Corpus-based methods in linguistic research
VẤN ĐỀ HÔM NAY
CORPUS-BASED METHODS
IN LINGUISTIC RESEARCH
Nguyen Thi Hong Ha MA *
Abstract: Corpus-based methods in linguistic research was initiated by Paul
Baker in 2006 through the publication ‘Using Corpora in Discourse Analysis’
(London: Continuum). Since then, the method has been applied by many linguists
in great variety of research, especially in making dictionaries, teaching languages,
and comparing translations. With a number of advantages, the method enables
researchers to quantify linguistic patterns, and to come to solid conclusions. Thanks
to the technology development, user-friendly software has motivated the method to
move rapidly, promising to bring many useful linguistic applications to our life.
Key words: corpus, method, research, linguistics, application.
1. Introduction
Corpus linguistics is a popular field of
linguistics involving ‘the analysis of very
large collections of electronically stored
texts aided by computer software’ (Paul
Baker in Litosseliti, 2010: 93). The word
‘corpus’inLatinmeansbody,soacorpuscan
be understood a ‘body’ of texts. According
to Mc Enery and Wilson (1996:1), corpus
linguistics features a ‘methodology’ rather
than a traditional branch of linguistics like
semantics, grmammar, or phonetics.
Corpuslinguisticswithcharacteristics
of empirical, inductive forms of analysis
relies on instances of language in real-life
use. The aims of corpus linguistics are to
find out patterns, rules or explore trends
about the ways people actually use their
language in everyday life.
Asfarasweknow,quantitativeresearch
concentrate on how much or how many
there is/are of a particular characteristic or
item of a linguistic aspect. The strong point
of this method is that it enables researchers
to compare ralatively large numbers of
things, people by applying a comparatively
easy index. Quantative data use statistical
methods for analysis, that is, particular
methametics tools allow researchers to
conduct on numeric data. Corpus-based
methods are regarded as effective research
methods in linguistics, and they ‘should
not be considered as only quantitative, but
ratherthananapproachwhichcancombine
both qualitative and quantitative processes’
(Paul Baker in Litosseliti, 2010: 93).
This article mentions corpus-based
methods in details, from basic concepts,
their applications, to building and
annotating corpora in linguistic research.
2. Theoretical Concepts of Corpus
Linguistics
2.2. Advantages of Corpus Linguistics
There are several great advantages
of corpus-based methods in linguistics
as follow:
(i) Corpus-based methods enable
researchers to restate or refect hypotheses
about language use;
1.1. Definition
Tạp chí
* Dean of English Faculty B,
Hanoi University of Business and Technology
23
Kinh doanh và Công nghệ
No 05E/2019
VẤN ĐỀ HÔM NAY
3. Types of corpora
(ii) Corpus-based methods allow
researchers to bring out new questions
and theories about language;
(iii) Corpus-based methods help
researchers to quantify linguistic patterns,
reaching more solid conclusions on
language;
Corpora in existence are divided
into three main pairs (Baker, P. 2006
in Litosseliti, 2010), based on their
characteristics and aims:
3.1. General corpora and specialized
corpora
A general corpus aims to represent
a particular language like the British
National Corpus BNC or The Bank
of English BoE., which is extremely
large and takes a long time to build and
annotate. More importantly, they are
very useful resources for a wide range
of research purposes. They play the role
of ‘benchmark’ about a typical language
in comparison with a specilized corpus.
Meanwile, a specialized corpus is much
smaller and has more limited sets of
texts in restrictions on time, genre or
place/language variety. For instance, a
specialized corpus of just newspapers
published in October, 2019 in Vietnam.
Speacilized corpora are generally easier
to collect and for specific research
questions.
3.2. Written corpora and spoken corpora
Written corpora contain computer-
mediated texts such as e-mails, text
messages or websites or mixture of all
three while spoken corpora are usually
smaller due to complexities surrounding,
gathering and transcribing data. Written
corpora with the access to the Internet are
easier to build because most of the texts
are already electronically coded. This type
of corpora is expected to be increasingly
popular when societies more frequently
use electronic forms of communication.
3.3. Multilingual corpora and
Parallel Corpora
(iv) Corpus-based methods with
large corpora lead researchers to obvious
evidence of rare or unusual instances
of language, and confirmation on very
common phenomena of language.
2.3. Research questions of corpus-
based methods
The overarching questions of corpus-
based methods is ‘how do people really use
language’, around which many research
questions related to different fields in
linguistics are raised. For example, in
the field of language teaching it can be
seen: ‘Is the language used in textbooks
actually reflected the language that people
encounter in everyday life?’ (Mindt, 1996);
or in study on language genres: ‘Has
writeen language become more informal
over recent years?’(Kennedy, 1998).
Moreover, the comparative trend also
appears in research questions within studies
on corpus-based linguistics, such as ‘How
does the use of linguistic feature X differ in
usage between language varieties A and B
intermsoffrequencyand/ortypicalusage?’
or ‘What associations are triggered by the
use of linguistic item X, based on its typical
uses?’ Not only to discover similarities
between the features of languages, but also
do corpus-based methods help find out
differences, in spite of small difference,
or even no difference which is worth
researching. In addition, corpus linguistic
research approaches language patterns that
people are unaware of, but they may still
strongly influence users.
Thecomparisonofdifferentlanguages
has emerged as a growing area of corpus
Tạp chí
Kinh doanh và Công nghệ
24
No 05E/2019
VẤN ĐỀ HÔM NAY
linguistics. This area is useful in fields of use a corpus and the others used a corpus.
language teaching, language testing, and These are the findings of the study
translation. A multilingual corpus usually (Hunston, 2002: 97):
involves equal amounts of texts from
- The 1987 Longman Dictionary of
a number of differents languages in the ContemporaryEnglish(withoutacorpus):
same genre. There is no need to translate 20 senses;
directlyfromonelanguageintoanotherfor
- The 1995 Longman Dictionary
such texts. Meanwhile, a parallel corpus (with a corpus): over 40 senses;
is a special type of miltilingual corpus, in
- The 1995 COBUILD Dictionary
which texts are exact equivalents of each (with a corpus): over 30 senses.
other. Parallel corpus are often carefully
Obviously, corpus-base methods
designed, sentence-aligned, that means showtheiradvantagesofenhancingsenses
tags are added to the corpus to identify of meaning of words in dictionaries,
which sentences are translations of each reflecting the reality of diversely using
other. With the aid of the right software, language in everyday life.
researchers view translations of sentecnes
side by side, and recognise the differences
between translations and the original.
3.4. Learner Corpora
4.2. Aplication in translation studies
Corpus-base methods, especially
with the types of multilingual and
parallel corpora show their strengths in
A learner corpus is produced by comparative transaltion and interpreting
learners of that language, that is very studies. When conducting research on
useful for teachers to realize common punctuation in Hans Christian Andersen’s
errors at different stages of development, stories and in their translations into
and indicate over- and underuses of English, Malmkjer (1997) concludes that
vocabulary or grammar of learners in in translations, punctuation tends to be
comparison with an equivalent corpus of strenghtened, with commas aften being
native speakers. The Longman Learner replaced with semicolons or full stops,
Corpus and the International Corpus of and semicolons being transferred to full
LearnerEnglishbothreceivecontributions stops, too. This leads to long, complex
from a great variety of learners all over sentences being divided into shorter and
the world, helping researchers find out the simpler clauses in translations to reduce
extent to which a student’s first language the complexity of sentence structures in
is likely to influence the way they acquire the original.
English.
4. Applications of corpus-based reveals that translators usually make
methods in linguistics optional cohesive markers explicit in the
In another study, Mauranen (2000)
4.1. Application in linguistic description translations, even though they are not
Corpus-based methods can aid available in the original text. This results
researchers in making dictionary with in a tendency to spell things out rather
real-life examples of words in use. than make them implicit.
Hunston (2002) researches the senses of
the verb ‘KNOW’ shown in examples in
4.3. Application in forensic linguistics
Coulthard (1993) carries out his study
three dictionaries, one of which did not on witness statements used as evidence in
Tạp chí
Kinh doanh và Công nghệ
25
No 05E/2019
VẤN ĐỀ HÔM NAY
the trial of Derk Bentley (who was excuted verbs, which were less frequent in use in
in Britian in 1953 for his involvement the corpus. Such studies are very useful for
in a policeman’s death). He compares writing textbooks and designing teaching
the frequencies of words in Bentley’s syllabus in language teaching.
statements with that in general written and
spokenEnglish,andthatofotherpolicemen
4.6. Application in stylistics
Corpus methods of analysis have
and witnesses. He notes that Bentley had been used in stylistics in order to enhance
a higher frequency of using word ‘then’ systematicity and decrease subjectivity.
than others. However, this word is a very For example, Malhberg (2009) examines
typical feature of the police. This, together stylistics in writing literary works by
with other corpus-based eveidence, the Charles Dickens and states that the writer
researcher argues that Bentley aged 11 often mentions the ways characters use
had not made his own statement, but it had household objects as a way of drawing
been written for him.
readers’ attention to the characters’
4.4.ApplicationinCriticalDiscourse emotions. Indicating a number of
Analysis CDA
examples in the works related to objects
In the area of CDA, Baker (2006) like a watering-pot or a knife and fork,
demonstrates how corpus-based methods the researcher through her corpus-based
can be used to express the ‘incremental analysis concludes that in Dickens’novels
effect of discourse’. He states that an these objects are consistently used to
association between two words, occuring emphasize characters’ emotional states.
many times in naturally occurring
language is much better evidence for an building corpora
underlying hegemonic discourse made Difference
5. Corpus-based research tools and
across
space/genres
explicit through the word combination (variation) and over time (change) is the
than a single case. Furthermore, Mautner most commonly applied in corpus- based
(2007) examines a corpus from a wide research. According to Baker, among them
range of language sources to see how frequencydataare‘indicatorofmarkedness’
the elderly construct their discourse. The (Baker 2010: p.125); wordlists are the most
researcher recognizes that their discourse basic ‘points of entry’(Baker 2010: p.133)
is expressed as ill-health victims who to analyse; keywords are ‘somewhat
are in need of care more often than as more sophisticated’ (Baker 2010: p.134)
empowered or independent citizens.
means of research and concordances with
4.5. Application in language teaching associated information are involved in
Corpora can also help language collocates and clusters.
teaching be more effective. Mindt (1996)
studies a corpus of spoken English and
The most frequently used data include:
(i) Corpus, text and sentence (average
realized that native speakers use the model word length);
verb ‘will’most frequently for referring to
(ii) Standardized type/ token ration
future time. Yet, in German textbooks used (STTR) and standard deviation (SD);
for teaching English, ‘will’was introduced
to students in the middle of the second year
(iii) Significance (p-value).
Theoretically, any text and collection
when they had already learnt other modal oftextsareconsideredtobeacorpusandthe
Tạp chí
Kinh doanh và Công nghệ
26
No 05E/2019
VẤN ĐỀ HÔM NAY
analysis can be conducted on the corpus of to make sure that the corpus must be as a
veryshorttexts.AccordingtoMcEneryand whole. Perhaps, it is not necessary to take
Wilson (1996), a corpus usually contains the whole text, but parts of it if they are
a sample ‘maximally representative of too long novels. Equal-sized samples from
the variety under examination’ is ‘of a texts also need to be considered carefully to
finte size’, exists in ‘machine readable’ assure the balance. For instance, different
and ‘constitutes a standard reference for parts (beginnings, middles, ends) of texts
the language variety which it represents’ are equally sampled for the corpus. If texts
(McEnery and Wilson, 1996: 22-23). It is are quite short and from one writer, the
clear that a corpus must be large enough whole texts might be included rather than
to show some feature about frequencies of different sections.
linguistic phenomena, helping researchers
to discover the common as well as unusual the assistance of analysis software doing
things in language. counting, sorting and presenting language
Corpora are usually applied with
Baker in Litosseliti (2010: p.95) characteristics. The current software such
suggests three criteria for identifying the as WordSmith Tools, Xaira, Wmatrix, and
size of a corpus: aspects of language, AntConc can be employed in conjuction
type of language, and pratical reasons. with a range of corpora. First, each text
Kennedy (1998: 68) states that ‘a corpus within a corpus is usually saved on a
of 100,000 words will usually big enough separate file containing a ‘header’ which
for the sudy of prosody’, and an analysis has information about its author, date of
of verb-form morphology will need half publication,genre,etc.Itallowsresearchers
a million words. Meanwhile, according to concentrate on specific types of texts
to Biber (1993) confirms that a million or to compare different types. Second, a
words will be enough for a study on corpus is annotated or tagged with further
grammar. Also, the more various the information for more complex calculations
language is, the larger the corpus requires. carried out on them. Sometimes, standard
The British National Corpus involving a generalized mark-up language (SGML)
very wide range of written and spoken where tags are presented in codes
language genres and as a standard (elements) inside matching angle brackets
reference for British English, has the size (< >) is applied. Finally, certain features
of 100 million words while a corpus of of the language variety are represented
weather forecast with restricted language with other codes (entities) starting with an
just needs a much more smaller size. The ampersand character (&) and ending in a
practical conditions affecting the building semi-colon.
a corpus could be the avalability of texts,
the amount of money and time spent on although tagging can be automatically
the study and permission from authors. done by computer as tagging software
Hand-checking is often required
The key theoretical concepts in corpus is not always 100% accurate. Computer
linguistic include sampling, balcance and programs usually work best on texts having
representativeness. A corpus must be grammatically predictable sentences and
representative of a particular language relatively familiar words. Apart from
variety, sothetextsneedchoosingcarefully checking, only human beings can interpret
Tạp chí
Kinh doanh và Công nghệ
27
No 05E/2019
VẤN ĐỀ HÔM NAY
the results of calculations from computer, to build a corpus, a continuing need
no computer software can cover this work. to update balanced reference corpora,
6. Conclusion
researcher’s ability to use computer
In summary, corpus-based methods fluently, identifying certain type of
are regared as potential to produce language patterns, it is worth applying
interesting findings about language, to linguistic studies where comparisons
but as with many other methods, it is must be done. The advantage of the
researchers’ task to provide explantion corpus-based methods lies in employing
for the findings. Corpus-based methods fast and accurate techniques to discover
cannot cover all fields in linguistics, so it patterns that human reserarchers would
could be combined with other methods to not recognize. Corpus analysis uses large
maximize strengths and avoid limitations. amount of natural data, so a high degree
Although it contains its weak points such of reliability and validity of linguistic
as being time and money-consuming research is usually achieved.
References
1. Baker, P. (2006). Using Corpora in Discourse Analysis. London: Continuum.
2. Baker, P. (2010). Sociolinguistics and Corpus Linguistics. Edinburgh
Sociolinguistics. Edinburgh: Edinburgh University Press.
3. Biber, D. (1993). ‘Representativess in corpus design’ Literary and linguistics
Computing 8,4: 243-57.
4.Coulthard,M.(1993).‘Onbeginningthestudyofforensictexts:corpusconcordance
collocation’. In M.Hoey (ed.). Data, Description, Discourse. London: Harper Collins.
5. Hunston, S. (2002). Corpora in Applied Linguistics. Cambridge: Cambridge
University Press.
6. Kennedy, G. (1998), An Introduction to Corpus Linguistics. London: Longman.
7. Litosseliti, L. (2010). Research Methods in Linguistics. Continuum.
8. Malhberg, M. (2009). ‘Corpus stylistics and the Pickwickian watering-pot’ in
Baker (ed.). Contemporary Approaches to Corpus Linguistics. London: Continuum.
9. Malmkjer, K. (1997). ‘Punctuation in Hans Christian Andersen’s stories and
in their translations into English’, in F.Poyatos (ed.). Nonverbal Communication and
Translation. New Perspectives and Challenges in Literature, Interpretation and the
Media. Amsterdam and Philadelphia: Benjamins.
10. Mauranen, A. (2000). ‘Strange strings in translated language: a study on
corpora’, in M. Olohan (ed.). Intercultural Faultlines. Research Models in Translation
Studies 1: Textual and Cognitive Aspects. Manchester: St. Jerome Publishing.
11. Mautner, G. (2007). ‘Mining large corpora for social information: the case of
elderly’, Language in Society, 36 (1), 51-72.
12. Mindt, D. (1996). ‘English corpus linguistics and the foreign language teaching
syllabus’, in J. Thomas and M. Shorty (eds), Using Corpora for Language Research.
London: Longman, 232-247.
13. Mc Enery and Wilson (1996). Corpus Linguistics. Edinburgh: Edinburgh
University Press.
Tạp chí
Kinh doanh và Công nghệ
28
No 05E/2019
VẤN ĐỀ HÔM NAY
PHƯƠNG PHÁP DỰA VÀO KHỐI LIỆU TRONG NGHIÊN CỨU NGÔN NGỮ
ThS. Nguyễn Thị Hồng Hà *
Phương pháp dựa vào khối liệu trong phong phú: từ việc mô tả ngôn ngữ trong
nghiên cứu ngôn ngữ được học giả Paul làm từ điển, đến hỗ trợ dạy tiếng, ngôn
Baker khởi xướng vào năm 2006 qua ấn ngữ pháp lý, phong cách học, đến nghiên
bản ‘Sử dụng Khối liệu trong Phân tích cứu so sánh trong dịch thuật,…
Diễn ngôn’ (Nxb London: Comtinuum).
Đây là một phương pháp mới trong
Kể từ đó, phương pháp này được nhiều nghiên cứu ngôn ngữ ở Việt Nam, nên tác
nhà ngôn ngữ học áp dụng trong các giả muốn giới thiệu nó đến bạn đọc. Cùng
nghiên cứu của mình, đặc biệt trong việc với sự phát triển của công nghệ, các phần
biên soạn từ điển, dạy ngoại ngữ, và so mềm hỗ trợ tính toán hiện đại đã làm cho
sánh dịch thuật.
phương pháp này ngày càng phát triển
Phương pháp này có nhiều lợi thế, mạnh, mang lại nhiều ứng dụng thiết thực
giúp nhà nghiên cứu định lượng hóa các của ngôn ngữ cho đời sống đương đại.
mẫu ngôn ngữ, từ đó đi đến những kết luận
Từ khóa: Phương pháp, nghiên cứu,
đầy thuyết phục. Ứng dụng của phương khối liệu, ngôn ngữ học, ứng dụng.
pháp này trong nghiên cứu ngôn ngữ rất
Ngày nhận bài: 15/10/2019
Tạp chí
* Chủ nhiệm Khoa Tiếng Anh B,
Trường ĐH KD&CN Hà Nội
29
Kinh doanh và Công nghệ
No 05E/2019
7 trang
19720
Free
Bạn đang xem tài liệu "Corpus-based methods in linguistic research", để tải tài liệu gốc về máy hãy click vào nút Download ở trên
File đính kèm:
- corpus_based_methods_in_linguistic_research.pdf
