Sociolinguistics Laboratory holdings:

Sociolinguistics Laboratory Corpus Holdings:

Type of Data

Register

Demographic Data

Name

*=requires special software (provided)

# CD-ROMs

Language(s)

Speech

Synt Phrase

Broadcast News Speech

Soundfile only

Soundfile (Waveform)

Textual

Transcribed

WL (carrier)

Sent

Cas

None

Gender

Age

Region

Education

CSR-IV Hub 4

English

√

CELEX2*

Dutch,

Eng,

Ger

√

CallFriend-English*

AmEng

√

CallFriend-Farsi*

Farsi

√

CallFriend-Hindi*

Hindi

√

CallFriend-Japanese*

Japanese

√

CallFriend-Korean*

Korean

√

CallFriend-Tamil*

Tamil

√

Switchboard-2 Phase II

AmEng (InldN)

√

TIMIT Acoustic-Phonetic Continuous Speech Corpus

Am Eng

√

OR/PH

√

Treebank-2

Eng

√

SC/AN

√

Urban Voices

Brit Eng

√

Notes regarding table:

1. Format of soundfiles is typically .wav or proprietary

2. CallFriend recordings are of speakers in dyads

Other places you can find corpora:

1. EE Library

2. Online – the Linguistics Data Consortium now makes a few of their most popular corpora available online (in searchable form) to guests who have no subscription to LDC otherwise.

a. a great place to begin is with Emily Bender’s NWAVE31 webpage:

"Corpus Methods for Sociolinguists" http://faculty.washington.edu/ebender/links.shtml

Typical Sociolinguistic Practice

1. Sociolinguist acquires her own or someone else’s data (large texts or soundfiles)

2. Transcribes or obtains a transcription of the data (e.g., entire corpus or extracted words)

3. Manually codes these data (phones, phonemes, part of speech, constituent tag) so they may be linked to a timestamp on a recording and to a category of interest (vowel identity, F1, F2) following the researcher’s dependent and independent variables

3. Sets up a database program to contain the measures taken from these data (phonetic, sociolinguistic, syntactic, phonological, etc.) as well as codes for speaker variables (age, gender, class, network score, etc.)

4. Sorts the data so particular forms of interest may be tallied

5. Creates summary tallies of the forms of interest (average frequencies of each speaker’s usage of (-t/d) deletion, for example)

6. Compares tallies obtained for each speaker or group of speakers

Corpora can provide the data...

Special software included in the cd’s can allow for querying, which can help you

...identify forms in the data

...find out the frequencies of forms in the data

...check preceding and following contexts for forms

...and much more!

Demonstrations

So, what’s on the corpus cd’s that we have?

CallFriend

Q: Why would I use CallFriend? A: You would find CallFriend useful if you desired unscripted conversational information. You will find that there are two channels—one for each speaker in the conversation. To separate these channels, you need to use speech analysis software, such as Praat, that allows the data from multiple channels to be read separately.

Q: Where do the CallFriend data come from? A: Recorded conversations between native speakers of the relevant language, recorded using telephones in the USA or overseas.

1. CD is in ISO 9660 format (both MAC and PC accessible format)

2. Go to a folder on the CD. There are three choices: “training”, “test”, and “evaluation”. It doesn’t really matter which you use for most purposes, as all three contain unscripted conversation. Note: the conversations are NOT transcribed.

3. Open DOC folder to learn about the contents of the cd

4. Opening the files in the file folder list:

.tbl files may be opened in Excel