jStart Flash

What makes up Watson?

It's a question we get asked more often than not. Watson is an incredibly complex system--the one which played Jeopardy! consisted of over 100 components written in over a dozen computer programming languages. But in essence, Watson can be broken down into some major sub-systems:

• DeepQA: Many people make the mistake of thinking that Watson is simply a search engine on steroids. It's not. In fact, Watson leverages sophisticated technology called DeepQA to provide answers that search engines, simply, can't: natural language processing, text analytics, and an interaction with the questioner to "hone in" to the answer that the questioner is looking for...as opposed to most search engines which rely on the keyword search paradigm.


• Learning Models: Watson employs a series of learning models to help it combine disparate data together (associate them) as well as weigh the evidence. This helps Watson understand that when someone asks "why do I feel so blue?" the person is referring to an emotional state instead of a color, for instance. The learning models also help Watson learn from its mistakes, since as part of its DeepQA process, feedback is provided that then affects the weight of future queries.


• Machine Learning and Reasoning: Watson employs highly sophisticated machine learning and reasoning algorithms to help it generate hypothesis as well as score the evidence it gathers with regards to a particular query. It does this by comparing multiple interpretations of a question, generate many plausible answers based, partially, on inference paths, and evaluate the answers based on evidence it has within its sources. In other words, there are rudimentary "thinking" components (injest information, analyze and compare with sources, generate answers, weigh those by reasoning, put forth a potential answer based on that reasoning which generates a "best guess", if the system has a high enough degree of confidence in that guess) within Watson that helps it come up with a correct answer, given the nuances in human language and human thought associations.

All of which is great, but what about the 'real world'?

The primary mission of jStart is to take these nacent technologies and test them in the field--using real live clients with real live challenges. And the team has been busy doing just that with Watson. Here are some scenarios we've come up with, working with clients and customers:

1. diagnostic medicine

Imagine a medical center which leveraged Watson to help its practioners answer critical patient care questions in as short a time as possible? The jStart team, along with the Watson Solutions Group has been tackling just such a scenario, leveraging the unique qualities of Watson to help a large healthcare provider determine the best treatment for its patients. In this case, Watson is leveraged as a kind of hyper accelerated research assistant, able to provide obscure (and potentially critical) information to the provider to make the best diagnosis and treatment plan possible.

2. media forecasting

What will make the next movie into a blockbuster hit? It's a question the movie studios have been asking since the advent of, well, movies. And the reason is simple: modern movies can be an incredible investment--hundreds of millions of dollars are "bet" on movie becoming a success. But what if Watson could provide some insights as to what might make a movie a hit--given not only the history, but based partially on current popular sentiment? Watson could never predict the exact ingredients for a successful movie--but it might be able to help the studio understand when things have gone horribly wrong...or right...with a movie.

3. financial analysis

How about a system which provides deep analysis into financial scenarios, helping analysts understand obscure or oblique correlations to existing conditions? Watson's strength is comparing disparate data--which on the face of it may seem completely unrelated--into an overall picture. "Connecting the dots", if you will. Could Watson be leveraged to act as a canary in the coal mine, giving advanced warning of upcoming market turbulence...or market opportunities? These questions are currently being examined by the jStart team in conversations with some of the world's largest financial institutions and governing bodies.

It's a new frontier. And we've just begun.

Throughout 2012, jStart will be working closely with the Watson Solutions Group as we engage additional clients, learn about their challenges, and discover new and unique ways to apply Watson technology to address those needs. In fact, the team will be working on initial engagements for brand new Watson technologies, including scaling Watson for different environments, as well as alternative methods to deliver Watson technologies to those companies which may not need Watson 24/7. The team is working on new ideas for injesting data into Watson, so that the system can "stay current" by processing data in real-time (something that Watson, today, can not do). In fact, there are a number of enhancements that are being considered for Watson, all of which are being weighed by the voice of the market as jStart continues to engage, collaborate, and discover requirements from our clients and customers.

Where are we concentrating our efforts?

It might be easier to say where we are not looking at potential uses of Watson. Although some of the early work on Watson was highly focused on healthcare, the team is looking far beyond healthcare for potential solutions: the financial services industry, media and entertainment, and government are all industry verticals of interest. But, when you think about it, Watson has so many practical applications, that the potential is only limited by the amount of resources jStart (and IBM) can put into the mix. With that in mind, the team has been feverishly working to identify the initial markets to deploy Watson technology, with an eye towards stepping aside and allowing the dedicated Watson Solutions Group expand those markets as Watson matures. In essence, it's exactly what jStart normally does for a new and emerging technology--and the very reason why the team exists as part of IBM's Emerging Technologies group.

SUBSCRIBE TODAY!

Has this e-zine been useful? You can sign up for future editions of jStart's quarterly Flash:

Keep tabs on jStart

About jStart's Flash

jStart's Flash is a newsletter to keep you informed of the latest with IBM Emerging Technology's client engagement team, jStart. Our mission within IBM software group is to explore emerging internet technologies in partnership with our clients and customers.