2.2 Comparative overview of the corpora quoted

The following features some additional information on the British National Corpus (BNC) and the Freiburg-Lancaster-Oslo-Bergen Corpus of British English (F-LOB), which are used by Hoffmann and Smith & Seoane respectively.

	BNC	FLOB
Language variety	British English	British English
Size (words)	100 million	1 million
Period	End of compilation: 1991-1995 End of data collection: 1993	End of compilation: 1991-1996 End of data collection: 1991
Written/spoken proportion	written (90%), spoken (10%)	written (100%)
Linguistic annotation	POS-tagged	POS-tagged
Manual check	no	checked manually
Corpus software	XAIRA BNCweb CQP	Wordsmith*

*the Freiburg-Lancaster-Oslo-Bergen Corpus can be used with independent concordancers, such as Wordsmith

PART-OF-SPEECH TAGGING

Part-of-speech tagging (see Glossary for definitions of the main terms) is the most common kind of grammatical annotation, normally performed by automatic POS-taggers. The POS-tagging tools and tagsets used for the BNC and the FLOB are different.

	BNC	FLOB
Tagset	CLAWS 5	CLAWS 8
Number of POS-tags	61 POS-tags and 30 ambiguity tags	141 tags

The different number of POS-tags already implies that the tagsets are not similar to one another, though the basic word class categories are normally marked in the same way. Note that both the BNC and the FLOB are annotated with CLAWS taggers. Compare the following POS-tags for general adverb and singular common noun:

	BNC	FLOB
General adverb	AV0	RR
Singular common noun	NN1	NN1

These peculiarities are especially important when carrying out a search in different corpora (one for training and one for testing the query), because additional conversion of queries consisting of part-of-speech tags is required.

The part-of-speech tagsets for these two corpora can be accessed at:

BNC tagset http://www.natcorp.ox.ac.uk/docs/c5spec.html
FLOB tagset http://research.ncl.ac.uk/necte/clawstags_c8.pdf

Created with the Personal Edition of HelpNDoc: Write eBooks for the Kindle