Datasets:
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 | ๅญฆ็ใใใใใใผใซ้ธๆใใใใใฆใใ | ใใใใใผใซ้ธๆใๅญฆ็ใใใใใฆใใ | 0entailment
| overlap-full | 2 | scrambling |
1 | ่ไบบใๅฅณใฎๅญใ่ฟฝใๆใฃใ | ๅฅณใฎๅญใ่ไบบใ่ฟฝใๆใฃใ | 0entailment
| overlap-full | 2 | scrambling |
2 | ็ทๆงใๅคงไบบใ่ฟฝใๅใใ | ๅคงไบบใ็ทๆงใ่ฟฝใๅใใ | 0entailment
| overlap-full | 2 | scrambling |
3 | ๅฅณใฎๅญใๅคงไบบใ่ฆใฆใใ | ๅคงไบบใๅฅณใฎๅญใ่ฆใฆใใ | 0entailment
| overlap-full | 2 | scrambling |
4 | ็ทๆงใใใใใใผใซ้ธๆใ่ฟฝใๆใฃใ | ใใใใใผใซ้ธๆใ็ทๆงใ่ฟฝใๆใฃใ | 0entailment
| overlap-full | 2 | scrambling |
5 | ใใใน้ธๆใใซใใใซใ่ฟฝใๅใใ | ใซใใใซใใใใน้ธๆใ่ฟฝใๅใใ | 0entailment
| overlap-full | 2 | scrambling |
6 | ในใใผใใผใใผใๅญฆ็ใใซใใใงใใ | ๅญฆ็ใในใใผใใผใใผใใซใใใงใใ | 0entailment
| overlap-full | 2 | scrambling |
7 | ็ทๆงใใฉใคใใผใๅฉใใ | ใฉใคใใผใ็ทๆงใๅฉใใ | 0entailment
| overlap-full | 2 | scrambling |
8 | ็ทๆงใๅญไพใใใใใฆใใ | ๅญไพใ็ทๆงใใใใใฆใใ | 0entailment
| overlap-full | 2 | scrambling |
10 | ใตใผใใกใผใๅคงไบบใๆผใใ | ๅคงไบบใใตใผใใกใผใๆผใใ | 0entailment
| overlap-full | 2 | scrambling |
11 | ใตใผใใกใผใใซใใใซใใซใใใงใใ | ใซใใใซใใตใผใใกใผใใซใใใงใใ | 0entailment
| overlap-full | 2 | scrambling |
12 | ่ฅ่
ใใใใใใผใซ้ธๆใ่ฆใฆใใ | ใใใใใผใซ้ธๆใ่ฅ่
ใ่ฆใฆใใ | 0entailment
| overlap-full | 2 | scrambling |
13 | ใใใใใผใซ้ธๆใไผ็คพๅกใๅฉใใ | ไผ็คพๅกใใใใใใผใซ้ธๆใๅฉใใ | 0entailment
| overlap-full | 2 | scrambling |
14 | ใตใผใใกใผใใฉใคใใผใ่นดใฃใ | ใฉใคใใผใใตใผใใกใผใ่นดใฃใ | 0entailment
| overlap-full | 2 | scrambling |
15 | ็ทใฎๅญใๅฅณๆงใ่ฟฝใใใใฆใใ | ๅฅณๆงใ็ทใฎๅญใ่ฟฝใใใใฆใใ | 0entailment
| overlap-full | 2 | scrambling |
16 | ใตใผใใกใผใใฉใคใใผใใซใใใงใใ | ใฉใคใใผใใตใผใใกใผใใซใใใงใใ | 0entailment
| overlap-full | 2 | scrambling |
17 | ใใใฑใผ้ธๆใใใใน้ธๆใ่ฆใฆใใ | ใใใน้ธๆใใใใฑใผ้ธๆใ่ฆใฆใใ | 0entailment
| overlap-full | 2 | scrambling |
18 | ็ทใฎๅญใๅฅณใฎๅญใ่ฟฝใๅใใ | ๅฅณใฎๅญใ็ทใฎๅญใ่ฟฝใๅใใ | 0entailment
| overlap-full | 2 | scrambling |
19 | ใซใใใซใ่ไบบใ่ฆใคใใฆใใ | ่ไบบใใซใใใซใ่ฆใคใใฆใใ | 0entailment
| overlap-full | 2 | scrambling |
20 | ๅญไพใ็ทใฎๅญใใใใใฆใใ | ็ทใฎๅญใๅญไพใใใใใฆใใ | 0entailment
| overlap-full | 2 | scrambling |
21 | ใซใใใซใ่ฅ่
ใ่ฆใฆใใ | ่ฅ่
ใใซใใใซใ่ฆใฆใใ | 0entailment
| overlap-full | 2 | scrambling |
22 | ๅฅณๆงใใใใน้ธๆใ่ฟฝใใใใฆใใ | ใใใน้ธๆใๅฅณๆงใ่ฟฝใใใใฆใใ | 0entailment
| overlap-full | 2 | scrambling |
24 | ๅฅณๆงใ็ทใฎๅญใใใใใฆใใ | ็ทใฎๅญใๅฅณๆงใใใใใฆใใ | 0entailment
| overlap-full | 2 | scrambling |
25 | ใใใฑใผ้ธๆใใใใใใผใซ้ธๆใ่ฆใฆใใ | ใใใใใผใซ้ธๆใใใใฑใผ้ธๆใ่ฆใฆใใ | 0entailment
| overlap-full | 2 | scrambling |
26 | ่ไบบใใใใน้ธๆใๅฉใใ | ใใใน้ธๆใ่ไบบใๅฉใใ | 0entailment
| overlap-full | 2 | scrambling |
27 | ใใใฑใผ้ธๆใใฉใคใใผใใซใใใงใใ | ใฉใคใใผใใใใฑใผ้ธๆใใซใใใงใใ | 0entailment
| overlap-full | 2 | scrambling |
28 | ใตใผใใกใผใใใใใใผใซ้ธๆใๆผใใ | ใใใใใผใซ้ธๆใใตใผใใกใผใๆผใใ | 0entailment
| overlap-full | 2 | scrambling |
30 | ่ไบบใๅฅณใฎๅญใใใใใฆใใ | ๅฅณใฎๅญใ่ไบบใใใใใฆใใ | 0entailment
| overlap-full | 2 | scrambling |
31 | ใซใใใซใๅฅณๆงใ่ฆใฆใใ | ๅฅณๆงใใซใใใซใ่ฆใฆใใ | 0entailment
| overlap-full | 2 | scrambling |
32 | ่ไบบใในใใผใใผใใผใใใใใฆใใ | ในใใผใใผใใผใ่ไบบใใใใใฆใใ | 0entailment
| overlap-full | 2 | scrambling |
33 | ๅญไพใๅคงไบบใ่ฟฝใๆใฃใ | ๅคงไบบใๅญไพใ่ฟฝใๆใฃใ | 0entailment
| overlap-full | 2 | scrambling |
34 | ๅญฆ็ใใฉใคใใผใใใใใฆใใ | ใฉใคใใผใๅญฆ็ใใใใใฆใใ | 0entailment
| overlap-full | 2 | scrambling |
35 | ใใใน้ธๆใใตใผใใกใผใๅฉใใ | ใตใผใใกใผใใใใน้ธๆใๅฉใใ | 0entailment
| overlap-full | 2 | scrambling |
36 | ๅฅณใฎๅญใใซใใใซใๆผใใ | ใซใใใซใๅฅณใฎๅญใๆผใใ | 0entailment
| overlap-full | 2 | scrambling |
37 | ใใใฑใผ้ธๆใใฉใคใใผใๅฉใใ | ใฉใคใใผใใใใฑใผ้ธๆใๅฉใใ | 0entailment
| overlap-full | 2 | scrambling |
38 | ใตใผใใกใผใๅญฆ็ใๆใใใฆใใ | ๅญฆ็ใใตใผใใกใผใๆใใใฆใใ | 0entailment
| overlap-full | 2 | scrambling |
39 | ๅฅณๆงใใซใใใซใ่ฆใคใใฆใใ | ใซใใใซใๅฅณๆงใ่ฆใคใใฆใใ | 0entailment
| overlap-full | 2 | scrambling |
40 | ไผ็คพๅกใๅคงไบบใ่ฆใคใใฆใใ | ๅคงไบบใไผ็คพๅกใ่ฆใคใใฆใใ | 0entailment
| overlap-full | 2 | scrambling |
41 | ใใใน้ธๆใๅญฆ็ใ่นดใฃใ | ๅญฆ็ใใใใน้ธๆใ่นดใฃใ | 0entailment
| overlap-full | 2 | scrambling |
42 | ใใใน้ธๆใในใใผใใผใใผใใซใใใงใใ | ในใใผใใผใใผใใใใน้ธๆใใซใใใงใใ | 0entailment
| overlap-full | 2 | scrambling |
43 | ใใใน้ธๆใในใใผใใผใใผใ่ฆใคใใฆใใ | ในใใผใใผใใผใใใใน้ธๆใ่ฆใคใใฆใใ | 0entailment
| overlap-full | 2 | scrambling |
44 | ใใใน้ธๆใ่ไบบใ่ฆใคใใฆใใ | ่ไบบใใใใน้ธๆใ่ฆใคใใฆใใ | 0entailment
| overlap-full | 2 | scrambling |
45 | ๅญไพใ่ไบบใๆใใใฆใใ | ่ไบบใๅญไพใๆใใใฆใใ | 0entailment
| overlap-full | 2 | scrambling |
46 | ๅคงไบบใใตใผใใกใผใใใใใฆใใ | ใตใผใใกใผใๅคงไบบใใใใใฆใใ | 0entailment
| overlap-full | 2 | scrambling |
47 | ๅฅณๆงใใตใผใใกใผใ่ฟฝใๆใฃใ | ใตใผใใกใผใๅฅณๆงใ่ฟฝใๆใฃใ | 0entailment
| overlap-full | 2 | scrambling |
48 | ใใใใใผใซ้ธๆใใใใน้ธๆใ่ฟฝใใใใฆใใ | ใใใน้ธๆใใใใใใผใซ้ธๆใ่ฟฝใใใใฆใใ | 0entailment
| overlap-full | 2 | scrambling |
49 | ใฉใคใใผใๅฅณๆงใ่ฟฝใใใใฆใใ | ๅฅณๆงใใฉใคใใผใ่ฟฝใใใใฆใใ | 0entailment
| overlap-full | 2 | scrambling |
50 | ใใใน้ธๆใๅฅณๆงใๆผใใ | ๅฅณๆงใใใใน้ธๆใๆผใใ | 0entailment
| overlap-full | 2 | scrambling |
51 | ใฉใคใใผใๅญฆ็ใๆผใใ | ๅญฆ็ใใฉใคใใผใๆผใใ | 0entailment
| overlap-full | 2 | scrambling |
52 | ไผ็คพๅกใในใใผใใผใใผใๅฉใใ | ในใใผใใผใใผใไผ็คพๅกใๅฉใใ | 0entailment
| overlap-full | 2 | scrambling |
53 | ไผ็คพๅกใๅฅณๆงใ่ฆใฆใใ | ๅฅณๆงใไผ็คพๅกใ่ฆใฆใใ | 0entailment
| overlap-full | 2 | scrambling |
54 | ๅฅณๆงใใฉใคใใผใ่ฆใคใใฆใใ | ใฉใคใใผใๅฅณๆงใ่ฆใคใใฆใใ | 0entailment
| overlap-full | 2 | scrambling |
55 | ๅญฆ็ใใใใฑใผ้ธๆใ่ฟฝใๅใใ | ใใใฑใผ้ธๆใๅญฆ็ใ่ฟฝใๅใใ | 0entailment
| overlap-full | 2 | scrambling |
56 | ใใใใใผใซ้ธๆใ่ฅ่
ใใใใใฆใใ | ่ฅ่
ใใใใใใผใซ้ธๆใใใใใฆใใ | 0entailment
| overlap-full | 2 | scrambling |
57 | ใซใใใซใ็ทใฎๅญใๆผใใ | ็ทใฎๅญใใซใใใซใๆผใใ | 0entailment
| overlap-full | 2 | scrambling |
58 | ใใใน้ธๆใๅญไพใๆผใใ | ๅญไพใใใใน้ธๆใๆผใใ | 0entailment
| overlap-full | 2 | scrambling |
59 | ๅคงไบบใ็ทใฎๅญใ่ฆใคใใฆใใ | ็ทใฎๅญใๅคงไบบใ่ฆใคใใฆใใ | 0entailment
| overlap-full | 2 | scrambling |
60 | ใใใน้ธๆใใตใผใใกใผใใใใใฆใใ | ใตใผใใกใผใใใใน้ธๆใใใใใฆใใ | 0entailment
| overlap-full | 2 | scrambling |
61 | ใตใผใใกใผใๅญฆ็ใ่นดใฃใ | ๅญฆ็ใใตใผใใกใผใ่นดใฃใ | 0entailment
| overlap-full | 2 | scrambling |
62 | ในใใผใใผใใผใใใใน้ธๆใ่นดใฃใ | ใใใน้ธๆใในใใผใใผใใผใ่นดใฃใ | 0entailment
| overlap-full | 2 | scrambling |
63 | ใตใผใใกใผใ็ทๆงใๅฉใใ | ็ทๆงใใตใผใใกใผใๅฉใใ | 0entailment
| overlap-full | 2 | scrambling |
64 | ๅฅณๆงใ็ทใฎๅญใ่นดใฃใ | ็ทใฎๅญใๅฅณๆงใ่นดใฃใ | 0entailment
| overlap-full | 2 | scrambling |
66 | ใฉใคใใผใในใใผใใผใใผใ่ฟฝใใใใฆใใ | ในใใผใใผใใผใใฉใคใใผใ่ฟฝใใใใฆใใ | 0entailment
| overlap-full | 2 | scrambling |
67 | ๅญไพใใฉใคใใผใๆผใใ | ใฉใคใใผใๅญไพใๆผใใ | 0entailment
| overlap-full | 2 | scrambling |
68 | ใใใฑใผ้ธๆใใซใใใซใๆผใใ | ใซใใใซใใใใฑใผ้ธๆใๆผใใ | 0entailment
| overlap-full | 2 | scrambling |
69 | ไผ็คพๅกใๅฅณๆงใๆใใใฆใใ | ๅฅณๆงใไผ็คพๅกใๆใใใฆใใ | 0entailment
| overlap-full | 2 | scrambling |
70 | ใฉใคใใผใไผ็คพๅกใใใใใฆใใ | ไผ็คพๅกใใฉใคใใผใใใใใฆใใ | 0entailment
| overlap-full | 2 | scrambling |
71 | ๅฅณๆงใใซใใใซใใใใใฆใใ | ใซใใใซใๅฅณๆงใใใใใฆใใ | 0entailment
| overlap-full | 2 | scrambling |
72 | ใใใใใผใซ้ธๆใไผ็คพๅกใใใใใฆใใ | ไผ็คพๅกใใใใใใผใซ้ธๆใใใใใฆใใ | 0entailment
| overlap-full | 2 | scrambling |
73 | ๅญไพใ็ทๆงใใใใใฆใใ | ็ทๆงใๅญไพใใใใใฆใใ | 0entailment
| overlap-full | 2 | scrambling |
74 | ใใใใใผใซ้ธๆใใซใใใซใใใใใฆใใ | ใซใใใซใใใใใใผใซ้ธๆใใใใใฆใใ | 0entailment
| overlap-full | 2 | scrambling |
75 | ็ทๆงใใซใใใซใ่ฟฝใๆใฃใ | ใซใใใซใ็ทๆงใ่ฟฝใๆใฃใ | 0entailment
| overlap-full | 2 | scrambling |
76 | ่ฅ่
ใในใใผใใผใใผใ่นดใฃใ | ในใใผใใผใใผใ่ฅ่
ใ่นดใฃใ | 0entailment
| overlap-full | 2 | scrambling |
77 | ใซใใใซใใใใน้ธๆใ่ฆใคใใฆใใ | ใใใน้ธๆใใซใใใซใ่ฆใคใใฆใใ | 0entailment
| overlap-full | 2 | scrambling |
78 | ใตใผใใกใผใๅฅณใฎๅญใ่นดใฃใ | ๅฅณใฎๅญใใตใผใใกใผใ่นดใฃใ | 0entailment
| overlap-full | 2 | scrambling |
79 | ใฉใคใใผใๅคงไบบใ่ฆใคใใฆใใ | ๅคงไบบใใฉใคใใผใ่ฆใคใใฆใใ | 0entailment
| overlap-full | 2 | scrambling |
80 | ๅคงไบบใๅญไพใ่ฟฝใๅใใ | ๅญไพใๅคงไบบใ่ฟฝใๅใใ | 0entailment
| overlap-full | 2 | scrambling |
81 | ๅญไพใใใใใใผใซ้ธๆใ่ฟฝใใใใฆใใ | ใใใใใผใซ้ธๆใๅญไพใ่ฟฝใใใใฆใใ | 0entailment
| overlap-full | 2 | scrambling |
82 | ็ทใฎๅญใๅฅณใฎๅญใ่ฟฝใๅใใ | ๅฅณใฎๅญใ็ทใฎๅญใ่ฟฝใๅใใ | 0entailment
| overlap-full | 2 | scrambling |
83 | ็ทๆงใใใใใใผใซ้ธๆใๆใใใฆใใ | ใใใใใผใซ้ธๆใ็ทๆงใๆใใใฆใใ | 0entailment
| overlap-full | 2 | scrambling |
84 | ็ทใฎๅญใใฉใคใใผใ่ฟฝใๅใใ | ใฉใคใใผใ็ทใฎๅญใ่ฟฝใๅใใ | 0entailment
| overlap-full | 2 | scrambling |
86 | ใตใผใใกใผใใใใฑใผ้ธๆใใใใใฆใใ | ใใใฑใผ้ธๆใใตใผใใกใผใใใใใฆใใ | 0entailment
| overlap-full | 2 | scrambling |
87 | ใฉใคใใผใ่ฅ่
ใใซใใใงใใ | ่ฅ่
ใใฉใคใใผใใซใใใงใใ | 0entailment
| overlap-full | 2 | scrambling |
88 | ๅฅณๆงใ็ทๆงใ่นดใฃใ | ็ทๆงใๅฅณๆงใ่นดใฃใ | 0entailment
| overlap-full | 2 | scrambling |
89 | ๅฅณใฎๅญใใซใใใซใใใใใฆใใ | ใซใใใซใๅฅณใฎๅญใใใใใฆใใ | 0entailment
| overlap-full | 2 | scrambling |
90 | ๅญไพใๅฅณๆงใใซใใใงใใ | ๅฅณๆงใๅญไพใใซใใใงใใ | 0entailment
| overlap-full | 2 | scrambling |
91 | ็ทๆงใๅฅณใฎๅญใ่ฟฝใๆใฃใ | ๅฅณใฎๅญใ็ทๆงใ่ฟฝใๆใฃใ | 0entailment
| overlap-full | 2 | scrambling |
92 | ใซใใใซใในใใผใใผใใผใๅฉใใ | ในใใผใใผใใผใใซใใใซใๅฉใใ | 0entailment
| overlap-full | 2 | scrambling |
93 | ๅญฆ็ใ่ฅ่
ใใใใใฆใใ | ่ฅ่
ใๅญฆ็ใใใใใฆใใ | 0entailment
| overlap-full | 2 | scrambling |
94 | ใฉใคใใผใๅฅณใฎๅญใๅฉใใ | ๅฅณใฎๅญใใฉใคใใผใๅฉใใ | 0entailment
| overlap-full | 2 | scrambling |
95 | ๅญไพใใใใฑใผ้ธๆใ่นดใฃใ | ใใใฑใผ้ธๆใๅญไพใ่นดใฃใ | 0entailment
| overlap-full | 2 | scrambling |
96 | ใใใน้ธๆใใใใใใผใซ้ธๆใ่ฆใฆใใ | ใใใใใผใซ้ธๆใใใใน้ธๆใ่ฆใฆใใ | 0entailment
| overlap-full | 2 | scrambling |
97 | ๅญไพใๅคงไบบใ่ฟฝใๆใฃใ | ๅคงไบบใๅญไพใ่ฟฝใๆใฃใ | 0entailment
| overlap-full | 2 | scrambling |
98 | ใใใใใผใซ้ธๆใๅฅณใฎๅญใๆผใใ | ๅฅณใฎๅญใใใใใใผใซ้ธๆใๆผใใ | 0entailment
| overlap-full | 2 | scrambling |
99 | ๅฅณๆงใใใใใใผใซ้ธๆใ่ฆใคใใฆใใ | ใใใใใผใซ้ธๆใๅฅณๆงใ่ฆใคใใฆใใ | 0entailment
| overlap-full | 2 | scrambling |
101 | ่ฅ่
ใไผ็คพๅกใๆใใใฆใใ | ไผ็คพๅกใฏ่ฅ่
ใๆใใใฆใใ | 0entailment
| overlap-nonorder | 2 | scrambling |
102 | ๅฅณใฎๅญใ่ไบบใๆผใใ | ่ไบบใฏๅฅณใฎๅญใๆผใใ | 0entailment
| overlap-nonorder | 2 | scrambling |
103 | ใซใใใซใๅฅณๆงใๆใใใฆใใ | ๅฅณๆงใฏใซใใใซใๆใใใฆใใ | 0entailment
| overlap-nonorder | 2 | scrambling |
104 | ่ฅ่
ใใใใใใผใซ้ธๆใ่ฆใฆใใ | ใใใใใผใซ้ธๆใฏ่ฅ่
ใ่ฆใฆใใ | 0entailment
| overlap-nonorder | 2 | scrambling |
105 | ใซใใใซใๅญไพใใซใใใงใใ | ๅญไพใฏใซใใใซใใซใใใงใใ | 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
ใใใใใใผใซ้ธๆใ่ฆใฆใใ',
'hypothesis': 'ใใใใใผใซ้ธๆใ่ฅ่
ใ่ฆใฆใใ',
'label': 0,
'heuristics': 'overlap-full',
'number_of_NPs': 2,
'semtag': 'scrambling'
}
original
An example of looks as follows:
{
'id': 12,
'sentence_A_Ja': '่ฅ่
ใใใใใใผใซ้ธๆใ่ฆใฆใใ',
'sentence_B_Ja': 'ใใใใใผใซ้ธๆใ่ฅ่
ใ่ฆใฆใใ',
'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
- verypluming/JaNLI
- Assessing the Generalization Capacity of Pre-trained Language Models through Japanese Adversarial Natural Language Inference
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
- 82