Want to make creations as awesome as this one?


Coimbra Group Summer School on European Multilingualism, June 24th to June 28th, 2024Prof. Dr. Raúl Sánchez Prieto, raulsanchez@usal.es

National stereotypes across centuries


0. Debunking nations and names1. National stereotypes2. Text linguistic opinion mining for detecting and analysing national stereotypes (using the example of Spain)3. Critical Discourse Analysis and text linguistic opinion mining for historical texts

Interesting facts about nations and names:

  • Germany (Deutschland, Germania, Alemania, Alemanha, Alamannen
  • English, Sasanach, Saxon
  • *þiudisk (deutsch, tedesco) - *walhisk (welsch-Welsch)
  • Dutch - Holland/Olanda - Low Countries/Paesi Bassi - Nederland/De Lage Landen
  • German - nemec - чуждый
  • French - Franken
  • Spain - Visigothia?
  • Rus - Russia - Ruotsi - Vänäjä - Sweden - Svenskarna
  • Imperium Romanum - Βασιλεία τῶν Ῥωμαίων/Ῥωμανία - Romania
  • Hispania - Hespanha - Portugal - Porto

0. Debunking nations and names

For Greenland (2000: 15) stereotypes are “probabalistic, generalised representations of any social group”. Stereotypes play an important role in social cognition. Their effects on perception, thought and behaviour are considerable. Bias and prejudice are an integral part of stereotyping leading to the frequent denigration of ‘others’.The more social groups suffer from anxiety about other groups, the more they will practice negative stereotyping (Berger 2021: 13).

1. National stereotypes

A close analysis of national histories leads to the conclusion that they are often characterised by positive auto-stereotypes and negative stereotypes about ‘others’, which can be both external and internal.Fritzsche (2008): “The relationship between victimhood and violence is embedded in most national historiographies […]. [The national idea, S.B.] is first conjured up as being under threat. And it is this state of alarm that produces the energy to override competing identities, often violently. Violence is inscribed in the national narrative because the nation imagines itself first and foremost as a collective good that is incomplete and imperilled. In many ways, the national narrative must sustain itself by reproducing its own state of jeopardy. National histories tremble as a result.”

Example of Germany after unification of 1871.Internal enemies:

  • National histories contributed to constructing a whole string of internal and external enemies in order to strengthen an altogether insecure national ‘we’ group
  • Catholics and Socialists were scapegoats of German nationalism
  • “Inferioritätsdebatte”: Protestants vs. Catholics (Max Weber)
External enemies:
  • France
  • England
  • Russia and the Slavs
The role of historians and historiography in cementing stereotypes.

2. Text linguistic opinion mining for detecting and analysing national stereotypes (using the example of Spain)

National stereotypes, nation branding, social media and text linguistics: an unusual but useful connectionText linguistic and media linguistic approach presented here > practical and straightforward method for the empirical analysis of national stereotyping Main goal  > to provide a working tool for evaluating national stereotypes on social media.Since “l’image de l’Autre est construite à travers un discours où le stéréotype règne en maître glorieux” (Laamiri/Ouasti 2001: 117), national stereotypes are hard to erase and need to be actively opposed.

This is one of the most important reasons countries engage in nation branding.Main objectives of Southern European crisis countries in nation branding: “helping restore international credibility and investor confidence” and “reverse international ratings downgrades” > The cultural dimension of national stereotyping (Blum 2004: 252) and the economic facets of a nation brand are closely related.Most studies on nation branding are conceptual:

  • empirical research is limited (Papadopoulos et al. 2016: 459)
  • mostly based on sociological description models (v.g. Kohut 2012, a Pew Research Center report on the reputation of European countries)
  • Social media research is being incorporated into this sociological approach to nation branding and national stereotyping.

Linguistic studies on national stereotypes not very common (Quasthoff/ Hallsteindóttir 2016: 347) > several research streams (Vilinbahova 2014):

  • Semantic and lexicographic stereotype research (see Ossenberg/Baur 2016 for an overview)
  • Critical Discourse Analysis: Van Dijk, Wodak, etc.
  • Text and media linguistics: very scarce work, v.g. Dąbrowska (1999), PümpelMader (2010).
Social media have been neglected in text linguistic research concerning national stereotyping.In contrast, in Critical Discourse Analysis there is an ongoing discussion on how to approach social media data (Unger/Wodak/KhosraviNik 2016).As social media are the new “digital public sphere” (Valtysson 2012: 77), the empirical analysis of social media conversations is vital for understanding how stereoypes on any particular European nation are built up and spread.

An empirical approach to text linguistic opinion mining for detecting and analysing national stereotypesThe proposed approach builds on previous research on text actions (≈ text acts). The description of the actions displayed in online conversations:

  • could reveal the textual intentions of the commenting user
  • will detect the topics internet users comment on
  • will expose how the users characterise a given nation
  • will incorporate polarity
  • will take the collocational behaviour of terms that refer directly to the nation
  • brand into account.

Text actions and sentiment analysisText actions (or text acts) are the main structuring element in texts (Rothkegel 1992: 675). Rothkegel’s theory of text action (Rothkegel 1993), based on Rehbein (1977), Gülich (1986) and Brandt/Rosengren (1992) is modelled on Speech Act Theory.Text actions “tie together functional, thematic, text-organizing and linguistic aspects” (Rothkegel 1993: 59).Contrary to speech acts and similarly to Dynamic Speech Act Theory (Geis 1995), text acts cannot be associated “with the use of particular linguistic cons-tructions” (Geis 1995: 12) > they are communicative actions that constitute the text’s thematic structure (Rothkegel 1984: 240ff).

There are several theoretical and practical approaches on text acts (Czachur 2000: 66ff for an overview).Our own is based on Von der Lage-Müller’s “Handlungsmodell” (Von der LageMüller 1995: 50ff).Three different hierarchical levels of internal organisation for text actions.

Similar approaches based on Von der Lage-Müller: Schröder (2003), Janich (2005), Golonka (2009), Ortner (2014), Sánchez Prieto (2017). No attempt so far made to link text actions with sentiment analysis.Sentiment analysis “Task of identifying positive or negative opinions, emotions and evaluations” (Stock/Stock 2013: 435). Its main purpose is “to determine the contextual polarity of a given sentiment” (Hollander et al. 2016: 8).Computational linguistics has done most of the research in this respect (see Taboada 2016). Sentiment analysis can also be applied to other linguistic subfields and “polarity” is a common concept in imagology (e.g. Leerssen 2007: 344).

Sentiment analysis will be used to determine the polarity of secondary text actions. This method will allow us:

  • tto determine the national stereotypes social media users comment o
  • to establish the polarity of those comments.
Thus, here national stereotypes will be detected and analysed in conjunction with the text actions they shape and their polarity.

Hoping to find a NLP method for automatically recognizing/classifying text acts.Evaluation of the performance of some speech act classifiers:

  • Maximum Entropy Model, best tool, but accuracy 62% (far below Choi 2005)
  • Support Vector Machines, accuracy far below Moldovan et al. (2011)
  • Naïve Bayes classifiers, accuracy far below Moldovan et al. (2011)
  • SentiWordNet, accuracy far below Singh at al. (2013)
  • Semantic Orientation Calculator, accuracy far below Taboada (2011)
All insufficient for our descriptive needs.

These (disappointing) results are consistent with Dashtipour (2016: 768) > the applicability of these tools is limited to the subject domains they were designed for (Moreno Ortiz/Pérez Hernández 2013: 98) and there is not always a clear relation between text act and their morphosyntactic and semantic surface.Our procedure: 1) Corpus breakdown into units of meaning (UoM) by adding sentence boundaries with Python Natural Language Toolkit2) Manual revision and correction of the automatic segmentation3) Categorisation of the UoM into one of the secondary text actions considered4) Determination of polarity.

Collocational behaviour of key words that refer to the nation brand Collocations are understood here as the recurrent co-occurrence of words, i.e., the “tendency for words to occur together repeatedly” (Saeed 2009: 60). The study of collocation patterns is widely used in Critical Discourse Analysis (Fairclough 2003: 131).It may help reveal stereotyping schemes in a text linguistic study but generally large bodies of text are needed for obtaining significant results.

Spain in the eyes of Northern, Southern and Eastern European Facebook usersCorpus descriptionConversations posted on Facebook pages of different newspapers across Europe:

  • Search terms: “Spain”, “Spanish” in the respective language
  • Period: June to December 2016 (six months).
  • Criteria for inclusion in the corpus: comments in Spain-related threads
  • published beneath Facebook posts about Spain (no sports and showbiz news).
One right- and one left-leaning newspaper with the highest circulation in each case per country. Circulation figures by the national circulation bureaus.

Text actions in the corpus: results

Assessing Spanish society(1) Spanien sollte lieber die tierquäler bestrafen, die stierkämpfe abschaffen und die toten windhunde von den bäumen schneiden, auf denen die spanischen jäger sie am saisonende aufknüpfen (A-STAND 52)(2) “La giustizia spagnola, peggiore della.nostra, se possibile...... (I-CORRIERE 42)(3) Je trouve que l'Espagne est un pays beaucoup plus tolérant avec l'homosexualité que la France. J'ai vécu à Madrid et personne ne m’a jamais insulté pour embrasser un garçon dans le métro où me balader main dans la main avec lui dans la rue de n'importe quel quartier. JAMAIS. (F-LE MONDE 119).

Table 3. STA and PSTA assessing Spanish society in Northern, Southern and Eastern European comments about Spain Austrian Flemish French Italian Polish Bulgarian STA Describing Spanish society 81,8% (n=9) 50% (n=4) 26% (n=20) 22,8% (n=24) 0 0

  • characterising Spain as a modern country
0 0 50% (n= 10) 45,8% (n= 11)
  • characterising Spain as a traditional country
77.7% (n= 7) 100% (n= 4) 35% (n= 7) 45,8% (n= 11)
  • neutral
22.3% (n= 2) 0 15% (n= 3) 8,4% (n= 2) STA Assessing the Spanish on trustworthiness 18,2% (n=2) 50% (n=4) 28,6% (n=22) 64,8% (n=68) 100% (n=4) 0
  • stating that the Spanish are trustworthy
0 0 45,4% (n= 10) 7,4% (n= 5) 75% (n= 3)
  • stating that the Spanish are not to be trusted
100% (n= 2) 100% (n= 4) 54,6% (n= 12) 70,6% (n= 48) 25% (n= 1)
  • neutral
0 0 0 22% (n= 15) 0 STA Assessing the Spanish on amicability 0 0 28,6% (n=22) 10,5% (n=11) 0 0
  • expressing that the Spanish are friendly
68,2% (n=15) 45,4% (n= 5)
  • describing the Spanish as unfriendly
18,2% (n=4) 54,6% (n= 6)
  • neutral
13,6% (n=3) 0 0 STA Assessing the Spanish on tolerance 0 0 16,8% (n=13) 1,9% (n=2) 0 0
  • describing Spanish society as tolerant
76,9% (n= 10) 50% (n=1)
  • describing Spanish society as intolerant
23,1% (n= 3) 50% (n=1)
  • neutral
0 0 Total 100% (n=11, 7.9% from total STAs) 100% (n=8, 6.5% from total STAs) 100% (n=77, 13.2% from total STAs) 100% (n=105, 13.1% from total STAs) 100% (n=4, 4.8% from total STAs) 100% (n=0)

Assessing the Spanish economy(4) Ein Land, in dem Arbeitslosigkeit, Schulden und sinkende Löhne einen Linksruck bringt... (A-PRESSE 46)(5) senza governo crescono del 3,5 %. noi con il bomba, pdioti alfano e verdini , 0.00000000000000 periodico (I-REPUBBLICA 331)(6) Le plus grave c'est que la corruption est completement normalisée dans la société espagnole (F-MONDE 636)(7) Hij heeft, en nog, de Spanjaarden uitgemolken en de slechte resultaren zijn er, 22,8 % werklozen, 28,6 % vd bevolking is arm, 50 % vd jongeren hebben geen werk (BE-DM 4)(8) Waar wachten jullie nog op, Brussel en de Vlaamse rand, om net als rond Madrid vier concentrische ringwegen aan te leggen! (BE-STAN 39)

Table 4. STA and PSTA assessing Spanish economy in Northern, Southern and Eastern European comments about Spain Austrian Flemish French Italian Polish Bulgarian STA Assessing the shape of the Sp. economy 47.4% (n=18) 33.3% (n=14) 55% (n=78) 68.8% (n=123) 0 0

  • describing the Spanish economy as strong
5.6% (n=1) 0 7.7% (n=6) 26% (n=32)
  • describing the Spanish economy as weak
94.4% (n=17) 85.7% (n=12) 83.3% (n=65) 58.5% (n=72)
  • neutral
0 14.3% (n=2) 9% (n=7) 15.5% (n=19) Assessing the working ethos of the Spanish 10.5% (n=4) 11.9% (n=5) 4.9% (n=7) 0 0 0
  • describing the Spanish as hard-working
0 0 28.6% (n=2)
  • describing the Spanish as lazy
100% (n=4) 100% (n=5) 57.2% (n=4)
  • neutral
0 0 14.2% (n=1) Commenting on corruption in Spain 26.3% (n=10) 28.6% (n=12) 4.2% (n=6) 10% (n=18) 0 0
  • characterising the Spanish as honest
0 8.3% (n=1) 33.3% (n=2) 33.3% (n=6)
  • characterising the Spanish as corrupt
80% (n=8) 83.4% (n=10) 67.7% (n=4) 67.7% (n=12)
  • neutral
20% (n=2) 8.3% (n=1) 0 0 Assessing affluence and material prosperity in Spain 15.8% (n=6) 26.2% (n=11) 35.9% (n=51) 21.2% (n=38) 0 0
  • describing Spain as an affluent and prosperous country
0 27.3 (n= 3) 3.9% (n=2) 7.9% (n=3)
  • describing Spain as an impoverished and declining country
83.3% (n=5) 72.7% (n=8) 92.2% (n=47) 92.1% (n=35)
  • neutral
16.7% (n= 1) 0 3.9% (n=2) 0 Total 100% (n=38, 27.1% from total STAs) 100% (n=42, 33.9% from total STAs) 100% (n=142, 24.7% from total STAs) (n=179, 22.2% from all STAs) 0 0

Assessing Spanish politics

Table 5. STA and PSTA assessing Spanish politics in Northern, Southern and Eastern European comments about Spain Austrian Flemish French Italian Polish Bulgarian Describing the Spanish political system 8.1% (n=5) 13% (n=7) 11.6% (n=36) 5.2% (n=12) 11.8% (n=4) 0

  • stating that Spain is a stable democracy
40% (n=2) 28.6% (n=2) 41.7% (n=15) 66.7% (n=8) 25% (n=1)
  • commenting on the lack of democracy in Spain
60% (n=3) 57.1% (n=4) 58.3% (n=21) 25% (n=3) 75% (n=3)
  • neutral
0 14.3% (n=1) 0 8.3% (n=1) 0 Describing Spanish political parties 80.6% (n=50) 64.8% (n=35) 72.3% (n=224) 88% (n=206) 17.6% (n=6) 100% (n=7)
  • in a positive manner
10% (n=5) 5.7% (n=2) 18.3% (n=41) 11.7% (n=24) 50% (n=3) 71.4% (n=5)
  • in a negative manner
82% (n=41) 85.7% (n=30) 65.2% (n=146) 62.1% (n=128) 16.6% (n=1) 28.6% (n=2)
  • neutral
8% (n=4) 8.6% (n=3) 16.5% (n=37) 26.2% (n=54) 33.4% (n=2) 0 Commenting on regional nationalism 11.3% (n= 7) 22.2 (n= 12) 16.1% (n=50) 6.8% (n=16) 70.6% (n=24) 0
  • aligning him or herself with Spanish constitutio-nalism
42.9% (n=3) 25% (n=3) 50% (n=25) 56.2% (n=9) 75% (n=18)
  • aligning him or herself with regional nationalism
57.1% (n=4) 75% (n=9) 30% (n=15) 43.8% (n=7) 8.3% (n=2)
  • neutral
0 0 20% (n=10) 0 16.7% (n=4) Total 100% (n=62, 44.3% from total STAs) 100% (n=54, 43.5% from total STAs) 100% (n=310, 54.8% from all STAs) 100% (n=234, 29.2% from all STAs) 100% (n=34, 41% from all STAs) 100% (n=7, 88% from all STAs)

(9) De man die Spanje kapotbespaarde, wordt nu de premier van een minderheidsregering... (BE-DM 32) (10) Sie [Die Spanier] sind empört über die Günstlingswirtschaft und Korruption in den beiden großen Volksparteien (A-PRESSE 45)(11) Най-после, някой да го каже на 'ясен български'! (BG-KAPITAL 1)(12) Catalonië onafhankelijk! (BE-DM 4)(13) Debiły! Wszyscy separatyscy są debiłami (PL- RZECZPOSPOLITA 15)

Collocations in the corpus: results

Table 6. STA and PSTA assessing Spanish on other topics in Northern, Southern and Eastern European comments about Spain Austrian Flemish French Italian Polish Bulgarian Commenting on Spanish tourist destinations 34.5% (n=10) 35% (n=7) 16.6% (n=7) 8.8% (n=25) 15.6% (n=7) 0

  • making a positive comment on a tourist destination
80% (n=8) 42.8% (n=3) 28.6% (n=2) 36% (n=9) 100% (n=7)
  • making a negative comment on a tourist destination
20% (n=2) 28.6 (n=2) 42.8% (n=3) 52% (n=13) 0
  • neutral
0 28.6% (n=2) 28.6% (n=2) 12% (n=3) 0 Commenting on Spanish culture 10.3% (n=3) 0 7.2% (n=3) 11.2% (n=32) 35.6% (n=16) 0
  • making a positive comment on Spanish culture
66.7% (n=2) 100% (n=3) 9.4% (n=3) 75% (n=12)
  • making a negative comment on Spanish culture
33.3% (n=1) 0 40.6% (n=13) 12.5% (n=2)
  • neutral
0 50% (n=16) 12.5% (n=2) Commenting on other topics 55.2% (n=16) 65% (n=13) 76.2% (n=32) 80% (n=228) 48.8% (n=22) 12% (n=1)
  • making a positive comment
43.7% (n=7) 38.4% (n=5) 65.6% (n=21) 25.4% (n=58) 9.1 (n=2) 0
  • making a negative comment
43.7% (n=7) 30.8% (n=4) 15.6% (n=5) 50% (n=114) 13.6% (n=3) 100% (n=1)
  • neutral
12.6% (n=2) 30.8 (n=4) 18.8% (n=6) 24.6% (n=56) 77.3% (n=17) 0 Total 100% (n=29, 20.7% from all STAs) 100% (n=20, 16.1% from all STAs) 7.3% (n=42) 100% (n=285, 35.5% from all STAs) 100% 54.2% (n=45, 54,2% from all STAs) 100% (n=1, 12% from all STAs)

Conclusion and comments on this approach

This part of the presentation is based on Sánchez Prieto (2021). Click on this link to read it. You will also find there the quoted references.

The perception of foreign nations in the past (example: Middle Ages)In the Middle Ages the vision of the Other was determined by territory, religion and language. Characteristics:

  • Only manuscripts are available
  • Historical languages are a problem
  • Difficult interpretation
Study of historical perceptions within the framework of CDA:Perceptions and attitudes are imaginary abstractions that are observable in texts. Attitudes towards other human groups are overt manifestations of an opaque and/or implicit ideology (Fairclough 1995: 132).

3. Critical Discourse Analysis and text linguistic opinion mining for historical texts

Concept "perception (or attitude) of peoples/nations"Adaptation of Baker's (1992: 29) scheme on the different types of existing attitudes:

  • Perception of territory: positive autoperception in the case of Spain and the Netherlands. Always negative for non-Western "nationes"
  • Perception of Western Christianity: Christians vs. infidels
  • Perception of languages

Corpus and research designStarting point: Black Legend. Is there something prior to the Black Legend that explains the success of Hispanophobia in the Netherlands?Corpus

  • Spiegel Historiael, a Middle Dutch chronicle from the 13th century.
  • Commissioned by Count Florentius V of Holland, a count with ambitions to the Germanic throne
  • Lexematic search criterion: "spaen-" and "span-" as radical morphemes.

Critical Discourse Analysis for historical texts in actionAuthors: Scollon (1998), Van Dijk (2001), Reisigl/Wodak (2001), Jäger (2004). Discourse-Historical Approach: Wodak (2001).Analytical toolkit i three steps: 1) Context determination and contextualization 2) Topical structure 3) Argumentative schemes.

Historical Context and Contextualization of the PassageExposing at least the following characteristics of each fragment:

  • Historical figures that appear in the passage (if any). Their identification is key to the understanding of the fragment.
  • Event narrated or referred to in the passage.
  • Relationship between the historical figure and the event narrated.
Want een Griec, hiet Macedo, // Wilde up die Drievoudichede Seggen sware ongelovechede, // Ende seide dat al openbare, Dattie Heilege Gheest geen God en ware. // In Spaengen begonste upgaen Een meester, hiet Prissiliaen. Het ware al een, seide die gone, // Vader, Heilech Gheest ende Sone, Een persoen wert ende God. (SH Derde Partie. Van Maximuse ende ander dinc. XXVII, 61)

Topical structureWodak's principle of sequential analysis (Wodak 2001: 85) +Thematic or topical-actional analysis

Argumentative schemesBased on Argumentative Theory (Walton et al. 2008), not on CDA:

  • Argument ad verecundiam.
  • Witness testimony argument.
  • Argument ad populum.
  • Argument ad antiquitatem..
  • Argument ad exemplum.
  • Argument per analogiam.
  • Composition/division argument.
  • Argument ad crumenam.
  • Ad hominem argument.
  • Ethicist argument.
  • Need for aid argument.

Steps and operators I

Steps and operators II


  • Negative perception of Spain and the Spanish
  • The Spanish are characterized as a people only partly Christia
  • But the Spanish are Western people
  • Spain is wealthy

Why these results?The Holy Empire and the Dutch problably saw the peninsular kingdoms as a possible opponent to the Germanic throne... ....even more so when Portugal, Leon, Castile and Aragon were the only European kingdoms that reconquered territories in a lasting way from Islam, unlike the Central European crusaders.

This part of the presentation is based on Sánchez Prieto (2022). Click on this link to read it. You will also find there the quoted references.