and critical discourse studies Context, content, computation Costas - - PowerPoint PPT Presentation

and critical discourse studies
SMART_READER_LITE
LIVE PREVIEW

and critical discourse studies Context, content, computation Costas - - PowerPoint PPT Presentation

Siena English Language and Linguistics Seminars (SELLS), University of Siena, 9 November 2009 Corpus-based methodology and critical discourse studies Context, content, computation Costas Gabrielatos Lancaster University Main criticisms of CL


slide-1
SLIDE 1

Corpus-based methodology and critical discourse studies

Context, content, computation

Costas Gabrielatos

Lancaster University

Siena English Language and Linguistics Seminars (SELLS), University of Siena, 9 November 2009

slide-2
SLIDE 2

Main criticisms of CL

CL does not take account of the relevant context CL does not examine sufficient amount of (co-)text

lists of words (frequency, keywords, collocates) short concordance lines

Also Nature / definition of CL Objectivity – Subjectivity Replicability

(see also Marchi & Taylor, forthcoming; Partington, 2009; Taylor, 2008)

slide-3
SLIDE 3

Projects

Current: The representation of Islam and Muslims in the UK press, 1998-2008. ESRC. Sept. 2009 – Aug. 2010.

  • PI: Paul Baker; CI: Tony McEnery; RA: Costas Gabrielatos.
  • Corpus: 200,000 articles; 140 million words (and counting)

Completed: Discourses of refugees and asylum seekers in the UK press, 1996-2006. ESRC. Oct. 2005 – March 2007.

  • PI:

Paul Baker; CIs: Tony McEnery, Ruth Wodak; RAs: Costas Gabrielatos (CL), Majid KhosraviNik (CDA), Michal Krzyzanowski (CDA).

  • Corpus: 175,000 articles; 140 million words
  • http://ucrel.lancs.ac.uk/projects/rasim/
slide-4
SLIDE 4

CL does not take account of the relevant context

CL researchers have no less access to sources of relevant contextual information than CDS researchers. A non-linguistic quantitative analysis of a corpus reveals patterns which ... … pinpoint periods/sources/texts that can be usefully examined in detail. … uncover helpful contextual elements.

slide-5
SLIDE 5

Revealing contextual elements 1: Reaction to trigger events

Islam Corpus Query

  • Alah OR Allah OR ayatolah OR ayatollah OR burka* OR burqa* OR chador* OR

fatwa* OR hejab* OR imam* OR islam* OR Koran OR Mecca OR Medina OR Mohammedan* OR Moslem* OR Muslim* OR mosque OR mufti* OR mujaheddin* OR mujahedin* OR mullah* OR muslim* OR Prophet Mohammed OR Q'uran OR rupoush OR rupush OR sharia OR shari'a OR shia! OR shi-ite* OR Shi'ite* OR sunni* OR the Prophet OR wahabi OR yashmak* AND NOT Islamabad AND NOT shiatsu AND NOT sunnily

  • Graph depicting number of articles per month
  • Establishing events coinciding with / triggering spikes.
  • Extent of change in number of articles due to trigger events.

– UK press – Individual newspapers

slide-6
SLIDE 6

50 100 150 200 250 300 350

9/11 7/7 Veil C2 Bali C

1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009

Iraq invasion Iraq 2 Madrid Somalia

National UK newspapers: average number of articles /month

slide-7
SLIDE 7

Is the general picture representative of all newspapers?

slide-8
SLIDE 8

Yes

  • Four spikes shared by

at least two-thirds of the 12 newspapers:

– 9/11, 7/7: 12 – Veil + Cartoons 2: 11 – Bali bombings + Cartoons: 9

  • All newspapers but one

show an upward trend.

  • 19 spikes collectively – only 5

shared by more than half!

  • Different relative importance
  • f primary and/or secondary

spikes

  • Five groups in terms of

primary spikes:

– 9/11 & 7/7 – 9/11 – 7/7 + other – Other + 9/11 & 7/7 – Other

No

slide-9
SLIDE 9

9/11 & 7/7

3 newspapers

slide-10
SLIDE 10

50 100 150 200 250 300 350 400 450

Guardian

9/11 7/7 Bali C Veil C 2

1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009

Iraq inv. O in T Iran Elect.

slide-11
SLIDE 11

50 100 150 200 250 300 350 400 450 500 550 600 650 700

Independent

9/11 7/7 Bali C Veil C2

1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009

Iraq inv.

slide-12
SLIDE 12

50 100 150 200 250 300 350 400 450

Mirror

9/11 7/7 Bali C Veil C2 Iraq inv.

1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009

slide-13
SLIDE 13

9/11

4 newspapers

slide-14
SLIDE 14

2 4 6 8 10 12 14 16 18 20 22

Business

9/11 7/7 Bali C Veil C 2

1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009

Phil. Indon. Phil.

slide-15
SLIDE 15

50 100 150 200 250 300 350 400 450 500 550 600

Telegraph

9/11 7/7 Veil C 2

1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009

slide-16
SLIDE 16

50 100 150 200 250 300 350 400 450 500 550 600 650 700 750 800 850

Times

9/11 7/7 Veil C2 Bali C

1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009

Iraq inv. Iran elect.

slide-17
SLIDE 17

5 10 15 20 25 30 35 40

People

9/11 Iraq inv.

1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009

Madrid Iraq2 7/7 Bali C Somalia

slide-18
SLIDE 18

7/7 +

  • ther

2 newspapers

slide-19
SLIDE 19

50 100 150 200 250 300

Express

9/11 7/7 Bali C Veil C 2

1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009

Iraq 2 Madrid

slide-20
SLIDE 20

50 100 150 200

Observer

9/11 Iraq inv. 7/7 Veil C2 C protests

1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009

Archbi shop Flogging Thailand Somalia

slide-21
SLIDE 21

Other + 9/11 & 7/7

2 newspapers

slide-22
SLIDE 22

50 100 150 200 250 300 350

Mail

9/11 7/7 Bali C Veil C2 C protests ???

1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009

slide-23
SLIDE 23

50 100 150 200 250 300

1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009

9/11 7/7 Veil C2

Sun

??? Iraq inv. C protests Archbi shop ??? O in T Iran elect.

slide-24
SLIDE 24

Other

1 newspaper

slide-25
SLIDE 25

50 100 150 200 250 300 350

Star

1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009

9/11 7/7 Bali C Veil C2 Somalia ??? ???

slide-26
SLIDE 26

Can we measure a newspaper’s response to a given trigger event?

slide-27
SLIDE 27

Change in number of topic-related articles Trigger event: 9/11

Average of 12 months pre-9/11 9/11 spike (avg. Sept.-Oct. 2001) Average of 12 months pre-9/11 Average of 12 months post-9/11

  • Diff. % of number of articles

Reaction

The extent to which the number

  • f articles changed immediately

after a trigger event

Sustain

The extent to which the change was sustained a year after the trigger event

slide-28
SLIDE 28

9/11 and Islam/Muslims: Reaction and Sustain

Business Express Guardian Independent Mail Mirror Observer People Star Sun Telegraph Times

50 100 150 200 250 300 350 400 450 500 550 600 650 700 750 800 850

  • 50

50 100 150 200 250

Reaction: Diff.% pre-spike 12 vs. 9/11 spike Sustain: Diff.% pre-spike 12 vs. post-spike 12

slide-29
SLIDE 29

What about the broadsheets-tabloids distinction?

slide-30
SLIDE 30

9/11, reaction and sustain: clusters

slide-31
SLIDE 31

9/11, reaction and sustain: clusters

slide-32
SLIDE 32

9/11, reaction and sustain: clusters

slide-33
SLIDE 33

9/11, reaction and sustain: clusters

slide-34
SLIDE 34

9/11, reaction and sustain: clusters

slide-35
SLIDE 35

Revealing contextual elements 2: Use of loaded terms

RASIM Corpus Query

  • refugee OR asylum OR deport* OR immigr* OR emigr* OR migrant*

OR illegal alien* OR illegal entry OR leave to remain AND NOT deportivo AND NOT deportment

(see Gabrielatos, 2007)

Nonsensical (= loaded) terms

illegal/legal refugee*/asylum seeker* bogus/genuine immigrant*/migrant*

definitions

  • Which newspapers use them?
  • How frequently?

– Per million words – Per thousand articles

(see also Baker et al., 2008)

slide-36
SLIDE 36

Definitions

Longman Dictionary of Contemporary English (2003) Refugee Council refugee Someone who has been forced to leave their country, especially during a war, or for political or religious reasons. Someone whose asylum application has been successful and who is allowed to stay in another country having proved they would face persecution back home. asylum seeker Someone who leaves their own country because they are in danger, especially for political reasons, and who asks the government of another country to allow them to live there. Someone who has fled persecution in their homeland, has arrived in another country, made themselves known to the authorities and exercised the legal right to apply for asylum. immigrant Someone who enters another country to live there permanently.

  • migrant

Someone who goes to live in another area

  • r country, especially in order to find work.

[economic migrant] Someone who has moved to another country to work.

International Association for the Study of Forced Migration Forced migration: refugees and asylum seekers Voluntary migration: immigrants and (economic) migrants

back

slide-37
SLIDE 37

RASIM: Frequency of nonsensical terms

Business Express Guardian Independent Mail Mirror Observer People Star Sun Telegraph Times

2 4 6 8 10 12 14 16 18 20 22 24 26 28 1 2 3 4 5 6 7 8 9 10 11

  • Freq. per million words
  • Freq. per 1000 articles
slide-38
SLIDE 38

Nonsensical terms: clusters

slide-39
SLIDE 39

Nonsensical terms: clusters

slide-40
SLIDE 40

Nonsensical terms: clusters

slide-41
SLIDE 41

CL does not examine sufficient amount of (co-)text

lists of words (frequency, keywords, collocates) short concordance lines Corpus research can involve the close analysis of longer stretches of text, up to whole texts ... ... while explicit annotation of features (e.g. stance) enables the quantification of emerging patterns ... ... and replication of the analysis.

slide-42
SLIDE 42

CL taking a closer look

Examination of uses of suffocat* and drown* in relation to RASIM.

Why?

  • Some forms corpus-wide collocates of RASIM
  • Not key in broadsheet-tabloid comparison
  • Investigation of (what was expected to be)

sympathetic reporting on RASIM

  • illegal shared collocate of many forms of suffocat* and drown*

Presentation was negative in almost 50% of instances Negative presentation was direct or indirect.

slide-43
SLIDE 43

Direct

Through attribution by

– modification of victims with adjectives such as illegal, clandestine, etc. – co-reference to the victims using illegal, etc. – reference to their attempts to enter as illegal, etc.

slide-44
SLIDE 44

In June, 58 illegal immigrants from China suffocated in the back of a lorry in Dover, after a journey across Europe. [The Express, Nov. 2000] A Dutch lorry driver was jailed for 14 years for killing 58 Chinese immigrants who suffocated in his trailer as he tried to smuggle them into Britain. Perry Wacker, 33, closed an air vent during the Channel crossing so that the ferry crew could not hear his illegal cargo [The Daily Mail, June 2001]

slide-45
SLIDE 45

Indirect

Through framing the report within …

– general references to illegal immigration – indirect references to the ‘illegality’ of RASIM

(e.g. suspected asylum seeker, sneak across the perilous straits)

– references to smuggling, trafficking, illegal entry/transport etc. – references to problems with immigration/asylum or RASIM – references to problems with, or laxity of, the existing immigration / asylum system

slide-46
SLIDE 46

A SUSPECTED asylum seeker drowned and another was seriously ill with hypothermia last night after they tried to cross the Channel in a 12ft kayak. [The Daily Mail, June 2002] China is among the top four countries whose citizens are sneaking in. It does not want a repeat

  • f

such tragedies as the drowning last year in Morecambe Bay or in 2000 when 58 Chinese suffocated in the back of a lorry heading for Dover. [The Times, Sept. 2005] The risks

  • f

trafficking were highlighted last summer by the deaths of 58 Chinese immigrants found suffocated in a Dutch-owned truck which arrived in Dover from Belgium. Illegal immigration is also expected to be high

  • n

the agenda

  • f

an Anglo- French summit in Cahors, southern France … [The Guardian, Feb. 2001]

slide-47
SLIDE 47

Comparison of tabloids and broadsheets

suffocat* + drown* T (n=250) B (n=409) T % B % LL Negative attribution [Direct] 75 93 30.0 22.7 3.15 Negative framing [Indirect] 46 80 18.4 19.6 0.11 Negative presentation [TOTAL] 121 173 48.4 42.3 1.28

LL = 6.63  probability of results being due to chance 1% LL = 3.84  probability of results being due to chance 5%

slide-48
SLIDE 48

suffocat* / drown*

 Negative presentation is unexpectedly high in both groups.  There is no statistically significant difference between B and

T in the proportion of negative presentation.

 Indirect negative presentation is equally favoured.  Tabloids seem to prefer direct negative presentation …  … but difference is not statistically significant.  Both B and T make sure to project sympathy when

reporting tragedies involving RASIM during their journey to the destination country ...

 ... but they very frequently communicate the notion that

the victims were party to an illegal act, and, consequently, were somehow responsible for their fate.

slide-49
SLIDE 49

suffocat* / drown* Following the CL leads and digging deeper

slide-50
SLIDE 50

The coherence of racism

  • r

Racism sandwich

slide-51
SLIDE 51

[‘Interview’ with ‘the man in the street’, The Sun, June 2000]

Fifty-eight Chinese immigrants suffocating to death in a lorry after paying £18,000 each to be smuggled into Dover? WVM: Absolutely tragic and I hope in some sort of way it's a lesson to the rest of them not to take such huge risks with their lives just to get into Britain. But it's the ones behind the smuggling that should be made to pay the price not the poor souls desperate to flee.

[Letter, The Sun, June 2001]

WHAT a horrific, callous man Perry Wacker is to let those 58 Chinese migrants suffocate in the rear of his truck. If

  • ur

Government had stood firm and made it difficult to enter Britain

  • turning

migrants back instead

  • f

looking after them

  • they

would not try to smuggle themselves here. Then this tragic waste

  • f

life and the anguish

  • f

the people who found them might not have happened. The manslaughter charge should have been shared by the Government for not sorting out the problem.

slide-52
SLIDE 52

Sting in the tail

(see also Morley, 2004)

slide-53
SLIDE 53

HEADLINE: Sun, sea sand and corpses: It's an idyllic scene: a young couple enjoy a picnic on the beach. But a few yards along the sand lies a body - one of many lives lost in the struggle to reach Europe. In our second special report on immigration, Peter Lennon visits Zahara de los Atunes in southern Spain: Real Lives Zahara de los Atunes, situated on part of the Spanish coastline that has become one of the most popular windsurfing areas in Europe, recently achieved unwelcome notoriety when a photographer took a picture of a young couple sunbathing on the beach within yards of the body of a drowned immigrant.

slide-54
SLIDE 54

It was a little unlucky for Zahara which, at 41km west of Tarifa, is not the preferred destination of the immigrants. They make for the closer beaches of Punta Paloma and Bolonia, seven and 15km out from the town. But crossing in increasing numbers in light craft, at the mercy of winds and tides, Zahara is now getting its share. [The Guardian, 13 December 2000]

slide-55
SLIDE 55

Slow-release racism

slide-56
SLIDE 56

ANYONE returning to these shores after a few years abroad will look in awe at how Ireland has changed beyond recognition. We have on this tiny island an incredible mix of Africans, East Europeans, Scandanavians and people from the Middle-East who have chosen to make Ireland their home. Most are here to work, to make money and raise a family. Others are here to escape the horror of their blood- soaked native lands, where they have seen their mothers, fathers, sisters and brothers butchered before their eyes. They came here wearing nothing but the clothes on their back, probably having handed over obscene money to ruthless people traffickers - the price of their freedom. Dozens of them, some pregnant, never get this far, drowning on flimsy rafts on the Strait of Gibraltar, crossing from North Africa to the promised land of Europe. Others, like the 58 Chinese immigrants found in Dover in 2001, suffocate in the back of articulated trucks. Under European law, we owe these people shelter, food and protection once they arrive here.

slide-57
SLIDE 57

But there is one word which gives politicians a huge headache - and poses the greatest challenge in how we deal with immigrants: Deportation. Two Irish cases in particular spring to mind - Nigerian student Kunle Elukanlo, the teenager who begged to be allowed to sit his Leaving Cert here - and Nimota Bamidele, who claimed she would be stoned to death if returned to the same country. These people clearly pose no threat to our national

  • safety. But soon we will have to deal with people who do.

And that's why Bertie Ahern, Mary Harney and Michael McDowell must look across the water at British Home Secretary Charles Clarke's rules on what should constitute deportation. In the wake of the London bombings last month, he has promised to kick people out of Britain if they: SUPPORT terrorists and justify or celebrate violence FOSTER hatred which could lead to inter-community violence, or WRITE, produce, publish or distribute material inciting violence. Sounds sensible, doesn't it?

slide-58
SLIDE 58

And, yet, we have to listen to the warped bile of civil rights groups questioning the legality of giving the prophets

  • f hate their marching orders.

We have to put up with militant Islamic groups crying foul – saying objections to the new rules were ignored. Maybe they should try selling that argument to Eileen and Paddy Tallon, whose fireman son Sean died in the 9/11 attacks

  • n New York in 2001 doing what he loved - rescuing people.

Or John Falding, who was speaking to his girlfriend Anat Rosenberg on a mobile phone when she was killed in the No30 bus bomb in London's Tavistock Square. The Government should photocopy Clarke's deportation rules and rewrite it for the statute book here. Genuine foreign nationals who want to make Ireland their home, who want to embrace our way of life and enrich it with theirs, have nothing to fear from such legislation. But those who preach or drum up support for violence - round them up and kick them out. Now. [The Mirror, 26 August 2005]

slide-59
SLIDE 59

Conclusions (1)

  • The quality of reporting can be quantified by examining

(among other aspects) ... – the link between the reaction+sustain in entity-specific articles and ‘trigger’ events. – the frequency of use of explicitly loaded terms in relation to entities in focus (semantic prosodies). – negative attitudes hidden in / sandwiched within superficially / partially ‘sympathetic’ reporting (discourse prosodies).

  • Such targeted analysis ...

– provides a means of principled and transparent text selection. – expands the current family of CL techniques by ‘raiding’ compatible methodologies.

slide-60
SLIDE 60

Conclusions (2)

  • The broadsheet-tabloid distinction is better seen as a cline.
  • Some tabloids may show some broadsheet features

– and vice versa (see also Duguid, forthcoming).

  • Some newspapers show B/T features more consistently

(e.g. The Business, The Sun).

  • The same newspaper may vary its approach to reporting -

and its stance towards the topic/entities involved - according to the reported event.

  • Studies of groups of newspapers may miss important

individual differences (see also Marchi & Taylor, forthcoming).

  • Studies of particular newspapers can safely generalise only

about the particular newspaper.

slide-61
SLIDE 61

Questions

How helpful / generalisable are conclusions drawn from the analysis of a small number of articles … … particularly when they have been selected because they are ‘interesting’ or ‘telling’? Can depth of analysis compensate for lack of representativeness?

slide-62
SLIDE 62

Suggestions (1)

  • Objective vs. subjective distinction not useful – misleading.
  • Objective processes involve subjective decisions at some point:

– Setting thresholds for statistical test results (why are we happy with 1% probability of chance/error- but not 1.1%?) – Lexis-based research is not a-theoretical (McEnery & Gabrielatos,

2006).

slide-63
SLIDE 63

Suggestions (2)

  • Useful distinctions:

– explicit annotation implicit annotation – discrete categories non-discrete categories – precise counting approximate/fuzzy counting – stat. testing general impression – total accountability selective treatment  replicability non-replicability

  • Replicability applies to analytical categories, procedures and

results - not interpretation.

slide-64
SLIDE 64

What corpus linguists tend to do What corpus linguistics can do