We've seen proof that robo-journalists can outperform their human sportswriter counterparts, so moving into new professions like commentary seems like a natural progression of sorts. That's the mission Greg Lee has chosen to undertake with a new artificial intelligence system called Scores. Thankfully, rather than carrying the awkward potential of putting real people out of work, Scores is meant to help those in the commentary booth.
Machines have an endless memory bank
Lee believes that when commentators insert stories or anecdotes into their coverage, it makes for a better listening experience. But even your most experienced play-by-play analyst can only recall so much in a given moment — and that's where Scores comes in. The system works by assessing any live sporting event for what's important: which teams are involved, standout players, the current score, and so on. It parses this information against a massive database of past stories in the hopes of finding an interesting comparison to share with those tuning in. Scores narrows the choices down to those that most resemble what's happening on the field before suggesting the results to a human commentator.
Processing all of that data and coming up with an appropriate match is the biggest challenge facing Lee's system; such connections are often obvious to us, but machines don't share the same logic. Yet Scores is making progress. Researches have tested the technology with pre-recorded sports broadcasts and gotten positive feedback from volunteer listeners. It's only a matter of time before your amateur commentator starts sounding more and more like Vin Scully.