D3.1b: An Amharic corpus, sized 20 million words, morphologically annotated

Building the Amharic Web corpus

The Building of the corpus is described at AmharicCorpus.

Corpus properties

Basic properties of corpus sources are summarised below.

The size of corpus structures:

Document count	33,542
Paragraph count	341,327
Sentence count	1,208,926
Token count	20,287,250
Ge'ez lexicon size	955,628
Sera lexicon size	948,553

Morphological annotation

The corpus was tagged by TreeTagger with a model trained on the cleaned version of the WIC Corpus [1].

The average accuracy of the tagger is 87.4 %.

Tag-set

Basic Class	Definition of the tag	Tag
Noun	Verbal/infinitival noun, formed from any verb	VN
	Noun attached with a preposition	NP
	Noun attached with conjunction	NC
	Noun with a proclitic preposition and an enclitic conjunction	NPC
	Any other noun	N
Pronoun	Pronoun attached with preposition	PRONP
	Pronoun attached with conjunction	PRONC
	Pronoun with a proclitic preposition and an enclitic conjunction	PRONPC
	Any other Pronoun	PRON
Verb	Auxiliary verb	AUX
	Relative verb	VREL
	Verb attached with preposition	VP
	Verb attached with conjunction	VC
	Verb with a proclitic preposition and an enclitic conjunction	VPC
	Verb (all other)	V
Adjective	Adjective attached with preposition	ADJP
	Adjective attached with conjunctions	ADJC
	Adjective with a proclitic preposition and an enclitic conjunction	ADJPC
	Any other Adjective	ADJ
Preposition	Preposition	PREP
Conjunction	Conjunction	CONJ
Adverb	Adverb	ADV
Numeral	Cardinal	NUMCR
	Ordinal	NUMOR
	Numeral attached with preposition	NUMP
	Numeral attached with conjunction	NUMC
	Numeral with aproclitic preposition and an enclitic conjunction	NUMPC
Interjection	Interjections	INT
Punctuation	Punctuation	PUNC
Unclassified	Unclassified	UNC

Tag frequencies

The most frequent parts of speech in both corpora are nouns and verbs. The most frequent part of speech tags:

Part of speech tag	Token count
N	7,386,470
NP	2,660,200
VP	1,601,728
V	1,331,531
SENT	946,905
VREL	920,223
PUNC	741,439
PREP	729,404
NUMCR	687,686
ADJ	647,608
PRON	391,243
VN	389,152
AUX	373,346
NC	322,592
CONJ	292,046
PRONP	204,243
ADV	173,772
NPC	140,109
ADJP	126,138

Corpus query interface

The corpus has been indexed by corpus manager and query system Sketch Engine [2]. The corpus can be searched at ​http://corpora.fi.muni.cz/habit/.

References

• [1] -- RYCHLÝ, Pavel and Vít SUCHOMEL. Annotated Amharic Corpora. In Proceedings of Text, Speech, and Dialogue 19th International Conference, TSD 2016 Brno
• [2] -- 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.