Having seen how journalists will blow hundreds of dollars on proprietary but basic scraping and processing software, I wouldn't be surprised if the software was an adroit use of NLP with OCR, but nothing out-of-this-world fancy. Management of documents and data is important too, so perhaps this software provided a good front end for it?

edit: in response to the downvotes, I'm not saying journalists are dumb, but that the data/technical problems they face are myriad and they often don't get the grounding they need to tackle them, so they, in my opinion, are too quick to pay for commercial software that only solves part of the problem they need to fix. Not just "opinion" here, but actual experience with colleagues and consulting on projects with outside groups. That said, the software they allude to sounds like a system that can mass-process scanned documents, OCR them, and I would assume, use NLP to reduce the amount of work needed to hand clean. But I'm betting that there was still an incredible amount of hand-cleaning that had to be done.