=Paper=
{{Paper
|id=Vol-547/paper-75
|storemode=property
|title=ELECTRONIC CORPORA: AS POWERFUL TOOLS IN COMPUTATIONAL LINGUISTIC ANALYSES
|pdfUrl=https://ceur-ws.org/Vol-547/93.pdf
|volume=Vol-547
|dblpUrl=https://dblp.org/rec/conf/ciia/Grazib09
}}
==ELECTRONIC CORPORA: AS POWERFUL TOOLS IN COMPUTATIONAL LINGUISTIC ANALYSES==
https://ceur-ws.org/Vol-547/93.pdf
ELECTRONIC CORPORA: AS POWERFUL TOOLS
IN COMPUTATIONAL LINGUISTIC ANALYSES.
Mohamed Grazib
Djillali Liabes University: Computer science department Sidi Bel Abbes
E-mail: mfgrazib@hotmail.com
Abstract: Technology has emerged almost all the domains in our daily life. In
computational linguistics, the uses of electronic corpora are very important.
Nowadays it is possible to study linguistic phenomena by using statistical
analyses: Concordances, collocations and frequencies have great influence in
making linguistic researches more available, more adequate and more accurate.
Keywords: corpus, computational linguistics, corpus linguistics, Concordances,
collocations and frequencies
1 Introduction:
Electronic Corpora are indispensable for computational linguistics; in addition to the
availability and the accuracy the tasks can be done in few minutes. Nowadays both
the qualitative and quantitative analyses of language are possible by the uses of
Electronic corpora and computers. This article is an attempt to show the benefits of
corpora in the English applied linguistic studies.
2. What is an electronic corpus?
The word Corpus plural (corpora) or (corpuses) is derived from the Latin word
“corpus” which means:” Body” in French “corps”; a corpus is a large set of texts
(electronically stored and processed) , it may be used to refer to any text in written or
spoken form that can be available on computers as software or via internet. G. Cook
(2003:73) suggests that the word corpus refers to a databank of language which has
actually occurred-whether written, spoken or a mixture of the two. The written texts
are originally from magazines, books, diaries, newspapers, letters, popular
fictions……; however the spoken texts can be any recorded formal or informal
conversations: Telephone conversations, dialogues, radio-shows, political
meetings…….
3. What is computational linguistics?
The Association for Computational Linguistics defines computational linguistics as
the scientific study of language from a computational perspective. Computational
linguistics is a discipline between linguistics and computer science .It is a part of the
cognitive sciences and it has a strong relation with artificial intelligence.
Computational linguistics originated from the 1950s, where the United States used
computers to translate automatically texts from foreign languages into English,
particularly Russian scientific journals. Traditionally, computational linguistics was
usually performed by computer scientists who had specialized in the application of
computers to the processing of a natural language.
4. What is Corpus linguistics?
Corpus linguistics is the study and analysis of data obtained from a corpus. The main
task of the corpus linguist is not only to find the data but to analyse it. Computers are
useful, and sometimes indispensable, tools used in this process. Corpus linguistics is
based on two main software objects: a corpus, which is the body of data to be
investigated, and a concordancer, a tool for searching that corpus. Corpus Linguistics
is now seen as the study of linguistic phenomena through large collections of
machine-readable texts: corpora. Biber et al (1998:23) said that: “Corpus linguistics
makes it possible to identify the meanings of words by looking at their occurrences in
natural contexts, rather than relying on intuitions about how a word is used or on
incomplete citation collections”.
5. Size of corpora:
Corpora come in many shapes and sizes, because they are built to serve different
purposes. Nowadays 1 million words is fairly small in terms of corpora. We can
make a distinction between reference1 and monitor corpora 2: The following list
shows a very limited sample of corpora’s sizes.
• Bank of English: about 400 million words.
• COBUILD/Birmingham Corpus: More than 200 million words.
• Longman Lancaster corpus: 30 million words.
____________________________
1
Reference corpora have a fixed size (e.g., the British National Corpus).
2
Monitor corpora are expandable (e.g., the Bank of English).
• British National Corpus (BNC):100 million words.
• American National Corpus (ANC): 11.5 million words.
• Brown corpus: 1million words.
• Lancaster-Oslo/Bergen (LOB) corpus: 1 million words.
• Northern Ireland Transcribed Corpus: 400,000 words.
• Corpus of Spoken American English (CSAE):200,000 words.
What is evident is that the size of any corpus depends mainly on the purposes it was
Created for, and that this size can vary from some hundred words to some million
words.
6. Concordance, Collocation and Frequency.
In reality, a corpus by itself can do nothing at all; it is nothing other than a store of
used language. A corpus does not contain new information about language but the
software offers us new perspectives. Most readily available software packages process
data from a corpus in three ways: showing, frequency, phraseology and collocations.
G. Cook (2003:111).
6.1. Concordance:
A concordance is a screen display or printout of a chosen word or phrase in its
different contexts, with that word or phrase arranged down the centre of the display
along with the text that comes before and after it.
John Sinclair (1991:32) defines a concordance as a collection of the occurrences of a
word-form each in its own textual environment. In the same context S. Hunston
(2002:39) says that it is a programme that searches a corpus for a selected word or
phrase and presents every instance of that word or phrase in the centre of the
computer screen with the words that come before and after it to the left and right. The
selected word appearing in the centre of the screen is known as the “node word”.
The following example illustrates the 10 concordances of the word computer from
Web Concordancer LOB.txt.
1 etition between the analogue computer and the digital computer. To a
2 g made on a Ferranti Mercury Computer at Meteorological Office, Duns
3 unnecessary devices that the computer can be made an economic propos
4 ouch with manufacturers about computer developments of special signi
5 he {0PIW} are compiled by the computer from data sheets (dictionary
6 seen that the problem of the computer is in no way related to the p
7 racy is required the digital computer is the only one to use and ele
8 he {0ACE} digital electronic computer of our laboratory and, further
9 with the help of the digital computer of the University of Toronto.
10 al purpose electronic digital computer to do the job. It is therefore
6.2. Collocation:
Firth (1957) stated that “you shall know the word by the company it keeps”. The
meaning of Firth’s citation here is to classify words not only on the basis of their
meanings, but also on the basis of their co- occurrence with other words.
S. Hunston (2002:12) defines collocation as the statistical tendency of words to co-
occur.
Collocation investigations can be a preliminary step for other research questions:
investigating the distribution of word senses and uses, and comparing the use of
seemingly synonymous words, because languages have many words that are similar,
and dictionary definitions often characterise such words as identical or synonymous in
meaning, however investigating the use and distribution of synonyms in a corpus
allows us to determine their contextual preferences associated with other collocates or
associated with register differences. Biber et al (1998:24).
The following table shows the five most time’s collocations with nouns, verbs, and
adjectives.
Nouns Verbs Adjectives
# TIMES # TIMES # TIMES
WORD WORD WORD
NEARBY NEARBY NEARBY
1 YEARS 1933 WAS 17846 LONG 4850
2 YEAR 1703 IS 12614 GOOD 1587
3 PERIOD 1360 HAD 8128 SHORT 1522
4 PEOPLE 1334 BE 8023 OTHER 1202
5 DAY 1139 WERE 4298 RIGHT 1111
Table 1: collocations of the word time with (nouns, verbs, and adjectives) from
view.byu.edu :
6.3. Frequency:
Frequency list tells us what words and phrases are used most often. Biber et al
(1998:23) argue that frequency investigations tell us how often different words are
used; allowing us to identify particularly common and uncommon words. Based on
the evidence of the billion-word Oxford English Corpus, the 100 commonest English
words found in writing around the world are as follows:
1 the 26 they 51 when 76 come
2 be 27 we 52 make 77 its
3 to 28 say 53 can 78 over
4 of 29 her 54 like 79 think
5 and 30 she 55 time 80 also
6 a 31 or 56 no 81 back
7 in 32 an 57 just 82 after
8 that 33 will 58 him 83 use
9 have 34 my 59 know 84 two
10 I 35 one 60 take 85 how
11 it 36 all 61 people 86 our
12 for 37 would 62 into 87 work
13 not 38 there 63 year 88 first
14 on 39 their 64 your 89 well
15 with 40 what 65 good 90 way
16 he 41 so 66 some 91 even
17 as 42 up 67 could 92 new
18 you 43 out 68 them 93 want
19 do 44 if 69 see 94 because
20 at 45 about 70 other 95 any
21 this 46 who 71 than 96 these
22 but 47 get 72 then 97 give
23 his 48 which 73 now 98 day
24 by 49 go 74 look 99 most
25 from 50 me 75 only 100 us
Table2: The first 100 frequent English words.
As seen in the table above many of the most frequently used words are grammatical
words (articles, auxiliaries, prepositions….); however the first noun position (time) is
the 55th. We can also explore frequencies according to the main word classes: The
frequencies of the main word classes in 1 million-word computer corpora of written
English are given in the table bellow:
% % %
Total 1 million Informative prose Imaginative prose
words in each sections (fiction) sections
corpus
US UK US UK US UK
NOUNS 26.80 25.2 28.50 26.9 21.77 20.0
VERBS 18.20 17.8 17.02 16.4 21.69 21.9
DETERMINERS 14.16 14.2 14.84 15.2 12.11 11.4
PREPOSITIONS 12.04 12.2 12.77 13.1 9.87 9.6
ADJECTIVES 7.07 7.3 7.65 7.8 5.35 5.7
PRONOUNS 6.56 7.1 4.75 5.0 11.94 13.1
CONJUCTIONS 5.92 5.5 5.94 5.5 5.86 5.4
ADVERBS 5.23 5.5 4.73 5.0 6.72 7.2
1
OTHERS 4.02 5.2 3.80 5.2 8.49 5.8
TOTAL 100 100 100 100 100 100
Table 3: The frequencies of the main word classes
By making a brief analysis2, based on the information taken from the table above, we
can notice that:
- The nouns are the most frequent used words
- Verbs are more frequent in conversation and in fiction.
- Pronouns are more frequent in spoken English and in fiction that in
informative writing.
- Conjunctions have almost the same frequency in both corpora
7. Using electronic corpora in computational linguistics:
Many words have meanings that are similar, and yet the words are not able to be
substituted one for the other. Dictionaries, which deal with words separately rather
than comparatively, can be of little help, but observing typical usages of near
synonyms can clarify differences in meaning. S. Hunston (2002:45).
____________________________
1. Includes wh- words, foreign words, numerals………
2. The analyses are from the Brown Corpus of American English
(Francis&Kucera.1982:547) and the Lancaster-Oslo-Bergen (LOB) Corpus of
written British English (Johansson&Hofland.1989:15).
The study is about the following synonyms (Sheer, pure, complete, utter and
absolute). The first analysis is upon a traditional research (using dictionaries)
In this context Partington (1998:33-46) gives examples of intensifying adjectives:
“sheer, pure, complete, utter, and absolute”. He points out that dictionaries tend to
define those words in similar ways, and even give them as synonymous of each other:
- The Collins COBUILD English Dictionary (CCED), suggests that
“complete” and “pure” are synonyms of “sheer”.
- The Longman Dictionary of Contemporary English (LDOCE) gives
“pure” as a synonym of “sheer”.
- The earlier Collins COBUILD English Language Dictionary (CCELD)
gives “absolute” as a super ordinate of “sheer”.
In spite of this apparent similarity in meanings, the typical collocates of each
adjective differ to quite a considerable degree. For example “sheer” is used with
nouns of degree or magnitude (sheer weight, sheer number) often in the pattern (the
sheer noun + of noun); e.g. (the sheer weight of noise).
The other adjectives do not collocate with these nouns. In addition, “sheer” alone is
often used in expressions indicating causality (though sheer insistence; by sheer hard
work; because of sheer hard work; his sheer integrity got him though; his enthusiasm
and sheer hard work meant that things moved quickly). Partington (1998:36). He
ends this analysis by making some statements: “Complete”, is used with nouns
indicating:
- Absence:(complete ban)
- Change:( complete revamping)
- Destruction: (complete collapse)
- Absolute is used with what Partington calls “hyperbolic” nouns, such as
(chaos, disgrace, genius……). Ibid (1998:43)
7.1. The corpora analyses.
7.1.1. By frequency analyses
The table bellow shows the frequencies of the adjectives (complete, absolute,
sheer, and utter):
DISTRIB WORD/PHRASE TOKENS PER MIL IN REG1
REG1 [100,000,000 WORDS]
1 COMPLETE 12594 125.94
1 ABSOLUTE 3432 34.32
1 PURE 3305 33,05
1 SHEER 2028 20.28
1 UTTER 652 6.52
Table 4: Frequencies from (view.byu.edu )
By using corpora we can see immediately that complete is the most used word
(12594 times), followed by the adjective absolute (3432 times) , in the 3rd position
we can find that pure is used (3305 times), sheer is used (2028 times) ; however
utter is the last position with only a frequency of (652).
7.1.2. By register analyses
The following five tables show the frequencies of the adjectives concerned by
registers:
REGISTER SPOKEN FICTION NEWS ACADEM NONFIC OTHER
IC MISC MISC
TOKENS 23 31 30 144 139 217
SIZE
10.33 16.19 10.64 15.43 16.63 28.39
(MW)
PER MIL 2.2 1.9 2.8 9.3 8.4 7.6
Table 5: The frequency of “Complete” by registers from (view.byu.edu )
REGISTER SPOKEN FICTION NEWS ACADEMIC NONFIC OTHER
MISC MISC
TOKENS 56 422 218 177 259 585
SIZE
10.33 16.19 10.64 15.43 16.63 28.39
(MW)
PER MIL 5.4 26.1 20.5 11.5 15.6 20.6
Table6: The frequency of “Sheer” by registers from (view.byu.edu )
REGISTER SPOKEN FICTION NEWS ACADEM NONFIC OTHER
IC MISC MISC
TOKENS 87 444 152 380 355 706
SIZE
10.33 16.19 10.64 15.43 16.63 28.39
(MW)
PER MIL 8.4 27.4 14.3 24.6 21.3 24.9
Table7: The frequency of “pure” by registers from (view.byu.edu )
REGISTER SPOKEN FICTION NEWS ACADE NONFIC OTHER
MIC MISC MISC
TOKENS 22 196 37 22 57 120
SIZE (MW) 10.33 16.19 10.64 15.43 16.63 28.39
PER MIL 2.1 12.1 3.5 1.4 3.4 4.2
Table8: The frequency of “Utter” by registers from (view.byu.edu ) .
REGISTER SPOKEN FICTION NEWS ACADE NONFIC OTHER
MIC MISC MISC
TOKENS 318 370 223 925 676 920
SIZE (MW) 10.33 16.19 10.64 15.43 16.63 28.39
PER MIL 30.8 22.8 21.0 59.9 40.6 32.4
Table9: The frequency of “absolute” by registers from (view.byu.edu ) .
The adjective absolute is used mainly in the academic register by a frequency of 59.9
per million words; however the adjective pure is also used in the fiction register by a
frequency of 27.4 per million words, and in the other registers by a frequency of 24.9
per million words. The adjective sheer is used mainly in fiction 422 times which
means 26.1 per million words, it is also used in news register by a frequency of 20.5
per million words; but in what concerns the adjective utter, it is mainly used in
fiction register( only 12.1 per million words); however its use in the other registers is
less important. The adjective complete, is less used if compared with the other
adjectives, we can distinguish that it reaches only a frequency of 9.3 per million
words
7.1.3. By collocation analyses.
The following table shows us the adjectives with their top 20th most frequent
collocations.
DISTRIB WORD/PHRASE TOKENS
REG1
1 PURE WHITE 104
2 COMPLETE NEW 29
3 SHEER HARD 25
4 PURE NEW 18
5 COMPLETE UNIFIED 17
6 SHEER PHYSICAL 17
7 PURE PUBLIC 12
8 ABSOLUTE BEST 11
9 COMPLETE SUPRACONAL 11
10 COMPLETE PHYSICAL 10
11 COMPLETE POLITICAL 9
12 COMPLETE SHORT 9
13 PURE ORAL 9
14 COMPLETE FINANCIAL 8
15 COMPLETE HUMAN 8
16 COMPLETE MENTAL 8
17 PURE ECONOMIC 8
18 ABSOLUTE MINIMUM 7
19 ABSOLUTE MORAL 7
20 COMPLETE MONETARY 7
th
Table 10: adjectives with the top 20 most frequent collocations from
(view.byu.edu ) .
In analysing collocation’s table we can notice immediately that the words: (white
104, new 18 and public 12) are the most frequent words that collocate with the
adjective pure; the word (new 29) collocates most frequently with complete. The
first frequent word that collocates with absolute is (best 11); however the adjective
utter does not exist in the top 20 words that collocate with the adjectives listed before.
8. CONCLUSION:
Corpus Linguistics has developed considerably in the last decades due to the great
possibilities offered by the natural language processing with computers. The
availability of computers and machine-readable texts has made it possible to get data
quickly and easily. Linguistic domains are investigated by the use of computers; the
results are very amazing if compared with the traditional research methods.
References:
1. Biber, D., Conrad, S. & Reppen, R. Corpus Linguistics. Investigating language
structure and use. 1998. Cambridge: CUP. (1998).
2. Cook. G. Applied linguistics: Oxford. OUP. (2003)
3. Firth, J.R. 1957. Papers in Linguistics 1934-1951. Oxford: OUP(1957).
4.Francis, W.N & Kucera, H: Frequency analysis of English usage . Boston ,
MA: Houghton Mifflin. (1982)
5. Hunston, S. Corpora in applied linguistics. Cambridge: Cambridge University
Press. (2002).
6. Johannsson S. and Hofland K.. Frequency Analysis of English vocabulary and
grammar: based on the LOB corpus. Clarendon Press, Oxford.( 1989).
7. Partington, Alan. Patterns and meanings: Using corpora for English language
research and teaching. Amsterdam: John Benjamins PublishingCompany(1998).
8. Sinclair, John McH.. Corpus, concordance, collocation. Oxford: Oxford
University Press. (1991).
9.view.byu.edu Mark Davies, Professor of Corpus Linguistics at Brigham Young
University.
10.Web Concordancer LOB.txt.