mathiascreutz commited on
Commit
10f8929
1 Parent(s): 4228fac

Minor modifications

Browse files
Files changed (1) hide show
  1. README.md +18 -5
README.md CHANGED
@@ -128,14 +128,27 @@ to specify the level of quality you desire, such as:
128
  The entries in the training sets have been ranked automatically by how
129
  likely they are paraphrases, best first, worst last. The quality
130
  parameter indicates the estimated proportion (in percent) of true
131
- paraphrases in the set. Allowed quality values range between 60 and
132
- 100, in increments of 5 (60, 65, 70, ..., 100). A value of 60 means
133
- that 60% of the sentence pairs in the downloaded set are estimated to
134
- be true paraphrases (and the remaining 40% are not). A higher value
135
  produces a smaller but cleaner set. The smaller sets are subsets of
136
- the larger sets, such that the `quality=95`set is a subset of
137
  `quality=90`, which is a subset of `quality=85`, and so on.
138
 
 
 
 
 
 
 
 
 
 
 
 
 
 
139
 
140
  ### Data Instances
141
 
 
128
  The entries in the training sets have been ranked automatically by how
129
  likely they are paraphrases, best first, worst last. The quality
130
  parameter indicates the estimated proportion (in percent) of true
131
+ paraphrases in the training set. Allowed quality values range between
132
+ 60 and 100, in increments of 5 (60, 65, 70, ..., 100). A value of 60
133
+ means that 60% of the sentence pairs in the training set are estimated
134
+ to be true paraphrases (and the remaining 40% are not). A higher value
135
  produces a smaller but cleaner set. The smaller sets are subsets of
136
+ the larger sets, such that the `quality=95` set is a subset of
137
  `quality=90`, which is a subset of `quality=85`, and so on.
138
 
139
+ The default `quality` value, if omitted, is 100. This matches no
140
+ training data at all, which can be convenient, if you are only
141
+ interested in the validation and test sets, which are considerably
142
+ smaller, but manually annotated.
143
+
144
+ Note that an alternative to typing the parameter values explicitly,
145
+ you can use configuration names instead. The following commands are
146
+ equivalent to the ones above:
147
+
148
+ `data = load_dataset("GEM/opusparcus", "de.100")`
149
+
150
+ `data = load_dataset("GEM/opusparcus", "fr.90")`
151
+
152
 
153
  ### Data Instances
154