Transcript Where have we been and where are we going?

Outline: Where have we been
and were are we going?
•
•
We’re making consistent progress, or
We’re running around in circles, or
–
–
–
–
ACL Meeting
Fred Jelinek
2005
2000
1995
Bob Moore
1990
of data lifts
all boats
1985
No matter what
happens, it’s goin’
be great!
% Statistical
Papers
•
1950s: Empiricism (Information Theory, Behaviorism)
1970s: Rationalism (AI, Cognitive Psychology)
1990s: Empiricism (Data Mining, Statistical NLP, Speech)
2010s: Rationalism (TBD)
100%
We’re going off a cliff…
80%
–
Don’t worry; be happy
60%
40%
20%
Rising tide 0%
Rising Tide of Data Lifts All Boats
If you have a lot of data, then you don’t need a lot of methodology
• 1985: “There is no data like more data”
– Fighting words uttered by radical fringe elements
(Mercer at Arden House)
• 1995: The Web changes everything
• All you need is data (magic sauce)
–
–
–
–
–
–
–
No linguistics
No artificial intelligence (representation)
No machine learning
No statistics
No error analysis
No data mining
No text mining
“It never pays to think until you’ve
run out of data” – Eric Brill
Moore’s Law Constant:
Banko & Brill: Mitigating the Paucity-of-Data Problem (HLT 2001)
Data Collection Rates  Improvement Rates
More
data is
better
data!
Fire everybody and
spend the money on data
Quoted out of context
No consistently
best learner
The rising tide of data will lift all boats!
TREC Question Answering & Google:
What is the highest point on Earth?
The rising tide of data will lift all boats!
Acquiring Lexical Resources from Data:
Dictionaries, Ontologies, WordNets, Language Models, etc.
http://labs1.google.com/sets
England
Japan
Cat
cat
France
Germany
Italy
Ireland
China
India
Indonesia
Malaysia
Dog
Horse
Fish
Bird
more
ls
rm
mv
Spain
Scotland
Belgium
Korea
Taiwan
Thailand
Rabbit
Cattle
Rat
cd
cp
mkdir
Canada
Austria
Australia
Singapore
Australia
Bangladesh
Livestock
Mouse
Human
man
tail
pwd
Applications
•
What good is word sense disambiguation (WSD)?
–
Information Retrieval (IR)
•
Salton: Tried hard to find ways to use NLP to help IR
–
•
•
–
•
•
but failed to find much (if anything)
Croft: WSD doesn’t help because IR is already using those
methods
Sanderson (next two slides)
Machine Translation (MT)
•
•
Original motivation for much of the work on WSD
But IR arguments may apply just as well to MT
What good is POS tagging? Parsing? NLP? Speech?
Commercial Applications of Natural Language
Processing, CACM 1995
–
$100M opportunity (worthy of government/industry’s attention)
1.
2.
•
Don’t worry;
Be happy
Search (Lexis-Nexis)
Word Processing (Microsoft)
Warning: premature commercialization is risky
Sanderson (SIGIR-94)
http://dis.shef.ac.uk/mark/cv/publications/papers/my_papers/SIGIR94.pdf
Not much?
5 Ian Andersons
F
• Could WSD help IR?
• Answer: no
– Introducing ambiguity
by pseudo-words
doesn’t hurt (much)
Query Length (Words)
Short queries matter most, but hardest for WSD
Sanderson (SIGIR-94)
http://dis.shef.ac.uk/mark/cv/publications/papers/my_papers/SIGIR94.pdf
Soft WSD?
F
• Resolving ambiguity
badly is worse than not
resolving at all
– 75% accurate WSD
degrades performance
– 90% accurate WSD:
breakeven point
Query Length (Words)
Some Promising Suggestions
(Generate lots of conference papers, but may not support the field)
• Two Languages are
Better than One
– For many classic hard NLP
problems
• Word Sense
Disambiguation (WSD)
• PP-attachment
• Conjunction
• Predicate-argument
relationships
• Japanese and Chinese
Word breaking
– Parallel corpora  plenty
of annotated (labeled)
testing and training data
– Don’t need unsupervised
magic (data >> magic)
• Demonstrate that NLP is good
for something
– Statistical methods (IR &
WSD) focus on bags of nouns,
• Ignoring verbs, adjectives,
predicates, intensifiers, etc.
– Hypothesis: Ignored because
perceptrons can’t model XOR
– Task: classify “comments” into
“good,” “bad” and “neutral”
• Lots of terms associated with
just one category
• Some associated with two
– Depending on argument
• Good & Bad, but not neutral:
Mickey Mouse, Rinky Dink
– Bad: Mickey Mouse(us)
– Good: Mickey Mouse(them)
– Current IR/WSD methods
don’t capture predicateargument relationships
Web Apps: Document Language
Model ≠ Query Language Model
• Documents
–
–
–
–
Function Words
Adjectives
Verbs
Predicates
• Queries
–
–
–
–
–
Typos
Brand Names
Celebrities
Named Entities
Slower Vocab Growth
Technical Op: Reduce IR to Translation
Promising Apps: Web Spam, Frame Problem
Speech Data Mining
& Call Centers:
An Intelligence Bonanza
• Some companies are collecting
information with technology
designed to monitor incoming calls
for service quality.
• Last summer, Continental Airlines
Inc. installed software from
Witness Systems Inc. to monitor
the 5,200 agents in its four
reservation centers.
• But the Houston airline quickly
realized that the system, which
records customer phone calls and
information on the responding
agent's computer screen, also was
an intelligence bonanza, says
André Harris, reservations training
and quality-assurance director.
Speech Data Mining
• Label calls as success or failure based on
some subsequent outcome (sale/no sale)
• Extract features from speech
• Find patterns of features that can be used
to predict outcomes
• Hypotheses:
– Customer: “I’m not interested”  no sale
– Agent: “I just want to tell you…”  no sale
Inter-ocular effect (hits you between the eyes);
Don’t need a statistician to know which way the wind is blowing
Outline
• We’re making consistent progress, or
• We’re running around in circles, or
– Don’t worry; be happy
• We’re going off a cliff…
According to unnamed sources:
Speech Winter  Language Winter
Dot Boom  Dot Bust
Sample of 20 Survey Questions
(Strong Emphasis on Applications)
• When will
– More than 50% of new PCs have dictation on them, either at
purchase or shortly after.
– Most telephone Interactive Voice Response (IVR) systems
accept speech input.
– Automatic airline reservation by voice over the telephone is the
norm.
– TV closed-captioning (subtitling) is automatic and pervasive.
– Telephones are answered by an intelligent answering machine
that converses with the calling party to determine the nature and
priority of the call.
– Public proceedings (e.g., courts, public inquiries, parliament,
etc.) are transcribed automatically.
• Two surveys of ASRU attendees: 1997 & 2003
$
Hockey Stick
Business Case
2003
Last
Year
2004
This
Year
t
2005
Next
Year
2003 Responses ≈ 1997 Responses + 6 Years
(6 years of hard work  No progress)
Wrong Apps?
• New Priorities
– Increase demand for
space >> Data entry
• New Killer Apps
– Search >> Dictation
• Speech Google!
– Data mining
• Old Priorities
– Dictation app dates back to
days of dictation machines
– Speech recognition has not
displaced typing
• Speech recognition has
improved
• But typing skills have
improved even more
– My son will learn typing in
1st grade
– Sec rarely take dictation
– Dictation machines are history
• My son may never see one
• Museums have slide rulers
and steam trains
– But dictation machines?
Borrowed Slide: Jelinek (LREC)
Great Strategy  Success
Great Challenge: Annotating Data
• Produce annotated data with minimal
supervision Self-organizing “Magic” ≠ Error Analysis
• Active learning
– Identify reliable labels
– Identify best candidates for annotation
• Co-training
• Bootstrap (project) resources from one
application to another
Grand Challenges
ftp://ftp.cordis.lu/pub/ist/docs/istag040319-draftnotesofthemeeting.pdf
Roadmaps: Structure of a Strategy
(not the union of what we are all doing)
•
Goals
– Example: Replace keyboard with
microphone
– Exciting (memorable) sound bite
– Broad grand challenge that we
can work toward but never solve
•
Metrics
– Examples:
•
– Quantity is not a good thing
– Awareness
– 1-slide version
• if successful, you get maybe 3
more slides
•
– Easy to measure
•
• Mostly for next year: Q1-4
• Plus some for years 2, 5, 10 & 20
Milestones
– Should be no question if it has
been accomplished
– Example: reduce WER on task x
by y% by time t
Accomplishments v. Activities
– Accomplishments are good
– Activity is not a substitute for
accomplishments
– Milestones look forward whereas
accomplishments look backward
• Serendipity is good!
Size of container
– Goal: 1-3
– Metrics: 3
– Milestones: a dozen
• WER: word error rate
• Time to perform task
•
Small is beautiful
– Accomplishments: a dozen
•
Broad applicability & illustrative
– Don’t cover everything
– Highlight stuff that
• Applies to multiple groups
• Forward-Looking / Exciting
Goals:
1. The multilingual companion
2. Life log
Grand Challenges
Goal: Produce NLP apps
that improve the way
people communicate
with one another
Goal: Reduce
barriers to entry
€€€
Apps &
Techniques
Resources
Evaluation
Substance: Recommended if…
Summary: What Worked
and What Didn’t? What’s the right
answer?
•
Data
–
Stay on msg: It is the data, stupid!
•
•
WVLC (Very Large) >> EMNLP (Empirical Methods)
If you have a lot of data,
–
•
•
Then you don’t need a lot of methodology
Rising Tide of Data Lifts All Boats
Methodology
–
Empiricism means different things to different people
1.
2.
3.
–
There’ll be a
quiz at the end
of the decade…
Machine Learning (Self-organizing Methods)
Exploratory Data Analysis (EDA)
Corpus-Based Lexicography
Lots of papers on 1
•
•
Magic: Recommended if…
EMNLP-2004 theme (error analysis)  2
Senseval grew out of 3
Promise: Recommended if…
Short term ≠ Long term
Lonely