Datasets:
Add further dataset stats
Browse files
README.md
CHANGED
@@ -85,15 +85,7 @@ Antonyms and Negation Using the New SemAntoNeg Benchmark_ (see
|
|
85 |
|
86 |
<br>
|
87 |
|
88 |
-
|
89 |
-
sentences has been added by paraphrasing them with the pre-trained model
|
90 |
-
[`🤗tuner007/pegasus_paraphrase`](https://huggingface.co/tuner007/pegasus_paraphrase).
|
91 |
-
|
92 |
-
Finally, the swapped version of each pair (premise ⇋ hypothesis) has also been
|
93 |
-
included, and any duplicates have been removed.
|
94 |
-
|
95 |
-
The contribution of each of these individual datasets to the final CANNOT
|
96 |
-
dataset is:
|
97 |
|
98 |
| Dataset | Samples |
|
99 |
|:--------------------------------------------------------------------------|-----------:|
|
@@ -102,10 +94,32 @@ dataset is:
|
|
102 |
| Automated Fact-Checking of Claims from Wikipedia | 14,970 |
|
103 |
| From Group to Individual Labels Using Deep Features | 2,110 |
|
104 |
| It Is Not Easy To Detect Paraphrases | 8,597 |
|
105 |
-
| <
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
106 |
|
107 |
-
|
108 |
-
datasets
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
109 |
|
110 |
|
111 |
## Additional Information
|
|
|
85 |
|
86 |
<br>
|
87 |
|
88 |
+
Once processed, the number of remaining samples in each of the datasets above are:
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
89 |
|
90 |
| Dataset | Samples |
|
91 |
|:--------------------------------------------------------------------------|-----------:|
|
|
|
94 |
| Automated Fact-Checking of Claims from Wikipedia | 14,970 |
|
95 |
| From Group to Individual Labels Using Deep Features | 2,110 |
|
96 |
| It Is Not Easy To Detect Paraphrases | 8,597 |
|
97 |
+
| <div align="right"><b>Total</b></div> | **25,949** |
|
98 |
+
|
99 |
+
<br>
|
100 |
+
|
101 |
+
Additionally, for each of the negated samples, another pair of non-negated
|
102 |
+
sentences has been added by paraphrasing them with the pre-trained model
|
103 |
+
[`🤗tuner007/pegasus_paraphrase`](https://huggingface.co/tuner007/pegasus_paraphrase).
|
104 |
+
|
105 |
+
Finally, the swapped version of each pair (premise ⇋ hypothesis) has also been
|
106 |
+
included, and any duplicates have been removed.
|
107 |
|
108 |
+
With this, the number of premises/hypothesis in the CANNOT dataset that appear
|
109 |
+
in the original datasets are:
|
110 |
+
|
111 |
+
| <div align="left"><b>Dataset</b></div> | <div align="center"><b>Sentences</b></div> |
|
112 |
+
|:--------------------------------------------------------------------------|----------------------:|
|
113 |
+
| Not another Negation Benchmark | 552 (0.36 %) |
|
114 |
+
| GLUE Diagnostic Dataset | 586 (0.38 %) |
|
115 |
+
| Automated Fact-Checking of Claims from Wikipedia | 89,728 (59.98 %) |
|
116 |
+
| From Group to Individual Labels Using Deep Features | 12,626 (8.16 %) |
|
117 |
+
| It Is Not Easy To Detect Paraphrases | 17,198 (11.11 %) |
|
118 |
+
| <div align="right"><b>Total</b></div> | **120,690** (77.99 %) |
|
119 |
+
|
120 |
+
The percentages above are in relation to the total number of premises and
|
121 |
+
hypothesis in the CANNOT dataset. The remaining 22.01 % (34,062 sentences) are
|
122 |
+
the novel premises/hypothesis added through paraphrase and rule-based negation.
|
123 |
|
124 |
|
125 |
## Additional Information
|