Updating description of fields and statistics
Browse files
README.md
CHANGED
@@ -14,7 +14,7 @@ task_categories:
|
|
14 |
|
15 |
### Dataset Summary
|
16 |
|
17 |
-
This dataset is a set of samples for testing the spell
|
18 |
|
19 |
The dataset contains two splits:
|
20 |
|
@@ -23,16 +23,14 @@ test.json contains samples hand-selected to evaluate the quality of models.
|
|
23 |
train.json contains synthetic samples generated in various ways.
|
24 |
|
25 |
The purpose of creating the dataset was to test an internal spellchecker for [a generative poetry project](https://github.com/Koziev/verslibre), but it can also be useful in other projects, since it does not have an explicit specialization for poetry.
|
26 |
-
You can consider this dataset as an extension of [RuCOLA](https://huggingface.co/datasets/RussianNLP/rucola).
|
27 |
-
In addition, for some samples there is a corrected version of the text ("fixed_sentence" field), so it can be used as an extension of datasets in [ai-forever/spellcheck_benchmark](https://huggingface.co/datasets/ai-forever/spellcheck_benchmark).
|
28 |
|
29 |
### Example
|
30 |
|
31 |
```
|
32 |
{
|
33 |
"id": 1483,
|
34 |
-
"
|
35 |
-
"
|
36 |
"label": 0,
|
37 |
"error_type": "Tokenization",
|
38 |
"domain": "prose"
|
@@ -45,38 +43,25 @@ The test split contains only examples of mistakes made by people. There are no s
|
|
45 |
|
46 |
The examples of errors in the test split come from different people in terms of gender, age, education, context, and social context.
|
47 |
|
48 |
-
The input and output text can be not only one sentence, but also 1) part of a sentence, 2) several sentences - a paragraph, 3) a fragment of a poem, usually a quatrain or two quatrains.
|
49 |
|
50 |
-
The texts may include offensive
|
|
|
|
|
51 |
|
52 |
One sample may contain several errors of different types.
|
53 |
|
54 |
|
55 |
-
|
56 |
-
### Uncensoring samples
|
57 |
-
|
58 |
-
A number of samples contain text with explicit obscenities:
|
59 |
-
|
60 |
-
```
|
61 |
-
{
|
62 |
-
"id": 1,
|
63 |
-
"sentence": "Но не простого - с лёгкой еб@нцой.",
|
64 |
-
"fixed_sentence": "Но не простого - с лёгкой ебанцой.",
|
65 |
-
"label": 0,
|
66 |
-
"error_type": "Misspelling",
|
67 |
-
"domain": "prose"
|
68 |
-
}
|
69 |
-
```
|
70 |
-
|
71 |
### Poetry samples
|
72 |
|
73 |
-
|
|
|
74 |
|
75 |
```
|
76 |
{
|
77 |
"id": 24,
|
78 |
-
"
|
79 |
-
"
|
80 |
"label": 0,
|
81 |
"error_type": "Grammar",
|
82 |
"domain": "poetry"
|
@@ -88,9 +73,9 @@ A few poetry samples are included in this version:
|
|
88 |
### Dataset fields
|
89 |
|
90 |
**id** (int64): the sentence's id, starting 1.
|
91 |
-
**
|
92 |
-
**
|
93 |
-
**label** (str): the target class. "1" for "
|
94 |
**error_type** (str): the violation category: Spelling, Grammar, Tokenization, Punctuation, Mixture, Unknown.
|
95 |
**domain** (str): domain: "prose" or "poetry".
|
96 |
|
@@ -101,8 +86,8 @@ A few poetry samples are included in this version:
|
|
101 |
```
|
102 |
{
|
103 |
"id": 6,
|
104 |
-
"
|
105 |
-
"
|
106 |
"label": 0,
|
107 |
"error_type": "Tokenization",
|
108 |
"domain": "prose"
|
@@ -114,8 +99,8 @@ A few poetry samples are included in this version:
|
|
114 |
```
|
115 |
{
|
116 |
"id": 5,
|
117 |
-
"
|
118 |
-
"
|
119 |
"label": 0,
|
120 |
"error_type": "Punctuation",
|
121 |
"domain": "prose"
|
@@ -127,8 +112,8 @@ A few poetry samples are included in this version:
|
|
127 |
```
|
128 |
{
|
129 |
"id": 38,
|
130 |
-
"
|
131 |
-
"
|
132 |
"label": 0,
|
133 |
"error_type": "Spelling",
|
134 |
"domain": "prose"
|
@@ -140,8 +125,8 @@ A few poetry samples are included in this version:
|
|
140 |
```
|
141 |
{
|
142 |
"id": 61,
|
143 |
-
"
|
144 |
-
"
|
145 |
"label": 0,
|
146 |
"error_type": "Grammar",
|
147 |
"domain": "prose"
|
@@ -151,16 +136,50 @@ A few poetry samples are included in this version:
|
|
151 |
Please note that error categories are not always set accurately, so you should not use
|
152 |
the "error_type" field to train classifiers.
|
153 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
154 |
|
155 |
### Statistics
|
156 |
|
157 |
Statistics for test split.
|
158 |
|
159 |
```
|
160 |
-
|
161 |
-
| Domain |
|
162 |
-
|
163 |
-
| prose |
|
164 |
-
| poetry |
|
165 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
166 |
```
|
|
|
|
|
|
|
|
|
|
|
|
|
|
14 |
|
15 |
### Dataset Summary
|
16 |
|
17 |
+
This dataset is a set of samples for training and testing the spell checking, grammar error correction and ungrammatical text detection models.
|
18 |
|
19 |
The dataset contains two splits:
|
20 |
|
|
|
23 |
train.json contains synthetic samples generated in various ways.
|
24 |
|
25 |
The purpose of creating the dataset was to test an internal spellchecker for [a generative poetry project](https://github.com/Koziev/verslibre), but it can also be useful in other projects, since it does not have an explicit specialization for poetry.
|
|
|
|
|
26 |
|
27 |
### Example
|
28 |
|
29 |
```
|
30 |
{
|
31 |
"id": 1483,
|
32 |
+
"text": "Разучи стихов по больше",
|
33 |
+
"fixed_text": "Разучи стихов побольше",
|
34 |
"label": 0,
|
35 |
"error_type": "Tokenization",
|
36 |
"domain": "prose"
|
|
|
43 |
|
44 |
The examples of errors in the test split come from different people in terms of gender, age, education, context, and social context.
|
45 |
|
46 |
+
The input and output text can be not only one sentence, but also 1) a part of a sentence or incomplete dialog responses, 2) several sentences - a paragraph, 3) a fragment of a poem, usually a quatrain or two quatrains.
|
47 |
|
48 |
+
The texts may include offensive phrases, phrases that offend religious or political feelings, fragments that contradict moral standards, etc.
|
49 |
+
Such samples are only needed to make the corpus as representative as possible for the tasks of processing messages
|
50 |
+
in various media such as blogs, comments, etc.
|
51 |
|
52 |
One sample may contain several errors of different types.
|
53 |
|
54 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
55 |
### Poetry samples
|
56 |
|
57 |
+
The texts of the poems are included in the test part of the dataset, which makes it unique among similar
|
58 |
+
datasets for the Russian language:
|
59 |
|
60 |
```
|
61 |
{
|
62 |
"id": 24,
|
63 |
+
"text": "Чему научит забытьё?\nСмерть формы д'арует литьё.\nРезец мгновенье любит стружка...\nСмерть безобидная подружка!",
|
64 |
+
"fixed_text": null,
|
65 |
"label": 0,
|
66 |
"error_type": "Grammar",
|
67 |
"domain": "poetry"
|
|
|
73 |
### Dataset fields
|
74 |
|
75 |
**id** (int64): the sentence's id, starting 1.
|
76 |
+
**text** (str): the original text (part of sentence, whole sentence or several sentences).
|
77 |
+
**fixed_text** (str): the corrected version of original text.
|
78 |
+
**label** (str): the target class. "1" for "not defects", "0" for "contains defects".
|
79 |
**error_type** (str): the violation category: Spelling, Grammar, Tokenization, Punctuation, Mixture, Unknown.
|
80 |
**domain** (str): domain: "prose" or "poetry".
|
81 |
|
|
|
86 |
```
|
87 |
{
|
88 |
"id": 6,
|
89 |
+
"text": "Я ��одбираю по проще слова",
|
90 |
+
"fixed_text": "Я подбираю попроще слова",
|
91 |
"label": 0,
|
92 |
"error_type": "Tokenization",
|
93 |
"domain": "prose"
|
|
|
99 |
```
|
100 |
{
|
101 |
"id": 5,
|
102 |
+
"text": "И швырнуть по-дальше",
|
103 |
+
"fixed_text": "И швырнуть подальше",
|
104 |
"label": 0,
|
105 |
"error_type": "Punctuation",
|
106 |
"domain": "prose"
|
|
|
112 |
```
|
113 |
{
|
114 |
"id": 38,
|
115 |
+
"text": "И ведь что интересно, русские официально ни в одном крестовом позоде не участвовали.",
|
116 |
+
"fixed_text": "И ведь что интересно, русские официально ни в одном крестовом походе не участвовали.",
|
117 |
"label": 0,
|
118 |
"error_type": "Spelling",
|
119 |
"domain": "prose"
|
|
|
125 |
```
|
126 |
{
|
127 |
"id": 61,
|
128 |
+
"text": "на него никто не польститься",
|
129 |
+
"fixed_text": "на него никто не польстится",
|
130 |
"label": 0,
|
131 |
"error_type": "Grammar",
|
132 |
"domain": "prose"
|
|
|
136 |
Please note that error categories are not always set accurately, so you should not use
|
137 |
the "error_type" field to train classifiers.
|
138 |
|
139 |
+
### Uncensoring samples
|
140 |
+
|
141 |
+
A number of samples contain text with explicit obscenities:
|
142 |
+
|
143 |
+
```
|
144 |
+
{
|
145 |
+
"id": 1,
|
146 |
+
"text": "Но не простого - с лёгкой еб@нцой.",
|
147 |
+
"fixed_text": "Но не простого - с лёгкой ебанцой.",
|
148 |
+
"label": 0,
|
149 |
+
"error_type": "Misspelling",
|
150 |
+
"domain": "prose"
|
151 |
+
}
|
152 |
+
```
|
153 |
|
154 |
### Statistics
|
155 |
|
156 |
Statistics for test split.
|
157 |
|
158 |
```
|
159 |
+
+--------+----------+--------------+----------+-------------+---------+---------+---------+-----------+--------+-------+
|
160 |
+
| Domain | Spelling | Tokenization | No error | Punctuation | Unknown | Mixture | Grammar | Semantics | Gender | TOTAL |
|
161 |
+
+--------+----------+--------------+----------+-------------+---------+---------+---------+-----------+--------+-------+
|
162 |
+
| prose | 2127 | 2111 | 18033 | 1562 | 658 | 178 | 203 | 20 | 1 | 24893 |
|
163 |
+
| poetry | 209 | 339 | 693 | 475 | 629 | 153 | 2 | 0 | 0 | 2500 |
|
164 |
+
+--------+----------+--------------+----------+-------------+---------+---------+---------+-----------+--------+-------+
|
165 |
+
```
|
166 |
+
|
167 |
+
Statistics on the number of edits required to obtain a corrected version of the text:
|
168 |
+
```
|
169 |
+
+-----------------+-------------------+------------------+
|
170 |
+
| Number of edits | Number of samples | Share of samples |
|
171 |
+
+-----------------+-------------------+------------------+
|
172 |
+
| 1 | 5824 | 0.75 |
|
173 |
+
| 2 | 1137 | 0.15 |
|
174 |
+
| 3 | 348 | 0.04 |
|
175 |
+
| 4 | 184 | 0.02 |
|
176 |
+
| 5 | 97 | 0.01 |
|
177 |
+
| >5 | 183 | 0.02 |
|
178 |
+
+-----------------+-------------------+------------------+
|
179 |
```
|
180 |
+
|
181 |
+
## See also
|
182 |
+
|
183 |
+
[RuCOLA](https://huggingface.co/datasets/RussianNLP/rucola)
|
184 |
+
[ai-forever/spellcheck_benchmark](https://huggingface.co/datasets/ai-forever/spellcheck_benchmark)
|
185 |
+
|