amWaC: Amharic corpus from the web

The Amharic web corpus (amWaC) is an Amharic corpus made up of texts collected from the Internet. The corpus was prepared according to the standards described in the document A Corpus Factory for Many Languages (Kilgarriff et al. at LREC 2010).

Data was crawled by the SpiderLing web spider three times, in August 2013, October 2015, January 2016 and 2017 with a final size of almost 26 million words. Texts are in the Ge’ez script with matching SERA transliteration (The system for Ethiopic representation in ASCII).

Transliteration of selected Ge’ez characters into SERA system (Latin script).

Amharic corpus - Transliteration of selected Ge'ez characters into SERA system

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

Top level domains Web domains Domain size distribution
com 33,503 gov.et 21,224 At least 1000 documents 12
et 21,731 blogspot.com 10,328 At least 500 documents 6
org 18,631 jw.org 6,512 At least 100 documents 29
net 1,295 addisadmassnews.com 3,751 At least 50 documents 19
va 106 wikipedia.org 3,228 At least 10 documents 69
others 243 ethiopiazare.com 3,045 At least 1 document 202

The content of news/political and religious sites has a significant presence in the corpus sources.

The corpus was created in the framework of the HaBiT project (Harvesting big text data for under-resourced languages), see more on https://habit-project.eu/wiki/AmharicCorpus

Part-of-speech tagset

The AmharicWaC corpus was tagged with the TreeTagger based on manual annotation of Amharic 1065 news items containing 210,000 prosodic words. See the Amharic part-of-speech tag legend.

Tools to work with the Amharic corpus

A complete set of tools is available to work with this Amharic corpus to generate:

  • word sketch – Amharic collocations categorized by grammatical relations
  • thesaurus – synonyms and similar words for every word
  • keywords – terminology extraction of one-word units
  • word lists – lists of Amharic nouns, verbs, adjectives etc. organized by frequency
  • n-grams – frequency list of multi-word units
  • concordance – examples in context
  • text type analysis – statistics of metadata in the corpus

version 1 (21st April 2017)

  • created word sketches
  • added attribute “sera”

initial version (5th April 2017)

  • size 17 million words

Amharic web corpus

Rychlý, P., & Suchomel, V. (2016, September). Annotated Amharic Corpora. In International Conference on Text, Speech, and Dialogue (pp. 295-302). Springer International Publishing.

Corpus factory method

Kilgarriff, A., Reddy, S., Pomikálek, J., & Avinesh, P. V. S. (2010, May). A corpus factory for many languages. In LREC.

Search the Amharic corpus

Sketch Engine offers a range of tools to work with this Amharic corpus.

Other text corpora

Sketch Engine offers 800+ language corpora.

Use Sketch Engine in minutes

Generate collocations, frequency lists, examples in contexts, n-grams or extracting terms. Use our Quick Start Guide to learn it in minutes.