Guest blog post by Douglas McGarrie, CTO IBM Scotland.
The US Open Tennis Championships have just started. IBM CTO for Scotland, Douglas McGarrie, discusses some of the things IBM is doing with data and technology in tennis… and in particular, how to answer the question, ‘Can a Computer Understand Love?’
Like many, I’ll be settling down over the next couple of weeks to watch the US Open Tennis. I like my tennis (ok, watching rather than playing these days). Impressive athletes for sure, unblemished technique on display, but for me it’s more about the fact that once on court the players have nowhere to hide. So much relies on preparation, practice, hard work . . . and increasingly so on technology. Players and coaches use technology and data both leading up to tournaments and during them. At the championships themselves, the staff, the media, and the watching public all take advantage of the latest and greatest technology to successfully run the events and enrich the overall experience. IBM has been hugely involved in the sport of tennis for the last 25 years, and it’s a great way for us to showcase all of the cool technology that you can use at major sporting events.
Considering data for a moment: IBM will be highlighting new ways to use data at major live sporting events. These events produce rich data streams that can be analysed to provide new insights to fans and this year we are using these streams of data to point out ‘moments of achievement, in real time, such as the player’s fastest serve during the tournament. This information can be shared not only by the broadcasters, calling it out on TV, but also by the social content teams, driving further interest and engagement from fans. Tennis fans love data. One of the applications we have evolved over the last few years is the IBM Slamtracker which has collected 41 million data points, from eight years of grand slam tennis and has identified three ‘keys to the match’ or indicators that most likely to affect a players ability to succeed.
Switching gears just a little, at the other end of the summer, back in June, there was Wimbledon, where up here in Scotland, we watched with envy as people passed out in the heat, wishing that we could have just a little . . . please?
One of the things I liked hearing about at Wimbledon was how IBM was looking at applying its Watson system to the context of tennis. Its still early days in terms of how to use this type of technology for sports, but there are interesting possibilities for the future.
Watson is a cognitive system, so called because of the way it works, being ‘human-like’ and capable of interacting on our terms, rather than those of a computer. Often this requires the system to perform what we call, natural language processing (NLP) this then means the computer can understand language written by a human, to answer questions, or perhaps enter into a conversation.
Last year, we used Watson technologies at Wimbledon to analyse tweets about the championships. To do this, we ingested massive volumes of twitter data in real-time and then applied Watson’s natural language understanding to determine the topics and sentiment (positive or negative comment) of those topics. In this way we were able to see what the global audience was saying about the championships in real-time.
This year we decided to look at something a bit different by first of all training a Watson system on the whole topic of tennis and the championships, such that members of the team could ask questions and to get answers and information. This was known as the ‘Watson Digital Assistant’. It was only used by members of the events team but we anticipate providing it to others in the future as Watson learns more about tennis.
Natural Language Processing was the heart of the original Watson system. When Watson beat the world’s best Jeopardy! players we demonstrated a new phase of sophistication in the way that a computer can understand the English language. The game show brought its own language challenges it’s famous for its unique style of questioning that makes it challenging, even for us humans. Just like Jeopardy!, Wimbledon has its own particular form of the English language, with all of the tennis acronyms and terminology making sure that Watson understands this and learns the language of tennis has been one of our tasks this year. For example, Watson needs to know that in the world of tennis, the word love often doesn’t mean an over-abundance of affection between tennis players, but instead the score of zero.
When we first started analysing the requirements for a Watson Wimbledon system, it quickly became clear that we have two related but distinct sources of information that Watson needed to understand. On the one hand we have some rich sources of what might be considered ‘general knowledge’ about the championships Who was the last English player to win a Singles title? might be an example question (hopefully Watson will not provide the answer, ‘Andy Murray’!). On the other hand, Watson also needs to know and to manipulate a lot of complex statistical data in order to answer questions like In the last 2 matches, what was Roger Federer’s average first serve speed?.
To meet these differing needs, we built a Watson solution in two halves; one half deals with the general knowledge piece, the other half on statistics and numbers. We like to think of this as Watson’s right and left brains. A common user interface across these two halves means the user of the system just enter a question of any form and Watson seamlessly handles the decision on whether to answer with its left or right brain without the user being aware, similar to the way that our human brains operate.
Watson is a hypothesis-based system, which means it first forms a set of hypothesis about what the possible answers to a question might be, then determines its confidence for each answer by examining the supporting evidence for that answer. For Wimbledon, our Watson system passes the user’s question to both the right-brained statistical Watson and the left-brained language Watson and uses the answer confidence levels to determine what is given back to the user.
From answering questions in a quiz show, to understanding the world of healthcare and financial services, Watson is proving its application in business . . . its possibilities are only really limited by our imagination.
And more recently we’ve added a wide variety of other types of data to Watson’s repertoire, perhaps most significantly, imagesincluding photos, medical images and videos. Simply put, we’re teaching Watson to see.
So, for now . . . can a computer understand love? You bet.
Douglas McGarrie, CTO IBM Scotland
An enthusiastic senior IT leader with 25 years experience within the IT industry. Responsible for the technical strategy, solutions and in leading the whole technical community for the benefit of our clients across Scotland. A passion for technology with rich and varied global experience across many industries. Consulting with clients on their IT Strategy whilst delivering the vision of the future of IT Engaging IBM Research in new and exciting ventures with our clients.
Duncan Anderson, CTO, IBM Watson Group