1 David Chen & Raymond Mooney Department of Computer Sciences University of Texas at Austin Learning...

1

David Chen & Raymond MooneyDepartment of Computer Sciences

University of Texas at Austin

Learning to Sportscast: A Test of Grounded Language Acquisition

2

Motivation

• Constructing annotated corpora for language learning is difficult

• Children acquire language through exposure to linguistic input in the context of a rich, relevant, perceptual environment

3

Goals

• Learn to ground the semantics of language

Block

• Learn language through correlated linguistic and visual inputs

4

Challenge

5

Challenge

6

Challenge

A linguistic input may correspond to many possible events

Block ?

??

7

Overview

• Sportscasting task

• Tactical generation

• Strategic generation

• Human evaluation

8

Learning to Sportscast

• Robocup Simulation League games

• No speech recognition– Record commentaries in text form

• No computer vision– Ruled-based system to automatically extract

game events in symbolic form

• Concentrate on linguistic issues

9

Robocup Simulation League

10

Robocup Simulation League

Pink4’s pass was intercepted by Purple6

11

Learning to Sportscast

• Learn to sportscast by observing sample human sportscasts

• Build a function that maps between natural language (NL) and meaning representation (MR)– NL: Textual commentaries about the game– MR: Predicate logic formulas that represent

events in the game

12

Mapping between NL/MR

NL: “Purple3 passes the ball to Purple5”

MR: Pass ( Purple3, Purple5 )

Semantic Parsing (NL MR)

Tactical Generation (MR NL)

13

Robocup Sportscaster TraceNatural Language Commentary Meaning Representation

Purple goalie turns the ball over to Pink8

badPass ( Purple1, Pink8 )

Pink11 looks around for a teammate

Pink8 passes the ball to Pink11

Purple team is very sloppy today

Pink11 makes a long pass to Pink8

Pink8 passes back to Pink11

turnover ( Purple1, Pink8 )

pass ( Pink11, Pink8 )


ballstopped


kick ( Pink11 )

kick ( Pink8)

kick ( Pink11 )

kick ( Pink11 )

kick ( Pink8 )

14












ballstopped


kick ( Pink11 )

kick ( Pink8)

kick ( Pink11 )

kick ( Pink11 )

kick ( Pink8 )

15












ballstopped


kick ( Pink11 )

kick ( Pink8)

kick ( Pink11 )

kick ( Pink11 )

kick ( Pink8 )

16



P6 ( C1, C19 )






P5 ( C1, C19 )

P2 ( C22, C19 )

P2 ( C19, C22 )

P0

P2 ( C19, C22 )

P1 ( C22 )

P1( C19 )

P1 ( C22 )

P1 ( C22 )

P1 ( C19 )

17

Robocup 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).

18

Overview





19

Tactical Generation

• Learn how to generate NL from MR

• Example:

• Two steps1. Disambiguate the training data

2. Learn a language generator

Pass(Pink2, Pink3) “Pink2 kicks the ball to Pink3”

20

System Overview

Purple7 loses the ball to Pink2

Sportscaster Robocup Simulator

Ambiguous Training Data

Pink2 kicks the ball to Pink5


Pink8 shoots the ball

Turnover ( purple7 , pink2 )

Pass ( pink5 , pink8)

Pass ( Purple5, Purple7 )

Kick ( pink2 )Pass ( pink2 , pink5 )Kick ( pink5 )

BallstoppedKick ( pink8 )

21

System Overview









Pass ( Purple5, Purple7 )



Semantic Parser Learner

Initial SemanticParser

22

System Overview









Pass ( purple5, purple7 )



Initial SemanticParser


Unambiguous Training Data

Pink2 kicks the ball to Pink5Pink5 makes a long pass

to Pink8Pink8 shoots the ball Kick ( pink8 )

Pass ( pink2 , pink5 )

Kick ( pink2 )

Kick ( pink5 )

23

System Overview












SemanticParser






Kick ( pink2 )

Kick ( pink5 )


24

System Overview












SemanticParser






Kick ( pink5 )



25

System Overview












SemanticParser






Kick ( pink5 )



26

System Overview












SemanticParser









27

Semantic Parser Learners

• Learn a function from NL to MR

NL: “Purple3 passes the ball to Purple5”

MR: Pass ( Purple3, Purple5 )

Semantic Parsing (NL MR)

Tactical Generation (MR NL)

•We experiment with two semantic parser learners–WASP (Wong & Mooney, 2006; 2007)–KRISP (Kate & Mooney, 2006)

28

WASP: Word Alignment-based 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)

• Capable of both semantic parsing and tactical generation

29

KRISP: Kernel-based Robust Interpretation by Semantic Parsing

• Productions of MR language are treated like semantic concepts

• SVM classifier is trained for each production with string subsequence kernel

• These classifiers are used to compositionally build MRs of the sentences

• More resistant to noisy supervision but incapable of tactical generation

30

Matching

• Ability to find correct NL/MR pair• 4 Robocup championship games from 2001-2004.

– Avg # events/game = 2,613– Avg # sentences/game = 509

• Leave-one-game-out cross-validation• Metric:

– Precision: % of system’s annotations that are correct– Recall: % of gold-standard annotations correctly

produced– F-measure: Harmonic mean of precision and recall

31

Systems

Learner

KRISPER

(Kate & Mooney, 2007)

KRISP

WASPER WASP

WASPER-GEN WASP’s language generator

32

KRISPER and WASPER












SemanticParser






Kick ( pink5 )


(KRISP/WASP)


33

Systems

Learner

KRISPER


KRISP

WASPER WASP


34

WASPER-GEN












TacticalGenerator






Kick ( pink5 )

Tactical Generator Learner(WASP)


35

Matching Results

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

Average results on leave-one-game-out cross-validation

F-m

easu

re randomKRISPERWASPERWASPER-GEN

36

Overview





37

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.

38

Example of Strategic Generation

pass ( purple7 , purple6 )

ballstopped

kick ( purple6 )


ballstopped

kick ( purple2 )


kick ( purple3 )

badPass ( purple3 , pink9 )

turnover ( purple3 , pink9 )

39

Example of Strategic Generation


ballstopped

kick ( purple6 )


ballstopped

kick ( purple2 )


kick ( purple3 )

badPass ( purple3 , pink9 )


40

Strategic Generation

• For each event type (e.g. pass, kick) estimate the probability that it is described by the sportscaster.

• Requires correct NL/MR matching– Use estimated matching from tactical

generation– Iterative Generation Strategy Learning

41

Iterative Generation Strategy Learning (IGSL)

• Directly estimates the likelihood of an event being commented on

• Self-training iterations to improve estimates

• Uses events not associated with any NL as negative evidence

42

Strategic Generation Performance

• Evaluate how well the system can predict which events a human comments on

• Metric:– Precision: % of system’s annotations that are correct

– Recall: % of gold-standard annotations correctly produced

– F-measure: Harmonic mean of precision and recall

43

Strategic Generation Results

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

Average results on leave-one-game-out cross-validation

F-m

easu

re

inferred fromWASPinferred fromKRISPERinferred fromWASPERinferred fromWASPER-GENIGSL

inferred fromgold matching

44

Overview





45

• 4 fluent English speakers as judges• 8 commented game clips

– 2 minute clips randomly selected from each of the 4 games

– Each clip commented once by a human, and once by the machine

• Presented in random counter-balanced order• Judges were not told which ones were human or

machine generated

Human Evaluation (Quasi Turing Test)

46

Demo Clip

• Game clip commentated using WASPER-GEN with IGSL, since this gave the best results for generation.

• FreeTTS was used to synthesize speech from textual output.

47

Human Evaluation

ScoreEnglish 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

48

Human Evaluation

CommentatorEnglishFluency


SportscastingAbility

Human 3.94 4.25 3.63

Machine 3.44 3.56 2.94

Difference 0.5 0.69 0.69

ScoreEnglish Fluency


Sportscasting Ability

5 Flawless Always Excellent

4 Good Usually Good

3 Non-native Sometimes Average

2 Disfluent Rarely Bad

1 Gibberish Never Terrible

49

Future Work

• Expand MRs to beyond simple logic formulas• Apply approach to learning situated language in a

computer video-game environment (Gorniak & Roy, 2005)

• Apply approach to captioned images or video using computer vision to extract objects, relations, and events from real perceptual data (Fleischman & Roy, 2007)

50

Conclusion

• Current language learning work uses expensive, unrealistic training data.

• We have developed a language learning system that can learn from language paired with an ambiguous perceptual environment.

• We have evaluated it on the task of learning to sportscast simulated Robocup games.

• The system learns to sportscast almost as well as humans.

51

Backup Slides

52

Systems

Disambiguation Learning language generator

WASP Random WASP

KRISPER


KRISP N/A

WASPER WASP WASP

KRISPER-WASP KRISP WASP


WASP

WASP with gold matching

N/A WASP

53


WASP Random WASP

KRISPER


KRISP N/A

WASPER WASP WASP



WASP


N/A WASP

Lower baseline

Upper baseline

Systems

54

KRISPER and WASPER












SemanticParser






Kick ( pink5 )


(KRISP/WASP)


55


WASP Random WASP

KRISPER


KRISP N/A

WASPER WASP WASP



WASP


N/A WASP

Lower baseline

Upper baseline

Systems

56

WASPER-GEN












TacticalGenerator






Kick ( pink5 )

Tactical Generator Learner(WASP)


57


WASP Random WASP

KRISPER


KRISP N/A

WASPER WASP WASP



WASP


N/A WASP

Lower baseline

Upper baseline

Systems

58


WASP Random WASP

KRISPER


KRISP N/A

WASPER WASP WASP



WASP


N/A WASP

Lower baseline

Upper baseline

Systems

Matching

59

Matching

• 4 Robocup championship games from 2001-2004.– Avg # events/game = 2,613

– Avg # sentences/game = 509

• Leave-one-game-out cross-validation• Metric:

– Precision: % of system’s annotations that are correct

– Recall: % of gold-standard annotations correctly produced

– F-measure: Harmonic mean of precision and recall

60

Matching Results

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

Average results on leave-one-outexperiments

F-m

easu

re WASPKRISPERWASPERWASPER-GEN

61


WASP Random WASP

KRISPER


KRISP N/A

WASPER WASP WASP



WASP


N/A WASP

Lower baseline

Upper baseline

Systems

62


WASP Random WASP

KRISPER


KRISP N/A

WASPER WASP WASP



WASP


N/A WASP

Lower baseline

Upper baseline

SystemsTactical

Generation

63

Tactical Generation

• 4 Robocup championship games from 2001-2004.– Avg # events/game = 2,613– Avg # sentences/game = 509

• Leave-one-game-out cross-validation• NIST score

– Evaluate the quality of machine translations based on matching n-grams

– BLEU metric with some modifications– More weight for rarer n-grams– Less sensitive to translation length

64

Tactical Generation Results

0

1

2

3

4

5

6

Average results on leave-one-out experiments

NIS

T

WASP

WASPER

KRISPER-WASPWASPER-GEN

WASP with goldmatching

65

Parsing Results

010203040

5060708090

Average results on leave-one-out experiments

F-m

easu

re

WASP

KRISPER

WASPER

KRISPER-WASPWASPER-GEN

WASP with goldmatching

66

Matching Results

67

Parse Results

68

Generation Results

69

Strategic Generation Results

70

Grounded Language Learning in Robocup Robocup Simulator

Sportscaster

Simulated Perception

Perceived Facts

Score!!!!Grounded

Language LearnerLanguageGenerator

SemanticParser

SCFG Score!!!!

71

“Mary is on the phone”

???

72 72

“Mary is on the phone”???

73 73


Ironing(Mommy, Shirt)

74 74



Working(Sister, Computer)

75 75




Carrying(Daddy, Bag)

76 76




Carrying(Daddy, Bag)

Talking(Mary, Phone)

Sitting(Mary, Chair)

77

Grounding Language

“Spanish goalkeeper Iker Casillas blocks the ball”

78

Grounding Language

“Spanish goalkeeper Iker Casillas blocks the ball”

79

Grounding Language

Blocks

• WordNet - (v) parry, block, deflect (impede the movement of (an opponent or a ball)) "block an attack“

• Merriam-Webster - intransitive verb: to block an opponent in sports

80

Grounding Language

Blocks

• WordNet - (v) parry, block, deflect (impede the movement of (an opponent or a ball)) "block an attack“

• Merriam-Webster - intransitive verb: to block an opponent in sports

81

Grounding Language

Blocks

82

Robocup Sportscaster Trace

purple6 passes to purple2

purple3 loses the ball to pink9

purple2 makes a short pass to purple3

ballstopped

kick ( purple6 )



kick ( purple2 )


kick ( purple3 )

Natural Language Commentary Meaning Representation

83








84





kick ( purple6 )

kick ( purple2 )

kick ( purple3 )


85





kick ( purple 3 )

ballstopped

kick ( purple6 )


kick ( purple2 )


kick ( purple2 )


kick ( purple3 )

kick (purple 3


86





kick ( purple 3 )

kick ( purple6 )

kick ( purple2 )

kick ( purple2 )

kick ( purple3 )

kick (purple 3


87





kick ( purple 3 )

kick ( purple6 )

kick ( purple2 )

kick ( purple2 )

kick ( purple3 )

kick (purple 3 )


Negative Evidence

1 David Chen & Raymond Mooney Department of Computer Sciences University of Texas at Austin Learning...

Documents

Transcript of 1 David Chen & Raymond Mooney Department of Computer Sciences University of Texas at Austin Learning...