Available via license: CC BY 4.0
Content may be subject to copyright.
e Spoken BNC2014
Designing and building a spoken corpus
of everyday conversations
Robbie Love
i, Claire Dembry
ii, Andrew Hardie
i,
Vaclav Brezina
i and Tony McEnery
i
i Lancaster University / ii Cambridge University Press
is paper introduces the Spoken British National Corpus 2014, an
11.5-million-word corpus of orthographically transcribed conversations
among L1 speakers of British English from across the UK, recorded in the years
2012–2016. Aer showing that a survey of the recent history of corpora of spo-
ken British English justies the compilation of this new corpus, we describe
the main stages of the Spoken BNC2014’s creation: design, data and metadata
collection, transcription, XML encoding, and annotation. In doing so we aim
to (i) encourage users of the corpus to approach the data with sensitivity to the
many methodological issues we identied and attempted to overcome while com-
piling the Spoken BNC2014, and (ii) inform (future) compilers of spoken corpora
of the innovations we implemented to attempt to make the construction of cor-
pora representing spontaneous speech in informal contexts more tractable, both
logistically and practically, than in the past.
Keywords: Spoken BNC2014, transcription, corpus construction,
spoken corpora
. Introduction
e ESRC Centre for Corpus Approaches to Social Science (CASS)
1 at Lancaster
University and Cambridge University Press have compiled a new, publicly-
accessible corpus of present-day spoken British English, gathered in informal con-
texts, known as the Spoken British National Corpus 2014 (Spoken BNC2014). is
. e research presented in this paper was supported by the ESRC Centre for Corpus Approaches
to Social Science, ESRC grant reference ES/K002155/1.
International Journal of Corpus Linguistics : (), –. ./ijcl...lov
- / - - © John Benjamins Publishing Company
is is an open access article under a OA CC BY license
Robbie Love et al.
is the rst publicly-accessible corpus of its kind since the spoken component of
the original British National Corpus, completed in 1994, which, despite its age,
is still used as a proxy for present-day English in research today (e.g. Hadikin
2014, Rühlemann & Gries 2015). e new corpus contains data from the years
2012 to 2016. It is publicly available via Lancaster University’s CQPweb server as
of September 2017, with the underlying XML les downloadable from late 2018.
It will subsequently form the spoken component of the larger British National
Corpus 2014, whose written component is also under development. e main
source of recordings for the project was a national participation campaign. In
total, the research team has amassed 11.5 million words of transcribed content,
featuring 668 speakers in 1,251 recordings.
is paper describes how we designed and built the Spoken BNC2014. Section2
presents a review of a number of existing corpora of spoken English. We start by
introducing our own corpus’s predecessor, the spoken component of the original
British National Corpus (hereaer Spoken BNC1994), and then consider other
notable English corpora that consist wholly or partly of spoken data compiled
aer 1994. is review demonstrates that no subsequent corpus fulls the criteria
which appear to have led to the Spoken BNC1994’s great utility and longevity in
linguistic research. In Section3, we address decisions relating to corpus design and
the collection of both the data (audio recordings) and metadata (speaker and re-
cording information). is design necessarily represents a compromise between the
ideally representative corpus and the constraints of what is realistically possible. e
compromise we adopt is employing an opportunistic approach to data collection
rather than attempting to adhere to a predetermined sampling frame, as discussed
in detail in Section3.1. Likewise, the design of the metadata scheme for the corpus
represents a compromise between (i) the need for comparability with the Spoken
BNC1994 and (ii) the desirability of improving upon some of the decisions made
during the compilation of the Spoken BNC1994. Finally, we discuss the participant
recruitment method and our decision to use smartphones as recording devices to
collect our data. In Section4, we describe the development of a bespoke transcrip-
tion scheme for the Spoken BNC2014, as well as decisions made with regards to
XML conversion, POS-tagging and lemmatization.
. Similar existing corpora– why do we need a new one?
A well-known problem aicting corpus linguistics as a eld is its tendency to pri-
oritise written forms of language over spoken forms, in consequence of the much
greater diculty, higher cost and slower speed of collecting transcribed speech:
e Spoken BNC2014
A rough guess suggests that the cost of collecting and transcribing in electronic
form one million words of naturally occurring speech is at least 10 times higher
than the cost of adding another million words of newspaper text.
(Burnard 2002: 6)
Contemporary online access to newspaper material means that this disparity is
likely to be even greater today than in 2002. e resulting bias in corpus linguistics
towards a “very much written-biased view” (Lüdeling & Kytö 2008: vi) of language
is problematic if one takes the view that speech is the primary medium of commu-
nication (Čermák 2009: 113), containing linguistic variables that are important for
the accurate description of language, and yet inaccessible through the analysis of
corpora composed solely of written texts (Adolphs & Carter 2013: 1).
e Spoken BNC1994 is one of the few widely accessible corpora of spoken
British English, and is thus heavily relied upon in corpus research on spoken
English. However, that the “go-to” dataset is over twenty years old is, as we will show
in Section2.4, a problem for current and future research. Addressing this situation
is the main reason for our undertaking the development of a new spoken corpus.
. e Spoken British National Corpus 1994
e compilation of the Spoken BNC2014 is informed largely by the original BNC’s
spoken component (see Crowdy 1993, 1994, 1995), which is “one of the biggest
available corpora of spoken British English” (Nesselhauf & Römer 2007: 297). e
goal of the BNC1994’s creators was “to make it possible to say something about
language in general” (Nesselhauf & Römer 2007: 5). us its spoken component was
designed to function as a representative sample of spoken British English (Burnard
2007). It was created between 1991 and 1994, and was designed in two parts: the
demographically-sampled (DS) part (c. 40%) and the context-governed part (c. 60%)
(Aston & Burnard 1998). e Spoken BNC1994DS, as we call it (also known as the
‘conversational part’, Leech etal. 2001: 2), contains informal, “everyday spontane-
ous interactions” (Leech etal. 2001: 2), and its contributors (the volunteers who
made the recordings of their interactions with other speakers) were “selected by
age group, sex, social class and geographic region” (Aston & Burnard 1998: 31). 124
adult contributors made recordings capturing the language of over 1,000 speakers
(Aston & Burnard 1998: 32). In total, the Spoken BNC1994DS contains 4.2 mil-
lion words of transcribed conversation. e context-governed part (also known
as the ‘task-oriented part’, Leech etal. 2001: 2) contains formal encounters from
particular institutional settings, which were “categorised by topic and type of inter-
action” (Aston & Burnard 1998: 31). e Spoken BNC1994DS also incorporates the
Bergen Corpus of London Teenage Language (COLT), a half-million-word sample of
Robbie Love et al.
spontaneous conversations among teenagers between the ages of 13–17, collected in
a variety of boroughs and school districts in London in 1993 (Stenström etal. 2002).
Despite certain weaknesses in design and metadata, which we discuss in
Section2.4, the Spoken BNC1994 has proven a highly productive resource for
linguistic research over the last two decades. It has been inuential in the areas
of grammar (e.g. Rühlemann 2006, Gabrielatos 2013, Smith 2014), sociolinguis-
tics (e.g. McEnery 2005, Säily 2011, Xiao & Tao 2007), conversation analysis (e.g.
Rühlemann & Gries 2015), pragmatics (e.g. Wang 2005, Capelle etal. 2015, Hatice
2015), and language teaching (e.g. Alderson 2007, Flowerdew 2009), among others.
Part of the reason for the widespread use of the BNC1994 is that it is an open-access
corpus; researchers from around the world can access the corpus at zero cost, ei-
ther by downloading the full text from the Oxford Text Archive (http://ota.ox.ac.
uk/desc/2554), or using the online interfaces provided by various institutions in-
cluding Brigham Young University (BNC-BYU, Davies 2004, http://corpus.byu.edu/
bnc/) and Lancaster University (BNCweb, Homann etal. 2008, http://bncweb.
lancs.ac.uk/bncwebSignup). Yet it is undoubtedly the unique access that the Spoken
BNC1994 has provided to large scale orthographic transcriptions of spontaneous
speech that has been the key to its success. Such resources are needed by linguists
but are expensive and time consuming to produce and hence are rarely accessible
as openly and easily as is the BNC1994.
. Other British English corpora containing spoken conversational data
Other corpora of spoken British English exist which are similarly conversational
and non-specialized in terms of context. Although they have the potential to be
just as inuential as the Spoken BNC1994DS, they are much harder to access for
several reasons. Some have simply not been made available to the public for com-
mercial reasons. e Cambridge and Nottingham Corpus of Discourse in English
(CANCODE), for example, forms part of the Cambridge English Corpus, which is a
commercial resource belonging to Cambridge University Press and is not accessible
to the wider research community (Carter 1998: 55). Other corpora are available
only aer payment of a license fee, which makes them generally less accessible. For
instance, Collins publishers’ WordBanks Online (https://www.collins.co.uk/page/
Wordbanks+Online) oers paid access to a 57-million-word subcorpus of the Bank
of English (containing data from British English and American English sources, 61
million words of which is spoken); the charges range, at time of this writing, from a
minimum of £695 up to £3,000 per year of access. Likewise, the British component
of the International Corpus of English (ICE-GB), containing one million words of
written and spoken data from the 1990s (Nelson etal. 2002: 3), costs over £400 for
a single, non-student license (http://www.ucl.ac.uk/english-usage/projects/ice-gb/
e Spoken BNC2014
iceorder2.htm). Some other corpora are generally available, but sample a more
narrowly dened regional variety of English than simply ‘British English’. For in-
stance, the Scottish Corpus of Texts and Speech (SCOTS) (Douglas 2003), while
it is free to use, contains only Scottish English and no other regional varieties of
English from the British Isles.
ese restrictions appear to have translated into a much lower level of research
output using these datasets. As a crude proxy for the academic impact of these
corpora, we searched for publications using them in Lancaster University’s online
library system. At the time of writing, a search for CANCODE retrieves 54 publica-
tions; WordBanks Online only 45; ICE-GB 300; and SCOTS 34. By contrast, search-
ing for the BNC1994 identies 3,000 publications. While an admittedly rough rule
of thumb, this quick search shows that even though conversational, non-specialized
spoken corpora that may be just as useful as the Spoken BNC1994DS have been
compiled since 1994, their limited availability, and/or the expense of accessing
them, has meant that the Spoken BNC1994 is almost certainly the most widely
used spoken corpus of British English to date.
. Justication for the Spoken BNC2014
It is clearly problematic that research into spoken English is still using a 23-year-old
corpus to explore ‘present-day’ English. e reason why no spoken corpus since
the Spoken BNC1994 has equalled its utility for research seems to be that no other
corpus has matched all four of its key strengths:
i. orthographically transcribed data
ii. large size
iii. general coverage of spoken British English
iv. (low or no cost) public access
Each of the other projects mentioned above fails to full one or more of these criteria.
e problem of the Spoken BNC1994’s continued use would be lessened if it
were not still treated as a proxy for present-day English– i.e. if its use were mainly
historical– but this is not the case. For researchers interested in spoken English
who do not have access to privately held spoken corpora this is unavoidable; the
Spoken BNC1994 is still clearly the best publicly-accessible resource for spoken
British English for the reasons outlined. Yet as time has passed, the corpus has been
used for purposes for which it is becoming increasingly less suitable. For example, a
recent study by Hadikin (2014), which investigates the behaviour of articles in spo-
ken Korean English, uses the Spoken BNC1994 as a reference corpus of present-day
English. Appropriately, Hadikin (2014) gives the following warning:
Robbie Love et al.
With notably older recordings [than the Korean corpora he compiled] […] one
has to be cautious about any language structures that may have changed, or may
be changing, in the period since then. (Hadikin 2014: 7)
In this respect, Hadikin’s (2014) work typies a range of recent research which, in the
absence of a suitable alternative, uses the Spoken BNC1994 as a sample of present-day
English. e dated nature of the Spoken BNC1994 is demonstrated by the presence
in the corpus of references to public gures, technology, and television shows that
were contemporary in the early 1990s; see Examples(1) to (3):
(1) Oh alright then, so if John Major gets elected then I’ll still [unclear]
2
(BNC, KCF 105)
(2) Why not just put a video on?
3 (BNC, KBC 18)
(3) Did you see e Generation Game?
4 (BNC, KCT 2546)
We see, then, the need for a new corpus of conversational British English to allow
researchers to continue the kinds of research that the Spoken BNC1994 has fos-
tered over the past two decades. is new corpus will also make it possible to turn
the ageing of the Spoken BNC1994 into an advantage– if it can be compared to a
comparable contemporary corpus, it could become a useful resource for exploring
recent change in spoken English. e Spoken BNC2014 project enables scholars to
realise these research opportunities as well as, importantly, allowing gratis public
access to the resulting corpus.
. Corpus design and data collection
A key decision we made early on in the creation of the Spoken BNC2014 was to col-
lect data which occurred only in informal contexts– i.e. data which would be com-
parable to the Spoken BNC1994DS. Our rationale for excluding context-governed
data is simply that we have noted there exists greater use of, and demand for, con-
versational data. Researchers who wish to study spoken British English occurring
in specic contexts are likely to collect their own, specialized corpora. Moreover,
some such specialized corpora have been released publicly by their creators and
. John Major was Prime Minister of the United Kingdom between 1990 and 1997.
. e VHS tape cassette, or ‘video’, was a popular medium for home video consumption in the
1980s and 1990s before the introduction of the DVD in the late 1990s.
. e Generation Game was a popular British television gameshow which was broadcast be-
tween 1971 and 2002.
e Spoken BNC2014
are available to researchers with an interest in the dened context in question; an
example is BASE, the British Academic Spoken English Corpus, which contains
university lectures and seminars (see ompson & Nesi 2001). So, researchers with
an interest in context-governed English speech already have options open to them.
A general corpus of informal speech is, however, harder to collect due to the re-
quirements of size and demographic spread– and therefore, in consequence, much
more in demand in the research community.
Another decision we faced was what we should do about the known short-
comings of the Spoken BNC1994DS. Most importantly, certain issues exist in
the Spoken BNC1994DS in terms of its speaker metadata; it has been criticised
for the “oen unhelpful” and inconsistent availability of speaker metadata (Lam
2009: 176). Indeed, Burnard (2002: 7) admits that the classications used to cate-
gorise speakers are sometimes “poorly dened” and “partially or unreliably pop-
ulated”. In Sections3.1 to 3.3, we describe the steps we took to attempt to improve
upon this in the Spoken BNC2014. By instructing contributors to use their own
devices (smartphones) to make recordings, rather than supplying recording equip-
ment, we could facilitate an opportunistic approach to data collection which re-
quired no training for contributors prior to recording. Furthermore, rather than
instructing contributors to provide metadata on behalf of all the speakers they re-
corded, the contributors gathered metadata directly from each speaker and passed
it on to the research team.
. Opportunistic data collection
Every corpus compilation project is, by definition, a sampling project (Biber
1993: 243). e appropriateness of the sample depends on factors including the
purpose of the research and the domain within which the data is being collected. In
the case of the Spoken BNC1994DS, the demographic categories were divided into
two types: “selection criteria” and “descriptive criteria” (Burnard 2002: 6). e selec-
tion criteria are the gender, age, socio-economic status and region of the speakers.
e descriptive criteria were those which were not controlled during the collection
of the data but which were recorded for information; these included the domain
and type of speech recorded (Burnard 2002: 6).
e aim of the compilers of the Spoken BNC1994 was to enable research “for a
wide variety of linguistic interests” (Wichmann 2008: 189). Part of this aim was to
assemble a corpus that was as representative as possible of the language variety un-
der investigation, i.e. “the language production of the population of British English
speakers in the United Kingdom” (Crowdy 1993: 259).
Before the Spoken BNC1994 was completed, Crowdy (1993) claimed that:
Robbie Love et al.
representativeness is achieved by sampling a spread of language producers in terms
of age, gender, social group, and region, and recording their language output over
a set period of time. (Crowdy 1993: 257, emphasis added)
However, a compromise was clearly made between what would maximize rep-
resentativeness and what was possible in practice. As Burnard (2002: 5) points
out, “no-one could reasonably claim that the corpus was statistically representative
of the whole language”, although he is clear that the combination of criteria for
selection and description would at least encourage proportionality between, and
variability within, the demographic categories (Burnard 2002: 6)– the component
categories of each selection criterion were predetermined, and target proportions
were assigned for each.
Table1 shows how each category of the BNC1994’s selection criteria was pop-
ulated, according to the number of words produced by speakers.
Table1. Proportions of words in Spoken BNC1994 assigned across each of the three
selection criteria (adapted from Burnard 2000: 13)
Selection criteria Category % ‘w-units’
Gender Male 41.14
Female 58.47
Unknown 0.38
Age 0–14 6.30
15–24 15.71
25–34 20.16
35–44 19.96
45–59 22.76
60+ 15.08
Socio-economic status AB 32.41
C1 26.08
C2 25.69
DE 14.91
Unknown 0.88
Particularly underrepresented at the time, it appears, were males, children, and the
elderly. Despite creating a sampling frame for the selection criteria, the priority in
practice seems to have been to collect as much data as possible and to accept the
consequent imbalances in the corpus across the demographic categories. is may
sound like a careless strategy, but we argue that this was the only reasonable ap-
proach given the costs associated with collecting spoken data. Furthermore, some
researchers would go on to cra smaller subcorpora of the data, which were more
balanced according to given metadata categories (e.g. BNC 64, Brezina & Meyerho
e Spoken BNC2014
2014). is means that, despite imbalances across the corpus as a whole, it was still
possible to analyse demographic categories of equal size if one was willing to work
with a smaller data set. Making use of a geological metaphor, the Spoken BNC1994
can be viewed as containing a small “core” of data with evenly balanced demo-
graphic categories, and a larger “mantle” of additional data which, when combined
with the core, produces a large but not balanced corpus.
For the Spoken BNC2014 we decided to adopt a similar approach to the
BNC1994: accepting the data that became available, while monitoring the levels of
the demographic categories to be alerted to any imbalances that were severe. is
is what we call an ‘opportunistic approach’ to data collection. If any such “holes” in
the data began to appear, we attempted to address these by targeting those specic
groups of people– variously through Facebook and Twitter advertisement cam-
paigns, student recruitment campaigns at universities, and press releases which
targeted speakers of a particular age, or from a certain geographical region. e
resulting data set (see Appendix), which is more than twice the size of the Spoken
BNC1994DS, represents an improvement in balance when compared to the Spoken
BNC1994– some categories are well balanced (e.g. northern vs. southern speakers)
and some categories are better populated than they would have been had we not
monitored the numbers and targeted specic social groups (e.g. elderly speakers).
However, there are some peaks and troughs– a major trough being the dearth
of speakers from Scotland, Wales and Northern Ireland. We accept this dearth
because spoken corpora of English spoken by people from these countries have
been collected and made available since the release of the Spoken BNC1994. e
previously mentioned SCOTS (Douglas 2003) contains approximately one million
words of Scottish English speech– most of which was collected in the 2000s. e
Bangor Siarad corpus (Deuchar etal. 2014) contains 450,000 words of bilingual
Welsh-English spontaneous speech collected between 2005 and 2008. ICE-Ireland
(Kallen & Kirk 2008) comprises approximately 300,000 words of spoken data col-
lected from speakers of Northern Irish English in the mid 1990s to early 2000s.
However, as discussed in Section2.2, no comparable corpus containing “English
English” has been made publicly available since the Spoken BNC1994DS; and so
we prioritized collecting data for England, as that is where the greatest need lay.
is prioritization of England does mean that the full Spoken BNC2014, though
not as imbalanced as the Spoken BNC1994DS, is not a properly balanced corpus if
taken as a whole. Yet, as noted, it was no more designed to be so than the Spoken
BNC1994 was. Our resolution is to explicitly facilitate the analysis of the Spoken
BNC2014 both as a full, unbalanced version (maximising the virtue of size), and
also as the “core” on its own (a smaller, balanced subcorpus derived from the whole
corpus). e core subcorpus contains an approximately equal number of tokens
within each category for each of the following criteria: gender, age, socio-economic
Robbie Love et al.
status, and English region. Users of the corpus in Lancaster University’s CQPweb
server are able to move between the entirety of the corpus and the core subcorpus as
they wish, so that they can select whichever ts better with the purpose at hand. e
core/non-core status of dierent segments of the corpus is also coded as metadata
in the XML-format release of the data.
e alternative, non-opportunistic approach– drawing up a sampling frame
and actively seeking out recordings from particular groups of speakers– might well
have produced a more representative or balanced corpus, but would, at the very
least, have undoubtedly taken much longer to produce.
5 at would have worked
against our aim to produce a corpus that can– for a while– be plausibly accepted as
a proxy for present-day British English. It would also have been prohibitively time
consuming to do this, which with a xed level of resource available would necessar-
ily lead to the end-result corpus being smaller by perhaps an order of magnitude.
. Recruitment of participants and audio recording
One of the most innovative features of the Spoken BNC2014 is the use of PPSR
(public participation in scientic research) for data collection (see Shirk etal. 2012).
Anyone who was interested in contributing recordings to the Spoken BNC2014 was
directed to a website which described the aims of the project and included a con-
tact form to allow them to register their interest in contributing data. People who
registered interest were contacted by the Cambridge team via email with further
instructions. Public attention was captured by a series of national media campaigns
in 2014 and 2015, as well as through our participation in public engagement events
such as the Cambridge University Festival of Ideas and the UK Economic and Social
Science Research Council’s Festival of Social Sciences. To incentivize participation,
we oered payment of £18 for every hour of recording of a sucient quality for
corpus transcription, and, importantly, submission of all associated consent forms
and full speaker metadata. All speakers were required to give informed consent
prior to recording, and contributors took responsibility for making recordings and
for gathering consent and metadata from all speakers they recorded. We used this
opportunity to gather metadata from each individual speaker directly, via the con-
tributors, since no contact was made between the research team and the speakers
with whom the contributors chose to converse. To ensure that all information and
. We found that that certain groups (e.g. NS-SEC groups 6 and 7) were less forthcoming with
data than others, despite contributors being paid for providing us with recordings. erefore, it
is not guaranteed that a non-opportunistic approach would produce a more balanced corpus, if
some groups of the population are largely unwilling to contribute, even with the oer of payment.
e Spoken BNC2014
consent was captured, no payments were made to contributors until all metadata,
consent forms and related documentation was fully completed for each recording.
6
Given the general availability of digital audio recorders as a built-in capability
of smartphones and other widely used consumer devices, our decision to use smart-
phones to gather the data meant that all recordings were made digitally. Specically,
contributors were instructed to use the built-in audio recording feature in their
smartphones to make recordings. is is in contrast to the Spoken BNC1994, which
used analogue recording devices, the recordings from which were subsequently
digitised (Crowdy 1994: 15). e use by contributors of their own recording equip-
ment greatly reduced the cost associated with arranging for the recordings, as we
did not need to purchase equipment, distribute it, train users to use it and collect
it back from them.
7
. Metadata categories in the Spoken BNC2014
Unlike the Spoken BNC1994, speakers in the Spoken BNC2014 provided their
own metadata. is gave us the exibility to collect a larger set of metadata than
was collected in the earlier corpus. e following sections introduce the items of
metadata that are recorded for each speaker in the corpus.
.. Name
is was retained only for the purpose of communication between the team at
Cambridge and the contributors. All names were converted into unique speaker
ID codes to maintain de-identication (the removal or coding of identiable in-
formation for public use, while retaining such information privately, Ribaric etal.
2016) before the transcripts were sent to Lancaster for processing (see Section4).
e term ‘de-identication’ refers to the same process that has hereforeto oen been
labelled ‘anonymization’.
. All data is stored and analysed in compliance with the UK Data Protection Act 1998.
. Although there is an increasing desire for spoken corpora that “move beyond text and lan-
guage as conventionally conceptualised” (Adolphs etal. 2015: 61), our goal was to create a corpus
that is comparable to the Spoken BNC1994DS. erefore, we made no attempt to record conver-
sations as video rather than just audio, or to record any other live contextual data– for example
the GPS position of the smartphones (cf. Adolphs etal. 2015). Video recording equipment that
does not heavily compromise the unobtrusiveness of the recording event (and, therefore, the
likelihood of the data being as natural as possible) has yet to become available at low expense
(Adolphs & Carter 2013: 147), and we did not have the time to develop a bespoke smartphone
application for recording any other data.
Robbie Love et al.
.. Age
For most of the speakers in the Spoken BNC2014 the exact age is available as free-
form speaker metadata (e.g. “27”). In addition, we have categorised speakers accord-
ing to two age-based schemes. e rst is the Spoken BNC1994 age categorisation
scheme: 0–14, 15–24, 25–34, 35–44, 45–59, 60+, and Unknown. is facilitates com-
parison between the two corpora. e second scheme is more ne-grained: 0–10,
11–18, 19–29, 30–39, 40–49, 50–59, 60–69, 70–79, 80–89, 90–99, and Unknown. We
increased the number of categories to facilitate more sophisticated apparent-time
analysis of the new data; the revised scheme starts with a primary division at 18/19
(18 being the latest age of school-leaving in the UK) and then subdivides the result-
ing juvenile/adult sections into decades (as closely as possible).
.. Gender
On the consent forms, speakers specied their gender via a free-text box (i.e. they
could write whatever they liked in their own words). All speakers self-reported their
gender as either “female” or “male”, which we code as F or M respectively; a third
classication, “n/a (multiple)”, is used only for groups of multiple speakers (e.g. in at-
tributing vocalisations such as laughter when produced by several speakers at once).
.. Accent/dialect
Speakers used a free-text box to enter a description of their own dialect (e.g. “Geordie”,
“Northern”, etc.). e self-reported dialect of speakers has then been coded accord-
ing to a four-level classication scheme (Table2), the fourth level of which is drawn
from the UK government’s Nomenclature of Territorial Units for Statistics (NUTS,
http://webarchive.nationalarchives.gov.uk/20160105160709/ http://www.ons.gov.uk/
ons/guide-method/geography/beginner-s-guide/eurostat/index.html). e scheme
therefore does not arise from considerations of linguistic classications of the UK
(cf. Trudgill 1999) but rather geopolitical ones. is choice of scheme reects our
principle that the pre-selection of categories for sociolinguistic analysis should not
impose assumptions of linguistic patterns upon the corpus but ought rather to allow
the data to reveal such patterns. While it might have been preferable for us to de-
velop a categorization and then train the speakers to use it, this would clearly have
been infeasibly time-consuming. But self-reported dialect data is not without its
own virtues: it is, for instance, of great value to researchers interested in perceptual
dialectology (e.g. Montgomery 2012). Moreover, it will be possible (in principle at
least) to assign regional-dialect classications to the recorded speakers according
to objective, linguistic criteria at some later point. But it is generally not possible to
facilitate perceptual dialectology research other than by asking the speakers what
variety of English they believe they speak. So, while one driver for our decision to
gather self-report data on dialect was practical, another was principled– we wanted
to gather from the speakers that information that could not be easily inferred, or
e Spoken BNC2014
inferred at all, from their data at a later date: the variety of British English they be-
lieved themselves to speak.
Table2. Classication scheme for speaker dialect in the Spoken BNC2014
(1) Global (2) Country (3) Supra-region (4) Region
UK English North North East
Yorkshire & Humberside
North West (not Merseyside)
Merseyside
Midlands East Midlands
West Midlands
South Eastern
South West
South East (not London)
London
Scottish Scottish Scottish
Welsh Welsh Welsh
Northern Irish Northern Irish Northern Irish
Non-UK Irish Irish Irish
Non-UK Non-UK Non-UK
Unspecied Unspecied Unspecied Unspecied
Based on speakers’ free-text answers to the question of what variety of English they
speak, each speaker is assigned to a category in each of the four levels in Table2
(“global”, “country”, “supraregion” and “region”). e assignments depend upon how
much could be inferred from their self-reported response. Our aim was to maximise
specicity (in other words, to “get as much out of” the metadata as possible while
allowing speakers to describe themselves in their own words). For example, a speaker
who entered “Geordie” would be assigned to: (Level 1– UK; Level 2– English; Level
3– North; Level 4– North East). A speaker who entered “Northern” would be as-
signed to: (Level 1– UK; Level 2– English; Level 3– North; Level 4– Unspecied).
us, a level 4 analysis would exclude a self-reported “northern” speaker and place
them in the “unspecied” category because the specic region of the north to which
they refer (if any) is not known. It should also be noted that analysing the data at the
third level (“supra-region”) facilitates comparison with the regional classication in
the Spoken BNC1994– albeit the latter is itself not unproblematic.
.. Occupation
Speaker occupation is coded according to the National Statistics Socio-economic
Classication (NS-SEC), which has been the government standard for the UK census
since 2001, and is also used in the UK Labour Force Survey. To complement this
scheme, the UK government makes available a free online tool (https://onsdigital.
Robbie Love et al.
github.io/dp-classication-tools/standard-occupational-classication/ONS_SOC_
occupation_coding_tool.html) which converts the name of any given occupation
into an NS-SEC code. NS-SEC is of course a dierent scheme to that used in the
Spoken BNC1994. at scheme, Social Grade, was never used by the government
and instead was (and is) popular in the market research sector. No formal standard
has been established for translating either of these schemes to the other, but in the
interests of comparability we have proposed an automatic mapping from NS-SEC
to Social Grade so that both schemes can be analysed in the Spoken BNC2014
(Table3).
Table3. Mapping between the NS-SEC and Social Grade assumed for Spoken BNC2014
speaker metadata
8
NS-SEC Description Social
Grade
Description
1 Higher managerial,
administrative and
professional occupations:
8
MAPS ON TO…
A Higher managerial,
administrative and professional
1.1 Large employers and
higher managerial and
administrative occupations
1.2 Higher professional
occupations
2 Lower managerial,
administrative and
professional occupations
B Intermediate managerial,
administrative and professional
3 Intermediate occupations C1 Supervisory, clerical and junior
managerial, administrative and
professional
4 Small employers and own
account workers
5 Lower supervisory and
technical occupations
C2 Skilled manual workers
6 Semi-routine occupations D Semi-skilled and unskilled
manual workers
7 Routine occupations
8 Never worked and
long-term unemployed
E State pensioners, casual
and lowest grade workers,
unemployed with state
benets only
* Students/unclassiable
. is is not in and of itself an analytic category; rather it comprises analytic categories 1.1 and 1.2
e Spoken BNC2014
.. Other metadata categories
e other speaker metadata categories we collected were:
i. Nationality
ii. Birthplace
iii. Current location
iv. Duration of stay there
v. Mother tongue
vi. Most inuential country on language
vii. Additional languages
viii. Education level
e metadata categories pertaining to the recordings were:
i. Number of speakers
ii. Recording location
iii. Conversations topics (assigned by contributor)
iv. Main conversation topic (assigned by contributor)
ese items of speaker and recording metadata are entirely self-reported; the form
in which the speakers provided this information is reproduced verbatim in the cor-
pus metadata and documentation without attempts to schematize or standardize.
. Transcribing the Spoken BNC2014
Transcription of the Spoken BNC2014 recordings was undertaken by a team of
transcribers employed by Cambridge University Press. ey were trained according
to a bespoke transcription scheme developed for this project (see Section4.1). In
Sections4.1 to 4.3, we discuss the decisions we made about the transcription of
the Spoken BNC2014 recordings, as well as the steps taken to convert the result-
ing transcripts into a suitable XML-based canonical format for distribution and
archiving of the corpus.
. Developing the transcription scheme
e rst two questions that Crowdy (1994) poses in his account of the Spoken
BNC1994’s transcription system are “who is the transcription for?” and “how will
it be used?” (Crowdy 1994: 25), foregrounding the importance of purpose when
transcribing spoken data. Our starting point was to employ a level of “standard
orthographic” transcription (Leech etal. 2001: 12) that was simple and easy to
Robbie Love et al.
implement. Like the Spoken BNC1994, the main aim of the Spoken BNC2014 is
to facilitate the quantitative study of “morphology, lexis, syntax, pragmatics, etc.”
(Atkins etal. 1992: 10), allowing users to search for “particular features or patterns”
and view them “in concordanced form” (Crowdy 1995: 228). An orthographic tran-
scription serves the needs of research in these areas. We explicitly exclude the study
of phonetics (segmental or prosodic) from the list of areas that the corpus caters for.
While we are of course entirely open to phoneticians making use of the corpus if they
wish and so far as they can, most phonetic research typically requires both (i) access
to high-quality audio recordings and (ii) full phonetic transcription, neither of which
was a possibility within the constraints of this project. Phonetics aside, transcribing
recordings in the form of an “idealized ‘script’ (like a screenplay or drama script) is
sucient for a wide variety of linguistic studies” (Atkins etal. 1992: 10).
e next decision to be made related to the precise nature of the scheme for
orthographic transcription to be employed. is can be a highly consequential
matter, as it aects the time taken for transcription, and thus the cost and there-
fore the possible size of the corpus (cf. Schmidt 2016). e key requirement is for
a robust transcription scheme that, critically, minimizes the level of transcriber
inference that is needed– that is, the number of decisions that a transcriber must
make about potentially ambiguous speech phenomena. Speech phenomena which
require a higher level of transcriber inference to be included in linguistic detail,
such as “false starts, hesitation, non-verbal signals” (Atkins etal. 1992: 10), take
more time to transcribe, and even more time to achieve consistency within each
transcriber’s work and across transcribers. We aimed, therefore, to normalize or
disregard these phenomena at the transcription stage as far as we could, while still
serving most of the needs of most of our intended users.
Dening such a robust scheme meant that all of the issues likely to be encoun-
tered by transcribers had to be explored, and decisions made about how to deal
with them, before full scale transcription commenced. We found that Atkins etal.
(1992: 11–12) provide a good starting point in terms of general recommendations
for approaching spoken corpus transcription. ese recommendations include: be-
ginning each turn with an identifying encoding of the speaker; marking inaudible
segments; normalizing numbers and abbreviations; and producing a “closed set of
permissible forms” for the transcription of dialect and non-standard words. Atkins
etal. (1992) also advise careful thought about the extent to which punctuation should
represent written conventions, and suggest that faithful and precise transcription of
overlapping speech is costly; thus, an evaluation of the value and utility of including
both punctuation and overlaps should be made before transcription begins.
Similarly, with regard to functional and non-functional sounds (also known as
lled pauses, or more informally ums and ahs), Atkins etal. (1992) note that clas-
sifying these speech sounds according to discourse function requires a high level
e Spoken BNC2014
of inference on the part of the transcriber. erefore “a large set of orthographic
representations” (Atkins etal. 1992: 12) of speech sounds, rather than their possible
functional mappings, should be added to the transcription scheme. at is, tran-
scribers should be instructed to select a transcription for each um or ah based only on
its sound form, and should not attempt to imbue meaning into the transcription of
these non-lexical sounds (e.g. by providing pragmatic annotation). Rather, as Atkins
etal. (1992) suggest, the interpretation of such sounds should be le to researchers
with access to the recordings who choose to investigate these phenomena at a later
date. is recommendation can be seen as a specic case of Crowdy’s (1994: 25)
more general principle that researchers should use the transcript as a “baseline”
and that analysis beyond the scope of a simple orthographic transcription should
be undertaken by those researchers who wish to “analyse the text in more detail”.
Admittedly, such additional analysis will not immediately be possible on release
of the corpus, because we have not been able to de-identify the audio recordings
from the Spoken BNC2014 within the scope of the present project (de-identication
being necessary to preserve speakers’ privacy). However, we will in future pursue
further funding to de-identify and release the original recordings, thus enabling
functional analysis of speech features currently transcribed without any pragmatic/
discourse classication. e benet of an approach which omits any features re-
quiring inferential decisions by the transcribers is not merely theoretical; rather,
we have practical evidence of its usefulness. During the pilot phase of our work,
we undertook an experiment in which we asked the transcribers to annotate any
segment of an utterance containing reported direct speech (that is, material that
the speaker is quoting from elsewhere) during their transcription of the audio. e
transcribers reported that this task was not dicult. However, when their work
was compared to a standardised transcript analysed by a linguist, they were found
to have marked less than a third of qualifying clauses. We thus see that requiring
transcribers to include detailed analytic distinctions either leads to low quality re-
sults, or necessitates a high level of post hoc correction by a linguist. Neither of these
outcomes was desirable or aordable. We were, therefore, convinced of the need to
make the transcription scheme exclude not only the annotation of quoted speech
but also any other type of additional annotation that would require the input of a
linguist– though, as noted, such additions to the basic transcription will of course
be welcome aer the release of the audio data.
Given the above points, our rst decision was to avoid simply re-using the
Spoken BNC1994 transcription scheme, as documented by Crowdy (1994). e
reason for this is that Crowdy’s (1994) account of the Spoken BNC1994 transcrip-
tion conventions is by no means comprehensive; only sixteen features are identied
and the entire scheme is less than two thousand words in length. Furthermore, not
enough examples are provided to eliminate ambiguity, and some of the examples
Robbie Love et al.
which are provided are transcribed inconsistently. For example, full stops and com-
mas are to be used to mark “a syntactically appropriate termination or pause in an
utterance, approximating to use in written text”, and an ellipsis to mark “a longer
pause– up to 5 seconds” (Crowdy 1994: 27). But in practice, the examples include
uses of full stops and commas in positions that would not license a punctuation
mark in written English, as shown in Example(4), suggesting that the full stop/
comma versus ellipsis rule was not followed by transcribers in a consistent manner:
(4) <2> I think it’s always, deceptive on days like this because its, overcast and [er]
[…]
<2> But, but er, he’s… just broken away from his girlfriend and [<unclear>]
<1> [Oh has] he, oh. Well he seemed happy enough when he called.
(Crowdy 1994: 28)
Furthermore, Crowdy’s (1994) scheme states that question marks are to be used to
indicate “questioning” utterances, but this is not done consistently in the examples
provided, as in Example(5):
(5) <1> It’s a funny old day isn’t it.
<2> Mm it’s not cold is it? (Crowdy 1994: 28)
It thus seemed appropriate not to apply the 1994 scheme again without thorough
review. is is not to imply that the original scheme is wrong; many of the rec-
ommendations, we believe, are sensible. However, considering examples such as
those above, we were concerned that the transcription scheme as it was did not
give enough detail about enough features to maximally ensure inter-transcriber
consistency. So, instead:
i. we conducted a critical evaluation of Crowdy’s (1994) scheme, identifying which
features should be retained, abandoned or adapted;
ii. we reviewed evidence from other work on spoken corpus transcription pub-
lished since the Spoken BNC1994’s compilation, with a particular focus on
spoken components of the Cambridge English Corpus as well as recent work
at Lancaster University on spoken corpora;
iii. we conducted a small pilot study (as mentioned above) to test some of the
proposed features in practice.
e resulting transcription scheme can be found in the user documentation (Love
et al. 2017) for the Spoken BNC2014 (which will be available to read online at
http://corpora.lancs.ac.uk/bnc2014 as well as included within the corpus distri-
bution download). Aer each recording was transcribed, the transcript was put
though two stages of checking at Cambridge University Press– audio-checking and
proofreading– before being sent to Lancaster for processing. At the audio-checking
stage, a 5% sample of the recording was checked against the transcript for linguistic
e Spoken BNC2014
accuracy. If errors were found, the entire recording was checked. Aer this, the
transcript was proofread for errors with regard to the transcription conventions
(without reference to the audio). Despite this checking, complete accuracy of tran-
scription cannot, of course, be assumed– even though the scheme has been limited
to a basic, orthographic level of transcription. It is unavoidable that the involve-
ment of over a dozen human transcribers (as was the case in the production of the
Spoken BNC2014) will lead to certain inconsistencies of transcription decisions.
Our extended and elaborated transcription scheme enabled us to minimize– but
we would not claim to eradicate– such inconsistency. Indeed, it would be naïve to
assume the latter. For example, let us consider the variant pronunciations of the
tag question isn’t it , as represented orthographically by isn’t it, ain’t it, innit, etc. e
transcription scheme lists these as permissible non-standard forms, and ideally we
would therefore expect each instance of the tag question to have been faithfully
transcribed using the spelling variant that matches the actual pronunciation. But
in practice, it is very unlikely that a match between non-standard orthography and
precise phonetic quality was achieved consistently, both within the transcripts of
a given transcriber and indeed between transcribers. As such, we encourage users
to consider the data not as a denitive representation of the original speech event,
but rather to bear in mind that the transcriptions have been produced under the
constraints of what we now believe to be the natural, terminal limit of consistency
between human transcribers.
. Speaker identication
None of the previous work that we consulted when developing the transcription
scheme had recognised one important consideration: the degree of condence with
which transcribers were able to identify the speaker responsible for each turn (es-
pecially in recordings which contain more than two speakers). Evidence from a
pilot study on this issue suggests that transcribers tend to assign their “best guess”
speaker to a given turn– resulting in inaccurate speaker identication in cases
where they guess incorrectly. To account for this, we introduced a new speaker ID
convention which allowed transcribers to indicate when they were not fully certain
in their choice of ID code. e purpose of this is to caution users of the corpus
against blindly assuming that all of the speaker ID codes in the corpus texts have
been assigned accurately. In Example(6), for instance, the transcriber has indicated
that they were not fully certain of which speaker produced the second turn, but that
their best guess is speaker S0514 (the [??] indicator of low condence shown here
represents an underlying XML attribute-value pair; see below).
(6) S0511: well what happens in the sessions?
S0514[??]: there was some watching videos and stu (BNCBB001)
Robbie Love et al.
ough this measure does not actually improve the accuracy of speaker identica-
tion, it does promote user awareness of potential issues with it. Furthermore, this
utterance-level attribute data makes it possible to restrict corpus queries to exclude
those turns with low condence in speaker identication. In total, 29,369 utter-
ances (2.45% of utterances; 170,806 tokens) fall into the low condence category.
In addition, we observed that transcriber condence decreased as the number of
speakers in the corpus texts increased, and that texts containing four speakers or
more (294 texts; 23.5%) are liable to high degrees (average 40%) of inaccuracy in
speaker identication. We therefore recommend that users who require speakers
to be attributed accurately use the restricted query function in CQPweb (or, equiv-
alently, appropriate preprocessing of a downloaded copy of the XML-formatted
corpus) to exclude texts containing four or more speakers.
. Converting the transcripts
e two established standard formats for corpus data interchange and archiving
are (i) plain text and (ii) plain text enhanced with markup using XML. Transcripts
of spoken data almost always include features beyond the actual words of the text
(e.g. indicators of utterance boundaries) and thus XML is the appropriate choice
of format. However, XML is a somewhat cumbersome format for direct data entry,
and is also rather dicult to teach to non-specialist audio transcribers. It is also
challenging to check for accuracy by eye. From the outset, then, we knew that while
our goal was to release a canonical version of the corpus in XML, this would not be
the system used for transcription. Instead, we designed the transcription scheme
to be human-friendly, while making sure that all of its elements could be unam-
biguously mapped to XML at a later stage. For that reason, the original transcripts
used short, easy-to-type codes for features such as utterance boundaries, speaker
labels, vocalisations, and de-identied elements (names of people, places, and other
potentially identifying information). As the recordings were transcribed by the
Cambridge team, the transcripts were sent in batches to the Lancaster team, who
used a set of automated conversion scripts to translate the transcripts into XML– at
the same time applying a series of further automatic checks on the correct use of
the transcription conventions that were not possible prior to conversion to a struc-
tured document. is approach was by no means an innovation– the transcription
scheme presented by Crowdy (1994) for the Spoken BNC1994 was likewise con-
verted to SGML (and, later, XML) in the released BNC1994.
In consequence, while (as previously noted) we do include a complete descrip-
tion of the human-friendly conventions used in transcription within the corpus
documentation (Love et al. 2017), these conventions are not used in the text of
the corpus itself; instead, the actual corpus text contains the canonical XML. e
e Spoken BNC2014
transcription scheme is, then, part of our record of how the corpus was created. It
is not exclusively a guide for users. We make it available to users of the corpus in
order to make the decisions discussed above absolutely transparent, but also in the
hope that it may prove useful as a point of departure for other researchers working
on the creation of spoken corpora of this kind.
A number of systems for the use of XML in corpus encoding have been pro-
posed as standards. ese include the Text Encoding Initiative (TEI; see Burnard &
Bauman 2013) and the Corpus Encoding Standard (CES; see Ide 1996). e former
of these was used for (and, in fact, originally developed alongside) the BNC1994.
However, as argued by Hardie (2014), these standards are fairly top-heavy and
require much more extensive and detailed XML markup than is either necessary
or useful for the vast majority of corpus linguistic research. For that reason, rather
than use TEI, we opted to follow the recommendations of Hardie (2014) for a “mod-
est” level of XML. We made use of the XML tags and attributes noted by Hardie
(2014: 94–101) as having become more-or-less established as de facto standard–
most of which are in fact also part of TEI and CES; we made additions to this set
of codes only where our transcription system required it. For instance, utterances
are marked up with <u> tags, and each utterance has a who attribute, containing
the unique ID code of the speaker. ese are exactly as described by Hardie (2014)
and originate in TEI. However, we also added a whoCondence attribute, which
records the transcriber’s level of condence in the speaker attribution (as per the
discussion above). e text headers in the corpus use a drastically simpler (and
more atly organised) set of metadata tags than TEI/XML, each element being
generated automatically, on a mostly one-to-one basis, from some column of the
metadata tables originally collected alongside the recordings. Both the header and
body tags are listed in full in the corpus documentation, which also includes a full
Document Type Denition (DTD) covering all elements and attributes.
e virtues of making standard types of analytic annotation available to all us-
ers of a corpus, by distributing a tagged version alongside the untagged text, are well
understood. In line with this principle, we have tagged the whole corpus for part-of-
speech (POS) and lemma using the same systems as the original BNC1994– most
notably the CLAWS tagger (see Garside & Smith 1997). However, in a departure
from the practice of the BNC1994, we use the C6 tagset instead of the simpler C5
tagset.
9 C5 tags were used in order to achieve a simpler (and thus more reliable)
system of POS tagging in the rst release of the BNC. However, later BNC1994
releases use a parallel system of simple tags, or major wordclasses, alongside the C5
tags. is system uses one single tag for all nouns, another single tag for all verbs,
and so on, and, in our view, addresses the need for a lower-complexity grammatical
. Both tagsets are available on the CLAWS website: http://ucrel.lancs.ac.uk/claws/
Robbie Love et al.
classication highly eectively. us, the combination of full-complexity C6 anno-
tation and low-complexity simple tags is the best way to address all the purposes
covered by the mid-complexity C5 tags. In the canonical form of the data, all three
annotations (C6 POS tags, simple POS tags, and lemmas) are coded as XML at-
tributes on the <w> element (which encloses each word: used only in the tagged
version of the corpus).
However, as noted above, while the XML release will represent the canonical
form of the corpus, the initial release was via Lancaster University’s CQPweb server.
CQPweb is the online interface component of the Corpus Workbench soware (see
Hardie 2012). CQPweb provides full support for a number of features which use of
the Spoken BNC2014 requires, namely (i) access to all layers of corpus annotation;
(ii) restricting analyses to utterances whose speakers full certain demographic
criteria of dialect, age, or gender; and (iii) limiting access only to users who have
signed the corpus licence. XML elements encoded within a CQPweb corpus can be
used to control the appearance of the text in concordance lines and other aspects of
the interface; on the Lancaster server, we congure the system to display utterance
boundaries and speaker ID codes in an easily readable format. So, for instance,
the underlying XML attribute-value pair whoCondence=“low”– which appears
on the <u> element– is rendered in the interface as [??], the notation shown in
Example(6) above. e display format that we use for such features in CQPweb
does not replicate the original codes as typed by the transcribers; the display codes
were instead devised afresh for maximal visual distinctiveness.
. Conclusion
In this paper, we have presented a general overview of the design and compilation
process of the Spoken BNC2014. Our aim has been to make clear the most im-
portant decisions we have made as we have collected, transcribed and processed
the data. e resulting corpus (and, eventually, its written counterpart) should be
of use to many researchers, educators and students in the corpus linguistics and
English language community and beyond. In the short term, we are pleased to note
the research presented in this special issue of IJCL, all of which uses the Spoken
BNC2014 Sample (see this issue’s Editorial), and we anticipate the publication of
ground breaking sociolinguistic research based upon this data in the forthcoming
year (Brezina etal. forthcoming). Furthermore, CASS has started a new project
addressing the creation of a balanced sociolinguistic core form both the Spoken
BNC2014 and the BNC1994DS (Brezina etal. 2016). e project combines the
expertise from the elds of corpus linguistics and variationist sociolinguistics to
develop subsamples of the two larger corpora that will allow sophisticated socio-
linguistic searches and analyses.
e Spoken BNC2014
References
Adolphs, S., & Carter, R. (2013). Spoken Corpus Linguistics: From Monomodal to Multimodal.
Abingdon: Routledge.
Adolphs, S., Knight, D., & Carter, R. (2015). Beyond modal spoken corpora: A dynamic approach
to tracking language in context. In P. Baker & T. McEnery (Eds.), Corpora and Discourse
Studies: Integrating Discourse and Corpora (pp. 41–62). Houndsmill: Palgrave Macmillan.
doi: 10.1057/9781137431738_3
Alderson, C. J. (2007). Judging the frequency of English words. Applied Linguistics, 28(3),
383–409. doi: 10.1093/applin/amm024
Aston, G., & Burnard, L. (1998). e BNC Handbook: Exploring the British National Corpus with
SARA. Edinburgh: Edinburgh University Press.
Atkins, A., Clear, J., & Ostler, N. (1992). Corpus design criteria. Literary and Linguistic Computing,
7(1), 1–16. doi: 10.1093/llc/7.1.1
Biber, D. (1993). Representativeness in Corpus Design. Literary and Linguistic Computing, 8(4),
243–257. doi: 10.1093/llc/8.4.243
Brezina, V., & Meyerho, M. (2014). Signicant or random? A critical review of sociolinguistic
generalisations based on large corpora. International Journal of Corpus Linguistics, 19(1),
1–28. doi: 10.1075/ijcl.19.1.01bre
Brezina, V., Gablasova, D., McEnery, T., & Meyerho, M. (2016). British National Corpus (BNC)
as a sociolinguistic dataset: Exploring individual and social variation. Retrieved from http://
gtr.rcuk.ac.uk/projects?ref=ES%2FP001599%2F1 (last accessed November 2016).
Brezina, V., Love, R., & Aijmer, K. (Eds.) (forthcoming). Corpus Approaches to Sociolinguistic
Variation in Contemporary British English: An Exploration of the Spoken BNC2014. New
York: Routledge.
Burnard, L. (2000). Reference Guide for the British National Corpus (World Edition). Oxford
University. Retrieved from http://www.natcorp.ox.ac.uk/archive/worldURG/urg.pdf (last
accessed December 2013).
Burnard, L. (2002). Where did we go wrong? A retrospective look at the British National Corpus.
In B. Kettemann & G. Markus (Eds.), Teaching and Learning by Doing Corpus Analysis (pp.
51–71). Amsterdam: Rodopi. doi: 10.1163/9789004334236_007
Burnard, L. (2007). Reference guide for the British National Corpus (XML Edition). Oxford
University. Retrieved from http://www.natcorp.ox.ac.uk/docs/URG/ (last accessed December
2013).
Burnard, L., & Bauman, S. (Eds.) (2013). TEI: P5 Guidelines. TEI Consortium. Retrieved from
http://www.tei-c.org/Guidelines/P5/ (last accessed June 2017).
Carter, R. (1998). Orders of reality: CANCODE, communication, and culture. ELT Journal, 52(1),
43–56. doi: 10.1093/elt/52.1.43
Cappelle, B., Dugas, E., & Tobin, V. (2015). An aerthought on let alone. Journal of Pragmatics,
80, 70–85. doi: 10.1016/j.pragma.2015.02.005
Čermák, F. (2009). Spoken corpora design: eir constitutive parameters. International Journal
of Corpus Linguistics, 14(1), 113–123. doi: 10.1075/ijcl.14.1.07cer
Crowdy, S. (1993). Spoken corpus design. Literary and Linguistic Computing, 8(4), 259–265.
doi: 10.1093/llc/8.4.259
Crowdy, S. (1994). Spoken corpus transcription. Literary and Linguistic Computing, 9(1), 25–28.
doi: 10.1093/llc/9.1.25
Robbie Love et al.
Crowdy, S. (1995). e BNC spoken corpus. In G. Leech, G. Myers & J. omas (Eds.), Spoken
English on Computer: Transcription, Mark-Up and Annotation (pp. 224–234). Harlow:
Longman.
Davies, M. (2004). BYU-BNC (Based on the British National Corpus from Oxford University
Press). Brigham Young University. Retrieved from http://corpus.byu.edu/bnc/ (last accessed
June 2017).
Deuchar, M., Davies P., Herring J., Parata Couto, M., & Carter D. (2014). Building bilingual
corpora. In E. M. omas & I. Mennen (Eds.), Advances in the Study of Bilingualism (pp.
93–111). Bristol: Multilingual Matters.
Douglas, F. (2003). e Scottish Corpus of Texts and Speech: Problems of corpus design. Literary
and Linguistic Computing, 18(1), 23–37. doi: 10.1093/llc/18.1.23
Flowerdew, J. (2009). Corpora in language teaching. In M. H. Long & C. J. Doughty (Eds.), e
Handbook of Language Teaching (pp. 327–350). Oxford: Wiley-Blackwell.
doi: 10.1002/9781444315783.ch19
Gabrielatos, C. (2013). If-conditionals in ICLE and the BNC: A success story for teaching or learn-
ing? In S. Granger, G. Gilquin & F. Meunier (Eds.), Twenty Years of Learner Corpus Research:
Looking back, Moving ahead (pp. 155–156). Louvain-la-Neuve: Presses Universitaires de
Louvain.
Garside, R., & Smith, N. (1997). A hybrid grammatical tagger: CLAWS4. In R. Garside, G. Leech &
A. McEnery (Eds.), Corpus Annotation: Linguistic Information from Computer Text Corpora
(pp. 102–121). London: Longman.
Hadikin, G. (2014). A, an and the environments in Spoken Korean English. Corpora, 9(1), 1–28.
doi: 10.3366/cor.2014.0049
Hardie, A. (2012). CQPweb– Combining power, exibility and usability in a corpus analysis
tool. International Journal of Corpus Linguistics, 17(3), 380–409. doi: 10.1075/ijcl.17.3.04har
Hardie, A. (2014). Modest XML for Corpora: Not a standard, but a suggestion. ICAME Journal
38, 73–103. doi: 10.2478/icame-2014-0004
Hatice, C. (2015). Impoliteness in Corpora: A Comparative Analysis of British English and Spoken
Turkish. Sheeld: Equinox.
Homann, S., Evert, S., Lee, D., & Ylva, B. (2008). Corpus Linguistics with BNCweb: A Practical
Guide. Frankfurt am Main: Peter Lang.
Ide, N. (1996). Corpus Encoding Standard. Expert Advisory Group on Language Engineering
Standards (EAGLES). Retrieved from http://www.cs.vassar.edu/CES/ (last accessed June
2017).
Kallen, J. L., & Kirk, J. (2008). ICE-Ireland: A User’s Guide Documentation to accompany the
Ireland Component of the International Corpus of English (ICE-Ireland). Belfast: Cló Ollscoil
na Banríona. Retrieved from http://www.johnmkirk.co.uk/johnmkirk/documents/003647.
pdf (last accessed June 2017).
Lam, P. (2009). e making of a BNC customised spoken corpus for comparative purposes.
Corpora, 4(1), 167–188. doi: 10.3366/E174950320900029X
Leech, G., Rayson, P., & Wilson, A. (2001). Word Frequencies in Written and Spoken English: Based
on the British National Corpus. Harlow: Pearson Education Limited.
Lüdeling, A., & Kytö, M. (2008). Introduction. In A. Lüdeling, & M. Kytö (Eds.), Corpus Linguistics:
An International Handbook (pp. i–xii). Berlin: Walter de Gruyter. doi: 10.1515/9783110211429
Love, R., Hawtin, A., & Hardie, A. (2017). e British National Corpus 2014: User Manual and
Reference Guide (version 1.0). Lancaster: ESRC Centre for Corpus Approaches to Social
Science.
e Spoken BNC2014
McEnery, T. (2005). Swearing in English: Bad Language, Purity and Power from 1586 to the Present.
New York, NY: Routledge.
Montgomery, C. (2012). e eect of proximity in perceptual dialectology. Journal of Soc iolinguistics,
16(5), 638–668. doi: 10.1111/josl.12003
Nelson, G., Wallis, S., & Aarts, B. (2002). Exploring Natural Language: Working with the
British Component of the International Corpus of English. Amsterdam/Philadelphia: John
Benjamins. doi: 10.1075/veaw.g29
Nesselhauf, N., & Römer, U. (2007). Lexical-grammatical patterns in spoken English: e case of
the progressive with future time reference. International Journal of Corpus Linguistics, 12(3),
297–333. doi: 10.1075/ijcl.12.3.02nes
Ribaric, S., Ariyaeeinia, A., & Pavesic, N. (2016). De-identication for privacy protection in
multimedia content: A survey. Signal Processing: Image Communication, 47, 131–151.
Rühlemann, C. (2006). Coming to terms with conversational grammar: ‘Dislocation’ and ‘dysu-
ency’. International Journal of Corpus Linguistics, 11(4), 385–409. doi: 10.1075/ijcl.11.4.03ruh
Rühlemann, C., & Gries, S. (2015). Turn order and turn distribution in multi-party storytelling.
Journal of Pragmatics, 87, 171–191. doi: 10.1016/j.pragma.2015.08.003
Säily, T. (2011). Variation in morphological productivity in the BNC: Sociolinguistic and meth-
odological considerations. Corpus Linguistics and Linguistic eory, 7(1), 119–141.
doi: 10.1515/cllt.2011.006
Schmidt, T. (2016). Good practices in the compilation of FOLK, the Research and Teaching
Corpus of Spoken German. International Journal of Corpus Linguistics, 21(3), 396–418.
doi: 10.1075/ijcl.21.3.05sch
Shirk, J. L., Ballard, H. L., Wilderman, C. C., Phillips, T., Wiggins, A., Jordan, R., McCallie, E.,
Minarchek, M., Lewenstein, B. V., Krasny, M. E., & Bonney, R. (2012). Public participation
in scientic research: A framework for deliberate design. Ecology and Society, 17(2), 29.
doi: 10.5751/ES-04705-170229
Smith, A. (2014). Newly emerging subordinators in spoken/written English. Australian Journal
of Linguistics, 34(1), 118–138. doi: 10.1080/07268602.2014.875458
Stenström, A. -B., Andersen, G., & Hasund, I. K. (2002). Trends in Teenage Talk: Corpus Compilation,
Analysis and Findings. Amsterdam/Philadelphia: John Benjamins. doi: 10.1075/scl.8
ompson, P., & Nesi, H. (2001). e British Academic Spoken English (BASE) Corpus Project.
Language Teaching Research, 5(3), 263–264.
Trudgill, P. (1999). e Dialects of England (2nd ed.). Oxford: Blackwell Publishing Ltd.
Wang, S. (2005). Corpus-based approaches and discourse analysis in relation to reduplication and
repetition. Journal of Pragmatics, 34(4), 505–540. doi: 10.1016/j.pragma.2004.08.002
Wichmann, A. (2008). Speech corpora and spoken corpora. In A. Lüdeling & M. Kytö (Eds.),
Corpus Linguistics: An International Handbook (pp. 187–206). Berlin: Walter de Gruyter.
Xiao, R., & Tao, H. (2007). A corpus-based sociolinguistic study of ampliers in British English.
Sociolinguistic Studies, 1(2), 231–273. doi: 10.1558/sols.v1i2.241
Robbie Love et al.
Authors’ addresses
Robbie Love
ESRC Centre for Corpus Approaches to Social Science (CASS)
Faculty of Arts and Social Sciences
Lancaster University
Lancaster, LA1 4YD
United Kingdom
r.m.love@lancaster.ac.uk
Claire Dembry
Cambridge University Press, University Printing House
Shaesbury Road
Cambridge, CB2 8BS
United Kingdom
cdembry@cambridge.org
Andrew Hardie
ESRC Centre for Corpus Approaches to Social Science (CASS)
Faculty of Arts and Social Sciences
Lancaster University
Lancaster, LA1 4YD
UK
a.hardie@lancaster.ac.uk
Vaclav Brezina
ESRC Centre for Corpus Approaches to Social Science (CASS)
Faculty of Arts and Social Sciences
Lancaster University
Lancaster, LA1 4YD
UK
v.brezina@lancaster.ac.uk
Tony McEnery
ESRC Centre for Corpus Approaches to Social Science (CASS)
Faculty of Arts and Social Sciences
Lancaster University
Lancaster, LA1 4YD
UK
a.mcenery@lancaster.ac.uk
