gene)ve rela)on Coreference Named En)ty Numerical Inference Lexical rela)on Spa)al reasoning Missing argument Event chain

Transcription

1

2 Coreference Lexical rela)on Missing argument Numerical Inference gene)ve rela)on Event chain Named En)ty Spa)al reasoning

3 Numerical Inference (We will show the distribu)on of these categories later)

4 Numerical Inference (We will show the distribu)on of these categories later)

5

6

7 114 pairs out of 4,351 TE pairs in two Japanese corpora [Shima et al 2011, Odani et al 2008] are inves)gated

8

9

10 is a Japanese actor born in Kagoshima Prefecture. He is 170 cm tall and weighs 62 kg. He specializes men's height

11 is a Japanese actor born in Kagoshima Prefecture. He is 170 cm tall and weighs 62 kg. He specializes men's height

12 is a Japanese actor born in Kagoshima Prefecture. He is 170 cm tall and He is 180 cm tall, has medium brown weighs 62 kg. He specializes hair, green eyes, and men's height

13 is a Japanese actor born in Kagoshima Prefecture. He is 170 cm tall and He is 180 cm tall, has medium brown weighs 62 kg. He specializes hair, green eyes, and men's height

14 He is 160 cm tall today but will is a Japanese actor born in Kagoshima Prefecture. probably He is 170 grow cm a bit tall more. and weighs 62 kg. He specializes hair, green eyes, and He is 180 cm tall, has medium brown men's height

15 He is 160 cm tall today but will is a Japanese actor born in Kagoshima Prefecture. probably He is 170 grow cm a bit tall more. and weighs 62 kg. He specializes hair, green eyes, and He is 180 cm tall, has medium brown men's height

16 He is 160 cm tall today but will is a Japanese actor born in Kagoshima Prefecture. probably He is 170 grow cm a bit tall more. and weighs 62 kg. He specializes hair, green eyes, and We do not men)on too tall (or too small) men so ozen He is 180 cm tall, has medium brown men's height

17 He is 160 cm tall today but will is a Japanese actor born in Kagoshima Prefecture. probably He is 170 grow cm a bit tall more. and weighs 62 kg. He specializes hair, green eyes, and We do not men)on too tall (or too small) men so ozen He is 180 cm tall, has medium brown men's height

18 He is 160 cm tall today but will is a Japanese actor born in Kagoshima Acquire Prefecture. probably common- sense He is 170 grow cm a bit tall more. and weighs 62 kg. He specializes hair, green eyes, and He is 180 cm tall, has medium brown knowledge from this distribu)on men's height

19 I earn $2,000 Oh a I earn month $6,000 and enjoy a month and I earn $5000 earn a $4,000 month. a My month. expenses my lifestyle I am a here s manager instead how!... of a erm store No! and I are about $3500 a month. I want to start earn $10,000 a month on profit... a savings plan Monthly salary

20

21 he is only 160 cm tall

22 he is only 160 cm tall

23 he is only 160 cm tall He is as tall as 204 cm????????

24 She is rather short, in part because her He is only 180 cm father is as well (he is only 160 cm He tall) is and as is tall surprisingly as 204 cm???????? He is as high as 170 cm

25 She is rather short, in part because her He is only 180 cm father is as well (he is only 160 cm He tall) is and as is tall surprisingly as 204 cm???????? He is as high as 170 cm

26

27 He is only 160 cm tall

28

29

30 iden)fy surface varia)on

31 convert to canonical units

32 recognize modifier (e.g., about, over more than, etc )

33 recognize modifier (e.g., about, over more than, etc )

34

35

36

37

38

39

40

41

42

43

44

45

46 He He He He is is is is cm cm cm cm tall tall tall tall

47 He He He He is is is is cm cm cm cm tall tall tall tall

48 He He He He is is is is cm cm cm cm tall tall tall tall normal rela)vely small normal normal

49 rela)vely small normal large I He He He lose is is is kg cm cm cm weight tall tall tall remove this sentence from dataset

50 He He He is is is I paid $100. cm cm cm tall tall tall unsure unsure unsure large

51

52

53 If I won 10- million- dollar lodery, non fact the amount of money that a person wins in a lodery Our shop have over 3,000 wonderful items!! boas)ng expected distribu)on distribu)on on the Web

54

55

56

57

58

59

60

61

62 small clue (e.g., only) large clue (e.g., as many as) 8% 6% 23 million numerical expressions 86% no textual clue

63

64