Basic structure of the corpus

FRED-S spans a subset of FRED texts not subject to copyright restrictions. More specifically, FRED-S covers:

1,011,396 running words
c. 123 hours of recorded speech
121 interviews
144 dialect speakers
57 different locations, in 18 different counties, in 5 major dialect areas:

the Southwest and Southeast of England, the Midlands, the North of England, and the Scottish Lowlands

(the full version covers 9 major dialect areas, including also the Scottish Highlands, the Hebrides, Wales and the Isle of Man; 43 different counties; c. 300 hours of recorded speech; 372 interviews with 431 speakers)

The recordings

All conversations in FRED-S were recorded between 1970 and 2000, the majority in the 1970s and 1980s (see table below). The recording date and/or recording decade for each text are specified in the text header as <RECDAT year> and <RECDEC 3 digits decade reference>. Except for a few texts where the tape was interrupted or the recording stops before the actual end of the conversation, most recordings are full-length interviews which last between 30 and 90 minutes. The longest interview is KEN_003 with almost 4.5 hours and roughly 47,000 words.

Table 1. Text distribution by recording date, FRED-S.

recording date

number of texts

% of textual material

1970 - 1979

47

42.2%

1980 - 1989

56

43.9%

1990 - 1999

15

10.3%

unknown

3

3.6%

Total

121

100.0%

Geographic coverage - dialect areas, counties and locations

FRED can be used to investigate single regional varieties as well as for the cross-regional comparison of linguistic features. On the uppermost geographic level, the texts are divided into 5 major dialect areas; for a more fine-grained analysis, there is also information on the county and location where each interview was recorded. The dialect area, county and location for each text are specified in the text header as <area value>, <county value> and <location value>. The county of a text can also be inferred from the file name, which starts with a three-letter 'Chapman county code' followed by underscore and a running number (DEV_001, DEV_002, DEV_003, etc. would all be texts from Devon; see www.genuki.org.uk for the Chapman county codes).

Tables 2 and 3 show a breakdown by dialect area and county of the amount of textual material in FRED-S (excluding interviewer utterances). The dialect area, county and location of each text are also specified in both the FRED-S text list (table 4) and the FRED-S speaker list (table 5) below.

Table 2. Text distribution by dialect area, FRED-S.

dialect area

number of texts

running words

% of textual material

Southwest (SW)

38

264,863

26.2%

Southeast (SE)

17

260,643

25.8%

Midlands (Mid)

16

152,535

15.1%

North (N)

30

266,955

26.4%

Scottish Lowlands (ScL)

20

66,400

6.6%

Total

121

1,011,396

100.0%

 

Table 3. Text distribution by county, FRED-S.

county

Chapman county code

dialect area

running words

% of textual material

Cornwall

CON

SW

26,535

2.6%

Devon

DEV

SW

79,870

7.9%

Oxfordshire

OXF

SW

13,801

1.4%

Somerset

SOM

SW

69,321

6.9%

Wiltshire

WIL

SW

75,336

7.4%

Kent

KEN

SE

155,192

15.3%

London

LND

SE

74,856

7.4%

Middlesex

MDX

SE

30,595

3.0%

Leicestershire

LEI

Mid

2,341

0.2%

Nottinghamshire

NTT

Mid

150,194

14.9%

Durham

DUR

N

26,507

2.6%

Lancashire

LAN

N

139,845

13.8%

Northumberland

NBL

N

27,777

2.7%

Westmorland

WES

N

21,304

2.1%

Yorkshire

YKS

N

51,522

5.1%

East Lothian

ELN

SCL

28,985

2.9%

Midlothian

MLN

SCL

21,068

2.1%

West Lothian

WLN

SCL

16,347

1.6%

 

FRED-S areal coverage: locations of samples in the corpus

Figure 1. FRED-S areal coverage: locations of samples in the corpus.

Genre(s) covered by the corpus

FRED is a spoken-language corpus (audio recordings with orthographic transcripts). It consists of oral history interviews with native speakers of BrE; the conversations are spontaneous and informal.