From games shows to the future of medicine

February 2012

IBM WatsonStep aside Deep Blue, Dr Dave Watson FBCS FIET, IBM Director of Emerging Technology, discusses Watson - a question and answer system that shows the future of unstructured data analytics, natural language processing and the design of workload optimised systems.

The massively parallel processing system with an eight-core processor design that has a combined total of 16 Terabytes of memory and can operate at over 80 Teraflops (trillions of operations per second) could be set to change the world of health.

Over the last century, IBM has achieved numerous scientific breakthroughs through its commitment to research and its tradition of Grand Challenges.

These Grand Challenges - such as Deep Blue, which was designed to rival world chess champion Gary Kasparov - work to push science in ways that weren’t thought possible before. IBM Watson is the latest IBM Research Grand Challenge, designed to further the science of natural language processing through advances in question and answer technology.

Jeopardy! The IBM challenge

In 1997, Deep Blue, the computer chess-playing system developed by IBM Research, captured worldwide attention by competing successfully against world chess champion Gary Kasparov. It was the culmination of a grand challenge to advance the science of computing in a way that created great popular interest.

Today, with companies increasingly capturing critical business information in natural language documentation, there is growing interest in workload optimised systems that deeply analyse the content of natural language questions and to answer those questions with precision.

Advances in question answering (QA) technology will increasingly help support professionals in critical and timely decision making in areas such as healthcare, business intelligence, knowledge discovery, enterprise knowledge management, and customer support.

With QA in mind, IBM settled on a challenge to build a computer system called “IBM Watson” (after Thomas J. Watson, the founder of IBM), which could compete at the human-champion level in real-time on the American TV quiz show Jeopardy!

The show, which has been broadcast in the United States for more than 25 years, pits three human contestants against one another to answer rich natural language questions over a broad range of topics, with penalties for wrong answers.

In this three-person competition, confidence, precision and answering speed are of critical importance.

Contestants usually come up with their answers in the few seconds it takes for the host to read a clue. To compete in this game at human-champion levels, a computer system would need to answer roughly 70 percent of the questions asked with greater than 80 percent precision in three seconds or less.

IBM Watson competed against two of the most well-known and successful Jeopardy! champions - Ken Jennings and Brad Rutter - in a two-match contest aired over three consecutive nights beginning on February 14, 2011.

IBM Watson represents an impressive leap forward in systems design and analytics. It runs IBM’s DeepQA technology, a new kind of analytics capability that can perform thousands of simultaneous tasks in seconds to provide precise answers to questions.

Powered by IBM processor technology, IBM Watson is an example of the complex analytics workloads that are becoming increasingly common and critical to business success and competitiveness in today’s data-intensive environment.


DeepQA is a massively parallel probabilistic evidence-based architecture. For the Jeopardy! challenge, more than 100 different techniques are used to analyse natural language, identify sources, find and generate hypotheses, find and score evidence, and merge and rank hypotheses.

Far more important than any particular technique is the way all these techniques are combined in DeepQA such that overlapping approaches can bring their strengths to bear and contribute to improvements in accuracy, confidence, or speed.

DeepQA is an architecture with an accompanying methodology, but it is not specific to the Jeopardy! challenge. IBM has begun adapting it to different business applications and additional exploratory challenge problems including medicine, enterprise search and gaming.

The overarching principles in DeepQA are:

  1. Massive parallelism: Exploit massive parallelism in the consideration of multiple interpretations and hypotheses.
  2. Many experts: Facilitate the integration, application and contextual evaluation of a wide range of loosely coupled probabilistic question and content analytics.
  3. Pervasive confidence estimation: No single component commits to an answer; all components produce features and associated confidences, scoring different question and content interpretations. An underlying confidence processing subsystem learns how to stack and combine the scores.
  4. Integrate shallow and deep knowledge: Balance the use of strict semantics and shallow semantics, leveraging many loosely formed ontologies.

Speed and scale

DeepQA is developed using Apache UIMA, a framework implementation of the unstructured information management architecture. UIMA was designed to support interoperability and scaleability of text and multi-modal analysis applications. All of the components in DeepQA are implemented as UIMA annotators.

These are components that analyse text and produce annotations or assertions about the text. Over time IBM Watson has evolved so that the system now has hundred of components. UIMA facilitated rapid component integration, testing and evaluation.

Early implementations of IBM Watson ran on a single processor, which required two hours to answer a single question. The DeepQA computation is massively parallel so it can be divided into a number of independent parts, each of which can be executed by a separate processor. UIMA-AS, part of Apache UIMA, enables the scaleability of UIMA applications using asynchronous messaging.

IBM Watson uses UIMA-AS to exploit the capabilities of 2,880 POWER7 cores in a cluster of 90 IBM Power 750 servers. Each cluster features 32 3.55 GHz cores running Linux.

The system has a combined total of 16 Terabytes of memory and can operate at over 80 Teraflops (trillions of operations per second). UIMA-AS manages all of the inter-process communication using the open JMS standard. The UIMA-AS deployment on these cores enabled IBM Watson to deliver answers in one to six seconds.

IBM Watson has roughly 200 million pages of natural language content (equivalent to reading 1 million books). IBM Watson uses the Apache Hadoop framework to facilitate pre-processing the large volume of data in order to create in-memory data sets used at run-time.

IBM Watson’s DeepQA UIMA annotators were deployed as mappers in the Hadoop map-reduce framework, which distributed them across processors in the cluster.Hadoop contributes to optimal CPU utilisation and also provides convenient tools for deploying, managing, and monitoring the data analysis process.

Designing IBM Watson on commercially available servers was a deliberate choice to ensure more rapid adoption of optimised systems in industries such as healthcare and financial services.

Key advances and exploitation

The project to build IBM Watson and compete successfully against the human champions of Jeopardy! led to some significant advances in the fields of unstructured data analytics, natural language processing, and the design of workload optimised systems.

It was clear that the technology could be adapted to help solve real-world problems - for example, diagnosing disease, handling on-line technical support questions, and parsing vast tracts of legal documents.

The next question to answer was: ‘Where next to direct IBM Watson's extraordinary capabilities?’

Watson and healthcare

Medicine has moved into the 21st century, but many aspects of healthcare have not. Many medical records are still on paper and are difficult to access.

Relevant research cannot be accessed at a moment’s notice since online content isn’t organised for efficient retrieval. At the same time, the rapidly increasing volume of medical data makes it almost impossible for doctors to keep up to date. (It's estimated that the amount of medical knowledge doubles every five years.)

One of the most difficult tasks a doctor faces is to diagnose conditions and diseases accurately and in a timely manner. For example, it often takes over 10 years before a patient is correctly diagnosed with celiac disease; it takes an average of four years to diagnose multiple sclerosis. Doctors fail to suspect these diseases because they have many symptoms that suggest more common problems.

Doctors have traditionally resisted computerised assistance in diagnosis and treatment because the technology has been awkward to use and questionnaire-based systems have been too rigid. Medical journals, clinical decision support systems and the Internet are helpful, but they have their limits.

The future of medicine

Although still probably two years away, IBM Watson stands to dramatically change the way doctors get information. Work is already underway, the medical faculties at Columbia University and University of Maryland in the US are helping develop an IBM Watson-type capability to assist medical staff.

IBM Watson can “understand” descriptions of a patient’s symptoms in natural language, and it can even scan years of medical records and doctors’ notes to determine what diagnostic and therapeutic options it might suggest. Doctors can ask it questions using the same terms they would use in an email to a colleague.

While this technology could never replace a doctor, it could serve as an invaluable tool for doctors to use. During a diagnosis, IBM Watson might say that the evidence it has points with 88% confidence to the patient having dermatitis, but that there’s also a 6% likelihood she has celiac disease, something the doctor might not have even considered.

IBM Watson’s ability to understand the meaning and context of human language, and rapidly process information to find precise answers to complex questions marks a major milestone for our industry. It holds enormous potential to transform how computers can help people, not just in healthcare, but across all sectors and aspects of our lives.

Comments (1)

Leave Comment
  • 1
    anonymous wrote on 24th Feb 2012

    A significant search engine advance, for professionals, moreover, should be very popular and, conserve effort, if Watson solves the SEO problem, in real time.

    Report Comment

Post a comment