The original ABI database (ABI-1) was developed by Aurix Ltd to investigate the performance of speech recognition systems with different speakers from the British Isles. Aurix currently markets a range of speech technology products including audio mining, speech detection, speech recognition and audio alignment

ABI-1 was collected by the University of Birmingham

The Accents of the British Isles (ABI-1) Speech Corpus

The 'Accents of the British Isles' (ABI-1) database is a collection of speech recordings of 280 speakers sampled throughout the British Isles.

*** ABI-1 is now available from The Speech Ark ***

The corpus consists of around 70 hours of recordings covering 14 distinct accent regions within the British Isles. The material has been designed to be suitable for training and evaluating speech technology systems or for phonetic study and analysis.

For each region, an average of 10 male and 10 female subjects each recorded around 15 minutes of read speech. The aim was that each speaker and both of their parents should have lived in their particular region all their lives. Speakers' ages ranged from 16 to 79, and all of the data was recorded using both a close-talking and far-field microphone in a range of natural environments.


Belfast, N. Ireland
Burnley, Lancashire
Denbigh, N. Wales
Dublin, S. Ireland
Elgin, Scottish Highlands
Glasgow, Scotland
Hull, East Yorkshire
Lowestoft, East Anglia
Standard British English
Tower Hamlets, Inner London
Truro, Cornwall

Material (per speaker)

Application oriented:
Catalogue Codes:
Equipment Control:
Game Commands:
PIN Numbers:

Phonetically motivated:
Careful Words:

10 sequences of 4 alphabetic characters (e.g. "BKUN")
40 command phrases (e.g. "play disk two track three")
60 one or two word phrases (e.g. "change view")
10 4-digit sequences (e.g. "two five eight three")

19 isolated "h-vowel-d" words repeated 5 times
20 phonetically balanced sentences
18 phonetically rich phrases
1 1/2 minute accent diagnostic passage


The full database containing both the microphone types is available to academic institutions for 'research purposes only' at 1,000 and for 'commercial use' at 20,000. (prices correct at time of going to press, 18th April 2006).

For further information please contact The Speech Ark - using our enquiry form