Building Watson: An Overview of DeepQA for the  Jeopardy! Challenge thumbnail
Pause
Mute
Subtitles
Playback speed
0.25
0.5
0.75
1
1.25
1.5
1.75
2
Full screen

Building Watson: An Overview of DeepQA for the Jeopardy! Challenge

Published on Sep 28, 20116838 Views

Computer systems that can directly and accurately answer peoples’ questions over a broad domain of human knowledge have been envisioned by scientists and writers since the advent of computers themsel

Related categories

Chapter list

Building Watson: An overview of the DeepQA Project00:00
A Grand Challenge Opportunity00:46
Want to Play Chess or Just Chat?02:07
Easy Questions?04:05
Hard Questions?06:11
The Jeopardy! Challenge07:32
The Possibilities Multiply10:36
Broad Domain13:18
Inducing Frames16:14
Evaluating Possibilities and Their Evidence17:24
Not Just for Fun19:56
Divide and Conquer (Typical in Final Jeopardy!)21:09
The Missing Link22:23
What It Takes to compete against Top Human Jeopardy! Players (1)24:24
What It Takes to compete against Top Human Jeopardy! Players (2)27:44
Early Experiments on Focused Content28:48
A Few Guiding Principles31:33
DeepQA: The Technology Behind Watson32:40
A Methodology for Rapid Advancement38:58
Methodology41:26
Grouping Features to produce Evidence Profiles47:25
Evidence: Time, Popularity, Source, Classification etc.48:07
Evidence: Puns48:54
In-Category Learning51:36
DeepQA: Incremental Progress in Answering Precision on the Jeopardy Challenge52:16
One Jeopardy! question can take 2 hours on a single 2.6Ghz Core53:48
Game Strategy54:59
With Precision, Accurate Confidence and Speed, the rest was History55:32
Run-Time Stack56:19
NLP Components Technology Performance56:20
Potential Business Applications57:00
Evidence Profiles from disparate data is a powerful idea57:23
DeepQA in Continuous Evidence-Based Diagnostic Analysis Symptoms57:46
The Core Technical Team58:17