Icelandic-English Parallel Sentence Extraction Dataset 21.10

  Authors

Steingrímsson, Steinþór

  Item identifier

http://hdl.handle.net/20.500.12537/151

 Date issued

2021-10-01

 Type

corpus, text

 Size

100000 sentences

 Language(s)

English , Icelandic

 Description

This is a BUCC-style dataset for testing the accuracy of parallel sentence extraction from comparable corpora. It has 100 thousand sentences, of which 2000 are parallel pairs, and the other 98 thousand for each language randomly selected sentences from the same domain (news). The parallel sentences are from the WMT 2021 news translation task, the randomly selected Icelandic sentences from the news subcorpus of the Icelandic Gigaword Corpus and the randomly selected English sentences from the English newscrawl. Þetta er gagnasett til að prófa nákvæmni aðferða við að veiða samhliða setningapör úr sambærilegum málheildum. Það inniheldur 100 þúsund setningar, af þeim eru 2000 samhliða.

 Publisher

Reykjavík University

 Subject(s)

parallel corpora comparable corpora sentence alignment sentence filtering aligned pairs aligned sentence pairs machine translation

 Collection(s)

Clarin IS

Show full item record