wiki:TigrinyaCorpus

Version 4 (modified by hales, 7 years ago) (diff)

--

D3.3a: A Tigrinya corpus, sized 3 million words

Building the Tigrinya web corpus

We have used the following steps to create a big Tigrinya Web corpus: First, adopting the Corpus factory method [1] bigrams of Tigrinya words from the Crúbadán database [2] were used to query Bing search engine for documents in Tigrinya. URLs of 9,034 documents found by the search engine were used as starting points for web crawler SpiderLing [3].

The following language models were created:

  • Character trigram model for language detection. 538 KB of text in the Ge`ez script from documents found by the search engine and manually checked was used to train the model.
  • Byte trigram model for character encoding detection. The model was trained using web pages obtained by the Corpus factory method.
  • 349 most frequent Tigrinya words from the manually checked bigrams of words from the Crúbadán database were used as a wordlist to check the language of a running text by boilerplate removal tool jusText [4].

The crawler was set to harvest web domains in national top level domains of Ethiopia, Eritrea, Somalia, Djibouti (et, er, so, dj) and other general TLDs (com, org, net, gov, info, edu). 42 GB of http responses was gathered in the process. HTML tags and boilerplate paragraphs were removed from the raw data by jusText. Tigrinya texts were separated using the character trigram language model.

Duplicate or near duplicate paragraphs were identified and removed using tool onion [4]. The final size of the corpus is 39 MB and 2.5 million tokens. The corpus is called tiWaC16 (Tigrinya `Web as Corpus' corpus, year 2016).

Corpus properties

Basic properties of corpus sources are summarised below.

The size of corpus structures:

Document count 1,907
Paragraph count 28,552
Sentence count 139,357
Token count 2,531,443
Ge'ez script lexicon size 225,132
Sera script lexicon size 220,935

Document count – the most frequent web domains and domain size distribution:

Top level domains Web domains Secon level domain size distribution
org 1,023 *.blogspot.com 349 At least 1000 documents 0
com 699 *.jw.org 307 At least 500 documents 0
net 55 tewahdo.org 174 At least 100 documents 4
harnnet.org 116 At least 50 documents 8
eritreantewahdo.org 97 At least 10 documents 28
mekaleh-eritra.org 78 At least 5 documents 42
mahberemariamisrael.com 76 At least 1 document 129
asmarino.com 76
fnoteatnatiewos.com 46
erena.org 41
forumeritrea.org 38
dehnet.org 32

Since the corpus is small, the domain variety is also limited. The content of politics, religious and blog sites has a significant presence in the corpus sources.

Corpus query interface

The corpus has been indexed by corpus manager and query system Sketch Engine [5]. The corpus can be searched at http://corpora.fi.muni.cz/habit/.

References

  • [1] -- Kilgarriff, Adam, Siva Reddy, Jan Pomikálek, and P. V. S. Avinesh. "A Corpus Factory for Many Languages." In LREC. 2010.
  • [2] -- Scannell, Kevin P. "The Crúbadán Project: Corpus building for under-resourced languages." In Building and Exploring Web Corpora: Proceedings of the 3rd Web as Corpus Workshop, vol. 4, pp. 5-15. 2007.
  • [3] -- Suchomel, Vít, and Jan Pomikálek. "Efficient web crawling for large text corpora." In Proceedings of the seventh Web as Corpus Workshop (WAC7), pp. 39-43. 2012.
  • [4] -- Pomikálek, Jan. "Removing boilerplate and duplicate content from web corpora." Disertační práce, Masarykova univerzita, Fakulta informatiky (2011).
  • [5] -- Kilgarriff, Adam, Vít Baisa, Jan Bušta, Miloš Jakubíček, Vojtěch Kovář, Jan Michelfeit, Pavel Rychlý, and Vít Suchomel. "The Sketch Engine: ten years on." Lexicography 1, no. 1 (2014): 7-36.