id
int64
0
14.4k
premise
stringlengths
9
49
hypothesis
stringlengths
8
33
label
class label
2 classes
heuristics
stringclasses
5 values
number_of_NPs
int32
1
4
semtag
stringclasses
16 values
0
ๅญฆ็”ŸใŒใƒ•ใƒƒใƒˆใƒœใƒผใƒซ้ธๆ‰‹ใ‚’ใŸใŸใ„ใฆใ„ใ‚‹
ใƒ•ใƒƒใƒˆใƒœใƒผใƒซ้ธๆ‰‹ใ‚’ๅญฆ็”ŸใŒใŸใŸใ„ใฆใ„ใ‚‹
0entailment
overlap-full
2
scrambling
1
่€ไบบใŒๅฅณใฎๅญใ‚’่ฟฝใ„ๆ‰•ใฃใŸ
ๅฅณใฎๅญใ‚’่€ไบบใŒ่ฟฝใ„ๆ‰•ใฃใŸ
0entailment
overlap-full
2
scrambling
2
็”ทๆ€งใŒๅคงไบบใ‚’่ฟฝใ„ๅ›žใ—ใŸ
ๅคงไบบใ‚’็”ทๆ€งใŒ่ฟฝใ„ๅ›žใ—ใŸ
0entailment
overlap-full
2
scrambling
3
ๅฅณใฎๅญใŒๅคงไบบใ‚’่ฆ‹ใฆใ„ใ‚‹
ๅคงไบบใ‚’ๅฅณใฎๅญใŒ่ฆ‹ใฆใ„ใ‚‹
0entailment
overlap-full
2
scrambling
4
็”ทๆ€งใŒใƒ•ใƒƒใƒˆใƒœใƒผใƒซ้ธๆ‰‹ใ‚’่ฟฝใ„ๆ‰•ใฃใŸ
ใƒ•ใƒƒใƒˆใƒœใƒผใƒซ้ธๆ‰‹ใ‚’็”ทๆ€งใŒ่ฟฝใ„ๆ‰•ใฃใŸ
0entailment
overlap-full
2
scrambling
5
ใƒ†ใƒ‹ใ‚น้ธๆ‰‹ใŒใ‚ซใƒƒใƒ—ใƒซใ‚’่ฟฝใ„ๅ›žใ—ใŸ
ใ‚ซใƒƒใƒ—ใƒซใ‚’ใƒ†ใƒ‹ใ‚น้ธๆ‰‹ใŒ่ฟฝใ„ๅ›žใ—ใŸ
0entailment
overlap-full
2
scrambling
6
ใ‚นใƒŽใƒผใƒœใƒผใƒ€ใƒผใŒๅญฆ็”Ÿใ‚’ใซใ‚‰ใ‚“ใงใ„ใ‚‹
ๅญฆ็”Ÿใ‚’ใ‚นใƒŽใƒผใƒœใƒผใƒ€ใƒผใŒใซใ‚‰ใ‚“ใงใ„ใ‚‹
0entailment
overlap-full
2
scrambling
7
็”ทๆ€งใŒใƒฉใ‚คใƒ€ใƒผใ‚’ๅŠฉใ‘ใŸ
ใƒฉใ‚คใƒ€ใƒผใ‚’็”ทๆ€งใŒๅŠฉใ‘ใŸ
0entailment
overlap-full
2
scrambling
8
็”ทๆ€งใŒๅญไพ›ใ‚’ใŸใŸใ„ใฆใ„ใ‚‹
ๅญไพ›ใ‚’็”ทๆ€งใŒใŸใŸใ„ใฆใ„ใ‚‹
0entailment
overlap-full
2
scrambling
10
ใ‚ตใƒผใƒ•ใ‚กใƒผใŒๅคงไบบใ‚’ๆŠผใ—ใŸ
ๅคงไบบใ‚’ใ‚ตใƒผใƒ•ใ‚กใƒผใŒๆŠผใ—ใŸ
0entailment
overlap-full
2
scrambling
11
ใ‚ตใƒผใƒ•ใ‚กใƒผใŒใ‚ซใƒƒใƒ—ใƒซใ‚’ใซใ‚‰ใ‚“ใงใ„ใ‚‹
ใ‚ซใƒƒใƒ—ใƒซใ‚’ใ‚ตใƒผใƒ•ใ‚กใƒผใŒใซใ‚‰ใ‚“ใงใ„ใ‚‹
0entailment
overlap-full
2
scrambling
12
่‹ฅ่€…ใŒใƒ•ใƒƒใƒˆใƒœใƒผใƒซ้ธๆ‰‹ใ‚’่ฆ‹ใฆใ„ใ‚‹
ใƒ•ใƒƒใƒˆใƒœใƒผใƒซ้ธๆ‰‹ใ‚’่‹ฅ่€…ใŒ่ฆ‹ใฆใ„ใ‚‹
0entailment
overlap-full
2
scrambling
13
ใƒ•ใƒƒใƒˆใƒœใƒผใƒซ้ธๆ‰‹ใŒไผš็คพๅ“กใ‚’ๅŠฉใ‘ใŸ
ไผš็คพๅ“กใ‚’ใƒ•ใƒƒใƒˆใƒœใƒผใƒซ้ธๆ‰‹ใŒๅŠฉใ‘ใŸ
0entailment
overlap-full
2
scrambling
14
ใ‚ตใƒผใƒ•ใ‚กใƒผใŒใƒฉใ‚คใƒ€ใƒผใ‚’่นดใฃใŸ
ใƒฉใ‚คใƒ€ใƒผใ‚’ใ‚ตใƒผใƒ•ใ‚กใƒผใŒ่นดใฃใŸ
0entailment
overlap-full
2
scrambling
15
็”ทใฎๅญใŒๅฅณๆ€งใ‚’่ฟฝใ„ใ‹ใ‘ใฆใ„ใ‚‹
ๅฅณๆ€งใ‚’็”ทใฎๅญใŒ่ฟฝใ„ใ‹ใ‘ใฆใ„ใ‚‹
0entailment
overlap-full
2
scrambling
16
ใ‚ตใƒผใƒ•ใ‚กใƒผใŒใƒฉใ‚คใƒ€ใƒผใ‚’ใซใ‚‰ใ‚“ใงใ„ใ‚‹
ใƒฉใ‚คใƒ€ใƒผใ‚’ใ‚ตใƒผใƒ•ใ‚กใƒผใŒใซใ‚‰ใ‚“ใงใ„ใ‚‹
0entailment
overlap-full
2
scrambling
17
ใƒ›ใƒƒใ‚ฑใƒผ้ธๆ‰‹ใŒใƒ†ใƒ‹ใ‚น้ธๆ‰‹ใ‚’่ฆ‹ใฆใ„ใ‚‹
ใƒ†ใƒ‹ใ‚น้ธๆ‰‹ใ‚’ใƒ›ใƒƒใ‚ฑใƒผ้ธๆ‰‹ใŒ่ฆ‹ใฆใ„ใ‚‹
0entailment
overlap-full
2
scrambling
18
็”ทใฎๅญใŒๅฅณใฎๅญใ‚’่ฟฝใ„ๅ›žใ—ใŸ
ๅฅณใฎๅญใ‚’็”ทใฎๅญใŒ่ฟฝใ„ๅ›žใ—ใŸ
0entailment
overlap-full
2
scrambling
19
ใ‚ซใƒƒใƒ—ใƒซใŒ่€ไบบใ‚’่ฆ‹ใคใ‚ใฆใ„ใ‚‹
่€ไบบใ‚’ใ‚ซใƒƒใƒ—ใƒซใŒ่ฆ‹ใคใ‚ใฆใ„ใ‚‹
0entailment
overlap-full
2
scrambling
20
ๅญไพ›ใŒ็”ทใฎๅญใ‚’ใ„ใ˜ใ‚ใฆใ„ใ‚‹
็”ทใฎๅญใ‚’ๅญไพ›ใŒใ„ใ˜ใ‚ใฆใ„ใ‚‹
0entailment
overlap-full
2
scrambling
21
ใ‚ซใƒƒใƒ—ใƒซใŒ่‹ฅ่€…ใ‚’่ฆ‹ใฆใ„ใ‚‹
่‹ฅ่€…ใ‚’ใ‚ซใƒƒใƒ—ใƒซใŒ่ฆ‹ใฆใ„ใ‚‹
0entailment
overlap-full
2
scrambling
22
ๅฅณๆ€งใŒใƒ†ใƒ‹ใ‚น้ธๆ‰‹ใ‚’่ฟฝใ„ใ‹ใ‘ใฆใ„ใ‚‹
ใƒ†ใƒ‹ใ‚น้ธๆ‰‹ใ‚’ๅฅณๆ€งใŒ่ฟฝใ„ใ‹ใ‘ใฆใ„ใ‚‹
0entailment
overlap-full
2
scrambling
24
ๅฅณๆ€งใŒ็”ทใฎๅญใ‚’ใŸใŸใ„ใฆใ„ใ‚‹
็”ทใฎๅญใ‚’ๅฅณๆ€งใŒใŸใŸใ„ใฆใ„ใ‚‹
0entailment
overlap-full
2
scrambling
25
ใƒ›ใƒƒใ‚ฑใƒผ้ธๆ‰‹ใŒใƒ•ใƒƒใƒˆใƒœใƒผใƒซ้ธๆ‰‹ใ‚’่ฆ‹ใฆใ„ใ‚‹
ใƒ•ใƒƒใƒˆใƒœใƒผใƒซ้ธๆ‰‹ใ‚’ใƒ›ใƒƒใ‚ฑใƒผ้ธๆ‰‹ใŒ่ฆ‹ใฆใ„ใ‚‹
0entailment
overlap-full
2
scrambling
26
่€ไบบใŒใƒ†ใƒ‹ใ‚น้ธๆ‰‹ใ‚’ๅŠฉใ‘ใŸ
ใƒ†ใƒ‹ใ‚น้ธๆ‰‹ใ‚’่€ไบบใŒๅŠฉใ‘ใŸ
0entailment
overlap-full
2
scrambling
27
ใƒ›ใƒƒใ‚ฑใƒผ้ธๆ‰‹ใŒใƒฉใ‚คใƒ€ใƒผใ‚’ใซใ‚‰ใ‚“ใงใ„ใ‚‹
ใƒฉใ‚คใƒ€ใƒผใ‚’ใƒ›ใƒƒใ‚ฑใƒผ้ธๆ‰‹ใŒใซใ‚‰ใ‚“ใงใ„ใ‚‹
0entailment
overlap-full
2
scrambling
28
ใ‚ตใƒผใƒ•ใ‚กใƒผใŒใƒ•ใƒƒใƒˆใƒœใƒผใƒซ้ธๆ‰‹ใ‚’ๆŠผใ—ใŸ
ใƒ•ใƒƒใƒˆใƒœใƒผใƒซ้ธๆ‰‹ใ‚’ใ‚ตใƒผใƒ•ใ‚กใƒผใŒๆŠผใ—ใŸ
0entailment
overlap-full
2
scrambling
30
่€ไบบใŒๅฅณใฎๅญใ‚’ใ„ใ˜ใ‚ใฆใ„ใ‚‹
ๅฅณใฎๅญใ‚’่€ไบบใŒใ„ใ˜ใ‚ใฆใ„ใ‚‹
0entailment
overlap-full
2
scrambling
31
ใ‚ซใƒƒใƒ—ใƒซใŒๅฅณๆ€งใ‚’่ฆ‹ใฆใ„ใ‚‹
ๅฅณๆ€งใ‚’ใ‚ซใƒƒใƒ—ใƒซใŒ่ฆ‹ใฆใ„ใ‚‹
0entailment
overlap-full
2
scrambling
32
่€ไบบใŒใ‚นใƒŽใƒผใƒœใƒผใƒ€ใƒผใ‚’ใŸใŸใ„ใฆใ„ใ‚‹
ใ‚นใƒŽใƒผใƒœใƒผใƒ€ใƒผใ‚’่€ไบบใŒใŸใŸใ„ใฆใ„ใ‚‹
0entailment
overlap-full
2
scrambling
33
ๅญไพ›ใŒๅคงไบบใ‚’่ฟฝใ„ๆ‰•ใฃใŸ
ๅคงไบบใ‚’ๅญไพ›ใŒ่ฟฝใ„ๆ‰•ใฃใŸ
0entailment
overlap-full
2
scrambling
34
ๅญฆ็”ŸใŒใƒฉใ‚คใƒ€ใƒผใ‚’ใŸใŸใ„ใฆใ„ใ‚‹
ใƒฉใ‚คใƒ€ใƒผใ‚’ๅญฆ็”ŸใŒใŸใŸใ„ใฆใ„ใ‚‹
0entailment
overlap-full
2
scrambling
35
ใƒ†ใƒ‹ใ‚น้ธๆ‰‹ใŒใ‚ตใƒผใƒ•ใ‚กใƒผใ‚’ๅŠฉใ‘ใŸ
ใ‚ตใƒผใƒ•ใ‚กใƒผใ‚’ใƒ†ใƒ‹ใ‚น้ธๆ‰‹ใŒๅŠฉใ‘ใŸ
0entailment
overlap-full
2
scrambling
36
ๅฅณใฎๅญใŒใ‚ซใƒƒใƒ—ใƒซใ‚’ๆŠผใ—ใŸ
ใ‚ซใƒƒใƒ—ใƒซใ‚’ๅฅณใฎๅญใŒๆŠผใ—ใŸ
0entailment
overlap-full
2
scrambling
37
ใƒ›ใƒƒใ‚ฑใƒผ้ธๆ‰‹ใŒใƒฉใ‚คใƒ€ใƒผใ‚’ๅŠฉใ‘ใŸ
ใƒฉใ‚คใƒ€ใƒผใ‚’ใƒ›ใƒƒใ‚ฑใƒผ้ธๆ‰‹ใŒๅŠฉใ‘ใŸ
0entailment
overlap-full
2
scrambling
38
ใ‚ตใƒผใƒ•ใ‚กใƒผใŒๅญฆ็”Ÿใ‚’ๆŒ‡ใ•ใ—ใฆใ„ใ‚‹
ๅญฆ็”Ÿใ‚’ใ‚ตใƒผใƒ•ใ‚กใƒผใŒๆŒ‡ใ•ใ—ใฆใ„ใ‚‹
0entailment
overlap-full
2
scrambling
39
ๅฅณๆ€งใŒใ‚ซใƒƒใƒ—ใƒซใ‚’่ฆ‹ใคใ‚ใฆใ„ใ‚‹
ใ‚ซใƒƒใƒ—ใƒซใ‚’ๅฅณๆ€งใŒ่ฆ‹ใคใ‚ใฆใ„ใ‚‹
0entailment
overlap-full
2
scrambling
40
ไผš็คพๅ“กใŒๅคงไบบใ‚’่ฆ‹ใคใ‚ใฆใ„ใ‚‹
ๅคงไบบใ‚’ไผš็คพๅ“กใŒ่ฆ‹ใคใ‚ใฆใ„ใ‚‹
0entailment
overlap-full
2
scrambling
41
ใƒ†ใƒ‹ใ‚น้ธๆ‰‹ใŒๅญฆ็”Ÿใ‚’่นดใฃใŸ
ๅญฆ็”Ÿใ‚’ใƒ†ใƒ‹ใ‚น้ธๆ‰‹ใŒ่นดใฃใŸ
0entailment
overlap-full
2
scrambling
42
ใƒ†ใƒ‹ใ‚น้ธๆ‰‹ใŒใ‚นใƒŽใƒผใƒœใƒผใƒ€ใƒผใ‚’ใซใ‚‰ใ‚“ใงใ„ใ‚‹
ใ‚นใƒŽใƒผใƒœใƒผใƒ€ใƒผใ‚’ใƒ†ใƒ‹ใ‚น้ธๆ‰‹ใŒใซใ‚‰ใ‚“ใงใ„ใ‚‹
0entailment
overlap-full
2
scrambling
43
ใƒ†ใƒ‹ใ‚น้ธๆ‰‹ใŒใ‚นใƒŽใƒผใƒœใƒผใƒ€ใƒผใ‚’่ฆ‹ใคใ‚ใฆใ„ใ‚‹
ใ‚นใƒŽใƒผใƒœใƒผใƒ€ใƒผใ‚’ใƒ†ใƒ‹ใ‚น้ธๆ‰‹ใŒ่ฆ‹ใคใ‚ใฆใ„ใ‚‹
0entailment
overlap-full
2
scrambling
44
ใƒ†ใƒ‹ใ‚น้ธๆ‰‹ใŒ่€ไบบใ‚’่ฆ‹ใคใ‚ใฆใ„ใ‚‹
่€ไบบใ‚’ใƒ†ใƒ‹ใ‚น้ธๆ‰‹ใŒ่ฆ‹ใคใ‚ใฆใ„ใ‚‹
0entailment
overlap-full
2
scrambling
45
ๅญไพ›ใŒ่€ไบบใ‚’ๆŒ‡ใ•ใ—ใฆใ„ใ‚‹
่€ไบบใ‚’ๅญไพ›ใŒๆŒ‡ใ•ใ—ใฆใ„ใ‚‹
0entailment
overlap-full
2
scrambling
46
ๅคงไบบใŒใ‚ตใƒผใƒ•ใ‚กใƒผใ‚’ใŸใŸใ„ใฆใ„ใ‚‹
ใ‚ตใƒผใƒ•ใ‚กใƒผใ‚’ๅคงไบบใŒใŸใŸใ„ใฆใ„ใ‚‹
0entailment
overlap-full
2
scrambling
47
ๅฅณๆ€งใŒใ‚ตใƒผใƒ•ใ‚กใƒผใ‚’่ฟฝใ„ๆ‰•ใฃใŸ
ใ‚ตใƒผใƒ•ใ‚กใƒผใ‚’ๅฅณๆ€งใŒ่ฟฝใ„ๆ‰•ใฃใŸ
0entailment
overlap-full
2
scrambling
48
ใƒ•ใƒƒใƒˆใƒœใƒผใƒซ้ธๆ‰‹ใŒใƒ†ใƒ‹ใ‚น้ธๆ‰‹ใ‚’่ฟฝใ„ใ‹ใ‘ใฆใ„ใ‚‹
ใƒ†ใƒ‹ใ‚น้ธๆ‰‹ใ‚’ใƒ•ใƒƒใƒˆใƒœใƒผใƒซ้ธๆ‰‹ใŒ่ฟฝใ„ใ‹ใ‘ใฆใ„ใ‚‹
0entailment
overlap-full
2
scrambling
49
ใƒฉใ‚คใƒ€ใƒผใŒๅฅณๆ€งใ‚’่ฟฝใ„ใ‹ใ‘ใฆใ„ใ‚‹
ๅฅณๆ€งใ‚’ใƒฉใ‚คใƒ€ใƒผใŒ่ฟฝใ„ใ‹ใ‘ใฆใ„ใ‚‹
0entailment
overlap-full
2
scrambling
50
ใƒ†ใƒ‹ใ‚น้ธๆ‰‹ใŒๅฅณๆ€งใ‚’ๆŠผใ—ใŸ
ๅฅณๆ€งใ‚’ใƒ†ใƒ‹ใ‚น้ธๆ‰‹ใŒๆŠผใ—ใŸ
0entailment
overlap-full
2
scrambling
51
ใƒฉใ‚คใƒ€ใƒผใŒๅญฆ็”Ÿใ‚’ๆŠผใ—ใŸ
ๅญฆ็”Ÿใ‚’ใƒฉใ‚คใƒ€ใƒผใŒๆŠผใ—ใŸ
0entailment
overlap-full
2
scrambling
52
ไผš็คพๅ“กใŒใ‚นใƒŽใƒผใƒœใƒผใƒ€ใƒผใ‚’ๅŠฉใ‘ใŸ
ใ‚นใƒŽใƒผใƒœใƒผใƒ€ใƒผใ‚’ไผš็คพๅ“กใŒๅŠฉใ‘ใŸ
0entailment
overlap-full
2
scrambling
53
ไผš็คพๅ“กใŒๅฅณๆ€งใ‚’่ฆ‹ใฆใ„ใ‚‹
ๅฅณๆ€งใ‚’ไผš็คพๅ“กใŒ่ฆ‹ใฆใ„ใ‚‹
0entailment
overlap-full
2
scrambling
54
ๅฅณๆ€งใŒใƒฉใ‚คใƒ€ใƒผใ‚’่ฆ‹ใคใ‚ใฆใ„ใ‚‹
ใƒฉใ‚คใƒ€ใƒผใ‚’ๅฅณๆ€งใŒ่ฆ‹ใคใ‚ใฆใ„ใ‚‹
0entailment
overlap-full
2
scrambling
55
ๅญฆ็”ŸใŒใƒ›ใƒƒใ‚ฑใƒผ้ธๆ‰‹ใ‚’่ฟฝใ„ๅ›žใ—ใŸ
ใƒ›ใƒƒใ‚ฑใƒผ้ธๆ‰‹ใ‚’ๅญฆ็”ŸใŒ่ฟฝใ„ๅ›žใ—ใŸ
0entailment
overlap-full
2
scrambling
56
ใƒ•ใƒƒใƒˆใƒœใƒผใƒซ้ธๆ‰‹ใŒ่‹ฅ่€…ใ‚’ใ„ใ˜ใ‚ใฆใ„ใ‚‹
่‹ฅ่€…ใ‚’ใƒ•ใƒƒใƒˆใƒœใƒผใƒซ้ธๆ‰‹ใŒใ„ใ˜ใ‚ใฆใ„ใ‚‹
0entailment
overlap-full
2
scrambling
57
ใ‚ซใƒƒใƒ—ใƒซใŒ็”ทใฎๅญใ‚’ๆŠผใ—ใŸ
็”ทใฎๅญใ‚’ใ‚ซใƒƒใƒ—ใƒซใŒๆŠผใ—ใŸ
0entailment
overlap-full
2
scrambling
58
ใƒ†ใƒ‹ใ‚น้ธๆ‰‹ใŒๅญไพ›ใ‚’ๆŠผใ—ใŸ
ๅญไพ›ใ‚’ใƒ†ใƒ‹ใ‚น้ธๆ‰‹ใŒๆŠผใ—ใŸ
0entailment
overlap-full
2
scrambling
59
ๅคงไบบใŒ็”ทใฎๅญใ‚’่ฆ‹ใคใ‚ใฆใ„ใ‚‹
็”ทใฎๅญใ‚’ๅคงไบบใŒ่ฆ‹ใคใ‚ใฆใ„ใ‚‹
0entailment
overlap-full
2
scrambling
60
ใƒ†ใƒ‹ใ‚น้ธๆ‰‹ใŒใ‚ตใƒผใƒ•ใ‚กใƒผใ‚’ใ„ใ˜ใ‚ใฆใ„ใ‚‹
ใ‚ตใƒผใƒ•ใ‚กใƒผใ‚’ใƒ†ใƒ‹ใ‚น้ธๆ‰‹ใŒใ„ใ˜ใ‚ใฆใ„ใ‚‹
0entailment
overlap-full
2
scrambling
61
ใ‚ตใƒผใƒ•ใ‚กใƒผใŒๅญฆ็”Ÿใ‚’่นดใฃใŸ
ๅญฆ็”Ÿใ‚’ใ‚ตใƒผใƒ•ใ‚กใƒผใŒ่นดใฃใŸ
0entailment
overlap-full
2
scrambling
62
ใ‚นใƒŽใƒผใƒœใƒผใƒ€ใƒผใŒใƒ†ใƒ‹ใ‚น้ธๆ‰‹ใ‚’่นดใฃใŸ
ใƒ†ใƒ‹ใ‚น้ธๆ‰‹ใ‚’ใ‚นใƒŽใƒผใƒœใƒผใƒ€ใƒผใŒ่นดใฃใŸ
0entailment
overlap-full
2
scrambling
63
ใ‚ตใƒผใƒ•ใ‚กใƒผใŒ็”ทๆ€งใ‚’ๅŠฉใ‘ใŸ
็”ทๆ€งใ‚’ใ‚ตใƒผใƒ•ใ‚กใƒผใŒๅŠฉใ‘ใŸ
0entailment
overlap-full
2
scrambling
64
ๅฅณๆ€งใŒ็”ทใฎๅญใ‚’่นดใฃใŸ
็”ทใฎๅญใ‚’ๅฅณๆ€งใŒ่นดใฃใŸ
0entailment
overlap-full
2
scrambling
66
ใƒฉใ‚คใƒ€ใƒผใŒใ‚นใƒŽใƒผใƒœใƒผใƒ€ใƒผใ‚’่ฟฝใ„ใ‹ใ‘ใฆใ„ใ‚‹
ใ‚นใƒŽใƒผใƒœใƒผใƒ€ใƒผใ‚’ใƒฉใ‚คใƒ€ใƒผใŒ่ฟฝใ„ใ‹ใ‘ใฆใ„ใ‚‹
0entailment
overlap-full
2
scrambling
67
ๅญไพ›ใŒใƒฉใ‚คใƒ€ใƒผใ‚’ๆŠผใ—ใŸ
ใƒฉใ‚คใƒ€ใƒผใ‚’ๅญไพ›ใŒๆŠผใ—ใŸ
0entailment
overlap-full
2
scrambling
68
ใƒ›ใƒƒใ‚ฑใƒผ้ธๆ‰‹ใŒใ‚ซใƒƒใƒ—ใƒซใ‚’ๆŠผใ—ใŸ
ใ‚ซใƒƒใƒ—ใƒซใ‚’ใƒ›ใƒƒใ‚ฑใƒผ้ธๆ‰‹ใŒๆŠผใ—ใŸ
0entailment
overlap-full
2
scrambling
69
ไผš็คพๅ“กใŒๅฅณๆ€งใ‚’ๆŒ‡ใ•ใ—ใฆใ„ใ‚‹
ๅฅณๆ€งใ‚’ไผš็คพๅ“กใŒๆŒ‡ใ•ใ—ใฆใ„ใ‚‹
0entailment
overlap-full
2
scrambling
70
ใƒฉใ‚คใƒ€ใƒผใŒไผš็คพๅ“กใ‚’ใ„ใ˜ใ‚ใฆใ„ใ‚‹
ไผš็คพๅ“กใ‚’ใƒฉใ‚คใƒ€ใƒผใŒใ„ใ˜ใ‚ใฆใ„ใ‚‹
0entailment
overlap-full
2
scrambling
71
ๅฅณๆ€งใŒใ‚ซใƒƒใƒ—ใƒซใ‚’ใ„ใ˜ใ‚ใฆใ„ใ‚‹
ใ‚ซใƒƒใƒ—ใƒซใ‚’ๅฅณๆ€งใŒใ„ใ˜ใ‚ใฆใ„ใ‚‹
0entailment
overlap-full
2
scrambling
72
ใƒ•ใƒƒใƒˆใƒœใƒผใƒซ้ธๆ‰‹ใŒไผš็คพๅ“กใ‚’ใ„ใ˜ใ‚ใฆใ„ใ‚‹
ไผš็คพๅ“กใ‚’ใƒ•ใƒƒใƒˆใƒœใƒผใƒซ้ธๆ‰‹ใŒใ„ใ˜ใ‚ใฆใ„ใ‚‹
0entailment
overlap-full
2
scrambling
73
ๅญไพ›ใŒ็”ทๆ€งใ‚’ใŸใŸใ„ใฆใ„ใ‚‹
็”ทๆ€งใ‚’ๅญไพ›ใŒใŸใŸใ„ใฆใ„ใ‚‹
0entailment
overlap-full
2
scrambling
74
ใƒ•ใƒƒใƒˆใƒœใƒผใƒซ้ธๆ‰‹ใŒใ‚ซใƒƒใƒ—ใƒซใ‚’ใŸใŸใ„ใฆใ„ใ‚‹
ใ‚ซใƒƒใƒ—ใƒซใ‚’ใƒ•ใƒƒใƒˆใƒœใƒผใƒซ้ธๆ‰‹ใŒใŸใŸใ„ใฆใ„ใ‚‹
0entailment
overlap-full
2
scrambling
75
็”ทๆ€งใŒใ‚ซใƒƒใƒ—ใƒซใ‚’่ฟฝใ„ๆ‰•ใฃใŸ
ใ‚ซใƒƒใƒ—ใƒซใ‚’็”ทๆ€งใŒ่ฟฝใ„ๆ‰•ใฃใŸ
0entailment
overlap-full
2
scrambling
76
่‹ฅ่€…ใŒใ‚นใƒŽใƒผใƒœใƒผใƒ€ใƒผใ‚’่นดใฃใŸ
ใ‚นใƒŽใƒผใƒœใƒผใƒ€ใƒผใ‚’่‹ฅ่€…ใŒ่นดใฃใŸ
0entailment
overlap-full
2
scrambling
77
ใ‚ซใƒƒใƒ—ใƒซใŒใƒ†ใƒ‹ใ‚น้ธๆ‰‹ใ‚’่ฆ‹ใคใ‚ใฆใ„ใ‚‹
ใƒ†ใƒ‹ใ‚น้ธๆ‰‹ใ‚’ใ‚ซใƒƒใƒ—ใƒซใŒ่ฆ‹ใคใ‚ใฆใ„ใ‚‹
0entailment
overlap-full
2
scrambling
78
ใ‚ตใƒผใƒ•ใ‚กใƒผใŒๅฅณใฎๅญใ‚’่นดใฃใŸ
ๅฅณใฎๅญใ‚’ใ‚ตใƒผใƒ•ใ‚กใƒผใŒ่นดใฃใŸ
0entailment
overlap-full
2
scrambling
79
ใƒฉใ‚คใƒ€ใƒผใŒๅคงไบบใ‚’่ฆ‹ใคใ‚ใฆใ„ใ‚‹
ๅคงไบบใ‚’ใƒฉใ‚คใƒ€ใƒผใŒ่ฆ‹ใคใ‚ใฆใ„ใ‚‹
0entailment
overlap-full
2
scrambling
80
ๅคงไบบใŒๅญไพ›ใ‚’่ฟฝใ„ๅ›žใ—ใŸ
ๅญไพ›ใ‚’ๅคงไบบใŒ่ฟฝใ„ๅ›žใ—ใŸ
0entailment
overlap-full
2
scrambling
81
ๅญไพ›ใŒใƒ•ใƒƒใƒˆใƒœใƒผใƒซ้ธๆ‰‹ใ‚’่ฟฝใ„ใ‹ใ‘ใฆใ„ใ‚‹
ใƒ•ใƒƒใƒˆใƒœใƒผใƒซ้ธๆ‰‹ใ‚’ๅญไพ›ใŒ่ฟฝใ„ใ‹ใ‘ใฆใ„ใ‚‹
0entailment
overlap-full
2
scrambling
82
็”ทใฎๅญใŒๅฅณใฎๅญใ‚’่ฟฝใ„ๅ›žใ—ใŸ
ๅฅณใฎๅญใ‚’็”ทใฎๅญใŒ่ฟฝใ„ๅ›žใ—ใŸ
0entailment
overlap-full
2
scrambling
83
็”ทๆ€งใŒใƒ•ใƒƒใƒˆใƒœใƒผใƒซ้ธๆ‰‹ใ‚’ๆŒ‡ใ•ใ—ใฆใ„ใ‚‹
ใƒ•ใƒƒใƒˆใƒœใƒผใƒซ้ธๆ‰‹ใ‚’็”ทๆ€งใŒๆŒ‡ใ•ใ—ใฆใ„ใ‚‹
0entailment
overlap-full
2
scrambling
84
็”ทใฎๅญใŒใƒฉใ‚คใƒ€ใƒผใ‚’่ฟฝใ„ๅ›žใ—ใŸ
ใƒฉใ‚คใƒ€ใƒผใ‚’็”ทใฎๅญใŒ่ฟฝใ„ๅ›žใ—ใŸ
0entailment
overlap-full
2
scrambling
86
ใ‚ตใƒผใƒ•ใ‚กใƒผใŒใƒ›ใƒƒใ‚ฑใƒผ้ธๆ‰‹ใ‚’ใŸใŸใ„ใฆใ„ใ‚‹
ใƒ›ใƒƒใ‚ฑใƒผ้ธๆ‰‹ใ‚’ใ‚ตใƒผใƒ•ใ‚กใƒผใŒใŸใŸใ„ใฆใ„ใ‚‹
0entailment
overlap-full
2
scrambling
87
ใƒฉใ‚คใƒ€ใƒผใŒ่‹ฅ่€…ใ‚’ใซใ‚‰ใ‚“ใงใ„ใ‚‹
่‹ฅ่€…ใ‚’ใƒฉใ‚คใƒ€ใƒผใŒใซใ‚‰ใ‚“ใงใ„ใ‚‹
0entailment
overlap-full
2
scrambling
88
ๅฅณๆ€งใŒ็”ทๆ€งใ‚’่นดใฃใŸ
็”ทๆ€งใ‚’ๅฅณๆ€งใŒ่นดใฃใŸ
0entailment
overlap-full
2
scrambling
89
ๅฅณใฎๅญใŒใ‚ซใƒƒใƒ—ใƒซใ‚’ใŸใŸใ„ใฆใ„ใ‚‹
ใ‚ซใƒƒใƒ—ใƒซใ‚’ๅฅณใฎๅญใŒใŸใŸใ„ใฆใ„ใ‚‹
0entailment
overlap-full
2
scrambling
90
ๅญไพ›ใŒๅฅณๆ€งใ‚’ใซใ‚‰ใ‚“ใงใ„ใ‚‹
ๅฅณๆ€งใ‚’ๅญไพ›ใŒใซใ‚‰ใ‚“ใงใ„ใ‚‹
0entailment
overlap-full
2
scrambling
91
็”ทๆ€งใŒๅฅณใฎๅญใ‚’่ฟฝใ„ๆ‰•ใฃใŸ
ๅฅณใฎๅญใ‚’็”ทๆ€งใŒ่ฟฝใ„ๆ‰•ใฃใŸ
0entailment
overlap-full
2
scrambling
92
ใ‚ซใƒƒใƒ—ใƒซใŒใ‚นใƒŽใƒผใƒœใƒผใƒ€ใƒผใ‚’ๅŠฉใ‘ใŸ
ใ‚นใƒŽใƒผใƒœใƒผใƒ€ใƒผใ‚’ใ‚ซใƒƒใƒ—ใƒซใŒๅŠฉใ‘ใŸ
0entailment
overlap-full
2
scrambling
93
ๅญฆ็”ŸใŒ่‹ฅ่€…ใ‚’ใŸใŸใ„ใฆใ„ใ‚‹
่‹ฅ่€…ใ‚’ๅญฆ็”ŸใŒใŸใŸใ„ใฆใ„ใ‚‹
0entailment
overlap-full
2
scrambling
94
ใƒฉใ‚คใƒ€ใƒผใŒๅฅณใฎๅญใ‚’ๅŠฉใ‘ใŸ
ๅฅณใฎๅญใ‚’ใƒฉใ‚คใƒ€ใƒผใŒๅŠฉใ‘ใŸ
0entailment
overlap-full
2
scrambling
95
ๅญไพ›ใŒใƒ›ใƒƒใ‚ฑใƒผ้ธๆ‰‹ใ‚’่นดใฃใŸ
ใƒ›ใƒƒใ‚ฑใƒผ้ธๆ‰‹ใ‚’ๅญไพ›ใŒ่นดใฃใŸ
0entailment
overlap-full
2
scrambling
96
ใƒ†ใƒ‹ใ‚น้ธๆ‰‹ใŒใƒ•ใƒƒใƒˆใƒœใƒผใƒซ้ธๆ‰‹ใ‚’่ฆ‹ใฆใ„ใ‚‹
ใƒ•ใƒƒใƒˆใƒœใƒผใƒซ้ธๆ‰‹ใ‚’ใƒ†ใƒ‹ใ‚น้ธๆ‰‹ใŒ่ฆ‹ใฆใ„ใ‚‹
0entailment
overlap-full
2
scrambling
97
ๅญไพ›ใŒๅคงไบบใ‚’่ฟฝใ„ๆ‰•ใฃใŸ
ๅคงไบบใ‚’ๅญไพ›ใŒ่ฟฝใ„ๆ‰•ใฃใŸ
0entailment
overlap-full
2
scrambling
98
ใƒ•ใƒƒใƒˆใƒœใƒผใƒซ้ธๆ‰‹ใŒๅฅณใฎๅญใ‚’ๆŠผใ—ใŸ
ๅฅณใฎๅญใ‚’ใƒ•ใƒƒใƒˆใƒœใƒผใƒซ้ธๆ‰‹ใŒๆŠผใ—ใŸ
0entailment
overlap-full
2
scrambling
99
ๅฅณๆ€งใŒใƒ•ใƒƒใƒˆใƒœใƒผใƒซ้ธๆ‰‹ใ‚’่ฆ‹ใคใ‚ใฆใ„ใ‚‹
ใƒ•ใƒƒใƒˆใƒœใƒผใƒซ้ธๆ‰‹ใ‚’ๅฅณๆ€งใŒ่ฆ‹ใคใ‚ใฆใ„ใ‚‹
0entailment
overlap-full
2
scrambling
101
่‹ฅ่€…ใŒไผš็คพๅ“กใ‚’ๆŒ‡ใ•ใ—ใฆใ„ใ‚‹
ไผš็คพๅ“กใฏ่‹ฅ่€…ใŒๆŒ‡ใ•ใ—ใฆใ„ใ‚‹
0entailment
overlap-nonorder
2
scrambling
102
ๅฅณใฎๅญใŒ่€ไบบใ‚’ๆŠผใ—ใŸ
่€ไบบใฏๅฅณใฎๅญใŒๆŠผใ—ใŸ
0entailment
overlap-nonorder
2
scrambling
103
ใ‚ซใƒƒใƒ—ใƒซใŒๅฅณๆ€งใ‚’ๆŒ‡ใ•ใ—ใฆใ„ใ‚‹
ๅฅณๆ€งใฏใ‚ซใƒƒใƒ—ใƒซใŒๆŒ‡ใ•ใ—ใฆใ„ใ‚‹
0entailment
overlap-nonorder
2
scrambling
104
่‹ฅ่€…ใŒใƒ•ใƒƒใƒˆใƒœใƒผใƒซ้ธๆ‰‹ใ‚’่ฆ‹ใฆใ„ใ‚‹
ใƒ•ใƒƒใƒˆใƒœใƒผใƒซ้ธๆ‰‹ใฏ่‹ฅ่€…ใŒ่ฆ‹ใฆใ„ใ‚‹
0entailment
overlap-nonorder
2
scrambling
105
ใ‚ซใƒƒใƒ—ใƒซใŒๅญไพ›ใ‚’ใซใ‚‰ใ‚“ใงใ„ใ‚‹
ๅญไพ›ใฏใ‚ซใƒƒใƒ—ใƒซใŒใซใ‚‰ใ‚“ใงใ„ใ‚‹
0entailment
overlap-nonorder
2
scrambling

Dataset Card for JaNLI

Dataset Summary

The JaNLI (Japanese Adversarial NLI) dataset, inspired by the English HANS dataset, is designed to necessitate an understanding of Japanese linguistic phenomena and to illuminate the vulnerabilities of models.

Languages

The language data in JaNLI is in Japanese (BCP-47 ja-JP).

Dataset Structure

Data Instances

When loading a specific configuration, users has to append a version dependent suffix:

import datasets as ds

dataset: ds.DatasetDict = ds.load_dataset("hpprc/janli")
print(dataset)
# DatasetDict({
#     train: Dataset({
#         features: ['id', 'premise', 'hypothesis', 'label', 'heuristics', 'number_of_NPs', 'semtag'],
#         num_rows: 13680
#     })
#     test: Dataset({
#         features: ['id', 'premise', 'hypothesis', 'label', 'heuristics', 'number_of_NPs', 'semtag'],
#         num_rows: 720
#     })
# })

dataset: ds.DatasetDict = ds.load_dataset("hpprc/janli", name="original")
print(dataset)
# DatasetDict({
#     train: Dataset({
#         features: ['id', 'sentence_A_Ja', 'sentence_B_Ja', 'entailment_label_Ja', 'heuristics', 'number_of_NPs', 'semtag'],
#         num_rows: 13680
#     })
#     test: Dataset({
#         features: ['id', 'sentence_A_Ja', 'sentence_B_Ja', 'entailment_label_Ja', 'heuristics', 'number_of_NPs', 'semtag'],
#         num_rows: 720
#     })
# })

base

An example of looks as follows:

{
  'id': 12,
  'premise': '่‹ฅ่€…ใŒใƒ•ใƒƒใƒˆใƒœใƒผใƒซ้ธๆ‰‹ใ‚’่ฆ‹ใฆใ„ใ‚‹',
  'hypothesis': 'ใƒ•ใƒƒใƒˆใƒœใƒผใƒซ้ธๆ‰‹ใ‚’่‹ฅ่€…ใŒ่ฆ‹ใฆใ„ใ‚‹',
  'label': 0,
  'heuristics': 'overlap-full',
  'number_of_NPs': 2,
  'semtag': 'scrambling'
}

original

An example of looks as follows:

{
  'id': 12,
  'sentence_A_Ja': '่‹ฅ่€…ใŒใƒ•ใƒƒใƒˆใƒœใƒผใƒซ้ธๆ‰‹ใ‚’่ฆ‹ใฆใ„ใ‚‹',
  'sentence_B_Ja': 'ใƒ•ใƒƒใƒˆใƒœใƒผใƒซ้ธๆ‰‹ใ‚’่‹ฅ่€…ใŒ่ฆ‹ใฆใ„ใ‚‹',
  'entailment_label_Ja': 0,
  'heuristics': 'overlap-full',
  'number_of_NPs': 2,
  'semtag': 'scrambling'
}

Data Fields

base

A version adopting the column names of a typical NLI dataset.

Name Description
id The number of the sentence pair.
premise The premise (sentence_A_Ja).
hypothesis The hypothesis (sentence_B_Ja).
label The correct label for the sentence pair (either entailment or non-entailment); in the setting described in the paper, non-entailment = neutral + contradiction (entailment_label_Ja).
heuristics The heuristics (structural pattern) tag. The tags are: subsequence, constituent, full-overlap, order-subset, and mixed-subset.
number_of_NPs The number of noun phrase in a sentence.
semtag The linguistic phenomena tag.

original

The original version retaining the unaltered column names.

Name Description
id The number of the sentence pair.
sentence_A_Ja The premise.
sentence_B_Ja The hypothesis.
entailment_label_Ja The correct label for this sentence pair (either entailment or non-entailment); in the setting described in the paper, non-entailment = neutral + contradiction
heuristics The heuristics (structural pattern) tag. The tags are: subsequence, constituent, full-overlap, order-subset, and mixed-subset.
number_of_NPs The number of noun phrase in a sentence.
semtag The linguistic phenomena tag.

Data Splits

name train validation test
base 13,680 720
original 13,680 720

Annotations

The annotation process for this Japanese NLI dataset involves tagging each pair (P, H) of a premise and hypothesis with a label for structural pattern and linguistic phenomenon. The structural relationship between premise and hypothesis sentences is classified into five patterns, with each pattern associated with a type of heuristic that can lead to incorrect predictions of the entailment relation. Additionally, 11 categories of Japanese linguistic phenomena and constructions are focused on for generating the five patterns of adversarial inferences.

For each linguistic phenomenon, a template for the premise sentence P is fixed, and multiple templates for hypothesis sentences H are created. In total, 144 templates for (P, H) pairs are produced. Each pair of premise and hypothesis sentences is tagged with an entailment label (entailment or non-entailment), a structural pattern, and a linguistic phenomenon label.

The JaNLI dataset is generated by instantiating each template 100 times, resulting in a total of 14,400 examples. The same number of entailment and non-entailment examples are generated for each phenomenon. The structural patterns are annotated with the templates for each linguistic phenomenon, and the ratio of entailment and non-entailment examples is not necessarily 1:1 for each pattern. The dataset uses a total of 158 words (nouns and verbs), which occur more than 20 times in the JSICK and JSNLI datasets.

Additional Information

Licensing Information

CC BY-SA 4.0

Citation Information

@InProceedings{yanaka-EtAl:2021:blackbox,
  author    = {Yanaka, Hitomi and Mineshima, Koji},
  title     = {Assessing the Generalization Capacity of Pre-trained Language Models through Japanese Adversarial Natural Language Inference},
  booktitle = {Proceedings of the 2021 EMNLP Workshop BlackboxNLP: Analyzing and Interpreting Neural Networks for NLP (BlackboxNLP2021)},
  url       = {https://aclanthology.org/2021.blackboxnlp-1.26/},
  year      = {2021},
}

Contributions

Thanks to Hitomi Yanaka and Koji Mineshima for creating this dataset.

Downloads last month
75
Edit dataset card

Models trained or fine-tuned on hpprc/janli