Last month, legal publisher ALM introduced Legal Radar, a “first-of-its-kind website and app” that uses artificial intelligence and natural language generation to offer faster and more personalized user experiences.
Legal Radar puts the reader in charge, allowing users to select the news they would like to see from a list of relevant industries, practice areas, law firms, companies, and geographic regions, then scrapes information from federal case database PACER to generate automated summaries (usually between 50-80 words) of key details about cases as well as pulling in original ALM content from other channels.
“The newsfeed is filled with short, easy-to-digest news briefs that are intended to be scanned, kind of like the experience you would have on a social media app like Twitter or a news app like Flipboard,” says Vanessa Blum,
Legal Radar represents a significant shift in the way that content is both generated and consumed. Connectiv spoke with Blum about the realities of building an AI-driven content product, how the customer content experience is changing in B2B media and what the rise of AI really means for editors and journalists.
Connectiv: Vanessa, how does the AI component of Legal Radar work?
Vanessa Blum: We start with a stream of raw data from the federal court system via PACER (Public Access to Court Electronic Records). We apply some data processing on the back end in order to normalize, structure and clean up that data. Then it’s converted into short summaries using natural language generation (NLG) technology from a platform called Automated Insights.
It goes in as structured data and it comes out as a readable summary. Then, as the final step, we have editors review the summary for accuracy and to make any edits that are necessary.
Connectiv: The release refers to a “first-of-a-kind website and app.” Can you talk more about what makes this first of a kind and how this offers a new customer experience?
Blum: I’ll talk about two things. First is that user experience. There’s never been a legal news product, certainly not a free legal news product, that is so easy to use on mobile, that can be personalized by user selection and is so seamless to digest information and respond to it. We think we nailed that UX in a way that hasn’t been done in legal media
The second part, which we are really excited about, is the way we are using technology and data processing to generate content for Legal Radar. It’s not the tech in itself, it’s that using technology allows us to be exponentially faster in delivering news to readers and also to deliver news across a wide array of topics and interest areas. I’m really excited about what the technology allows us to do, not only the tech in itself.
Connectiv: Talk about the interaction of the technology with editors. What’s this mean for an editor day-to-day?
Blum: I’ll start with the development process, and how closely our editors and developers worked together in building the back-end system. There are journalistic insights baked into every piece of the data processing engine—it’s the editors who devised how this data should be handled as well as the categories and the tagging that should be applied to it.
And then at the NLG level, these are templates that were created by editors to produce the kind of output that would be useful to readers. They account for over a dozen different fact patterns. It’s not a simple plug-and-play NLG engine, there is really this contribution of journalists and editors throughout the development of Legal Radar. Now that it’s up and running, we have editorial review of every item that’s created. We have staffing around the clock where an editor is looking over each and every item.
We thought that was necessary for two reasons—one is that the data set we are working with can be messy. We knew we needed something on the back end to protect against an error in the data producing an error in the content.
The other component is the ability of a human to enrich the content that we are putting out. These are very short, very fast-paced summaries but if something catches an editor’s interest, they will take an extra step—they will open a case, they will open a lawsuit and add a few key facts. We think it’s incredibly valuable to have the human judgment at the end of that process to resolve any questions or enrich what we are producing using the automated system.
Connectiv: A lot of publishers are taking a look at AI and trying to understand what they can do. As someone who’s successfully built an AI tool, what takeaways ca you share about working with AI and building and AI-driven product?
Blum: I have two main takeaways from this experience: first is to focus on the end user and not the tech. It’s easy to get enraptured by cool tech but the best practice is focus first on what you want to deliver and then focus on how the tool gets you to that result. In my role, learning about new tech and seeing how other companies are applying it is eye opening and can spark that creative process but it’s essential to stay user-focused.
The second thing is to build truly cross-functional teams. Creating Legal Radar required journalists, programmers, product designers and business strategists to all be around the table in a way that was really new for our organization. We tend to have content creators in one area and developers in another. For Legal Radar, content creation and technology are so intertwined that we had to break down the walls and get editors and programmers talking together to solve problems. Not only has that made our product better, it’s made our company better.
Connectiv: What was the biggest strategic takeaway from this experience?
Blum: Staying open minded. When we first started, we had a different data set in mind that we thought we’d be using to produce automated coverage. We learned early on that data set wasn’t workable for us, we had to pivot to something else.
One other thing that I’ll mention, we are working with Automated Insights and it’s a great product, but we found we had to build a lot of solutions at the front end before the data is fed into Automated Insights and at the back end before the content goes into the Legal Radar newsfeed. That’s not something we necessarily anticipated at the outset—how much thought and creativity we’d have to apply both to the data feed going into Automated Insights and how we would handle the content on the back-end.
Connectiv: As the head of newsroom innovation, what are you excited about with content and media? And conversely, what do you think is overrated?
Blum: I’m interested and excited in the combination of human and machine intelligence. I love watching how other news organizations are using technology, using algorithmic journalism, using AI and combining it with the expertise of their journalists to come up with solutions that are incredibly rich. That’s kind of the secret sauce in my view.
In terms of what I think is overhyped, I hate answering that because I’m sure I’ll be back talking about this a year from now, but I will say that smart speakers and developing news products for Alexa. I don’t get that one yet. I’m not convinced we’ll be receiving our information from smart speakers in the near future.
Connectiv: You’ve talked about journalists and AI working together. What’s your reaction to the idea of AI replacing editors and writers?
Blum: That’s the natural fear that people in our industry have as we begin learning about automated journalism. The more I’ve learned about it, the less that fear seems grounded. What technology is capable of is so different from what humans are capable of that it’s really through combining the two that we will see the most exciting advances. Technology is great at processing reams of data very fast, but in the business I’m in, which involves asking questions, exploring trends, talking to insiders, there’s no potential at this point that a machine will take over those functions.
When you combine the speed and data processing capabilities of the technology and turn that over to a human being to do the investigation and talk to real people, that’s where magic happens. I think journalist jobs will change–my own changed dramatically–and journalists will be forced to become more tech-savvy and be more open to using data processing in their work, but the core job of a journalist isn’t going away and cannot be replaced by a computer or an algorithm.