Transcript The MIT Artificial Intelligence Lab

The START Information
Access System
Boris Katz
http://www.ai.mit.edu/projects/infolab/
MIT Artificial Intelligence Laboratory — Research Directions
The Problem:
Finding information on line
Two Approaches:
1. Keyword search (search engines, e.g.,
AltaVista)
2. Natural language processing
MIT Artificial Intelligence Laboratory — Research Directions
What’s Wrong with Keyword Search?
MIT Artificial Intelligence Laboratory — Research Directions
What’s Right About Natural Language
Processing?
MIT Artificial Intelligence Laboratory — Research Directions
What’s Wrong with Natural Language
Processing (today)?
1. Too hard
Full-text NL understanding still beyond reach
•Intersentential reference
•Paraphrasing
•Summarization
•Common sense implication
2. Too slow
3. Not all information is language
Most Web resources are not textual
•Maps and Images
•Sound and Video
•Multimedia
Web resources are distributed across
numerous non-traditional databases
MIT Artificial Intelligence Laboratory — Research Directions
What is START?
START (SynTactic Analysis using Reversible Transformations)
provides multimedia information access using natural language.
Natural language
Natural language is human language. You don’t have to learn a special
language to use START. Ask your questions in English; enter information
using English.
Multimedia access using natural language annotations
START lets you use English to access any kind of information: text, pictures,
movies, and more.
“Just the right information”
START gives you the answer you want without including a thousand others.
Virtual collaboration
START retrieves information from its own knowledge base and from
databases all over the Web.
MIT Artificial Intelligence Laboratory — Research Directions
Natural Language
Natural language is human language. You don’t have to learn a special
language to use START. Ask your questions in English; enter information
using English
MIT Artificial Intelligence Laboratory — Research Directions
Multimedia Access Using Natural
Language Annotations
START lets you use English to access any kind of information: text,
pictures, movies, and more.
MIT Artificial Intelligence Laboratory — Research Directions
Just the Right Information
START gives you the answer you want without including a thousand other
answers.
MIT Artificial Intelligence Laboratory — Research Directions
Virtual Collaboration
START retrieves information from its own knowledge base and from
databases all over the Web.
MIT Artificial Intelligence Laboratory — Research Directions
Natural Language Annotations
Bridge the gap between our ability to analyze natural
language sentences and other information and our desire to
access the huge amount of data now available on the Web.
Annotations are collections of natural language sentences
and phrases that describe the content of various information
segments.
START
• analyzes these annotations
• creates the necessary representational structures
• produces special pointers to the information segments
summarized by the annotations.
MIT Artificial Intelligence Laboratory — Research Directions
Natural Language Annotations
Annotation
Document
Xxx xx
xx xxx
xxxx x
Xxx xx xxxx xx
xx xxxxx x xxx
xxx x xxx x xxx
+
Information
Provider
“Neptune was discovered
using mathematics.”
(negotiation)
START
START
START
START
Server
Server
Server
Server
Question
“How was Neptune discovered?”
(submitted)
Information
Seeker
(retrieved)
Document
Xxx xx
xx xxx
xxxx x
Xxx xx xxxx xx
xx xxxxx x xxx
xxx x xxx x xxx
MIT Artificial Intelligence Laboratory — Research Directions
Uniform Access
NL
questions
IMDb
Queries
START
Omnibase
U.S. Census
Data
Multimedia
responses
Fortune500
POTUS
• Local knowledge base of
ternary expressions
• Core vocabulary
• Uniform interface to multiple
database formats (Web, text, etc.)
• Extended lexicon
MIT Artificial Intelligence Laboratory — Research Directions
HPKB
How START Works
Web browser
Omnibase
START
(external
knowledge)
HTML
English
Parser
Input T-exps
English
Scripts
Scripts
Generator
IMDb
Matcher
Database
of T-exps
Annotations
T-exps from KB
Native
knowledge
Potus
U.S. Census
World Factbook
WWW
MIT Artificial Intelligence Laboratory — Research Directions
MIT Artificial Intelligence Laboratory — Research Directions