their images alone. This may very well be something that all content creators can use at some point.
Automatic center-point cropping of images
This is a great tool when it actually works out well. Jake mentioned that images are needed in various formats – thumbnails, featured images, social media previews, etc. A machine can learn how to better center crop them, meaning it uses the center to adjusts the image for different needs.
All these different places and networks have different dimensions for images. I use Adobe Spark to resize at times. While it’s easy, it still takes time. Often that works but a smart machine that can help us crop images better would be nice.
The cropping for now isn’t working well on social media. I shared my event coverage page before on Facebook and this was the thumbnail that was pulled in:
This commenter is right that is not a good crop.
Changing it was harder than it sounds as WordPress holds on to images once uploaded and even when deleted. I finally got it to look better:
Of course content gatherers will have to keep enough margin space so to speak on images when they take them. That could make the cropping easier. The Adobe Summit image is much easier to crop because it has more space around the meat of the image.
Wrap
These are a few more examples of how machine learning and technology can help us make content creation more efficient. As always, people wonder if advances can kill jobs. Sure it does, but… “did spellcheck kill jobs?” Jake asked. “Probably not. Maybe somebody has to do less editing now.”
It makes people’s lives easier and certainly we have to evolve with new technologies. Somebody still has to oversee the strategy and make sure the technology does what it’s supposed to do.
The same is true when it comes to automatic game stories for youth softball. Nobody else would be writing those stories anyway.
Discover more from Christoph’s Content Corner
Subscribe to get the latest posts sent to your email.