A Brief Intro to Corpus Techniques in ELT Research

By Erkan Karabacak

corpus: corpora: concordancer: keyword in context (KWIC):

corpus: a collection of texts

corpora: corpus in plural

concordancer: a search engine

keyword in context (KWIC): a list of words in context

•The Oxford Text Archivewww.ota.ox.ac.uk•Warwick Centre for Applied

Linguisticshttp://www2.warwick.ac.uk/fac/soc/al/•Open American National Corpushttp://americannationalcorpus.org/OANC/OANC-1.0.1-UTF8.zip

Monoconc Wordsmith Tools Concordance Simple Concordance Program WConcord TextStat AntConc

AntConc is our concordancer We will use BASE and texts from our

students as our corpora We will do some simple analyses to

answer some language related questions.

Open the read-me file online and read it.

http://www.antlab.sci.waseda.ac.jp/software/README_antconc3.2.1.txt

By Laurence Anthony, Waseda University, Tokyo

http://www.antlab.sci.waseda.ac.jp/software/antconc3.2.1w.exe

Open Google and search for “download AntConc”

We need a collection of texts (corpus) of an adequate size.

• developed at the Universities of Warwick and Reading

• a collection of transcripts of lectures and seminars recorded at two universities in the UK during the period 1998-2005.

• recorded in a variety of university departments. four broad disciplinary groups,

• each represented by 40 lectures and 10 seminars.

• Arts and Humanities • Life and Medical Sciences • Physical Sciences • Social Studies and Sciences.

• Arts and Humanities , 40 text files, untagged

• Life and Medical Sciences, 40 text files, untagged

• Physical Sciences • Social Studies and Sciences.

…now what are you reading now he asked as i put down the book and reached for my jacket i was labouring over Troilus and Criseyde reading an essay on Criseyde's character you love this rubbish eh he laughed you'll end up an old professor wanking by the fireside putting aside your pipe and warming up your hand first i should say this is not a autobiographical work [laughter] in any s-, in any way right [laughter] er

sm0003: you've said that before [laughter] nm0001: warming up your hand first [laughter] i looked at

him sternly only a joke man he said with mocking reassurance only a joke i sat on the bus deep in thought trying to work out why she should have betrayed him so easily why after all those pure shy exchanges the secret glances

Open AntConc FileOpen FilesSelect the files you

would like to analyze by ctrl+shift (or clicking with your mouse’s left button)Open

You will see the selected files in the left window (titled “corpus files”)

What is the size of the corpus? (How many words (tokens) are there?)

How many different words (types) are there?

Click “Word List”Make your selectionsStart

Let’s search for a single word.

Which lectures are the most fun? Which lectures did not have a lesson

plan? What part of speech mostly follows a

pause?

<struct type="tok" from="29" to="34">

  <feat name="base" value="right" />

  <feat name="msd" value="NN" />

  </struct>

<struct type="tok" from="34" to="35">

  <feat name="base" value="," />

  <feat name="msd" value="," />

  </struct>

<struct type="tok" from="36" to="40">  <feat name="msd" value="DT" />   <feat name="base" value="this" />   <feat name="affix" value="" />   </struct>

<struct type="tok" from="41" to="43">  <feat name="msd" value="VBZ" />   <feat name="base" value="be" />   <feat name="affix" value="s" />   </struct>

Let’s say we want to create a dictionary of medical terms.

Our analysis corpus is BAWE Life and Medical Sciences

What are the most frequently used 10 lexical bundles by American students?

What are the most frequently used 10 lexical bundles by Chinese students?

Of course, AntConc is not enough for every type of analysis.

An applied linguist who wishes to analyze large language data not only should know several application programs, but also learn a programming language; such as PERL

We can create a diachronic corpus from our students papers and observe their development.

We can tag texts for their part of speech or for other information.

We can automatically compile corpora from online sources.

We can do all of the above for other languages (Turkish, Chinese, Russian, and so on)

We can do EVERYTHING a linguist might need to do with texts.