Seman&c Learning. Hanna Hajishirzi. (Some slides taken from semantic parsing tutorial)

Size: px
Start display at page:

Download "Seman&c Learning. Hanna Hajishirzi. (Some slides taken from semantic parsing tutorial)"

Transcription

1 Seman&c Learning Hanna Hajishirzi (Some slides taken from semantic parsing tutorial)

2 Outline Semantics Semantic Parsing Grounded Language Acquisition

3 Seman&c Parsing Texas borders Kansas. next-to(tex,kan) Texas borders Kansas. What states border Texas? next-to(tex,kan) λx.state(x) next-to(x,tex)

4 Seman&c Parsing Seman&c Parsing: Transforming natural language (NL) sentences into computer executable complete meaning representa&ons (MRs) for domain- specific applica&ons Realis&c seman&c parsing currently entails domain dependence Example applica&on domains ATIS: Air Travel Informa&on Service CLang: Robocup Coach Language Geoquery: A Database Query Applica&on

5 ATIS: Air Travel Informa&on Service Interface to an air travel database [Price, 1990] Widely- used benchmark for spoken language understanding May I see all the Nlights from Cleveland to Dallas? Semantic Parsing Air- Transportation Show: (Flight- Number) Origin: (City "Cleveland") Destination: (City "Dallas") Query NA 1439, TQ 23,

6 CLang: RoboCup Coach Language In RoboCup Coach compe&&on teams compete to coach simulated players [h4p:// The coaching instruc&ons are given in a computer language called CLang [Chen et al. 2003] If the ball is in our goal area then player 1 should intercept it. Simulated soccer field Semantic Parsing (bpos (goal-area our) (do our {1} intercept)) CLang

7 Geoquery: A Database Query Applica&on Query applica&on for U.S. geography database containing about 800 facts [Zelle & Mooney, 1996] Which rivers run through the states bordering Texas? Semantic Parsing answer(traverse(next_to(stateid( texas )))) Arkansas, Canadian, Cimarron, Gila, Mississippi, Rio Grande Query Answer

8 Meaning Representa&on Languages Meaning representa&on language (MRL) for an applica&on is assumed to be present MRL is designed by the creators of the applica&on to suit the applica&ons needs independent of natural language MRL is unambiguous by design

9 Engineering Mo&va&on for Seman&c Parsing Applica&ons of domain- dependent seman&c parsing Natural language interfaces to compu&ng systems Communica&on with robots in natural language Personalized so_ware assistants Ques&on- answering systems Machine learning makes developing seman&c parsers for specific applica&ons more tractable

10 Dis&nc&ons from Other NLP Tasks Shallow seman&c processing Informa&on extrac&on Seman&c role labeling Intermediate linguis&c representa&ons Part- of- speech tagging Syntac&c parsing Seman&c role labeling Output meant for humans Ques&on answering Summariza&on Machine transla&on

11 Dis&nc&ons from Other NLP Tasks: Deeper Seman&c Analysis Show the long Alice sent me yesterday Semantic Parsing

12 Relations to Other NLP Tasks Tasks being performed within semantic parsing Word sense disambiguation Syntactic parsing as dictated by semantics Tasks closely related to semantic parsing Machine translation Natural language generation

13 Relations to Other NLP Tasks: Word Sense Disambiguation Semantic parsing includes performing word sense disambiguation Which rivers run through the states bordering Mississippi? Semantic Parsing State? River? answer(traverse(next_to(stateid( mississippi ))))

14 Rela&ons to Other NLP Tasks: Machine Transla&on The MR could be looked upon as another NL [Papineni et al., 1997; Wong & Mooney, 2006] Which rivers run through the states bordering Mississippi? answer(traverse(next_to(stateid( mississippi ))))

15 Rela&ons to Other NLP Tasks: Natural Language Genera&on Reversing a seman&c parsing system becomes a natural language genera&on system [Jacobs, 1985; Wong & Mooney, 2007a] Which rivers run through the states bordering Mississippi? Semantic Parsing NL Generation answer(traverse(next_to(stateid( mississippi ))))

16 Outline Semantics Semantic Parsing Grounded Language Acquisition

17 Seman&c Learning Texas borders Kansas. What states border Texas? next-to(tex,kan) λx.state(x) next-to(x,tex) Machine Learning Problem: Given: many input, output pairs Learn: a func&on that maps sentences to meaning representa&on language

18 Learning Semantic Parsers Training Sentences & Meaning Representa&ons Seman&c Parser Learner Novel sentence Seman&c Parser Meaning Representa&on

19 Semantic Parsing using CCG Zettlemoyer & Collins (2005)

20 Combinatory Categorial Grammar (CCG) Highly structured lexical entries A few general parsing rules (Steedman, 2000; Steedman & Baldridge, 2005) Each lexical entry is a word paired with a category Texas := NP borders := (S \ NP) / NP Mexico := NP New Mexico := NP

21 Parsing Rules (Combinators) Describe how adjacent categories are combined Functional application: A / B B A (>) B A \ B A (<) Forward composition Backward composition Texas borders New Mexico NP (S \ NP) / NP NP S S \ NP > <

22 CCG for Semantic Parsing Extend categories with semantic types (Lambda calculus expressions) Texas := NP : texas borders := (S \ NP) / NP : λx.λy.borders(y, x) Lexicon Lexical'Entry Category Text Syntax λ

23 Sample CCG Derivation Texas borders New Mexico NP texas (S \ NP) / NP λx.λy.borders(y, x) NP new_mexico S \ NP λy.borders(y, new_mexico) S borders(texas, new_mexico) > <

24 Another Sample CCG Derivation Texas touches New Mexico NP texas (S \ NP) / NP λx.λy.borders(y, x) NP new_mexico S \ NP λy.borders(y, new_mexico) S borders(texas, new_mexico) > <

25 'Probabilis=c'CCGs Probabilis&c CCGs Lexicon: Parameters: Λ= Texas NP tex, θ Probability'distribu=on:''sentence x, parse y,!logical'form z Log?linear'model: Parsing: ed as: P (y, z x;, )= e (x,y,z) (y,z ) e (x,y,z ) on 7 defines the features used in the ex f(x) = arg max p(z x;, ) z where p(z x;, )= y p(y, z x;, )

26 Learning Probabilistic CCG Training Sentences & Logical Forms Lexical Generation Lexicon L Parameter Estimation Sentences CCG Parser Feature weights w Logical Forms

27 Lexical Generation Input sentence: Texas borders New Mexico Output substrings: Texas borders New Mexico Texas borders borders New New Mexico Texas borders New Input logical form: borders(texas, new_mexico) Output categories:

28 Input Trigger constant c arity one predicate p arity one predicate p Category Rules Output Category NP : c N : λx.p(x) S \ NP : λx.p(x) arity two predicate p (S \ NP) / NP : λx.λy.p(y, x) arity two predicate p (S \ NP) / NP : λx.λy.p(x, y) arity one predicate p arity two predicate p and constant c arity two predicate p arity one function f arity one function f N / N : λg.λx.p(x) g(x) N / N : λg.λx.p(x, c) g(x) (N \ N) / NP : λx.λg.λy.p(y, x) g(x) NP / N : λg.argmax/min(g(x), λx.f(x)) S / NP : λx.f(x)

29 Input sentence: Texas borders New Mexico Output substrings: Texas borders New Mexico Texas borders borders New New Mexico Texas borders New Lexical Generation Input logical form: borders(texas, new_mexico) Output categories: NP : texas NP : new _mexico (S \ NP) / NP : λx.λy.borders(y, x) (S \ NP) / NP : λx.λy.borders(x, y) Take cross product to form an initial lexicon, include some domain independent entries What S/(S\NP)/N : λf.λg.λx.f(x) g(x)

30 Parameter Estimation A lexicon can lead to mul&ple parses, parameters decide the best parse Maximum condi&onal likelihood: Itera&vely find parameters that maximizes the probability of the training data Deriva&ons d are not annotated, treated as hidden variables Keep only those lexical items that occur in the highest scoring deriva&ons of training set

31 A Machine Transla&on Approach to Seman&c Parsing Wong & Mooney (2006) Based on a seman&c grammar of the natural language Uses machine transla&on techniques Synchronous context- free grammars (SCFG) Word alignments (Brown et al., 1993)

32 Seman&c Parsing Grammar rules are extracted from word alignments between training sentences and their meaning representa&ons Which rivers run through the states bordering Mississippi? answer(traverse(next_to(stateid( mississippi ))))

33 Semantic Parsing Semantic parsing maps NL sentences to completely formal MRs. Semantic parsers can be effectively learned from supervised corpora consisting of only sentences paired with their formal MRs (and possibly also SAPTs). The state-of-the-art in semantic parsing has been significantly advanced in recent years using a variety of statistical machine learning techniques and grammar formalisms

34 Domain- Specific Seman&c Parsing Generating training corpora is expensive Use domain knowledge to semantic parsing RoboCup Soccer domain (Hajishirzi et al., 2011)

35 Domain Knowledge Variables Team Player. Constants pink: team purple: team Pink1: player. Predicates Holding(Pink1, ball) AtCorner(ball) AtOffside(ball) Ac&ons Pass(Player1, Player2) BadPass(Player1, Player2) Kick(Player1) Dribble(Player1) Pass(Player1, Player2) Precondi&ons: Effects: Holding(Player1, ball) Holding(Player2, ball) 35

36 Seman&c Parsing with Domain Knowledge Map to events specinied in a logical format Pink7 makes a bad pass that was picked off by Purple7 Purple7 passes forward to Purple9 Purple9 kicks to Purple10 badpass kick kick Deeper understanding of narratives Uses domain knowledge in narrative understanding 36

37 Map Narra&ves to Events Challenge: Long narratives Inconsistent consecutive sentences No labeled data Solution: Use prior knowledge about event descriptions Use iterative learning Use consistency checking 37

38 Sentences sent1: P7 passes forward to P9. sent2: P9 kicks the ball to P10. Prior Knowledge Pass(x,y): Pre: holding(x) Eff:holding(y) Steal(x): Pre: ~holding(x) Eff: holding(x) <sent1, steal(p7,p9), true> Examples(sentence,event, state) <sent2, kick(p9), holding(p7)> Vote Generato r Check whether the event can happen in the current state sent1,steal(p7,p9),true + sent2,kick(p9),holding(p7) - Reasoning New weights ClassiNier P(event sentence,state) 38

39 Map Narra&ves to Event Sequences Find the most likely path badpass kick kick 39

40 Outline Semantics Semantic Parsing Grounded Language Acquisition

41 Grounded Language Acquisi&on Collecting Domain SpeciNic Knowledge is not always feasible Children acquire language through exposure to linguistic input in the context of a rich, relevant, perceptual environment

42 Grounded Seman&cs Learn to ground the semantics of language Block Learn language through correlated linguistic and world state

43 Grounded Seman&cs Game Events World Cup Live 94' 43 Essien is penalised for a challenge on Fabregas in midnield. Time! Type! Quali;ier! Team! Player! 2592! Pass! Throw In! Chelsea! Kalou! 2594! Tackle! Long Ball! Arsenal! Clichy! 2594! Disposses s! Alex Song celebrates giving Arsenal a first-ha Through Ball! Chelsea! Essien! 2596! foul! Through Ball! Chelsea! Essien! 2603! pass! Free Kick! Arsenal! Fabregas! 2605! Pass! Chipped! Arsenal! Clichy! 2609! Pass! Through Ball! Arsenal! Koscielny! 2615! Pass! Long Ball! Arsenal! Song! 2623! Pass! Head! Arsenal! Nasri! Grounded language acquisition: Learn to correspond text to a part of world state

44 Challenges 1. Not all events are mentioned in the sentence Commentary Game Events World Cup Live 94' 44 Song began the move, exchanged with Wilshere on the edge of the box before stealing the ball off Fabregas' foot and sliding it home with his left foot from ten yards out. What a fantastic goal. Time! Type! Quali;ier! Team! Player! Alex Song celebrates giving Arsenal a first-ha 2603! pass! Free Kick! Arsenal! Fabregas! 2605! Pass! Chipped! Arsenal! Clichy! 2609! Pass! Through Ball! Arsenal! Koscielny! 2615! Pass! Long Ball! Arsenal! Song! 2623! Pass! Head! Arsenal! Nasri! 2625! Pass! Through Ball! Arsenal! Song! 2627! Pass! Through Ball! Arsenal! Wilshere! 2629! Pass! Through Ball! Arsenal! Fabregas! 2630! Goal! Through Ball! Arsenal! Song!

45 Challenges 2. Some sentences correspond to no event Statistics, game analysis 44 Song began the move, exchanged with Wilshere on the edge of the box before stealing the ball off Fabregas' foot and sliding it home with his left foot from ten yards out. What a fantastic goal. Time! Type! Quali;ier! Team! Player! 2603! pass! Free Kick! Arsenal! Fabregas! 2605! Pass! Chipped! Arsenal! Clichy! 2609! Pass! Through Ball! Arsenal! Koscielny! 2615! Pass! Long Ball! Arsenal! Song! 2623! Pass! Head! Arsenal! Nasri! 2625! Pass! Through Ball! Arsenal! Song! 2627! Pass! Through Ball! Arsenal! Wilshere! 2629! Pass! Through Ball! Arsenal! Fabregas! 2630! Goal! Through Ball! Arsenal! Song!

46 Challenges Hand- annotated alignment is too expensive Commentary Game Events World Cup Live 94' 44 Song began the move [4], exchanged with Wilshere on the edge of the box [6,7] before stealing the ball off Fabregas' foot [8] and sliding it home with his left foot from ten yards out [9]. What a fantastic goal Time! Type! Quali;ier! Team! Player! Alex Song celebrates giving Arsenal a first-ha 2603! pass! Free Kick! Arsenal! Fabregas! 2605! Pass! Chipped! Arsenal! Clichy! 2609! Pass! Through Ball! Arsenal! Koscielny! 2615! Pass! Long Ball! Arsenal! Song! 2623! Pass! Head! Arsenal! Nasri! 2625! Pass! Through Ball! Arsenal! Song! 2627! Pass! Through Ball! Arsenal! Wilshere! 2629! Pass! Through Ball! Arsenal! Fabregas! 2630! Goal! Through Ball! Arsenal! Song!

47 Weak Supervision Commentary Game Events World Cup Live 94' 43 Essien penalised for a challenge on Fabregas in midnield. 44 Song began the move, exchanged with Wilshere on the edge of the box before stealing the ball off Fabregas' foot and sliding it home with his left foot from ten yards out. What a fantastic goal. Time! Type! Quali;ier! Team! Player! 2592! Pass! Throw In! Chelsea! Kalou! 2594! Tackle! Long Ball! Arsenal! Clichy! 2594! Disposses s! Alex Song celebrates giving Arsenal a first-ha Through Ball! Chelsea! Essien! 2596! foul! Through Ball! Chelsea! Essien! 2603! pass! Free Kick! Arsenal! Fabregas! 2605! Pass! Chipped! Arsenal! Clichy! 2609! Pass! Through Ball! Arsenal! Koscielny! 2615! Pass! Long Ball! Arsenal! Song! 2623! Pass! Head! Arsenal! Nasri! 2625! Pass! Through Ball! Arsenal! Song! 2627! Pass! Through Ball! Arsenal! Wilshere! 2629! Pass! Through Ball! Arsenal! Fabregas! 2630! Goal! Through Ball! Arsenal! Song!

48 Robocup Sportscaster Trace Natural Language Commentary Purple goalie turns the ball over to Pink8 Purple team is very sloppy today Pink8 passes the ball to Pink11 (Chen et.al 07,08) Meaning Representation badpass ( Purple1, Pink8 ) turnover ( Purple1, Pink8 ) kick ( Pink8) pass ( Pink8, Pink11 ) kick ( Pink11 ) Pink11 looks around for a teammate Pink11 makes a long pass to Pink8 Pink8 passes back to Pink11 kick ( Pink11 ) ballstopped kick ( Pink11 ) pass ( Pink11, Pink8 ) kick ( Pink8 ) pass ( Pink8, Pink11 )

49 Robocup Sportscaster Trace Natural Language Commentary Purple goalie turns the ball over to Pink8 Purple team is very sloppy today Pink8 passes the ball to Pink11 Meaning Representation badpass ( Purple1, Pink8 ) turnover ( Purple1, Pink8 ) kick ( Pink8) pass ( Pink8, Pink11 ) kick ( Pink11 ) Pink11 looks around for a teammate Pink11 makes a long pass to Pink8 Pink8 passes back to Pink11 kick ( Pink11 ) ballstopped kick ( Pink11 ) pass ( Pink11, Pink8 ) kick ( Pink8 ) pass ( Pink8, Pink11 )

50 System Overview Sportscaster Robocup Simulator Purple7 loses the ball to Pink2 Pink2 kicks the ball to Pink5 Pink5 makes a long pass to Pink8 Pink8 shoots the ball Pass ( Purple5, Purple7 ) Turnover ( purple7, pink2 ) Kick ( pink2 ) Pass ( pink2, pink5 ) Kick ( pink5 ) Pass ( pink5, pink8) Ballstopped Kick ( pink8 ) Ambiguous Training Data

51 System Overview Sportscaster Robocup Simulator Purple7 loses the ball to Pink2 Pink2 kicks the ball to Pink5 Pink5 makes a long pass to Pink8 Pink8 shoots the ball Ambiguous Training Data Pass ( Purple5, Purple7 ) Turnover ( purple7, pink2 ) Kick ( pink2 ) Pass ( pink2, pink5 ) Kick ( pink5 ) Pass ( pink5, pink8) Ballstopped Kick ( pink8 ) Initial Semantic Parser Semantic Parser Learner

52 System Overview Sportscaster Robocup Simulator Purple7 loses the ball to Kick ( pink2 ) Pink2 Pink2 kicks the ball to Pink5 Pass ( pink2, pink5 ) Pink5 makes a long pass to Pink8 Kick ( pink5 ) Pink8 shoots the ball Kick ( pink8 ) Unambiguous Training Data Purple7 loses the ball to Pink2 Pink2 kicks the ball to Pink5 Pink5 makes a long pass to Pink8 Pink8 shoots the ball Pass ( purple5, purple7 ) Turnover ( purple7, pink2 ) Kick ( pink2 ) Pass ( pink2, pink5 ) Kick ( pink5 ) Pass ( pink5, pink8) Ballstopped Kick ( pink8 ) Initial Semantic Parser Ambiguous Training Data

53 System Overview Sportscaster Robocup Simulator Purple7 loses the ball to Kick ( pink2 ) Pink2 Pink2 kicks the ball to Pink5 Pass ( pink2, pink5 ) Pink5 makes a long pass to Pink8 Kick ( pink5 ) Pink8 shoots the ball Kick ( pink8 ) Unambiguous Training Data Purple7 loses the ball to Pink2 Pink2 kicks the ball to Pink5 Pink5 makes a long pass to Pink8 Pink8 shoots the ball Ambiguous Training Data Pass ( purple5, purple7 ) Turnover ( purple7, pink2 ) Kick ( pink2 ) Pass ( pink2, pink5 ) Kick ( pink5 ) Pass ( pink5, pink8) Ballstopped Kick ( pink8 ) Semantic Parser Semantic Parser Learner

54 System Overview Sportscaster Robocup Simulator Purple7 loses the ball to Turnover ( purple7, Pink2 pink2 ) Pink2 kicks the ball to Pink5 Pass ( pink2, pink5 ) Pink5 makes a long pass to Pink8 Kick ( pink5 ) Pink8 shoots the ball Kick ( pink8 ) Unambiguous Training Data Purple7 loses the ball to Pink2 Pink2 kicks the ball to Pink5 Pink5 makes a long pass to Pink8 Pink8 shoots the ball Ambiguous Training Data Pass ( purple5, purple7 ) Turnover ( purple7, pink2 ) Kick ( pink2 ) Pass ( pink2, pink5 ) Kick ( pink5 ) Pass ( pink5, pink8) Ballstopped Kick ( pink8 ) Semantic Parser Semantic Parser Learner

55 System Overview Sportscaster Robocup Simulator Purple7 loses the ball to Turnover ( purple7, Pink2 pink2 ) Pink2 kicks the ball to Pink5 Pass ( pink2, pink5 ) Pink5 makes a long pass to Pink8 Kick ( pink5 ) Pink8 shoots the ball Kick ( pink8 ) Unambiguous Training Data Purple7 loses the ball to Pink2 Pink2 kicks the ball to Pink5 Pink5 makes a long pass to Pink8 Pink8 shoots the ball Ambiguous Training Data Pass ( purple5, purple7 ) Turnover ( purple7, pink2 ) Kick ( pink2 ) Pass ( pink2, pink5 ) Kick ( pink5 ) Pass ( pink5, pink8) Ballstopped Kick ( pink8 ) Semantic Parser Semantic Parser Learner

56 System Overview Sportscaster Robocup Simulator Purple7 loses the ball to Turnover ( purple7, Pink2 pink2 ) Pink2 kicks the ball to Pink5 Pass ( pink2, pink5 ) Pink5 makes a long pass to Pink8 Pass ( pink5, pink8) Pink8 shoots the ball Kick ( pink8 ) Unambiguous Training Data Purple7 loses the ball to Pink2 Pink2 kicks the ball to Pink5 Pink5 makes a long pass to Pink8 Pink8 shoots the ball Ambiguous Training Data Pass ( purple5, purple7 ) Turnover ( purple7, pink2 ) Kick ( pink2 ) Pass ( pink2, pink5 ) Kick ( pink5 ) Pass ( pink5, pink8) Ballstopped Kick ( pink8 ) Semantic Parser Semantic Parser Learner

57 (Hajishirzi et al., 2012) Professional Soccer Dataset Commentary Game Events World Cup Live 94' 43 Essien penalised for a challenge on Fabregas in midnield. 44 Song began the move, exchanged with Wilshere on the edge of the box before stealing the ball off Fabregas' foot and sliding it home with his left foot from ten yards out. What a fantastic goal. Time! Type! Quali;ier! Team! Player! 2592! Pass! Throw In! Chelsea! Kalou! 2594! Tackle! Long Ball! Arsenal! Clichy! 2594! Disposses s! Alex Song celebrates giving Arsenal a first-ha Through Ball! Chelsea! Essien! 2596! foul! Through Ball! Chelsea! Essien! 2603! pass! Free Kick! Arsenal! Fabregas! 2605! Pass! Chipped! Arsenal! Clichy! 2609! Pass! Through Ball! Arsenal! Koscielny! 2615! Pass! Long Ball! Arsenal! Song! 2623! Pass! Head! Arsenal! Nasri! 2625! Pass! Through Ball! Arsenal! Song! 2627! Pass! Through Ball! Arsenal! Wilshere! 2629! Pass! Through Ball! Arsenal! Fabregas! 2630! Goal! Through Ball! Arsenal! Song!

58 Challenges 1. Complex structure and paraphrases Commentary 44 Song began the move, exchanged with Wilshere on the edge of the box before stealing the ball off Fabregas' foot and sliding it home with his left foot from ten yards out. What a fantastic goal. Game Events Time! Type! Quali;ier! Team! Player! 2603! pass! Free Kick! Arsenal! Fabregas! 2605! Pass! Chipped! Arsenal! Clichy! 2609! Pass! Through Ball! Arsenal! Koscielny! 2615! Pass! Long Ball! Arsenal! Song! 2623! Pass! Head! Arsenal! Nasri! 2625! Pass! Through Ball! Arsenal! Song! 2627! Pass! Through Ball! Arsenal! Wilshere! 2629! Pass! Through Ball! Arsenal! Fabregas! 2630! Goal! Through Ball! Arsenal! Song!

59 Challenges 1. Complex structure and paraphrases Commentary 44 Song began the move, exchanged with Wilshere on the edge of the box before stealing the ball off Fabregas' foot and sliding it home with his left foot from ten yards out. What a fantastic goal. Game Events Time! Type! Quali;ier! Team! Player! 2603! pass! Free Kick! Arsenal! Fabregas! 2605! Pass! Chipped! Arsenal! Clichy! 2609! Pass! Through Ball! Arsenal! Koscielny! 2615! Pass! Long Ball! Arsenal! Song! 2623! Pass! Head! Arsenal! Nasri! 2625! Pass! Through Ball! Arsenal! Song! 2627! Pass! Through Ball! Arsenal! Wilshere! 2629! Pass! Through Ball! Arsenal! Fabregas! 2630! Goal! Through Ball! Arsenal! Song!

60 Challenges 1: Complex structure and paraphrases Commentary 44 Song began the move, exchanged with Wilshere on the edge of the box before stealing the ball off Fabregas' foot and sliding it home with his left foot from ten yards out. What a fantastic goal. Game Events Time! Type! Quali;ier! Team! Player! 2603! pass! Free Kick! Arsenal! Fabregas! 2605! Pass! Chipped! Arsenal! Clichy! 2609! Pass! Through Ball! Arsenal! Koscielny! 2615! Pass! Long Ball! Arsenal! Song! 2623! Pass! Head! Arsenal! Nasri! 2625! Pass! Through Ball! Arsenal! Song! 2627! Pass! Through Ball! Arsenal! Wilshere! 2629! Pass! Through Ball! Arsenal! Fabregas! 2630! Goal! Through Ball! Arsenal! Song!

61 Challenges 1. Complex structure and paraphrases Commentary 44 Song began the move, exchanged with Wilshere on the edge of the box before stealing the ball off Fabregas' foot and sliding it home with his left foot from ten yards out. What a fantastic goal. Game Events Time! Type! Quali;ier! Team! Player! 2603! pass! Free Kick! Arsenal! Fabregas! 2605! Pass! Chipped! Arsenal! Clichy! 2609! Pass! Through Ball! Arsenal! Koscielny! 2615! Pass! Long Ball! Arsenal! Song! 2623! Pass! Head! Arsenal! Nasri! 2625! Pass! Through Ball! Arsenal! Song! 2627! Pass! Through Ball! Arsenal! Wilshere! 2629! Pass! Through Ball! Arsenal! Fabregas! 2630! Goal! Through Ball! Arsenal! Song!

62 Challenges 2: Comments correspond to combination of events (Macro events) 44 Song began the move, exchanged with Wilshere on the edge of the box before stealing the ball off Fabregas' foot and sliding it home with his left foot from ten yards out. What a fantastic goal. Time! Type! Quali;ier! Team! Player! 2603! pass! Free Kick! Arsenal! Fabregas! 2605! Pass! Chipped! Arsenal! Clichy! 2609! Pass! Through Ball! Arsenal! Koscielny! 2615! Pass! Long Ball! Arsenal! Song! 2623! Pass! Head! Arsenal! Nasri! 2625! Pass! Through Ball! Arsenal! Song! 2627! Pass! Through Ball! Arsenal! Wilshere! 2629! Pass! Through Ball! Arsenal! Fabregas! 2630! Goal! Through Ball! Arsenal! Song! Arsenal is coming forward

63 Problem Commentary Game Events World Cup Live 94' 43 Essien penalised for a challenge on Fabregas in midnield. 44 Song began the move, exchanged with Wilshere on the edge of the box before stealing the ball off Fabregas' foot and sliding it home with his left foot from ten yards out. What a fantastic goal. Time! Type! Quali;ier! Team! Player! 2592! Pass! Throw In! Chelsea! Kalou! 2594! Tackle! Long Ball! Arsenal! Clichy! 2594! Disposses s! Alex Song celebrates giving Arsenal a first-ha Through Ball! Chelsea! Essien! 2596! foul! Through Ball! Chelsea! Essien! 2603! pass! Free Kick! Arsenal! Fabregas! 2605! Pass! Chipped! Arsenal! Clichy! 2609! Pass! Through Ball! Arsenal! Koscielny! 2615! Pass! Long Ball! Arsenal! Song! 2623! Pass! Head! Arsenal! Nasri! 2625! Pass! Through Ball! Arsenal! Song! 2627! Pass! Through Ball! Arsenal! Wilshere! 2629! Pass! Through Ball! Arsenal! Fabregas! 2630! Goal! Through Ball! Arsenal! Song!

64 Problem Commentary Game Events World Cup Live 94' 43 Essien penalised for a challenge on Fabregas in midnield. 44 Song began the move, exchanged with Wilshere on the edge of the box before stealing the ball off Fabregas' foot and sliding it home with his left foot from ten yards out. What a fantastic goal. Time! Type! Quali;ier! Team! Player! 2592! Pass! Throw In! Chelsea! Kalou! 2594! Tackle! Long Ball! Arsenal! Clichy! 2594! Disposses s! Alex Song celebrates giving Arsenal a first-ha Through Ball! Chelsea! Essien! 2596! foul! Through Ball! Chelsea! Essien! 2603! pass! Free Kick! Arsenal! Fabregas! 2605! Pass! Chipped! Arsenal! Clichy! 2609! Pass! Through Ball! Arsenal! Koscielny! 2615! Pass! Long Ball! Arsenal! Song! 2623! Pass! Head! Arsenal! Nasri! 2625! Pass! Through Ball! Arsenal! Song! 2627! Pass! Through Ball! Arsenal! Wilshere! 2629! Pass! Through Ball! Arsenal! Fabregas! 2630! Goal! Through Ball! Arsenal! Song!

65 How to Measure Correspondence? Model correspondence of a sentence and an event Pair every sentence with every event in the bucket Learn particular pattern to each pair 1: Chelsea looking for penalty as Malouda s header hits Koscielny. Pass Head Chelsea Malouda 1: Chelsea looking for penalty as Malouda s header hits Koscielny. Foul Head Arsenal Koschielny

66 How to Measure Correspondence? Model correspondence of a sentence and an event Pair every sentence with every event in the bucket Learn particular pattern to each pair 1: Chelsea looking for penalty as Malouda s header hits Koscielny. Pass Head Chelsea Malouda 1: Chelsea looking for penalty as Malouda s header hits Koscielny. Foul Head Arsenal Koschielny Intuition: A pair is good if the pattern of correspondence occurs frequently among the data

67 43 Essien penalised for a challenge on Fabregas in midnield. 44 Song began the move, exchanged with Wilshere on the edge of the box before stealing the ball off Fabregas' foot and sliding it home with his left foot from ten yards out. What a fantastic goal. Time! Type! Quali;ier! Team! Player! 2592! Pass! Throw In! Chelsea! Kalou! 2594! Tackle! Long Ball! Arsenal! Clichy! 2594! Disposs ess! Through Ball! Chelsea! Essien! 2596! foul! Through Ball! Chelsea! Essien! 2603! pass! Free Kick! Arsenal! Fabregas! 2605! Pass! Chipped! Arsenal! Clichy! 2609! Pass! Through Ball! Arsenal! Koscielny! 2615! Pass! Long Ball! Arsenal! Song! 2623! Pass! Head! Arsenal! Nasri! 2625! Pass! Through Ball! Arsenal! Song! 2627! Pass! Through Ball! Arsenal! Wilshere! 2629! Pass! Through Ball! Arsenal! Fabregas! 2630! Goal! Through Ball! Arsenal! Song! Pair- 1: Pair- 2: Pair- 3: Pair- 3: Pair- 4: Pair- 5: Pair- 6: Pair- 7: Pair- 8: Pair- 9: Pair- 10: S- 43 S- 43 S- 43 S- 43 S- 43 S- 43 S- 43 S- 44 S- 44 S- 44 S- 44 E- 1 E- 2 E- 3 E- 4 E- 5 E- 6 E- 7 E- 5 E- 6 E- 7 E- 8 For every sentence: Generate pairs of sentence and events in the bucket Pair- 11: Pair- 12: Pair- 13: Pair- 14: S- 44 S- 44 S- 44 S- 44 E- 9 E- 10 E- 11 E- 12

68 Measuring Correspondence Intuition: The pattern of correspondence in a pair appears more consistently across all games Consistency of a pair is measured by its popularity among other pairs Popularity of a pair requires a notion of similarity between pairs

69 Similarity between Pairs 1: Chelsea looking for penalty as Malouda s header hits Koscielny. Pass Head Chelsea Malouda 1: Chelsea looking for penalty as Malouda s header hits Koscielny. Foul Head Arsenal Koschielny

70 Similarity between Pairs Co- occurrence of words and arguments matter 1: Chelsea looking for penalty as Malouda s header hits Koscielny. Pass Head Chelsea Malouda 1: Chelsea looking for penalty as Malouda s header hits Koscielny. Foul Head Arsenal Koschielny

71 Similarity between Pairs Co- occurrence of words and arguments matter 1: Chelsea looking for penalty as Malouda s header hits Koscielny. Pass Head Chelsea Malouda 1: Chelsea looking for penalty as Malouda s header hits Koscielny. Foul Head Arsenal Koschielny Discriminative Similarity Learn the correspondence pattern in each pair Discriminate the pair from the rest

72 S- 43 E- 1 Pair- 1: S- 43 E- 2 Pair- 2: S- 43 E- 3 Pair- 3: S- 43 E- 4 Pair- 3: S- 43 E- 5 Pair- 4: S- 43 E- 6 Pair- 5: S- 43 E- 7 Pair- 6: S- 44 E- 5 Pair- 7: S- 44 E- 6 Pair- 8: S- 44 E- 7 Pair- 9: S- 44 E- 8 Pair- 10: S- 44 E- 9 Pair- 11: S- 44 E- 10 Pair- 12: S- 44 E- 11 Pair- 13: S- 44 E- 12 Pair- 14: S- 44 E- 13 Pair- 15: Popularity Ranking S- 43 E- 5 Pair- 5: S- 43 E- 2 Pair- 2: S- 43 E- 1 Pair- 1: S- 43 E- 3 Pair- 3: S- 43 E- 4 Pair- 4: S- 43 E- 7 Pair- 7: S- 43 E- 6 Pair- 6: S- 44 E- 10 Pair- 10: S- 44 E- 8 Pair- 8: S- 44 E- 15 Pair- 15: S- 44 E- 12 Pair- 12: S- 44 E- 11 Pair- 11: S- 44 E- 3 Pair- 3: S- 44 E- 9 Pair- 9: S- 44 E- 14 Pair- 14: S- 44 E- 7 Pair- 7: Pair Models Macro- event? Macro- event?

73 Qualita&ve Results 4: Cole is sent off for a lunge on Koscielny, it was poor, it was late but I'm not entirely sure that should have been red. Foul Long Ball Liverpool J. Cole Foul Through Ball Arsenal L. Koscielny Card None Liverpool J. Cole

74 Qualita&ve Results 5: First attack for Drogba, outmuscling Sagna and sending an effort in from the edge of the box which is blocked, and Song then brings down Drogba for a free kick. Take On Head Pass Chelsea D. Drogba Challenge Through Ball Arsenal B. Sagna Save Through Ball Arsenal A. Song Foul Through Ball Arsenal A. Song

75 Thank You! More questions:

Using Perceptual Context to Ground Language

Using Perceptual Context to Ground Language Using Perceptual Context to Ground Language David Chen Joint work with Joohyun Kim, Raymond Mooney Department of Computer Sciences, University of Texas at Austin 2009 IBM Statistical Machine Learning and

More information

BALLGAME: A Corpus for Computational Semantics

BALLGAME: A Corpus for Computational Semantics BALLGAME: A Corpus for Computational Semantics Ezra Keshet, Terry Szymanski, and Stephen Tyndall University of Michigan E-mail: {ekeshet,tdszyman,styndall}@umich.edu Abstract In this paper, we describe

More information

A THESIS SUBMITTED TO THE FACULTY OF GRADUATE SCHOOL OF THE UNIVERSITY OF MINNESOTA

A THESIS SUBMITTED TO THE FACULTY OF GRADUATE SCHOOL OF THE UNIVERSITY OF MINNESOTA Semantic Parsing for Automatic Generation of SQL Queries using Adaptive Boosting A THESIS SUBMITTED TO THE FACULTY OF GRADUATE SCHOOL OF THE UNIVERSITY OF MINNESOTA BY Rajesh Tripurneni IN PARTIAL FULFILLMENT

More information

Lecture 10: Generation

Lecture 10: Generation Lecture 10: Generation Overview of Natural Language Generation An extended example: cricket reports Learning from existing text Overview of Natural Language Generation Generation Generation from what?!

More information

Sharing the How (and not the What )

Sharing the How (and not the What ) Sharing the How (and not the What ) Percy Liang NIPS Workshop on Transfer Learning December 12, 2015 What is shared between tasks? Task A Task B 1 What is shared between tasks? Task A Task B 1 detecting

More information

Learning Latent Programs for Question Answering

Learning Latent Programs for Question Answering Learning Latent Programs for Question Answering Percy Liang Stanford University ICLR May 8, 2015 Goals Introduce a new challenge task for representation learning. 1 Goals Introduce a new challenge task

More information

The Champion UT Austin Villa 2003 Simulator Online Coach Team

The Champion UT Austin Villa 2003 Simulator Online Coach Team In Daniel Polani, Brett Browning, Andrea Bonarini and Kazuo Yoshida, editors, RoboCup-2003: Robot Soccer World Springer Verlag, Berlin, 2004. The Champion UT Austin Villa 2003 Simulator Online Coach Team

More information

Approximate Grammar for Information Extraction

Approximate Grammar for Information Extraction Approximate Grammar for Information Extraction Akshar Bharati, V.Sriram, B.Ravi Sekhar Reddy, Rajeev Sangal International Institute of Information Technology, Hyderabad {sriram,ravi_b}@gdit.iiit.net,{sangal}@iiit.net

More information

Scope asymmetry, QUDs, and exhaustification

Scope asymmetry, QUDs, and exhaustification Teodora Mihoc @ LingedIn, May 4, 2018 Scope asymmetry, QUDs, and exhaustification 1 Data Two/every-not scope ambiguities (1) Every horse didn t jump over the fence. a. every > not (surface scope) None

More information

CS472 Foundations of Artificial Intelligence. Final Exam December 19, :30pm

CS472 Foundations of Artificial Intelligence. Final Exam December 19, :30pm CS472 Foundations of Artificial Intelligence Final Exam December 19, 2003 12-2:30pm Name: (Q exam takers should write their Number instead!!!) Instructions: You have 2.5 hours to complete this exam. The

More information

INTRODUCTION TO PATTERN RECOGNITION

INTRODUCTION TO PATTERN RECOGNITION INTRODUCTION TO PATTERN RECOGNITION 3 Introduction Our ability to recognize a face, to understand spoken words, to read handwritten characters all these abilities belong to the complex processes of pattern

More information

USING COMPUTERIZED WORKFLOW SIMULATIONS TO ASSESS THE FEASIBILITY OF WHOLE SLIDE IMAGING: FULL ADOPTION IN A HIGH VOLUME HISTOLOGY LABORATORY

USING COMPUTERIZED WORKFLOW SIMULATIONS TO ASSESS THE FEASIBILITY OF WHOLE SLIDE IMAGING: FULL ADOPTION IN A HIGH VOLUME HISTOLOGY LABORATORY USING COMPUTERIZED WORKFLOW SIMULATIONS TO ASSESS THE FEASIBILITY OF WHOLE SLIDE IMAGING: FULL ADOPTION IN A HIGH VOLUME HISTOLOGY LABORATORY David McClintock, M.D. Fellow, Pathology Informatics Massachusetts

More information

D-Case Modeling Guide for Target System

D-Case Modeling Guide for Target System D-Case Modeling Guide for Target System 1/32 Table of Contents 1 Scope...4 2 Overview of D-Case and SysML Modeling Guide...4 2.1 Background and Purpose...4 2.2 Target System of Modeling Guide...5 2.3 Constitution

More information

Spatio-temporal analysis of team sports Joachim Gudmundsson

Spatio-temporal analysis of team sports Joachim Gudmundsson Spatio-temporal analysis of team sports Joachim Gudmundsson The University of Sydney Page 1 Team sport analysis Talk is partly based on: Joachim Gudmundsson and Michael Horton Spatio-Temporal Analysis

More information

Using Big Data to Model Football. Edward Griffiths Hong Kong Jockey Club 29 April 2015

Using Big Data to Model Football. Edward Griffiths Hong Kong Jockey Club 29 April 2015 Using Big Data to Model Football Edward Griffiths Hong Kong Jockey Club 29 April 2015 Outline of Presentation 1) Hong Kong Jockey Club Introduction 2) Why Big Data? 3) Case Study - The Football Betting

More information

Syntax and Parsing II

Syntax and Parsing II Syntax and Parsing II Dependency Parsing Slav Petrov Google Thanks to: Dan Klein, Ryan McDonald, Alexander Rush, Joakim Nivre, Greg Durrett, David Weiss Lisbon Machine Learning School 2015 Notes for 2016

More information

Statistical Machine Translation

Statistical Machine Translation Statistical Machine Translation Hicham El-Zein Jian Li University of Waterloo May 20, 2015 Hicham El-Zein Jian Li (UW) Statistical Machine Translation May 20, 2015 1 / 69 Overview 1 Introduction 2 Challenges

More information

Coupling distributed and symbolic execution for natural language queries. Lili Mou, Zhengdong Lu, Hang Li, Zhi Jin

Coupling distributed and symbolic execution for natural language queries. Lili Mou, Zhengdong Lu, Hang Li, Zhi Jin Coupling distributed and symbolic execution for natural language queries Lili Mou, Zhengdong Lu, Hang Li, Zhi Jin doublepower.mou@gmail.com Outline Introduction to neural enquirers Coupled approach of

More information

The Novice Coach s Guide to a Successful Soccer Season

The Novice Coach s Guide to a Successful Soccer Season The Novice Coach s Guide to a Successful Soccer Season Practical Advice on Coaching a Competitive Youth Soccer Team 1 The Novice Coach s Guide to a Successful Soccer Season The vast majority of soccer

More information

COMP 406 Lecture 05. Artificial. Fiona Yan Liu Department of Computing The Hong Kong Polytechnic University

COMP 406 Lecture 05. Artificial. Fiona Yan Liu Department of Computing The Hong Kong Polytechnic University COMP 406 Lecture 05 Artificial Intelligence Fiona Yan Liu Department of Computing The Hong Kong Polytechnic University Learning Outcomes of Intelligent Agents Search agents Uninformed search Informed search

More information

Spreading Activation in Soar: An Update

Spreading Activation in Soar: An Update Spreading Activation in Soar: An Update Steven Jones Computer Science and Engineering, University of Michigan, Ann Arbor scijones@umich.edu June 13, 2016 Steven Jones (U-M) Spreading Activation in Soar

More information

(8, 9 and Some 10 Year Olds) Shield and Steal Activity Description Coaching Objective. Coach divides the players into two

(8, 9 and Some 10 Year Olds) Shield and Steal Activity Description Coaching Objective. Coach divides the players into two Shield and Steal Activity Description Coaching Objective Coach sets up a 25x35 yard grid Dribbling for individual Coach divides the players into two possession and shielding. different colored teams. All

More information

Unit 10- Indoor Football

Unit 10- Indoor Football Unit 10- Indoor Football Indoor football was born as a spectator sport with the increase in indoor sport centres and with the public support of companies. It has many things in common with football, but

More information

Performance Indicators (examples) Responds correctly to questions about own name, sex and age

Performance Indicators (examples) Responds correctly to questions about own name, sex and age The Scholastic Early Childhood Program correlated to the Section 1: Speech and Language Development Language Development Communication Receptive Language Listens with understanding and interest to conversations,

More information

A Novel Approach to Predicting the Results of NBA Matches

A Novel Approach to Predicting the Results of NBA Matches A Novel Approach to Predicting the Results of NBA Matches Omid Aryan Stanford University aryano@stanford.edu Ali Reza Sharafat Stanford University sharafat@stanford.edu Abstract The current paper presents

More information

Commonsense Knowledge Acquisition and Applications

Commonsense Knowledge Acquisition and Applications Commonsense Knowledge Acquisition and Applications Towards Commonsense Enriched Machines Niket Tandon Ph.D. Supervisor: Gerhard Weikum Max Planck Institute for Informatics 2 Hard Rock property brown Hand,

More information

Cuts. (4 and 6 Gate)

Cuts. (4 and 6 Gate) Cuts (4 and 6 Gate) Plyometric/reactive cut combo drills are an excellent way to practice rapid changes in direction and landing stability. Variations include hurdles, ladders, medicine ball throws and

More information

Arsenal Vs Chelsea. 26 th April Get full match stats every week here

Arsenal Vs Chelsea. 26 th April Get full match stats every week here Arsenal Vs Chelsea 26 th April 2015 Hi, thanks for joining Premstats! If you haven't met us before we are a new football website that digs deeper into Premier League statistics than anyone else. We analyse

More information

PREDICTING the outcomes of sporting events

PREDICTING the outcomes of sporting events CS 229 FINAL PROJECT, AUTUMN 2014 1 Predicting National Basketball Association Winners Jasper Lin, Logan Short, and Vishnu Sundaresan Abstract We used National Basketball Associations box scores from 1991-1998

More information

9-11 YEAR OLD PLAYERS

9-11 YEAR OLD PLAYERS STAGE 3 ACTIVITIES 9-11 YEAR OLD PLAYERS NSCAA Foundations of Coaching Diploma NSCAA Foundations of Coaching Activities Stage 3: 9-11 Year Old Players 4V1 TO 4V2 IN THIRDS This game provides attackers

More information

Team Description: Building Teams Using Roles, Responsibilities, and Strategies

Team Description: Building Teams Using Roles, Responsibilities, and Strategies Team Description: Building Teams Using Roles, Responsibilities, and Strategies Simon Ch'ng and Lin Padgham (schng cs. trait, edu. au) (linpa cs. rrait, edu. au) Department of Computer Science Royal Melbourne

More information

Introduction to Pattern Recognition

Introduction to Pattern Recognition Introduction to Pattern Recognition Jason Corso SUNY at Buffalo 12 January 2009 J. Corso (SUNY at Buffalo) Introduction to Pattern Recognition 12 January 2009 1 / 28 Pattern Recognition By Example Example:

More information

Weekly Practice Schedule:

Weekly Practice Schedule: Coaching Ages 13 to 14 Years This section provides you with a twelve-week training program for children ages 13 through 14 years. Each practice features four fun and progressive drills and culminates with

More information

Season By Daniel Mattos GLSA Futsal Director FUTSAL

Season By Daniel Mattos GLSA Futsal Director FUTSAL 2016-17 Season By Daniel Mattos GLSA Futsal Director FUTSAL FA - Soul of Futsal US Soccer Federation - Futsal Benefits Soccer Pros born in Futsal Falcão - The Best Futsal Player in the World Repetitive,

More information

Mixture Models & EM. Nicholas Ruozzi University of Texas at Dallas. based on the slides of Vibhav Gogate

Mixture Models & EM. Nicholas Ruozzi University of Texas at Dallas. based on the slides of Vibhav Gogate Mixture Models & EM Nicholas Ruozzi University of Texas at Dallas based on the slides of Vibhav Gogate Previously We looked at -means and hierarchical clustering as mechanisms for unsupervised learning

More information

Spring 2010 Coaching Sessions U14

Spring 2010 Coaching Sessions U14 Spring 2010 Coaching Sessions U14 Coaching Ages 13 to 14 Years This section provides you with a eight week training program for children ages 13 through 14 years. Each practice features four fun and progressive

More information

Athletic Event Rules

Athletic Event Rules Athletic Event Rules Team Eligibility The teams admissible for the athletic events (soccer, ultimate frisbee on Saturday, volleyball and basketball on Sunday) are those teams that have registered their

More information

Northern SC U12 Playing Formats 8v8 (7 field players + 1 GK)

Northern SC U12 Playing Formats 8v8 (7 field players + 1 GK) Northern SC U12 Playing Formats 8v8 (7 field players + 1 ) This document outlines guidelines for increasing the consistency of playing formations and terminology we use with U12 players. As players of

More information

EVALITA 2011 The News People Search Task: Evaluating Cross-document Coreference Resolution of Named Person Entities in Italian News

EVALITA 2011 The News People Search Task: Evaluating Cross-document Coreference Resolution of Named Person Entities in Italian News EVALITA 2011 The News People Search Task: Evaluating Cross-document Coreference Resolution of Named Person Entities in Italian News L. Bentivogli, A. Marchetti, E. Pianta Outline Introduction and Related

More information

Bayesian Methods: Naïve Bayes

Bayesian Methods: Naïve Bayes Bayesian Methods: Naïve Bayes Nicholas Ruozzi University of Texas at Dallas based on the slides of Vibhav Gogate Last Time Parameter learning Learning the parameter of a simple coin flipping model Prior

More information

Coaching Players Ages 17 to Adult

Coaching Players Ages 17 to Adult Coaching Players Ages 17 to Adult This section provides you with a twelve-week training program for players ages 17 through Adult. Each practice features four fun and progressive drills and culminates

More information

Re-evaluating automatic metrics for image captioning

Re-evaluating automatic metrics for image captioning /mertkilickaya_ kilickayamert@gmail.com Re-evaluating automatic metrics for image captioning Mert Kilickaya, Aykut Erdem, Nazli Ikizler-Cinbis, Erkut Erdem Background source target f( = ( The cat sat on

More information

2 When Some or All Labels are Missing: The EM Algorithm

2 When Some or All Labels are Missing: The EM Algorithm CS769 Spring Advanced Natural Language Processing The EM Algorithm Lecturer: Xiaojin Zhu jerryzhu@cs.wisc.edu Given labeled examples (x, y ),..., (x l, y l ), one can build a classifier. If in addition

More information

Essen,al Probability Concepts

Essen,al Probability Concepts Naïve Bayes Essen,al Probability Concepts Marginaliza,on: Condi,onal Probability: Bayes Rule: P (B) = P (A B) = X v2values(a) P (A B) = P (B ^ A = v) P (A ^ B) P (B) ^ P (B A) P (A) P (B) Independence:

More information

Soccer Statistics Summary with Extended Coverage

Soccer Statistics Summary with Extended Coverage Soccer Statistics Summary with Extended Coverage 2018 1 SPORTRADAR SOCCER STATISTICS SUMMARY Sport & Category Information Category Country Code Category Name Sport Name Category Id Sport Id Tournament

More information

Using Spatio-Temporal Data To Create A Shot Probability Model

Using Spatio-Temporal Data To Create A Shot Probability Model Using Spatio-Temporal Data To Create A Shot Probability Model Eli Shayer, Ankit Goyal, Younes Bensouda Mourri June 2, 2016 1 Introduction Basketball is an invasion sport, which means that players move

More information

Decision Trees. Nicholas Ruozzi University of Texas at Dallas. Based on the slides of Vibhav Gogate and David Sontag

Decision Trees. Nicholas Ruozzi University of Texas at Dallas. Based on the slides of Vibhav Gogate and David Sontag Decision Trees Nicholas Ruozzi University of Texas at Dallas Based on the slides of Vibhav Gogate and David Sontag Announcements Course TA: Hao Xiong Office hours: Friday 2pm-4pm in ECSS2.104A1 First homework

More information

THE CANDU 9 DISTRffiUTED CONTROL SYSTEM DESIGN PROCESS

THE CANDU 9 DISTRffiUTED CONTROL SYSTEM DESIGN PROCESS THE CANDU 9 DISTRffiUTED CONTROL SYSTEM DESIGN PROCESS J.E. HARBER, M.K. KATTAN Atomic Energy of Canada Limited 2251 Speakman Drive, Mississauga, Ont., L5K 1B2 CA9900006 and M.J. MACBETH Institute for

More information

The International Coaches Association Advanced Passing Drills and Games

The International Coaches Association Advanced Passing Drills and Games The International Coaches Association Advanced Passing Drills and Games Manchester United Passing Drill Italian Passing Awareness Drill Liverpool Passing Game Juventus Reverse Passing Drill Burnley FC

More information

The soccer Pitch Goal kick Penalty Area Light green play colour Goal Line Dark green play colour Centre Circle Kick off Side line Side fence 2 3

The soccer Pitch Goal kick Penalty Area Light green play colour Goal Line Dark green play colour Centre Circle Kick off Side line Side fence 2 3 The soccer Pitch Goal kick Penalty Area Light green play colour Goal Line Dark green play colour Centre Circle Kick off Side line Side fence 2 3 Play method basics SHOOTING TECHNIQUES GOALKEEPING Kick

More information

Prentice Hall Realidades, Level B 2008

Prentice Hall Realidades, Level B 2008 Grades 5-8 Prentice Hall Realidades, Level B 2008 C O R R E L A T E D T O Colorado Model Content Standards for Foreign Language, Intermediate Level Grades 5-8 Colorado Model Content Standards for Foreign

More information

Latham Circle Soccer Club Fall Recreation and Skills Program Coach s and Referee s Guide (Players in Grades 3 and up)

Latham Circle Soccer Club Fall Recreation and Skills Program Coach s and Referee s Guide (Players in Grades 3 and up) Latham Circle Soccer Club Fall Recreation and Skills Program Coach s and Referee s Guide (Players in Grades 3 and up) Player equipment 1. All players are expected to bring their own soccer ball with them

More information

Investigating Content Selection for Language Generation using Machine Learning

Investigating Content Selection for Language Generation using Machine Learning Investigating Content Selection for Language Generation using Machine Learning Colin Kelly Computer Laboratory University of Cambridge 15 JJ Thomson Avenue Cambridge, UK Ann Copestake Computer Laboratory

More information

SOCCER PACKET # 6 INSTRUCTIONS

SOCCER PACKET # 6 INSTRUCTIONS SOCCER PACKET # 6 INSTRUCTIONS This Learning Packet has two parts: (1) text to read and (2) questions to answer. The text describes a particular sport or physical activity, and relates its history, rules,

More information

Joint Parsing and Translation

Joint Parsing and Translation Joint arsing and Translation Yang Liu and Qun Liu Institute of Computing Technology Chinese Academy of Sciences August 26, 2010 COLING 2010, Beijing, China 1 Separate arsing and Translation source target

More information

Sequence Tagging. Today. Part-of-speech tagging. Introduction

Sequence Tagging. Today. Part-of-speech tagging. Introduction Sequence Tagging Today Part-of-speech tagging Introduction Part of speech tagging There are 10 parts of speech, and they are all troublesome. -Mark Twain POS tags are also known as word classes, morphological

More information

RUGBY is a dynamic, evasive, and highly possessionoriented

RUGBY is a dynamic, evasive, and highly possessionoriented VISUALIZING RUGBY GAME STYLES USING SOMS 1 Visualizing Rugby Game Styles Using Self-Organizing Maps Peter Lamb, Hayden Croft Abstract Rugby coaches and analysts often use notational data describing match

More information

Standard League WRO Football Simple Simon EV3 Training Curriculum

Standard League WRO Football Simple Simon EV3 Training Curriculum Standard League WRO Football Simple Simon EV3 Training Curriculum Brian Thomas **LEGO, EV3 Mindstorms and Robolab are Registered Trademarks of the LEGO Corporation RoboCup Junior Australia Inc Copyright

More information

Parsimonious Linear Fingerprinting for Time Series

Parsimonious Linear Fingerprinting for Time Series Parsimonious Linear Fingerprinting for Time Series Lei Li, B. Aditya Prakash, Christos Faloutsos School of Computer Science Carnegie Mellon University VLDB 2010 1 L. Li, 2010 VLDB2010, 36 th International

More information

SYSTEMS AND TOOLS FOR MACHINE TRANSLATION

SYSTEMS AND TOOLS FOR MACHINE TRANSLATION Universitat Politècnica de València Master en Inteligencia Artificial, Renoconocimiento de Formas e Imagen Digital Curs 2006-2007 SYSTEMS AND TOOLS FOR MACHINE TRANSLATION Francisco Casacuberta fcn@iti.upv.es

More information

Know the direction of play

Know the direction of play U7 ATTACKING Know the direction of play ACTIVITY: DRIBBLE TO THE CONE Understand that to score the ball must go towards the opposition goal Set up a 5 yard square area with a red cone goal at one end and

More information

Iteration: while, for, do while, Reading Input with Sentinels and User-defined Functions

Iteration: while, for, do while, Reading Input with Sentinels and User-defined Functions Iteration: while, for, do while, Reading Input with Sentinels and User-defined Functions This programming assignment uses many of the ideas presented in sections 6 and 7 of the course notes. You are advised

More information

Rio Rapids Durango Soccer Club U13/14 Player Standards

Rio Rapids Durango Soccer Club U13/14 Player Standards Rio Rapids Durango Soccer Club U13/14 Player Standards Characteristics of the U13/14 Player -The more advanced U13/14 players are able to execute a range of skills, but most others are still developing

More information

A new AI benchmark. Soccer without Reason Computer Vision and Control for Soccer Playing Robots. Dr. Raul Rojas

A new AI benchmark. Soccer without Reason Computer Vision and Control for Soccer Playing Robots. Dr. Raul Rojas Soccer without Reason Computer Vision and Control for Soccer Playing Robots Dr. Raul Rojas A new AI benchmark -computer vision in real time - embodied intelligence: mechanics - energy management - local

More information

2009 Solutions. (G) Sk8r. Languages are everywhere even in places where you don t expect them.

2009 Solutions. (G) Sk8r. Languages are everywhere even in places where you don t expect them. Languages are everywhere even in places where you don t expect them. Consider the combo rules of P-Little s Triple-I XTreem Hyp0th3tica7 Sk8boarding Game. In it, players press a series of buttons (left,

More information

GEN II Robot Soccer EV3 Compass Training Curriculum Brian Thomas

GEN II Robot Soccer EV3 Compass Training Curriculum Brian Thomas GEN II Robot Soccer EV3 Compass Training Curriculum Brian Thomas RoboCup Junior Australia Inc Copyright 2015 1 GEN II EV3 Robot Soccer Training Overview Welcome! The aim of this curriculum is to guide

More information

GLOBAL PREMIER SOCCER

GLOBAL PREMIER SOCCER GLOBAL PREMIER SOCCER SESSIONS 7v7 SESSIONS - Defending 1v1 situation Players work in pairs with one ball. One player has a the ball and dribbles around inside the area. Player 2 must follow player 1 and

More information

Hazard analysis. István Majzik Budapest University of Technology and Economics Dept. of Measurement and Information Systems

Hazard analysis. István Majzik Budapest University of Technology and Economics Dept. of Measurement and Information Systems Hazard analysis István Majzik Budapest University of Technology and Economics Dept. of Measurement and Information Systems Hazard analysis Goal: Analysis of the fault effects and the evolution of hazards

More information

Building an NFL performance metric

Building an NFL performance metric Building an NFL performance metric Seonghyun Paik (spaik1@stanford.edu) December 16, 2016 I. Introduction In current pro sports, many statistical methods are applied to evaluate player s performance and

More information

SSA Under 14 to Under 19 Practice Activities

SSA Under 14 to Under 19 Practice Activities SSA Under 14 to Under 19 Practice Activities DRIBBLING MOVES AND ACTIVITIES TO INCLUDE IN WARM-UP (BEGINNING OF PRACTICE) CAN BE FOUND IN THE TECHNICAL WARM-UP DOCUMENT. Chain Tag: Set-Up: All players,

More information

Sentiment Analysis Symposium - Disambiguate Opinion Word Sense via Contextonymy

Sentiment Analysis Symposium - Disambiguate Opinion Word Sense via Contextonymy Sentiment Analysis Symposium - Disambiguate Opinion Word Sense via Contextonymy Guillaume Gadek, Josefin Betsholtz, Alexandre Pauchet, Stéphan Brunessaux, Nicolas Malandain and Laurent Vercouter Airbus,

More information

BASIC FUTSAL COACHING PREPARATION

BASIC FUTSAL COACHING PREPARATION BASIC FUTSAL COACHING PREPARATION I. Basics Futsal The priority in Futsal is to motivate players in an environment that is conducive to learning. The more pleasure kids derive from their participation,

More information

Football. English for THE GAMES

Football. English for THE GAMES English for THE GAMES Football When athletes come together from around the world to compete in the Olympic and Paralympic Games, they are fulfilling their dreams and competing at the highest level. Billions

More information

Introduction to Pattern Recognition

Introduction to Pattern Recognition Introduction to Pattern Recognition Jason Corso SUNY at Buffalo 19 January 2011 J. Corso (SUNY at Buffalo) Introduction to Pattern Recognition 19 January 2011 1 / 32 Examples of Pattern Recognition in

More information

Introduction Definition of decision-making: the capacity of the player to execute an action following some conscious tactical or strategical choice.

Introduction Definition of decision-making: the capacity of the player to execute an action following some conscious tactical or strategical choice. Decision Making in Rugby Here is a paper by Pierre Villepreux. He presented it at the 1993 Asian Pacific Congress in Calgary. Eleven years have passed, but this remains one of the benchmark papers on this

More information

Cuneiform. A Functional Workflow Language Implementation in Erlang. Jörgen Brandt Humboldt-Universität zu Berlin

Cuneiform. A Functional Workflow Language Implementation in Erlang. Jörgen Brandt Humboldt-Universität zu Berlin Cuneiform A Functional Workflow Language Implementation in Erlang Jörgen Brandt Humboldt-Universität zu Berlin 2015-12-01 Jörgen Brandt (HU Berlin) Cuneiform 2015-12-01 1 / 27 Cuneiform Cuneiform A Functional

More information

Common Core Standards for Language Arts. An Alignment with Santillana Español (Serie Amigos) Grades K-5

Common Core Standards for Language Arts. An Alignment with Santillana Español (Serie Amigos) Grades K-5 Common Core Standards for Language Arts An Alignment with Santillana Español (Serie Amigos) Grades K-5 College and Career Readiness Anchor Standards for Reading The K 5 standards on the following pages

More information

Adaptation of Formation According to Opponent Analysis

Adaptation of Formation According to Opponent Analysis Adaptation of Formation According to Opponent Analysis Jaroslav ZAJAC Slovak University of Technology Faculty of Informatics and Information Technologies Ilkovičova 3, 842 16 Bratislava, Slovakia ka zajacjaro@orangemail.sk

More information

STAGE 4 ACTIVITIES YEAR OLD PLAYERS

STAGE 4 ACTIVITIES YEAR OLD PLAYERS STAGE 4 ACTIVITIES 12-14 YEAR OLD PLAYERS NSCAA Foundations of Coaching Diploma NSCAA Foundations of Coaching Activities Stage 4: 12-14 Year Old Players 3V2 TO 6V5 SMALL SIDED GAME This activity provides

More information

TECHNICAL STUDY 2 with ProZone

TECHNICAL STUDY 2 with ProZone A comparative performance analysis of games played on artificial (Football Turf) and grass from the evaluation of UEFA Champions League and UEFA Cup. Introduction Following on from our initial technical

More information

Fredericton City Old Boys Soccer Club. Rules and Regulations As run by. Amended Page 1 of 14

Fredericton City Old Boys Soccer Club. Rules and Regulations As run by. Amended Page 1 of 14 & Rules and Regulations As run by Fredericton City Old Boys Soccer Club Amended February, 2018 Amended 04-28-15 Page 1 of 14 Table of Contents Topic Page Table of Contents --------------------------------------------------------------------

More information

Lab 1. Adiabatic and reversible compression of a gas

Lab 1. Adiabatic and reversible compression of a gas Lab 1. Adiabatic and reversible compression of a gas Introduction The initial and final states of an adiabatic and reversible volume change of an ideal gas can be determined by the First Law of Thermodynamics

More information

beestanbul RoboCup 3D Simulation League Team Description Paper 2012

beestanbul RoboCup 3D Simulation League Team Description Paper 2012 beestanbul RoboCup 3D Simulation League Team Description Paper 2012 Baris Demirdelen, Berkay Toku, Onuralp Ulusoy, Tuna Sonmez, Kubra Ayvaz, Elif Senyurek, and Sanem Sariel-Talay Artificial Intelligence

More information

Rugby Rules and Scoring

Rugby Rules and Scoring Rugby Rules and Scoring General: There are 3 types of Rugby o All three have different rules but they use the same field and the same ball. Rugby Union Rugby League Rugby 7 s Everything listed below is

More information

Meta-Knowledge as an engine in Classifier Combination

Meta-Knowledge as an engine in Classifier Combination 74 Meta-Knowledge as an engine in Classifier Combination Fredrick Edward Kitoogo 10 Department of Computer Science Faculty of Computing and IT, Makerere University Venansius Baryamureeba Department of

More information

GWYSL AYSO. Recreation Soccer. Experience Excellence in Soccer Education. U7-U8 Curriculum. The Soccer Education Specialists

GWYSL AYSO. Recreation Soccer. Experience Excellence in Soccer Education. U7-U8 Curriculum. The Soccer Education Specialists GWYSL AYSO Recreation Soccer U7-U8 Curriculum Experience Excellence in Soccer Education A division of USA Sport Group United Soccer Academy, Inc. 2 As the premier providers of soccer training on the East

More information

Predicting NBA Shots

Predicting NBA Shots Predicting NBA Shots Brett Meehan Stanford University https://github.com/brettmeehan/cs229 Final Project bmeehan2@stanford.edu Abstract This paper examines the application of various machine learning algorithms

More information

Pedestrian Dynamics: Models of Pedestrian Behaviour

Pedestrian Dynamics: Models of Pedestrian Behaviour Pedestrian Dynamics: Models of Pedestrian Behaviour John Ward 19 th January 2006 Contents Macro-scale sketch plan model Micro-scale agent based model for pedestrian movement Development of JPed Results

More information

The Universal Translator

The Universal Translator The Universal Translator David Cornfield dcornfield@apm.com Accellera Systems Initiative 1 Traditional Agents A Agent A Agent Chip ~ s m d A ~ s m d A B Agent B Agent m ~ s d B m ~ s d B Unit Level Chip

More information

Navigate to the golf data folder and make it your working directory. Load the data by typing

Navigate to the golf data folder and make it your working directory. Load the data by typing Golf Analysis 1.1 Introduction In a round, golfers have a number of choices to make. For a particular shot, is it better to use the longest club available to try to reach the green, or would it be better

More information

Northern SC U6 Playing Format 3v3

Northern SC U6 Playing Format 3v3 Northern SC U6 Playing Format 3v3 This document outlines guidelines for increasing the consistency of playing formations and terminology we use with U6 players. As players of this age may have different

More information

*This is a Recreational and Developmental league. The goal is to have fun and introduce them to soccer. WE DO NOT KEEP SCORE AT THIS AGE.

*This is a Recreational and Developmental league. The goal is to have fun and introduce them to soccer. WE DO NOT KEEP SCORE AT THIS AGE. U4, U5, & U6 Program Players on the field: 4v4 with NO goalies Roster Size: 6 8 Ball size: 3 Practices: no longer than 30 minutes per week Substitutions: Both Teams substitute at stoppage of play. Player

More information

Homewood Soccer Club Academy Curriculum Fall 2016

Homewood Soccer Club Academy Curriculum Fall 2016 Homewood Soccer Club Academy Curriculum Fall 2016 Overview and Philosophy Overview and Philosophy Homewood Soccer Club s Academy is the founda>on of our compe>>ve program. Our Academy Director is a USSF/USYS

More information

Using New Iterative Methods and Fine Grain Data to Rank College Football Teams. Maggie Wigness Michael Rowell & Chadd Williams Pacific University

Using New Iterative Methods and Fine Grain Data to Rank College Football Teams. Maggie Wigness Michael Rowell & Chadd Williams Pacific University Using New Iterative Methods and Fine Grain Data to Rank College Football Teams Maggie Wigness Michael Rowell & Chadd Williams Pacific University History of BCS Bowl Championship Series Ranking since 1998

More information

Data & Knowledge Engineering

Data & Knowledge Engineering Data & Knowledge Engineering 94 (2014) 110 131 Contents lists available at ScienceDirect Data & Knowledge Engineering journal homepage: www.elsevier.com/locate/datak Editorial Strategy planner: Graphical

More information

Ignition modelling Are our approaches aligned?

Ignition modelling Are our approaches aligned? Ignition modelling Are our approaches aligned? Lars Rogstadkjernet Gexcon Outline Experience from 3 rd party reviews Areas of diverging approach Dispersion and ignition modelling 3 rd party reviews 3 rd

More information

BASKETBALL PREDICTION ANALYSIS OF MARCH MADNESS GAMES CHRIS TSENG YIBO WANG

BASKETBALL PREDICTION ANALYSIS OF MARCH MADNESS GAMES CHRIS TSENG YIBO WANG BASKETBALL PREDICTION ANALYSIS OF MARCH MADNESS GAMES CHRIS TSENG YIBO WANG GOAL OF PROJECT The goal is to predict the winners between college men s basketball teams competing in the 2018 (NCAA) s March

More information

HOMEWORK BOOKLET DEVELOPMENT NAME: FORM: TEACHER:

HOMEWORK BOOKLET DEVELOPMENT NAME: FORM: TEACHER: HOMEWORK BOOKLET DEVELOPMENT NAME: FORM: TEACHER: There are seventeen laws of association football that must be followed. Each of these laws covers a different aspect of the game. This section will cover

More information

U6 Division Game Rules

U6 Division Game Rules U6 Division Game Rules Objective is to emphasize the skill of DRIBBLING. Minimize kicking without a purpose! Dribbling is the foundation and preparation for all the other fundamental skills of soccer,

More information