hpprc commited on
Commit
e2d5d60
1 Parent(s): 01fa168

:memo: Update documents

Browse files
Files changed (1) hide show
  1. README.md +22 -18
README.md CHANGED
@@ -126,25 +126,30 @@ An example of looks as follows:
126
 
127
  A version adopting the column names of a typical NLI dataset.
128
 
129
- - `id`: The number of the sentence pair.
130
- - `premise`: The premise (sentence_A_Ja).
131
- - `hypothesis`: The hypothesis (sentence_B_Ja).
132
- - `label`: The correct label for this sentence pair (either `entailment` or `non-entailment`); in the setting described in the paper, non-entailment = neutral + contradiction (entailment_label_Ja).
133
- - `heuristics`: The heuristics (structural pattern) tag. The tags are: subsequence, constituent, full-overlap, order-subset, and mixed-subset.
134
- - `number_of_NPs`: The number of noun phrase in a sentence.
135
- - `semtag`: The linguistic phenomena tag.
 
 
 
136
 
137
  #### original
138
 
139
  The original version retaining the unaltered column names.
140
 
141
- - `id`: The number of the sentence pair.
142
- - `sentence_A_Ja`: The premise.
143
- - `sentence_B_Ja`: The hypothesis.
144
- - `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
145
- - `heuristics`: The heuristics (structural pattern) tag. The tags are: subsequence, constituent, full-overlap, order-subset, and mixed-subset.
146
- - `number_of_NPs`: The number of noun phrase in a sentence.
147
- - `semtag`: The linguistic phenomena tag.
 
 
148
 
149
 
150
  ### Data Splits
@@ -155,7 +160,6 @@ The original version retaining the unaltered column names.
155
  | original | 13,680 | | 720 |
156
 
157
 
158
-
159
  ### Annotations
160
 
161
  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.
@@ -164,11 +168,11 @@ Additionally, 11 categories of Japanese linguistic phenomena and constructions a
164
 
165
  For each linguistic phenomenon, a template for the premise sentence P is fixed, and multiple templates for hypothesis sentences H are created.
166
  In total, 144 templates for (P, H) pairs are produced.
167
- 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.
168
 
169
  The JaNLI dataset is generated by instantiating each template 100 times, resulting in a total of 14,400 examples.
170
  The same number of entailment and non-entailment examples are generated for each phenomenon.
171
- 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.
172
  The dataset uses a total of 158 words (nouns and verbs), which occur more than 20 times in the JSICK and JSNLI datasets.
173
 
174
 
@@ -195,4 +199,4 @@ CC BY-SA 4.0
195
 
196
  ### Contributions
197
 
198
- Thanks to [Hitomi Yanaka](https://hitomiyanaka.mystrikingly.com/) and Koji Mineshima for creating this dataset.
 
126
 
127
  A version adopting the column names of a typical NLI dataset.
128
 
129
+ | Name | Description |
130
+ | ------------- | ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
131
+ | id | The number of the sentence pair. |
132
+ | premise | The premise (sentence_A_Ja). |
133
+ | hypothesis | The hypothesis (sentence_B_Ja). |
134
+ | 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). |
135
+ | heuristics | The heuristics (structural pattern) tag. The tags are: subsequence, constituent, full-overlap, order-subset, and mixed-subset. |
136
+ | number_of_NPs | The number of noun phrase in a sentence. |
137
+ | semtag | The linguistic phenomena tag. |
138
+
139
 
140
  #### original
141
 
142
  The original version retaining the unaltered column names.
143
 
144
+ | Name | Description |
145
+ | ------------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
146
+ | id | The number of the sentence pair. |
147
+ | sentence_A_Ja | The premise. |
148
+ | sentence_B_Ja | The hypothesis. |
149
+ | 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 |
150
+ | heuristics | The heuristics (structural pattern) tag. The tags are: subsequence, constituent, full-overlap, order-subset, and mixed-subset. |
151
+ | number_of_NPs | The number of noun phrase in a sentence. |
152
+ | semtag | The linguistic phenomena tag. |
153
 
154
 
155
  ### Data Splits
 
160
  | original | 13,680 | | 720 |
161
 
162
 
 
163
  ### Annotations
164
 
165
  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.
 
168
 
169
  For each linguistic phenomenon, a template for the premise sentence P is fixed, and multiple templates for hypothesis sentences H are created.
170
  In total, 144 templates for (P, H) pairs are produced.
171
+ 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.
172
 
173
  The JaNLI dataset is generated by instantiating each template 100 times, resulting in a total of 14,400 examples.
174
  The same number of entailment and non-entailment examples are generated for each phenomenon.
175
+ 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.
176
  The dataset uses a total of 158 words (nouns and verbs), which occur more than 20 times in the JSICK and JSNLI datasets.
177
 
178
 
 
199
 
200
  ### Contributions
201
 
202
+ Thanks to [Hitomi Yanaka](https://hitomiyanaka.mystrikingly.com/) and [Koji Mineshima](https://abelard.flet.keio.ac.jp/person/minesima/index-j.html) for creating this dataset.