Diana Oţăt

4 Corpus Linguistics Outcomes and Applications

in the Digital Era

4.1 Introduction

Prominent scholars highlight the key role of word use in the development of corpus
linguistics as the study of linguistic phenomena by means of extensive collections
of machine-readable texts, i.e., by means of corpora. The steady evolution of corpus
linguistics has been primarily motivated by the linguists’ need to understand how
words are actually used in natural languages, which most common words tend to be
used in certain contexts, what is common and what is uncommon for certain language
variations (including specialisms), thus leading to the first outcomes provided by
corpus-based approaches, i.e., words lists and synonymous terms. Hence, mainstream
literature pinpoints the emergence of corpus-based investigations as early as 1755,
when, as endorsed by Biber et al. (2006: 22), Johnson used a corpus of texts to gather
authentic uses of words that he then included as examples in his dictionary – a first
step made towards the understanding of the patterns of use associated with a word.
A less popular branch of linguistics, though widely explored in the 1940s
and 1950s, statistical linguistics, has also contributed to the development of what
nowadays corpus linguistics has become through innovative mathematical theories
of information. Yet, due to the lack of computer-assisted processing tools back at that
time, it proved to lack productivity and effectiveness.
Another branch of linguistics closely related to the current state-of the-art of
corpus linguistics was what ‘older linguists, of the heyday in the 1950s’, such as Harris,
Fries or Hill and other American structuralists regarded as ‘descriptive linguistics’
(Leech, 1992: 106), i.e., the scholars’ aim at describing the corpus under investigation.
Accordingly, endorsing the flexible typology of descriptive linguistics towards theory
construction, Leech (ibidem) highlights the less abstract nature of its outcomes,
particular to one language, where linguistic phenomena are more easily to localise,
observe and analyse. In this context, we grow aware of one main characteristic of
corpus linguistics, namely that most corpus-based analyses are applied to data
inquiries specific to individual languages.
In the 1980s, linguists registered a rebirth of corpus linguistics, which by then
had already indicated a close connection with quantitative linguistics, a specialised
research branch that promotes the need for quantitative methods in language
study, which, according to various linguists, are also frequently used in most other
disciplines as they can provide reliable outcomes when it comes to the description of
language in terms of frequency and infrequency rates.
38   Corpus Linguistics Outcomes and Applications in the Digital Era

However, this diachronic examination of the development of corpus linguistics

differs considerably from the status it has acquired in the digital era. For contemporary
researchers, corpus-based analysis does not only serve the mere purpose of dictionary
making. Theorists and practitioners alike go far beyond the normative function of corpus-
based analyses, concerned with simple inventories of linguistic structures, charting
new territories in the development of corpus linguistics oriented towards qualitative
and functional interpretations of quantitative research. It represents, as postulated by
Svartvik (1992: 8), a way ‘to take a look at real manifestation of language when discussing
linguistic problems’, for as ‘corpus linguistics is not the heaping of data for its own sake,
but rather the investigation of data for scientific purposes’ (ibidem).

4.2 Corpus Linguistics in the Digitalised Era

It is common knowledge that the dominant role in corpus linguistics is played by

modern technology. As previously mentioned, corpus linguistics has been regarded as
an operational framework in language study, rather than an isolated domain of study. It
is not a monolithic system providing fixed sets of homogenous methods and procedures
applied to language investigation, but, as defined by McEnery and Hardie (2012: 5),
‘a heterogeneous field’. Applied to the study of the language, corpus-based analysis
aims at investigating particular linguistic structures, the way they occur in different
contexts and the functions they acquire. The current perspectives in corpus linguistics
indicate the extensive use of machine-readable texts as the appropriate resources, the
raw material on which to study specific linguistic issues and phenomena.
As put forward by scholars, such as Kennedy (1998) or Biber et al. (2006), computer
technology advances not only have multiplied the investigation perspectives where to
apply corpus-based analysis, but have also provided novel benefits and backups in
comparison to, previous research studies. Undoubtedly, the first significant advance
that upgraded this hybridised field is the storage capacity of very large databases
of natural language that can be compiled from a wide range of sources. Today, we
can save, store and organise on our own computers ample writings or large chunks
of texts, being, thus, able to carry out thorough linguistic analyses that are no more
limited to sentence-length excerpts.
Modern software and computer-assisted tools secure comprehensive linguistic
analyses that are more accurate and reliable. In this climate of opinion, Biber et al.
(2006) also state that ‘unlike human readers that are likely to miss certain occurrences
of a word, computers can find all the instances of a word in a corpus and generate an
exhaustive list of them’. However, it is worth mentioning that corpus linguistics is
not only concerned with words frequency, for as other branches of linguistics, such
as phonetics, syntax and any another aspect of linguistics may be investigated via
corpora analysis while applying and combining a series of corpus-based investigation
methods. In line with the prominent figures in the field of linguistics, we can endorse
 Corpus Linguistics in the Digitalised Era   39

that computer-assisted analysis methods can be applied to various types of corpora,

aiming at investigating particular patterns of word associations on a far more complex
model than it is possible manually.
Among the core features of computer-assisted corpora analysis applied to
language study, we can mention the empirical character of corpus-based analysis, i.e.,
the investigation of actual language structures in authentic texts, namely in corpora,
which represent the basis for the analysis designed and implemented. As previously
indicated, dedicated software and computer-based analysis tools have been extensively
applied within this field of research, which led to the implementation and operation
of hybridised techniques, which rely heavily on quantitative and qualitative methods
alike. As endorsed by Biber et al., computer-assisted investigation has provided not
only reliable analysis methods, but also consistent outcomes, enabling individuals
actively interact in validating complex linguistic findings, while ‘the computer takes
care of record-keeping’ (ibid: 6).
Two main corpus-based research directions have been indicated by the
authors mentioned above, where linguistic features are not investigated as isolated
occurrences, but as systematic associations with other features. Accordingly, we
distinguish between ‘linguistic associations of the features’ that encompass lexical
and grammatical associations and ‘non-linguistic associations of the features’, i.e.,
distribution across registers, dialects and time periods (ibidem).
Concerned with the study of linguistic features by means of corpus-based analyses,
it is worth mentioning that corpus-based research can be applied to grammar on the
word level as well as on the sentence and discourse levels. Also, register and text
typology can be investigated by means of corpora study.
As far as grammar study is concerned, corpus-based analysis has recently
registered fruitful advances towards novel interdisciplinary areas of investigation.
Thus, even though former linguists highlighted the descriptive approach applied to the
grammar study, prescriptive methods tend to be used more frequently, for example,
in order to establish the variables framing the syntax of language. By applying
corpus-based analysis, new research perspectives envisaged the patterned use of
grammatical features in texts or the variation of language in use. The investigations
of natural language texts have enabled linguists to register and further investigate the
routinised ways that individuals tend to use the grammatical resources of a language.
Linguists postulated that, by investigating the distribution of various structures,
the association patterns between grammatical structures and other linguistic and
non-linguistic factors that influence the individuals’ linguistic choices, all of them
recorded in authentic texts, considerable outcomes and research advances have been
achieved. Such fruitful headways were not only reached at the grammatical level,
but also in terms of language evolution. Its dynamics and the implementation of the
most appropriate strategies to bridge socio-cultural and linguistic differences is an
enduring benefit for closely related fields of research, such as translation studies,
sociolinguistics or cultural studies.
40   Corpus Linguistics Outcomes and Applications in the Digital Era

4.3 Corpora Design and Compilation in the Digital Era

The central aim of the present paper is not only to highlight the considerably value-
added and reliable outcomes provided by computer-assisted corpus-based analysis,
but also to pinpoint the highly effective dedicated software and computer-assisted
tools applied in corpora design and compilation.
We should first mention that, paradoxically, the concept of corpus was initially
used to designate various non-linguistic collections, such as the Corpus Juris Civils
introduced by the emperor Justinian in the sixth century and regarded by numerous
scholars as a compilation of early Roman law and legal principles, which also
illustrated particular cases and provided clarifications of new laws and future
legislation to be put in effect (see Jan Svartvik, 1992). However, linguistics corpora
are regarded as collections of texts ‘assumed to be representative of a given language,
dialect, or other subset of a language, to be used for linguistic analysis’, as defined
by Francis (1992: 17). Thus, in language study, corpora have been primarily used for
linguistic analysis, a feature that according to Francis (ibid:19), differentiates them
from other types of corpora or large text collections, such as anthologies, for example,
the Oxford Book of English Verse, whose purpose is literary.
Other examples of corpora types are lexicographical corpora, compiled in the
process of making dictionaries, for example, the Oxford English Dictionary, edited
in the 19th century by Murray or the Merriam-Webster Dictionary edited in the 20th;
dialectological corpora compiled for the purpose of designing dialect atlases, for
example, in the Middle English Dialect Atlas, issued in 1981 by Benskin, or grammatical
corpora, among which Quirk’s Survey of English Usage published in the 20th century
is the most well known.
Starting with the early 1960s, modern concepts of corpora generally indicate the
use of large collections of texts available in machine-readable forms, of which the most
representative one is the Brown University Standard Corpus of Present-Day American
English compiled by Henry Kucera and W. Nelson Francis and which remains a sample
of present-day English for use with digital computers. In the next two decades, it was
followed by the creation of the British National Corpus (BNC), started in 1991.
Given the broad and multi-faceted research directions that emerged in the field of
linguistics, as well as in previously mentioned close related domains, where corpora
investigations are applied in order to achieve novel quantitative and qualitative
outcomes, McEnery and Hardie (2012: 12) put forward a series of criteria applied to
distinguish between the types of corpus-based investigations:
–– mode of communication
–– corpus-based versus corpus-driven communication
–– data collection regime
–– the use of annotated versus unannotated corpora
–– total accountability versus data selection
–– multilingual versus monolingual corpora
 Corpora Design and Compilation in the Digital Era   41

By establishing these criteria, the authors aimed at featuring a typology of corpus

linguistic research framed by the principles of corpora use.
Albeit corpus-based analyses have by no means been restricted to the English
language, it is common knowledge that corpora investigations applied to the study
of the English language have provided the most relevant and significant language
study perspectives in corpus linguistics during the past period, leading to a boost and
proliferation in English studies generally. Among the most frequently approached
research perspectives carried out by means of corpus analysis, we could mention
the study of language variation, dialect, register and style, where authentic samples
of different areas of language use have been compiled and investigated in order to
validate the diverse range of language users or the close analysis of frequency rates of
particular linguistic structures in different language varieties or in certain specialised
languages.
Thus, corpora are far from being considered mere language samples aimed to
provide useful illustrative examples, but genuine theoretical resources, which,
implemented in a series of applied fields of research, such as language teaching,
translation studies or even machine translation and dedicated processing software
(spelling, grammar and style), have provided essential sources of information leading
towards generalisations about the language and language use. Furthermore, online
freely available corpora are meant to ensure linguists from all over the world a user-
friendly access to language materials that would otherwise be difficult or impossible
to obtain. Conversely, linguists who are non-native speakers can use such corpora for
further research and practice.
Having established that corpora design and compilation in the context of new
cutting-edge technologies is a key concern of the present paper; it is also noteworthy
mentioning that corpora fall into two major categories, i.e., general corpora – designed
to investigate a given language as a whole – and specialised corpora – designed to
answer more specific research questions, mainly used in the study of special languages
or closely connected applied fields of research. Accordingly, while general corpora are
thoroughly organised in order to have a long ‘shelf-life’, as claimed by Lüdeling and
Kytö (2008: 154), specialised corpora aimed for the study of certain linguistic items
in certain contexts or circumstances are much more rapidly constructed. The authors
consider that most of the corpora can be understood as a collection of sub-corpora
which display relative homogenous features. Inevitably the current corpora design
is based on ‘a template of variables that creates a number of cells, each of which
constitutes a sub-corpus’ (ibidem). If a linguist aims at investigating the specialised
domain of the written legal language, s/he may start designing her/his corpus by
collecting, storing and organising various types of legal documents, such as contracts,
laws, treaties, etc. that may constitute the cells or the sub-corpora.
Aston and Burnard (1998: 29) put forward a series of references applied in
corpora design, i.e., field, tenor and mode variables, which, for example, can be easily
recognisable amongst the most representative criteria of the British National Corpus
42   Corpus Linguistics Outcomes and Applications in the Digital Era

(BNC) that includes field (the subject-matter of written texts), tenor (spoken texts),
and mode (books, periodicals).
A central issue in corpora design and particularly in the design of specialised
corpora is how to establish the most appropriate variables and how the dynamics
of the sub-corpora should function, considering the fact that most of the times such
specialised corpora reveal a multi-dimensional character. Nevertheless, as argued by
Lüdeling and Kytö (2008), in most situations it is obviously not possible to collect
all the texts that constitute the research object into a corpus. In this respect, the
authors advocate that the selection of a subset of the texts aimed for analysis stands
as a reliable method. Moreover, the scholars recommend that establishing of strict
sampling techniques may lead to reliable and operative corpus design. We would tend
to emphasise that the contents of a corpus designed for research purposes need to
be carefully considered, though some will argue that the design of the corpus will
depend more on what is freely available in an easily converted format than on other
criteria, thus pinpointing towards the benefits provided by the digital era.
Within this context, a series of other very important aspects that need to be
considered in the corpora design can be mentioned, i.e., software limitations,
copyright, and text availability.
It is a truism that dedicated software provides an overwhelming capacity to
store databases under different shapes (written, spoken, images, graphics, etc.);
various linguists have pointed out that readily available software packages, such as
WordSmith Tools (Scott, 2004), can deal with a corpus of tens of millions of tokens
in size. Lüdeling and Kytö (2008: 157) also mention the existence of larger corpora
that often work with software, which demands that the tokens are converted into
digits, which would facilitate the software to process more quickly. Be that as it may,
specialists from linguistics and IT alike advocate that complex operations on corpora
that may involve hundreds of millions of words can take some time to complete.
Furthermore, another aspect that needs to be taken into consideration is that the
storage and further use of the electronic version of published texts remains a crime in
many of the EU Member States and worldwide as well, if no copyright permissions are
legalised. Thus, due to the fact that such permissions are often difficult to gain, the
availability of some corpora may be restricted.
Another aspect highlighted also by Lüdeling and Kytö (ibid: 158) refers to the
availability of texts. In this respect, the authors mention the design limitations
imposed by the corpora designers who wish to include in their corpora ‘spoken texts
from an era before the invention of tape recorders’. The accuracy of the data as a
representation of the actual speech is always questionable. Even though written texts
are easier to obtain, there are some situations when extensive written texts are difficult
to be incorporated into a corpus, unless very large storage facilities are available for
scanning and keying.
 Application: A Model for Computer-Assisted Corpus Design and Analysis   43

4.4 Application: A Model for Computer-Assisted Corpus Design

and Analysis

In what follows, we aim at exemplifying a computer-assisted model for corpus

design and corpus analysis that can be applied to specialised corpora used for the
investigation of specialised languages as well as to other fields of research, such as
language teaching or translation training programmes.
The model proposed is based on the MAXQDA 11 for Windows professional
software dedicated to qualitative and mixed methods data analysis. As highlighted
by its promotors, MAXQDA provides a variety of research methods and approaches.
By means of this computer-assisted tool, we can organise, encode, annotate, and
interpret an array of data. Moreover, the analysis outcomes can be generated in easy-
to-read reports, visualisations, Excel sheets, while enabling the researchers to work
interactively and share the results among each other.
Applying MAXQDA to the study of language, we simulated first the possibilities
of the corpus design. Providing the fact that MAXQDA allows the import of various
format types, a possible corpus for language analysis may encompass a wide range
of written and spoken texts, as well as images, graphics or tables in TXT, RTF, DOC/X,
PDF, JPG, GIF, TIF, and PNG format. Media files can be also selected here, enabling the
designer(s) to set up even a multimodal corpus.
The software allows the users to design their corpus in a project-based format,
where each group member can actively participate, save memos and observations,
and apply them both to the design and the analysis of the corpus.
Moreover, the software allows the users design various corpora simultaneously, which
can be further structured in a sub-corpora, if organised in separate document groups.

Figure 4.1: Print Screen: Corpus organisation in MAXQDA 11.

44   Corpus Linguistics Outcomes and Applications in the Digital Era

For example, we could simultaneously design two specialised corpora used

for the study of different linguistic issues typical for legal language, i.e., a corpus
encompassing legal documents, which can be further organised in sub-corpora, such
as contracts, laws, regulations, etc., and a corpus encompassing specialised texts from
the automotive industry, i.e., user manuals, technical specifications, manufacturing
provisions, etc. Such corpora may be then individually investigated by each member
of the project group in accordance with the specific research objectives, or they can
even be contrastively analysed, if, for example, certain linguistic features are to be
characterised contrastively in particular registers or even along certain time periods,
i.e., time series.
By providing a user-friendly interface, MAXQDA reveals similar features to other
Windows programs, facilitating quick and effective processing on behalf of the users.
After completing the design stage of a corpus, which, as previously indicated, offers
innumerable compiling possibilities, we can embark on the linguistic analysis of the
corpus by means of the drop-down menus and the various toolbars with buttons that
offer quick access to the functions. It is worth mentioning that even at these stages, the
investigation possibilities provided by the software are numerous. A general overview
of a linguistic analysis model would imply the use of the following menus:
–– the Analysis Menu – provides a series of analysis options applied especially to
the lexical search and retrieval functions. Thus, the lexical search option enables
the user to search within the document, or just in the activated document sets,
memos, and retrieved segments. This function facilitates the search for certain
words, phrases, or combinations. Also, the keywords in a context can be searched
and automatically encoded. As indicated in the user manual, most of this menu
functions relate to retrieval. We can chose various criteria for the segments to be
found (e.g., OR, AND, logical combinations, or NEAR). Moreover, the retrieved
segments can also be filtered based on certain criteria in the ‘Retrieved Segments’
window.
–– the Codes Menu – enables the user to create and apply new codes on all the
documents or only on the activated ones, or even to create a complete index of all
codes assigned to all the document segments.
–– the Mixed Methods Menu – is used to process and combine qualitative and
quantitative data using documents and variables. Documents or document
groups can be investigated based on the assigned variables, limiting the retrievals
to certain document segments. The Quote Matrix and Crosstabs functions can be
applied in order to indicate connections between the encoded segments and the
selected variables.
–– the Visual Tools Menu – enables the users to visualise the outcomes by means of
seven different visualisation function options. MAXMaps, the tool for qualitative
modelling; the Code Matrix Browser; the Code Relations Browser; and the
Document Comparison Chart. The Document Portrait and the Codeline functions
 Application: A Model for Computer-Assisted Corpus Design and Analysis   45

also provide further visualisations which can be exported to Excel sheets as well
under the shape of tables or graphics.

Figure 4.2: Print Screen: document portrait generation according to the assigned codes.

–– The MAXDictio Menu – is an optional menu which offers a number of functions

for quantitative content analysis, e.g., coding according to created dictionaries
and viewing word frequencies.

Figure 4.3: Print Screen: Word frequency list in MAXQDA 11.

46   Corpus Linguistics Outcomes and Applications in the Digital Era

Figure 4.4: Print Screen: Tag cloud in MAXQDA 11.

As we can see, there are countless possibilities of analysis models, depending on

the investigated linguistic issues and the specificity of the research fields applied.
Of course, the design and implementation of simultaneous and multi-dimensional
language analyses is possible, providing reliable and consistent outcomes in just a
few minutes.

4.5 Conclusions

We can conclude that the computer-assisted approaches in corpus linguistics may

lead, nowadays, to the refining and redefining of a wide range of theories of language.
By means of dedicated software and computer-assisted tools, corpus linguistics has
broadened its research directions considerably, smoothing the path towards new
language explorations and theories.
As previously highlighted, within the context of advanced technologies, corpora
are steadily exploited by tools that enable users to search through them rapidly
and reliably. Most of such tools allow the production of frequency data, e.g., word
frequency lists, document portraits, or comparison charts.
Unquestionably, there is a close link between the current status of corpus
linguistics and modern technologies that brought to this field incredible speed, total
liability, statistical reliability, sustainable results, and the opportunity to manipulate
over considerably substantial and varied databases.

References
Aston. G. & Burnard, L. (1998). The BNC Handbook: Exploring the British National Corpus with Sara.
Edinburgh: Edinburgh University Press.
References   47

Benskin, M. (1981). The Middle English Dialect Atlas. In Benskin, M. & Samuels L. (Eds.). So Meny
People Longages and Tonges: philological essays in Scots and mediaeval English presented to
Angus McIntosh. Edinburgh. xxvi-xli.
Biber, D., Conrad S., & Reppen, R.. (2006). Corpus Linguistics: investigating language structure and
use. Cambridge: Cambridge University Press.
Francis. W. N. (1992). Language Corpora B.C. In Svartvik J. (Ed.). Directions in Corpus Linguistics:
Proceedings of Nobel Symposium 82 Stockholm, 4-8 August 1991. Berlin: Walter de Gruyter,
17-35.
Kennedy, G. (1998). An Introduction to Corpus Linguistics. London: Longman.
Leech, G. (1992). Corpora and theories of linguistic performance. In Svartvik J. (Ed.). Directions in
Corpus Linguistics: Proceedings of Nobel Symposium 82 Stockholm, 4-8 August 1991. Berlin:
Walter de Gruyter, 105-123.
Lüdeling A. & Kytö, M. (2008). Corpus Linguistics an International Handbook. Vol.1. Berlin: Walter de
Gruyter.
McEnery, T. & Hardie, A. (2012). Corpus Linguistics: Method, Theory and Practice. Cambridge:
Cambridge University Press.
Scott, M. (2007). Oxford WordSmith Tools Version 4.0. Oxford: Oxford University Press.
Svartvik, J. (1992). Corpus linguistics comes of age. In Svartvik J. (Ed.). Directions in Corpus
Linguistics: Proceedings of Nobel Symposium 82 Stockholm, 4-8 August 1991. Berlin: Walter de
Gruyter, 7-17.

Online resources
Maxqda for Windows. Reference Manual. Berlin: VERBI Software. Consult. Sozialforschung.
Available: http://www.maxqda.com [Accessed 2015, July, 30].