Datasets:

Modalities:
Text
Formats:
json
ArXiv:
Libraries:
Datasets
pandas
License:
s-conia commited on
Commit
b0c394f
1 Parent(s): f7b79f4

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +145 -3
README.md CHANGED
@@ -1,3 +1,145 @@
1
- ---
2
- license: apache-2.0
3
- ---
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ license: apache-2.0
3
+ task_categories:
4
+ - text-generation
5
+ language:
6
+ - it
7
+ - en
8
+ size_categories:
9
+ - 10K<n<100K
10
+ configs:
11
+ - config_name: default
12
+ data_files:
13
+ - split: train
14
+ path: piqa.train.json
15
+ - split: validation
16
+ path: piqa.validation.json
17
+ ---
18
+
19
+ # PIQA - Italian (IT)
20
+ This dataset is an Italian translation of [PIQA](https://arxiv.org/abs/1911.11641). PIQA stands for Physical Interaction Question Answering, a dataset of questions about common scenarios that require an understanding of the physical world.
21
+
22
+ ## Dataset Details
23
+ The dataset consists of questions about common scenarios that require an understanding of the physical world. Each question is associated with a correct answer and a distractor. The task is to predict the correct answer to the question.
24
+
25
+ The dataset includes the following splits:
26
+ * Train: 16,044 rows
27
+ * Validation: 1,830 rows
28
+
29
+ ### Differences with the original dataset
30
+ * **Did you know that not all questions in PIQA are questions?** In the original dataset, some instances are not questions but text completions, statements, or even single words. In this version, we categorize all instances so as to give the possibility to filter out non-question instances or treat them differently.
31
+ * The number of instances in this dataset is smaller than the original dataset due to the translation process, during which some instances were filtered out.
32
+
33
+ ### Languages
34
+ This dataset is **fully parallel** between English and Italian. This allows us to have comparable evaluation setups and results across the two languages.
35
+
36
+ ### Translation Process
37
+ The translation has been carried out using [🍱 OBenTO-LLM](https://github.com/c-simone/llm-data-translation), an open-source tool for LLM-based translation.
38
+ The main motivation for using an open-source LLM is to encourage free, open, reproducible, and transparent research in LLM evaluation.
39
+ See [🍱 OBenTO-LLM](https://github.com/c-simone/llm-data-translation) for more details on the translation process.
40
+
41
+ ### Other Information
42
+ - **Original dataset by:** [Bisk et al.](https://arxiv.org/abs/1911.11641)
43
+ - **Translation by:** [Simone Conia](https://scholar.google.com/citations?user=S1tqbTcAAAAJ)
44
+ - **Languages:** Italian, English
45
+ - **License:** Apache 2.0
46
+
47
+ ## Dataset Format
48
+ This is an example that shows the format of the dataset, where:
49
+ * `id`: a unique ID for each sample;
50
+ * `category`: type of task;
51
+ * `input_text`: the original English sentence in the dataset;
52
+ * `input_text_translation`: the translation of the sentence in Italian;
53
+ * `choices`: the original English choices;
54
+ * `choice_translations`: the translation of the choices in Italian;
55
+ * `gold_index`: the index of the correct answer.
56
+
57
+ #### Example of a question in PIQA
58
+ ```json
59
+ {
60
+ "id": "piqa_3",
61
+ "category": "question",
62
+ "input_text": "How do you shake something?",
63
+ "input_text_translation": "Come si fa a scuotere qualcosa?",
64
+ "choices": [
65
+ "Move it up and down and side to side quickly.",
66
+ "Stir it very quickly."
67
+ ],
68
+ "choice_translations": [
69
+ "Si deve muovere rapidamente avanti e indietro e da un lato all'altro.",
70
+ "Si mescola molto velocemente."
71
+ ],
72
+ "gold_index": 0
73
+ }
74
+ ```
75
+
76
+ #### Example of a text completion in PIQA
77
+ ```json
78
+ {
79
+ "id": "piqa_1",
80
+ "category": "text_completion",
81
+ "input_text": "To permanently attach metal legs to a chair, you can",
82
+ "input_text_translation": "Per fissare in modo permanente le gambe di metallo a una sedia, si può",
83
+ "choices": [
84
+ "weld the metal together to get it to stay firmly in place.",
85
+ "nail the metal together to get it to stay firmly in place."
86
+ ],
87
+ "choice_translations": [
88
+ "saldare il metallo per farlo rimanere saldamente in posizione.",
89
+ "incollare il metallo per farlo rimanere saldamente in posizione."
90
+ ],
91
+ "gold_index": 0
92
+ }
93
+ ```
94
+
95
+ #### Example of a "topic" in PIQA
96
+ ```json
97
+ {
98
+ "id": "piqa_29",
99
+ "category": "topic",
100
+ "input_text": "Soothe a painful sunburn.",
101
+ "input_text_translation": "Alleviare una scottatura solare dolorosa.",
102
+ "choices": [
103
+ "Wait until brewed tea bag is cool, then apply on burn.",
104
+ "Wait until brewed tea bag is hot, then apply on burn."
105
+ ],
106
+ "choice_translations": [
107
+ "Attendere fino a quando il sacchetto del tè in infusione è freddo, quindi applicarlo sulla scottatura.",
108
+ "Attendere fino a quando il sacchetto del tè in infusione è caldo, quindi applicarlo sulla scottatura."
109
+ ],
110
+ "gold_index": 0
111
+ }
112
+ ```
113
+
114
+ #### Example of a "property" in PIQA
115
+ ```json
116
+ {
117
+ "id": "piqa_855",
118
+ "category": "property",
119
+ "input_text": "Sleeves:",
120
+ "input_text_translation": "Maniche:",
121
+ "choices": [
122
+ "Can be cut by sciscors with ease.",
123
+ "Can be cut by a knife with ease."
124
+ ],
125
+ "choice_translations": [
126
+ "possono essere tagliate facilmente con le forbici.",
127
+ "possono essere tagliate con facilità con un coltello."
128
+ ],
129
+ "gold_index": 0
130
+ }
131
+ ```
132
+
133
+ ## License
134
+ The dataset is distributed under the Apache 2.0 license.
135
+
136
+ ## Acknowledgements
137
+ I would like to thank the authors of the original Winogrande dataset for making it available to the research community.
138
+ I would also like to thank [Future AI Research](https://future-ai-research.it/) for supporting this work and funding my research.
139
+
140
+ ### Special Thanks
141
+ My special thanks go to:
142
+ * Pere-Lluís Huguet Cabot and Riccardo Orlando for their help with [🍱 OBenTO-LLM](https://github.com/c-simone/llm-data-translation).
143
+
144
+ ## Dataset Card Authors
145
+ * [Simone Conia](https://scholar.google.com/citations?user=S1tqbTcAAAAJ): simone.conia@uniroma1.it