Datasets:

Modalities:
Tabular
Text
Formats:
csv
DOI:
Libraries:
Datasets
pandas
License:
fdelucaf commited on
Commit
e8592da
·
verified ·
1 Parent(s): 164d20e

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +68 -39
README.md CHANGED
@@ -41,7 +41,7 @@ license: cc-by-4.0
41
 
42
  ### Dataset Summary
43
 
44
- The CA-EN Parallel Corpus is a Catalan-English dataset of **14.385.296** parallel sentences. The dataset was created to support Catalan in NLP tasks, specifically
45
  Machine Translation.
46
 
47
  ### Supported Tasks and Leaderboards
@@ -57,28 +57,38 @@ The sentences included in the dataset are in Catalan (CA) and English (EN).
57
 
58
  ### Data Instances
59
 
60
- The dataset is a single tsv file where each row contains a parallel sentence pair and additional domain and text type information for each sentence.
61
- Datafields are separated by \,
62
- Text delimiter is \"
 
 
 
63
 
64
 
65
  ### Data Fields
66
 
67
  Each example contains the following 7 fields:
68
- * sentence_id: unique alphanumeric sentence identifier
69
- * en: ENGLISH
70
- * en_sentence: English sentence
71
- * ca: CATALAN
72
- * ca_sentence: Catalan sentence
73
- * domain: sentence domain
74
- * text_type: sentence text type
 
75
 
76
  #### Example:
77
 
78
  <pre>
79
  [
80
  {
81
- "00000a47-e4a5-8ab6-e0fa-3cbbeb596f34","en","As for the search engines, they also rely on the structure of your information content on the website to analyze and index your website.","ca","Pel que fa als motors de cerca, també es basen en l'estructura del seu contingut d'informació al lloc web per analitzar i indexar el seu lloc web.","MWM","SM"
 
 
 
 
 
 
82
 
83
  },
84
  ...
@@ -88,32 +98,52 @@ Each example contains the following 7 fields:
88
 
89
  List of domains
90
 
91
- * AUT: Automotive, transport, traffic regulations
92
- * LEG: legal, law, HR, certificates, degrees
93
- * MWM: Marketing, web, merchandising, customer support and service, e-commerce , advertising, surveys
94
- * LSM: Medicine, natural sciences, food/nutrition, biology, sexology, cosmetics, chemistry, genetics
95
- * ENV: Environment, agriculture, forestry, fisheries, farming, zoology, ecology
96
- * FIN: Finance, economics, business, entrepreneurship, business, competitions, labour, employment, accounting, insurance, insurance
97
- * POL: Politics, international relations, European Union, international organisations, defence, military
98
- * PRN: Porn, inappropriate content
99
- * COM: Computers, IT, robotics, domotics, home automation, telecommunications
100
- * ING: Pure engineering (mechanical, electrical, electronic, aerospace...), meteorology, mining, engineering, maritime, acoustics
101
- * ARC: Architecture, civil engineering, construction, public engineering
102
- * MAT: Mathematics, statistics, physics
103
- * HRM: History, religion, mythology, folklore, philosophy, psychology, ethics, anthropology, tourism
104
- * CUL: Art, poetry, literature, cinema, video games, theatre, theatre/film scripts, esotericism, astrology, sports, music, photography
105
- * GEN: General - generic cathegory with topics such as clothing, textiles, gastronomy, etc.
 
 
 
 
 
 
 
 
 
 
 
 
 
 
106
 
107
 
108
  List of text types
109
 
110
- * PAT: Patents
111
- * SM: Social Media (social networks, chats, forums, tweets...)
112
- * CON: Vernacular (transcription of conversations, subtitles)
113
- * EML: Emails
114
- * MNL: Manuals, data sheets
115
- * NEW: News, journalism
116
- * GEN: Prose, generic type of text
 
 
 
 
 
 
117
 
118
 
119
  ### Data Splits
@@ -135,10 +165,8 @@ domains and styles.
135
  The source data is Catalan authentic text translated to English or authentic English text translated to Catalan.
136
 
137
  The data was obtained through a combination of human translation and machine translation with human proofreading.
138
- After the translation process, the data was deduplicated and filtered to remove any sentence pairs with a cosine similarity of less than 0.75 in order
139
- to improve the data alignment quality.
140
- This was done using sentence embeddings calculated using [LaBSE](https://huggingface.co/sentence-transformers/LaBSE).
141
- The obtained cleaned corpus consists of **14.385.296** parallel sentences of human quality.
142
 
143
  #### Who are the source language producers?
144
 
@@ -171,7 +199,8 @@ Inherent biases may exist within the data.
171
 
172
  ### Other Known Limitations
173
 
174
- The dataset contains data of several specific domains. Application of this dataset in other domains would be of limited use.
 
175
 
176
  ## Additional Information
177
 
 
41
 
42
  ### Dataset Summary
43
 
44
+ The CA-EN Parallel Corpus is a Catalan-English dataset of **14.967.979** parallel sentences. The dataset was created to support Catalan in NLP tasks, specifically
45
  Machine Translation.
46
 
47
  ### Supported Tasks and Leaderboards
 
57
 
58
  ### Data Instances
59
 
60
+ The dataset is a single tsv file where each row contains a parallel sentence pair, as well as the following information per sentence:
61
+
62
+ * language probability score calculated with the language detector [lingua.py](https://github.com/pemistahl/lingua-py),
63
+ * alignment score calculated with [LaBSE](https://huggingface.co/sentence-transformers/LaBSE),
64
+ * domain
65
+ * text type.
66
 
67
 
68
  ### Data Fields
69
 
70
  Each example contains the following 7 fields:
71
+
72
+ * ca: Catalan sentence
73
+ * en: English sentence
74
+ * ca_prob: Language probability for the Catalan sentence calculated with the language detector [lingua.py](https://github.com/pemistahl/lingua-py)
75
+ * en_prob: Language probability for the English sentence calculated with the language detector [lingua.py](https://github.com/pemistahl/lingua-py)
76
+ * alignment: Sentence pair alignment score calculated with [LaBSE](https://huggingface.co/sentence-transformers/LaBSE)
77
+ * Domain: Domain (see List of domains)
78
+ * Type: Text type (see list of text types)
79
 
80
  #### Example:
81
 
82
  <pre>
83
  [
84
  {
85
+ Pel que fa als motors de cerca, també es basen en l'estructura del seu contingut d'informació al lloc web per analitzar i indexar el seu lloc web.
86
+ As for the search engines, they also rely on the structure of your information content on the website to analyze and index your website.
87
+ 0.9999799355804416
88
+ 0.9993718600460302
89
+ 0.91045034
90
+ MWM
91
+ SM
92
 
93
  },
94
  ...
 
98
 
99
  List of domains
100
 
101
+ AUT: Automotive, transport, traffic regulations
102
+
103
+ LEG: legal, law, HR, certificates, degrees
104
+
105
+ MWM: Marketing, web, merchandising, customer support and service, e-commerce , advertising, surveys
106
+
107
+ LSM: Medicine, natural sciences, food/nutrition, biology, sexology, cosmetics, chemistry, genetics
108
+
109
+ ENV: Environment, agriculture, forestry, fisheries, farming, zoology, ecology
110
+
111
+ FIN: Finance, economics, business, entrepreneurship, business, competitions, labour, employment, accounting, insurance, insurance
112
+
113
+ POL: Politics, international relations, European Union, international organisations, defence, military
114
+
115
+ PRN: Porn, inappropriate content
116
+
117
+ COM: Computers, IT, robotics, domotics, home automation, telecommunications
118
+
119
+ ING: Pure engineering (mechanical, electrical, electronic, aerospace...), meteorology, mining, engineering, maritime, acoustics
120
+
121
+ ARC: Architecture, civil engineering, construction, public engineering
122
+
123
+ MAT: Mathematics, statistics, physics
124
+
125
+ HRM: History, religion, mythology, folklore, philosophy, psychology, ethics, anthropology, tourism
126
+
127
+ CUL: Art, poetry, literature, cinema, video games, theatre, theatre/film scripts, esotericism, astrology, sports, music, photography
128
+
129
+ GEN: General - generic cathegory with topics such as clothing, textiles, gastronomy, etc.
130
 
131
 
132
  List of text types
133
 
134
+ PAT: Patents
135
+
136
+ SM: Social Media (social networks, chats, forums, tweets...)
137
+
138
+ CON: Vernacular (transcription of conversations, subtitles)
139
+
140
+ EML: Emails
141
+
142
+ MNL: Manuals, data sheets
143
+
144
+ NEW: News, journalism
145
+
146
+ GEN: Prose, generic type of text
147
 
148
 
149
  ### Data Splits
 
165
  The source data is Catalan authentic text translated to English or authentic English text translated to Catalan.
166
 
167
  The data was obtained through a combination of human translation and machine translation with human proofreading.
168
+
169
+ The obtained corpus consists of **14.967.979** parallel sentences of human quality.
 
 
170
 
171
  #### Who are the source language producers?
172
 
 
199
 
200
  ### Other Known Limitations
201
 
202
+ The dataset contains data of several specific domains. The dataset can be used as a whole or extracting subsets per domain or text types.
203
+ Applications of this dataset in domains other than the ones included in the domain list would be of limited use.
204
 
205
  ## Additional Information
206