= D3.4a: A Somali corpus, sized 10 million words =

== Building the Somali Web corpus ==
We have used the following steps to create a big Somali Web corpus: First, adopting the Corpus factory method [1] bigrams of Somali words from the Crúbadán database [2] were used to query Bing search engine for documents in Somali. URLs of 18,108 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. 292 KB of text 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.
 - 304 most frequent Somali 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. Somali 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 461 MB and 80 million tokens.
The corpus is called soWaC16 (Somali `Web as Corpus' corpus, year 2016).

== Corpus properties ==
Basic properties of corpus sources are summarised below.

The size of corpus structures:
||=Document count    =||    385,338||
||=Paragraph count   =||  1,937,758||
||=Sentence count    =||  2,643,336||
||=Token count       =|| 79,741,231||
||=Latin script lexicon size =||  1,399,350||

Document count – the most frequent web domains and domain size distribution:
||||= Top level domains =||||= Web domains =||||= Second level domain size distribution =||
|| net  || 295,358|| risaala.net          || 22,823|| At least 1000 documents  ||    73||
|| org  ||  75,860|| goolfm.net           || 22,544|| At least 500 documents   ||    96||
|| com  ||   7,397|| vidinfo.org          || 21,904|| At least 100 documents   ||   150||
|| info ||   4,577|| batalaalenews.net    || 17,079|| At least 50 documents    ||   181||
|| so   ||   1,930|| keydmedia.net        || 15,453|| At least 10 documents    ||   352||
|| ||             || alshahid.net         || 13,923|| At least 5 documents     ||   487||
|| ||             || daadmadheedhnews.net || 13,693|| At least 1 document      || 1,083||
|| ||             || somaliland.org       || 13,203|| || ||
|| ||             || vidoser.org          || 12,189|| || ||
|| ||             || somalilandpost.net   || 10,853|| || ||
|| ||             || radiodanan.net       || 10,196|| || ||
|| ||             || geeska.net           ||  8,378|| || ||
|| ||             || camuudnews.net       ||  8,218|| || ||
|| ||             || nogob.net            ||  7,045|| || ||
|| ||             || allsomali24.org      ||  6,755|| || ||
|| ||             || sagalradio.org       ||  6,154|| || ||
|| ||             || qarninews.net        ||  6,097|| || ||

Apart from other African languages represented in HaBiT project corpora, the Somalian corpus consists of texts from a broad number of Web domains. The content of news/politics sites has a significant presence in the corpus sources.

The most frequent words:
||=Word (Latin script) =||= Count =||
||oo	|| 2,130,200||
||ka	|| 1,808,365||
||ay	|| 1,470,184||
||ku	|| 1,445,719||
||iyo	|| 1,248,166||
||ee	|| 1,210,830||
||ah	|| 1,062,164||
||u	    || 1,041,418||
||in	|| 1,037,431||
||ayaa	||   985,020||
||uu	||   950,971||
||soo	||   794,868||
||la	||   720,451||
||lagu	||   397,822||
||ugu	||   365,182||

== 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/run.cgi/first_form?corpname=sowac16.

== 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.