tomaarsen HF staff commited on
Commit
37897b6
1 Parent(s): 6de3c20

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +27 -12
README.md CHANGED
@@ -140,9 +140,13 @@ Despite originally being intended for Natural Language Inference (NLI), this dat
140
  * Columns: "premise", "hypothesis", "label"
141
  * Column types: `str`, `str`, `class` with {"0": "entailment", "1": "neutral", "2", "contradiction"}
142
  * Examples:
143
- ```python
144
- {'premise': 'A person on a horse jumps over a broken down airplane.', 'hypothesis': 'A person is training his horse for a competition.', 'label': 1}
145
- ```
 
 
 
 
146
  * Collection strategy: Reading the premise, hypothesis and integer label from SNLI & MultiNLI datasets.
147
  * Deduplified: Yes
148
 
@@ -151,9 +155,13 @@ Despite originally being intended for Natural Language Inference (NLI), this dat
151
  * Columns: "sentence_1", "sentence_2", "label"
152
  * Column types: `str`, `str`, `float`
153
  * Examples:
154
- ```python
155
- {'premise': 'A person on a horse jumps over a broken down airplane.', 'hypothesis': 'A person is training his horse for a competition.', 'label': 1.0}
156
- ```
 
 
 
 
157
  * Collection strategy: Taking the `pair-class` subset and remapping "entailment", "neutral" and "contradiction" to 1.0, 0.5 and 0.0, respectively.
158
  * Deduplified: Yes
159
 
@@ -162,9 +170,12 @@ Despite originally being intended for Natural Language Inference (NLI), this dat
162
  * Columns: "anchor", "positive"
163
  * Column types: `str`, `str`
164
  * Examples:
165
- ```python
166
- {'anchor': 'A person on a horse jumps over a broken down airplane.', 'positive': 'A person is training his horse for a competition.'}
167
- ```
 
 
 
168
  * Collection strategy: Reading the SNLI & MultiNLI datasets and considering the "premise" as the "anchor" and the "hypothesis" as the "positive" if the label is "entailment". The reverse ("entailment" as "anchor" and "premise" as "positive") is not included.
169
  * Deduplified: Yes
170
 
@@ -173,8 +184,12 @@ Despite originally being intended for Natural Language Inference (NLI), this dat
173
  * Columns: "anchor", "positive", "negative"
174
  * Column types: `str`, `str`, `str`
175
  * Examples:
176
- ```python
177
- {'anchor': 'A person on a horse jumps over a broken down airplane.', 'positive': 'A person is outdoors, on a horse.', 'negative': 'A person is at a diner, ordering an omelette.'}
178
- ```
 
 
 
 
179
  * Collection strategy: Reading the SNLI & MultiNLI datasets, for each "premise" making a list of entailing and contradictory sentences using the dataset labels. Then, considering all possible triplets out of these entailing and contradictory lists. The reverse ("entailment" as "anchor" and "premise" as "positive") is not included.
180
  * Deduplified: Yes
 
140
  * Columns: "premise", "hypothesis", "label"
141
  * Column types: `str`, `str`, `class` with {"0": "entailment", "1": "neutral", "2", "contradiction"}
142
  * Examples:
143
+ ```python
144
+ {
145
+ 'premise': 'A person on a horse jumps over a broken down airplane.',
146
+ 'hypothesis': 'A person is training his horse for a competition.',
147
+ 'label': 1,
148
+ }
149
+ ```
150
  * Collection strategy: Reading the premise, hypothesis and integer label from SNLI & MultiNLI datasets.
151
  * Deduplified: Yes
152
 
 
155
  * Columns: "sentence_1", "sentence_2", "label"
156
  * Column types: `str`, `str`, `float`
157
  * Examples:
158
+ ```python
159
+ {
160
+ 'sentence_1': 'A person on a horse jumps over a broken down airplane.',
161
+ 'sentence_2': 'A person is training his horse for a competition.',
162
+ 'label': 1.0,
163
+ }
164
+ ```
165
  * Collection strategy: Taking the `pair-class` subset and remapping "entailment", "neutral" and "contradiction" to 1.0, 0.5 and 0.0, respectively.
166
  * Deduplified: Yes
167
 
 
170
  * Columns: "anchor", "positive"
171
  * Column types: `str`, `str`
172
  * Examples:
173
+ ```python
174
+ {
175
+ 'anchor': 'A person on a horse jumps over a broken down airplane.',
176
+ 'positive': 'A person is training his horse for a competition.',
177
+ }
178
+ ```
179
  * Collection strategy: Reading the SNLI & MultiNLI datasets and considering the "premise" as the "anchor" and the "hypothesis" as the "positive" if the label is "entailment". The reverse ("entailment" as "anchor" and "premise" as "positive") is not included.
180
  * Deduplified: Yes
181
 
 
184
  * Columns: "anchor", "positive", "negative"
185
  * Column types: `str`, `str`, `str`
186
  * Examples:
187
+ ```python
188
+ {
189
+ 'anchor': 'A person on a horse jumps over a broken down airplane.',
190
+ 'positive': 'A person is outdoors, on a horse.',
191
+ 'negative': 'A person is at a diner, ordering an omelette.',
192
+ }
193
+ ```
194
  * Collection strategy: Reading the SNLI & MultiNLI datasets, for each "premise" making a list of entailing and contradictory sentences using the dataset labels. Then, considering all possible triplets out of these entailing and contradictory lists. The reverse ("entailment" as "anchor" and "premise" as "positive") is not included.
195
  * Deduplified: Yes