Learning%20Language%20from%20its%20Perceptual%20Context

1
Learning Language from its Perceptual Context

• Ray Mooney
• Department of Computer Sciences
• University of Texas at Austin

Joint work with David Chen Rohit Kate Yuk Wah
Wong
2
Current State of Natural Language Learning

• Most current state-of-the-art NLP systems are
  constructed by training on large supervised
  corpora.
• Syntactic Parsing Penn Treebank
• Word Sense Disambiguation SenseEval
• Semantic Role Labeling Propbank
• Machine Translation Hansards corpus
• Constructing such annotated corpora is difficult,
  expensive, and time consuming.

3
Semantic Parsing

• A semantic parser maps a natural-language
  sentence to a complete, detailed semantic
  representation logical form or meaning
  representation (MR).
• For many applications, the desired output is
  immediately executable by another program.
• Two application domains
• GeoQuery A Database Query Application
• CLang RoboCup Coach Language

4
GeoQuery A Database Query Application

• Query application for U.S. geography database
  Zelle Mooney, 1996

DataBase
5
CLang RoboCup Coach Language

• In RoboCup Coach competition teams compete to
  coach simulated soccer players
• The coaching instructions are given in a formal
  language called CLang

Simulated soccer field
6
Learning Semantic Parsers

• Manually programming robust semantic parsers is
  difficult due to the complexity of the task.
• Semantic parsers can be learned automatically
  from sentences paired with their logical form.

NL?MR Training Exs
Meaning Rep
7
Our Semantic-Parser Learners

• CHILLWOLFIE (Zelle Mooney, 1996 Thompson
  Mooney, 1999, 2003)
• Separates parser-learning and semantic-lexicon
  learning.
• Learns a deterministic parser using ILP
  techniques.
• COCKTAIL (Tang Mooney, 2001)
• Improved ILP algorithm for CHILL.
• SILT (Kate, Wong Mooney, 2005)
• Learns symbolic transformation rules for mapping
  directly from NL to MR.
• SCISSOR (Ge Mooney, 2005)
• Integrates semantic interpretation into Collins
  statistical syntactic parser.
• WASP (Wong Mooney, 2006 2007)
• Uses syntax-based statistical machine translation
  methods.
• KRISP (Kate Mooney, 2006)
• Uses a series of SVM classifiers employing a
  string-kernel to iteratively build semantic
  representations.

?
?
8
WASPA Machine Translation Approach to Semantic
Parsing

• Uses statistical machine translation techniques
• Synchronous context-free grammars (SCFG) (Wu,
  1997 Melamed, 2004 Chiang, 2005)
• Word alignments (Brown et al., 1993 Och Ney,
  2003)
• Hence the name Word Alignment-based Semantic
  Parsing

9
A Unifying Framework for Parsing and Generation
Natural Languages
Machine translation
10
A Unifying Framework for Parsing and Generation
Natural Languages
Semantic parsing
Machine translation
Formal Languages
11
A Unifying Framework for Parsing and Generation
Natural Languages
Semantic parsing
Machine translation
Tactical generation
Formal Languages
12
A Unifying Framework for Parsing and Generation
Synchronous Parsing
Natural Languages
Semantic parsing
Machine translation
Tactical generation
Formal Languages
13
A Unifying Framework for Parsing and Generation
Synchronous Parsing
Natural Languages
Semantic parsing
Compiling Aho Ullman (1972)
Machine translation
Tactical generation
Formal Languages
14
Synchronous Context-Free Grammars (SCFG)

• Developed by Aho Ullman (1972) as a theory of
  compilers that combines syntax analysis and code
  generation in a single phase.
• Generates a pair of strings in a single
  derivation.

15
Synchronous Context-Free GrammarProduction Rule
Natural language
Formal language
QUERY ? What is CITY / answer(CITY)
16
Synchronous Context-Free Grammar Derivation
QUERY
QUERY
What is the capital of Ohio
answer(capital(loc_2(stateid('ohio'))))
STATE ? Ohio / stateid('ohio')
17
Probabilistic Parsing Model
d1
CITY
CITY
capital ( CITY )
capital
CITY
of
STATE
loc_2 ( STATE )
Ohio
stateid ( 'ohio' )
STATE ? Ohio / stateid('ohio')
18
Probabilistic Parsing Model
d2
CITY
CITY
capital ( CITY )
capital
CITY
of
RIVER
loc_2 ( RIVER )
Ohio
riverid ( 'ohio' )
RIVER ? Ohio / riverid('ohio')
19
Probabilistic Parsing Model
d1
d2
CITY
CITY
capital ( CITY )
capital ( CITY )
loc_2 ( STATE )
loc_2 ( RIVER )
stateid ( 'ohio' )
riverid ( 'ohio' )
0.5
0.5
?
?
0.3
0.05
0.5
0.5
STATE ? Ohio / stateid('ohio')
RIVER ? Ohio / riverid('ohio')
1.3
1.05
Pr(d1capital of Ohio) exp( ) / Z
Pr(d2capital of Ohio) exp( ) / Z
normalization constant
20
Overview of WASP
Unambiguous CFG of MRL
Lexical acquisition
Training set, (e,f)
Lexicon, L (an SCFG)
Parameter estimation
SCFG parameterized by ?
Training
Testing
Input sentence, e'
Output MR, f'
Semantic parsing
21
Tactical Generation

• Can be seen as inverse of semantic parsing

The goalie should always stay in our half
Semantic parsing
((true) (do our 1 (pos (half our))))
22
Generation by Inverting WASP

• Same synchronous grammar is used for both
  generation and semantic parsing.

Tactical generation
Semantic parsing
NL
MRL
QUERY ? What is CITY / answer(CITY)
23
Learning Language from Perceptual Context

• Children do not learn language from annotated
  corpora.
• Neither do they learn language from just reading
  the newspaper, surfing the web, or listening to
  the radio.
• Unsupervised language learning
• DARPA Learning by Reading Program
• The natural way to learn language is to perceive
  language in the context of its use in the
  physical and social world.
• This requires inferring the meaning of utterances
  from their perceptual context.

23
24
Language Grounding

• The meanings of many words are grounded in our
  perception of the physical world red, ball, cup,
  run, hit, fall, etc.
• Symbol Grounding Harnad (1990)
• Even many abstract words and meanings are
  metaphorical abstractions of terms grounded in
  the physical world up, down, over, in, etc.
• Lakoff and Johnsons Metaphors We Live By
• Its difficult to put my words into ideas.
• Most NLP work represents meaning without any
  connection to perception circularly defining the
  meanings of words in terms of other words or
  meaningless symbols with no firm foundation.

24
25
Mary is on the phone
26
Ambiguous Supervision for Learning Semantic
Parsers

• A computer system simultaneously exposed to
  perceptual contexts and natural language
  utterances should be able to learn the underlying
  language semantics.
• We consider ambiguous training data of sentences
  associated with multiple potential MRs.
• Siskind (1996) uses this type referentially
  uncertain training data to learn meanings of
  words.
• Extracting meaning representations from
  perceptual data is a difficult unsolved problem.
• Our system directly works with symbolic MRs.

27
Mary is on the phone
28
Mary is on the phone
29
Ironing(Mommy, Shirt)
Mary is on the phone
30
Ironing(Mommy, Shirt)
Working(Sister, Computer)
Mary is on the phone
31
Ironing(Mommy, Shirt)
Carrying(Daddy, Bag)
Working(Sister, Computer)
Mary is on the phone
32
Ambiguous Training Example
Ironing(Mommy, Shirt)
Carrying(Daddy, Bag)
Working(Sister, Computer)
Talking(Mary, Phone)
Sitting(Mary, Chair)
Mary is on the phone
33
Next Ambiguous Training Example
Ironing(Mommy, Shirt)
Working(Sister, Computer)
Talking(Mary, Phone)
???
Sitting(Mary, Chair)
Mommy is ironing a shirt
34
Ambiguous Supervision for Learning Semantic
Parsers (cont.)

• Our model of ambiguous supervision corresponds to
  the type of data that will be gathered from a
  temporal sequence of perceptual contexts with
  occasional language commentary.
• We assume each sentence has exactly one meaning
  in its perceptual context.
• Recently extended to handle sentences with no
  meaning in its perceptual context.
• Each meaning is associated with at most one
  sentence.

35
Sample Ambiguous Corpus
gave(daisy, clock, mouse)
ate(mouse, orange)
Daisy gave the clock to the mouse.
ate(dog, apple)
Mommy saw that Mary gave the hammer to the dog.
saw(mother, gave(mary, dog, hammer))
broke(dog, box)
The dog broke the box.
gave(woman, toy, mouse)
gave(john, bag, mouse)
John gave the bag to the mouse.
threw(dog, ball)
runs(dog)
The dog threw the ball.
saw(john, walks(man, dog))
Forms a bipartite graph
36
KRISPER KRISP with EM-like Retraining

• Extension of KRISP that learns from ambiguous
  supervision.
• Uses an iterative EM-like self-training method to
  gradually converge on a correct meaning for each
  sentence.

37
KRISPERs Training Algorithm
1. Assume every possible meaning for a sentence
is correct
gave(daisy, clock, mouse)
ate(mouse, orange)
Daisy gave the clock to the mouse.
ate(dog, apple)
Mommy saw that Mary gave the hammer to the dog.
saw(mother, gave(mary, dog, hammer))
broke(dog, box)
The dog broke the box.
gave(woman, toy, mouse)
gave(john, bag, mouse)
John gave the bag to the mouse.
threw(dog, ball)
runs(dog)
The dog threw the ball.
saw(john, walks(man, dog))
38
KRISPERs Training Algorithm
1. Assume every possible meaning for a sentence
is correct
gave(daisy, clock, mouse)
ate(mouse, orange)
Daisy gave the clock to the mouse.
ate(dog, apple)
Mommy saw that Mary gave the hammer to the dog.
saw(mother, gave(mary, dog, hammer))
broke(dog, box)
The dog broke the box.
gave(woman, toy, mouse)
gave(john, bag, mouse)
John gave the bag to the mouse.
threw(dog, ball)
runs(dog)
The dog threw the ball.
saw(john, walks(man, dog))
39
KRISPERs Training Algorithm
2. Resulting NL-MR pairs are weighted and given
to KRISP
gave(daisy, clock, mouse)
1/2
ate(mouse, orange)
Daisy gave the clock to the mouse.
1/2
ate(dog, apple)
1/4
1/4
Mommy saw that Mary gave the hammer to the dog.
saw(mother, gave(mary, dog, hammer))
1/4
1/4
broke(dog, box)
1/5
1/5
1/5
The dog broke the box.
gave(woman, toy, mouse)
1/5
1/5
gave(john, bag, mouse)
1/3
1/3
John gave the bag to the mouse.
threw(dog, ball)
1/3
1/3
runs(dog)
1/3
The dog threw the ball.
1/3
saw(john, walks(man, dog))
40
KRISPERs Training Algorithm
3. Estimate the confidence of each NL-MR pair
using the resulting trained parser
gave(daisy, clock, mouse)
ate(mouse, orange)
Daisy gave the clock to the mouse.
ate(dog, apple)
Mommy saw that Mary gave the hammer to the dog.
saw(mother, gave(mary, dog, hammer))
broke(dog, box)
The dog broke the box.
gave(woman, toy, mouse)
gave(john, bag, mouse)
John gave the bag to the mouse.
threw(dog, ball)
runs(dog)
The dog threw the ball.
saw(john, walks(man, dog))
41
KRISPERs Training Algorithm
4. Use maximum weighted matching on a bipartite
graph to find the best NL-MR pairs Munkres,
1957
gave(daisy, clock, mouse)
0.92
ate(mouse, orange)
Daisy gave the clock to the mouse.
0.11
ate(dog, apple)
0.32
0.88
Mommy saw that Mary gave the hammer to the dog.
saw(mother, gave(mary, dog, hammer))
0.22
0.24

broke(dog, box)
0.18
0.71
0.85
The dog broke the box.
gave(woman, toy, mouse)
0.14
0.95
gave(john, bag, mouse)
0.24
0.89
John gave the bag to the mouse.
threw(dog, ball)
0.33
0.97
runs(dog)
0.81
The dog threw the ball.
0.34
saw(john, walks(man, dog))
42
KRISPERs Training Algorithm
4. Use maximum weighted matching on a bipartite
graph to find the best NL-MR pairs Munkres,
1957
gave(daisy, clock, mouse)
0.92
ate(mouse, orange)
Daisy gave the clock to the mouse.
0.11
ate(dog, apple)
0.32
0.88
Mommy saw that Mary gave the hammer to the dog.
saw(mother, gave(mary, dog, hammer))
0.22
0.24
broke(dog, box)
0.18
0.71
0.85
The dog broke the box.
gave(woman, toy, mouse)
0.14
0.95
gave(john, bag, mouse)
0.24
0.89
John gave the bag to the mouse.
threw(dog, ball)
0.33
0.97
runs(dog)
0.81
The dog threw the ball.
0.34
saw(john, walks(man, dog))
43
KRISPERs Training Algorithm
5. Give the best pairs to KRISP in the next
iteration, and repeat until convergence
gave(daisy, clock, mouse)
ate(mouse, orange)
Daisy gave the clock to the mouse.
ate(dog, apple)
Mommy saw that Mary gave the hammer to the dog.
saw(mother, gave(mary, dog, hammer))
broke(dog, box)
The dog broke the box.
gave(woman, toy, mouse)
gave(john, bag, mouse)
John gave the bag to the mouse.
threw(dog, ball)
runs(dog)
The dog threw the ball.
saw(john, walks(man, dog))
44
Results on Ambig-ChildWorld Corpus
45
New ChallengeLearning to Be a Sportscaster

• Goal Learn from realistic data of natural
  language used in a representative context while
  avoiding difficult issues in computer perception
  (i.e. speech and vision).
• Solution Learn from textually annotated traces
  of activity in a simulated environment.
• Example Traces of games in the Robocup simulator
  paired with textual sportscaster commentary.

46
Grounded Language Learning in Robocup
Robocup Simulator
Sportscaster
Score!!!!
Score!!!!
47
Robocup Sportscaster Trace
Natural Language Commentary
Meaning Representation
badPass ( Purple1, Pink8 )
turnover ( Purple1, Pink8 )
Purple goalie turns the ball over to Pink8
kick ( Pink8)
pass ( Pink8, Pink11 )
Purple team is very sloppy today
kick ( Pink11 )
Pink8 passes the ball to Pink11
Pink11 looks around for a teammate
kick ( Pink11 )
ballstopped
kick ( Pink11 )
Pink11 makes a long pass to Pink8
pass ( Pink11, Pink8 )
kick ( Pink8 )
pass ( Pink8, Pink11 )
Pink8 passes back to Pink11
48
Robocup Sportscaster Trace
Natural Language Commentary
Meaning Representation
badPass ( Purple1, Pink8 )
turnover ( Purple1, Pink8 )
Purple goalie turns the ball over to Pink8
kick ( Pink8)
pass ( Pink8, Pink11 )
Purple team is very sloppy today
kick ( Pink11 )
Pink8 passes the ball to Pink11
Pink11 looks around for a teammate
kick ( Pink11 )
ballstopped
kick ( Pink11 )
Pink11 makes a long pass to Pink8
pass ( Pink11, Pink8 )
kick ( Pink8 )
pass ( Pink8, Pink11 )
Pink8 passes back to Pink11
49
Robocup Sportscaster Trace
Natural Language Commentary
Meaning Representation
badPass ( Purple1, Pink8 )
turnover ( Purple1, Pink8 )
Purple goalie turns the ball over to Pink8
kick ( Pink8)
pass ( Pink8, Pink11 )
Purple team is very sloppy today
kick ( Pink11 )
Pink8 passes the ball to Pink11
Pink11 looks around for a teammate
kick ( Pink11 )
ballstopped
kick ( Pink11 )
Pink11 makes a long pass to Pink8
pass ( Pink11, Pink8 )
kick ( Pink8 )
pass ( Pink8, Pink11 )
Pink8 passes back to Pink11
50
Robocup Sportscaster Trace
Natural Language Commentary
Meaning Representation
P6 ( C1, C19 )
P5 ( C1, C19 )
Purple goalie turns the ball over to Pink8
P1( C19 )
P2 ( C19, C22 )
Purple team is very sloppy today
P1 ( C22 )
Pink8 passes the ball to Pink11
Pink11 looks around for a teammate
P1 ( C22 )
P0
P1 ( C22 )
Pink11 makes a long pass to Pink8
P2 ( C22, C19 )
P1 ( C19 )
P2 ( C19, C22 )
Pink8 passes back to Pink11
51
Sportscasting Data

• Collected human textual commentary for the 4
  Robocup championship games from 2001-2004.
• Avg events/game 2,613
• Avg sentences/game 509
• Each sentence matched to all events within
  previous 5 seconds.
• Avg MRs/sentence 2.5 (min 1, max 12)
• Manually annotated with correct matchings of
  sentences to MRs (for evaluation purposes only).

52
WASPER

• WASP with EM-like retraining to handle ambiguous
  training data.
• Same augmentation as added to KRISP to create
  KRISPER.

53
KRISPER-WASP

• First iteration of EM-like training produces very
  noisy training data (gt 50 errors).
• KRISP is better than WASP at handling noisy
  training data.
• SVM prevents overfitting.
• String kernel allows partial matching.
• But KRISP does not support language generation.
• First train KRISPER just to determine the best
  NL?MR matchings.
• Then train WASP on the resulting unambiguously
  supervised data.

54
WASPER-GEN

• In KRISPER and WASPER, the correct MR for each
  sentence is chosen based on maximizing the
  confidence of semantic parsing (NL?MR).
• Instead, WASPER-GEN determines the best matching
  based on generation (MR?NL).
• Score each potential NL/MR pair by using the
  currently trained WASP-1 generator.
• Compute NIST MT score (alternative to BLEU score)
  between the generated sentence and the potential
  matching sentence.

55
Strategic Generation

• Generation requires not only knowing how to say
  something (tactical generation) but also what to
  say (strategic generation).
• For automated sportscasting, one must be able to
  effectively choose which events to describe.

56
Example of Strategic Generation
pass ( purple7 , purple6 ) ballstopped kick (
purple6 ) pass ( purple6 , purple2 )
ballstopped kick ( purple2 ) pass ( purple2 ,
purple3 ) kick ( purple3 ) badPass ( purple3 ,
pink9 ) turnover ( purple3 , pink9 )
57
Example of Strategic Generation
pass ( purple7 , purple6 ) ballstopped kick (
purple6 ) pass ( purple6 , purple2 )
ballstopped kick ( purple2 ) pass ( purple2 ,
purple3 ) kick ( purple3 ) badPass ( purple3 ,
pink9 ) turnover ( purple3 , pink9 )
58
Learning for Strategic Generation

• For each event type (e.g. pass, kick) estimate
  the probability that it is described by the
  sportscaster.
• Requires NL/MR matching that indicates which
  events were described, but this is not provided
  in the ambiguous training data.
• Use estimated matching computed by KRISPER,
  WASPER or WASPER-GEN.
• Use a version of EM to determine the probability
  of mentioning each event type just based on
  strategic info.

59
Iterative Generation Strategy Learning (IGSL)

• Directly estimates the likelihood of commenting
  on each event type from the ambiguous training
  data.
• Uses self-training iterations to improve
  estimates (à la EM).
• Uses events not associated with any NL as
  negative evidence for commenting on that event
  type.

60
Demo

• Game clip commentated using WASPER-GEN with
  EM-based strategic generation, since this gave
  the best results for generation.
• FreeTTS was used to synthesize speech from
  textual output.

61
Experimental Evaluation

• Generated learning curves by training on all
  combinations of 1 to 3 games and testing on all
  games not used for training.
• Baselines
• Random Matching WASP trained on random choice of
  possible MR for each comment.
• Gold Matching WASP trained on correct matching
  of MR for each comment.
• Metrics
• Precision of systems annotations that are
  correct
• Recall of gold-standard annotations correctly
  produced
• F-measure Harmonic mean of precision and recall

62
Evaluating Matching Accuracy

• Measure how accurately various methods assign MRs
  to sentences in the ambiguous training data.
• Use gold-standard matches to evaluate correctness.

63
Results on Matching
64
Evaluating Semantic Parsing

• Measure how accurately learned parser maps
  sentences to their correct meanings in the test
  games.
• Use the gold-standard matches to determine the
  correct MR for each sentence that has one.
• Generated MR must exactly match gold-standard to
  count as correct.

65
Results on Semantic Parsing
66
Evaluating Tactical Generation

• Measure how accurately NL generator produces
  English sentences for chosen MRs in the test
  games.
• Use gold-standard matches to determine the
  correct sentence for each MR that has one.
• Use NIST score to compare generated sentence to
  the one in the gold-standard.

67
Results on Tactical Generation
68
Evaluating Strategic Generation

• In the test games, measure how accurately the
  system determines which perceived events to
  comment on.
• Compare the subset of events chosen by the system
  to the subset chosen by the human annotator (as
  given by the gold-standard matching).

69
Results on Strategic Generation
70
Human Evaluation(Quasi Turing Test)

• Asked 4 fluent English speakers to evaluate
  overall quality of sportscasts.
• Randomly picked a 2 minute segment from each of
  the 4 games.
• Each human judge evaluated 8 commented game
  clips, each of the 4 segments commented once by a
  human and once by the machine when tested on that
  game.
• The 8 clips presented to each judge were shown in
  random counter-balanced order.
• Judges were not told which ones were human or
  machine generated.

71
Human Evaluation Metrics
Score English Fluency Semantic Correctness Sportscasting Ability
5 Flawless Always Excellent
4 Good Usually Good
3 Non-native Sometimes Average
2 Disfluent Rarely Bad
1 Gibberish Never Terrible
72
Results on Human Evaluation
Commentator English Fluency Semantic Correctness Sportscasting Ability
Human 3.94 4.25 3.63
Machine 3.44 3.56 2.94
Difference ?0.5 ?0.69 ?0.69
73
Immediate Future Directions

• Use strategic generation information to improve
  resolution of ambiguous training data.
• Improve WASPs ability to handle noisy training
  data.
• Improve simulated perception to extract more
  detailed and interesting symbolic facts from the
  simulator.

74
Machine Learning Research Direction

• Learning from ambiguous/weak supervision
  (Siskind,1996).
• Multiple Instance Learning assumes weak
  supervision for classification in the form of
  positive bags which contain at least on
  positive instance (Dietterich, et al., 1997)
• Learning with Structured Data assumes I/O are
  complex data structures (e.g. strings or graphs)
  rather than simple vectors and class labels
  (Bakir et al., 2007)
• Need, Structured Multiple Instance Learning
  where an input string is paired with a set of
  possible MRs, one of which is likely to be
  correct.

75
Longer Term Future Directions

• Apply approach to learning situated language in a
  computer video-game environment (Gorniak Roy,
  2005)
• Teach game AIs how to talk to you!
• Apply approach to captioned images or video using
  computer vision to extract objects, relations,
  and events from real perceptual data (Fleischman
  Roy, 2007)

76
Blatant Talk Advertisement

• Watch, Listen Learn Co-training on Captioned
  Images and Videos, S. Gupta, J. Kim, K. Grauman,
  and R. Mooney
• Semi-Supervised Learning session, Thursday,
  1140, R002

77
Conclusions

• Current language learning work uses expensive,
  unrealistic training data.
• We have developed language learning systems that
  can learn from sentences paired with an ambiguous
  perceptual environment.
• We have evaluated it on learning to sportscast
  simulated Robocup games where it learns to
  commentate games almost as well as humans.
• Learning to connect language and perception is an
  important and exciting research problem.