The distribution of neutral, negative and positive connotations of the word "guerra" in Italian daily newspapers across three time periods: 1901-1922, 1923-1945, 1946-1967

RINAUDO Stefano (MAT 1166089) CORPUS LINGUISTICS TERM PAPER The distribution of neutral, negative and positive connotations of the word “guerra” in Italian daily newspapers across three time periods: 1901-1922, 1923-1945, 1946-1967 1) Introduction 1.a) Interest in research This research project is grounded on the belief that applying corpus linguistics methods to historical studies can provide quantitative support to qualitative theories resulting in a less biased understanding of past societies. In fact, my study stemmed from the analysis of the positive Italian fascist rhetoric concerning war described by Emilio Gentile in Il culto del littorio (Gentile, 2001). My main goal was to provide quantitative data to Gentile’s theories on the fascist “myth of violence” and to chronologically expand his research scope. In his book, Gentile describes how, during the ventennio fascista, Mussolini succeeded in creating and consolidating a new positive myth of violence that was taught in schools and publicized in Italian newspapers controlled by PNF censorship (especially after 1924). The propaganda described war as the ultimate mission for the “Italian race” and promoted a positive view of the national conflicts through the glorification of the fascist martyrs of the First World War, the militarization of the society and the creation of the need of establishing a new Roman empire. In the fascist newspapers, violence and war were portrayed as the solution to most of the problems of the Italian people. The creation of Minculpop (May 1937) marked a turning point in war propaganda carried out through the newspapers “of the many organizations of the PNF or dependent on the PNF” with the aim of “spreading its ideology through a judicious orchestration of themes and interpretations of the past, present and future” 1 (Gentile, 2004, p. 40). 1.b) Research questions As previously stated, the aim of my research is to supply quantitative data to test Gentile’s theories on the fascist positive war propaganda in newspapers. To do so, I analyzed the distribution of the positive, negative and neutral connotations of the word “guerra” (Italian for “war”) in the daily Italian newspapers across three periods of time: the First World War and pre-fascist period (1901-1922); 1 I decided to translate the original Italian text into English to make the text more fluent and readable. from the ventennio fascista to the end of the Second World War (1923-1945); and the post Second World War period or the Cold War period (1946-1967). My first hypothesis is that the data will show that the (both positive and negative) connoted use of the word “guerra” was sensitively more frequent than the neutral use of the same word during the ventennio fascista. In fact, I expect the aggressive war propaganda in newspapers influenced the political discourse in both fascist and anti-fascist periodicals, resulting in a polarization of ideas about war. Secondly, I hypothesize that the data will show an appreciable increase of the distribution of the positive connotation of the word “guerra” between 1923 and 1945 followed by a significant decrease in the third time window analyzed. In fact, my assumption is that, during the ventennio fascista, the Minculpop censorship limited the voices of Italian people contesting the war. On the contrary, I expect an increase of the negative uses of “guerra” during the Cold War period characterized by the threat of the atomic bomb and the rise of the pacifist and hippie movements. I therefore aim to demonstrate a significant change in the distribution of positively and negatively connoted uses of the word across the three analyzed time periods. 1.c) DiaCORIS For retrieving the linguistic data, I chose the DiaCORIS (diachronic corpus of written Italian). This corpus was developed by the Alma Mater Studiorum – Università di Bologna together with the Accademia della Crusca and the Università di Modena e Reggio Emilia. Its aim is to integrate CORIS (reference corpus for contemporary written Italian) in a diachronic perspective: it contains texts from the Unification of Italy in 1861 until 2001, corresponding to the temporal beginning of the CORIS monitor corpora and thus constituting a temporal continuum that covers the historical evolution of the Italian language from Unification until today (Tamburini, 2022). All texts in the DiaCORIS corpus can be classified as "written-written" Italian, according to the categories established by Giovanni Nencioni (1983). The time interval covered by the DiaCORIS was divided into five sections, each containing 5 million words, resulting in a total of 25 million words. Each section reflects the structure of the textual macrovarieties in CORIS: press, fiction, essayistic prose, legal-administrative prose, and miscellanea. The proportions of these sections vary over time based on the significance of each textual macro-variety in the historical period being studied, as shown in Table 1. The growing size of the "press" section, for example, reflects the rising significance of mass media texts within Italian society during XX century, driven by an increase in both newspaper production and readership. Table 1 - (Tamburini, 2022) Thus, my motivations for selecting DiaCORIS included its well-researched division into periods that correspond to different stages of socio-cultural development in Italy, the extensive collection of newspaper articles, and the creators' goal of developing a balanced corpus that accurately reflects the complexities of XX century Italian press. In fact, in order to take into account the socio-cultural factors influencing each time period, the authors of the corpus included “texts from various geographic areas (e.g.: from newspapers of different cities, such as the Corriere della sera [Milano], La Stampa [Torino], La Nazione [Firenze], Il Mattino [Napoli], Il Giornale di Sicilia [Palermo]), or characterized by different ideological and political attitudes […]. [For example,] in the third [chronological] subcorpus (“Fascism”), particular attention has been given to obtaining representative texts for three balanced sub-sections: “Generic/Regime Press”, “Underground/Opposition Press” and “Catholic Press” (Onelli et alii, 2006). Another important feature of DiaCORIS is the possibility to show the searched word or lemma in a very large context. This was essential for my research because annotating the positive, negative or neutral connotation of the word “guerra” wouldn’t have been possible without context. 1.d) Methodology My investigation within the DiaCORIS followed these steps. I searched the "newspaper" sub-corpus, filtering my results for each chronological sub-corpus across the three time periods relevant to my study. Each time, I inserted the query "guerr[ae]" showing 100 random concordances for each period. Then, I annotated the positive, negative or neutral connotation of each token “guerra” following the criteria shown in Table 2: Table 2 - Annotation guidelines For each token, I also annotated the newspaper where the word “guerra” was found, the year and the author of the article. The result is a corpus of 300 instances of the word “guerra” (or its plural “guerre”). Each temporal sub-corpus thus consists of 100 annotated tokens. I opted to annotate the same number of tokens for each temporal sub-corpus, ensuring equal representation for each period. Once my corpus was created and annotated, I moved on to the statistical analysis of the data through the R programming language. Firstly, I conducted a statistical analysis of the changes in the proportions of neutral and non-neutral (positive and negative) uses of the word “guerra” across the three time periods. Subsequently, I addressed my second hypothesis, aiming to demonstrate a significant change in the distribution of positively and negatively connoted uses of “guerra” across the analyzed time periods. 2) Research outcomes 2.a) The distribution of neutral and non-neutral uses of “guerra” across the three periods of time In order to accept my first hypothesis, I needed to test the corresponding null hypothesis stating that the proportion of neutral and non-neutral uses of the “guerra” does not significantly change across the time periods under consideration. I started the statistical analysis creating a contingency table displaying the independent variable on the horizontal axis and the neutral and non-neutral connotations on the vertical axis. The table shows the frequency of the tokens for each non-overlapping category. 1901 – 1922 1923 – 1945 1946 – 1967 Neutral connotations 46 34 43 Non-neutral connotations 54 66 57 Total 100 100 100 Table 3 - Contingency table of neutral and non-neutral connotations For better visual understanding, I represented the data in two bar plots: one with stacked frequencies (Graph 1) and the other with non-stacked data (Graph 2). Having 100 annotated tokens available for each time period, calculation of proportions was not relevant. Legend: ■ Neutral uses of the word "guerra" ■ Non-neutral uses of the word "guerra" (positive and negative) Graph 1 – Stacked bar plot of neutral and non-neutral connotations Graph 2 – Non-stacked bar plot of neutral and non-neutral connotations The graphical representation suggested a slight increase in the frequency of non-neutral connotations in the second period of time, as predicted in the first hypothesis. Even if in all the periods the non- neutral uses are more frequent than the neutral ones, it’s during the ventennio fascista that the nonneutral connotations reach their peak with 66%. In fact, for 1901-1922 the data report a 54% of nonneutral uses and also for the Cold War period this figure stops at 57%. However, in order to assess the statistical relevance of the increase of non-neutral uses in the second time period, the effect size and the degree of dependency between the two variables had to be computed. In other words, I had to investigate whether the use of non-neutrally connoted “guerra” significantly varied across the three time periods. The effect size was thus computed through the assocstats() function of the vcd R library. The resulting Cramer's V value is 0.104 indicating a small effect size and thus a weak statistical association between the variables analyzed. Subsequently, in order to assess the relationship between the variables, a chisquared test was conducted through R chisq.test() function. The results indicated that there was no significant association between the variables (χ²(2) = 3.22, p = 0.1994). In fact, the p-value exceeded the conventional threshold of 0.05, suggesting that the null hypothesis couldn’t be rejected. These tests led to the conclusion that, although the texts in the corpus from 1922 to 1945 show a slight increase in the number of non-neutral uses of the word “guerra,” this trend is not statistically significant. An analysis of the potential reasons for this result will be addressed in the “Discussion of Results and Conclusions” chapter of this paper. 2.b) The distribution of positively and negatively connoted uses of “guerra” across the three periods of time. According to the second hypothesis, among the non-neutral uses of “guerra” in the Italian daily newspapers, I expected to encounter a significant increase of the positive connotations of this word during the years 1923-1945 and a significant increase of the negative connotations in the years 19461967. I therefore wanted to prove the presence of a significant change in the distribution of the positively and negatively connotated uses of the word across the three time periods analyzed. In order to do that, I started by creating a contingency table reporting the absolute frequencies of the positive and negative uses of “guerra” (Table 4). 1901 – 1922 1923 – 1945 1946 – 1967 Negative connotations 25 35 49 Positive connotations 29 31 8 Total 54 66 57 Table 4 - Contingency table of negative and positive connotations Subsequently, I created a table of proportions (Table 5) through the prop.table() function in R. This table shows the percentages of negative and positive connotations for each time period, making the data significantly more informative in addressing the research question. 1901 – 1922 1923 – 1945 1946 – 1967 Negative connotations 46,3% 53,0% 86,0% Positive connotations 53,7% 47,0% 14,0% Total 100% 100% 100% Table 5 – Table of proportions of negative and positive connotations Next, I decided to visually represent the data through two bar plots. In the first one (Graph 3), I displayed the the absolute frequencies of negatives and positives uses of “guerra” across the three time periods. The second is a stacked bar plot representing the variation of the percentages of the same data. Legend: ■ Negative uses of the word "guerra" ■ Positive uses of the word "guerra" Graph 3 - Bar plot of the absolute frequencies of negatives and positives uses of “guerra” Graph 4 - Stacked bar plot of the proportions of negatives and positives uses of “guerra” The graphical representations reveal a clear disproportion for the period 1946-1967, during which the word “guerra” was used with a negative connotation 86% of the time, compared to just 16% for positive connotations. On the other hand, we observe that in the fascist period the negative use of the word is slightly more frequent (53%) than the negative one (47%). This data becomes even more surprising if compared to the first time period: here the positively connoted tokens are slightly more frequent than the negative ones with a frequence of 53,7%. As for the data in chapter 2.a, it was essential to statistically evaluate the effect size and the degree of dependency between the variables. For this reason, I computed the effect size using the assocstats() function in R. This time, the resulting Cramer’s V value was 0.35. This indicated a moderate association between the connotation of “guerra” and the chronological independent variable. I then computed the chi-squared test through R chisq.test() function. The results revealed a significant relationship, with Chi-squared (X²) = 21.697, degrees of freedom (df) = 2, and a p-value < 0.0001 (p = 1.943e-05). The outcomes of these tests proved that the connotation of the “guerra” varies notably across the three time periods analyzed, implying that the null hypothesis had to be rejected. Having assessed a relation between the two variables, it was interesting to see which frequencies in the contingency table were greater than what the expected, and which were smaller. In order to have a clear vision of the outlayers, I checked the Pearson residuals computing the chisq.test()$residuals function on R. 1901 – 1922 1923 – 1945 1946 – 1967 Negative connotations -1.431375 -0.8853065 2.345837 Positive connotations 1.812226 1.1208627 -2.970001 Table 6 – Table reporting the Pearson residuals for each cell From Table 6, we can already notice that the data that stray furthest from the expected are the those of the third period of time. In fact, the table shows that the negatively connoted uses are sensibly more frequent than the expected, with a positive residual of 2.345837. Also, the frequence of the positively connoted uses of “guerra” in the third time period present an even greater discrepancy with respect to the expected frequency, with a negative residual of -2.970001. This data was visually represented in a mosaic plot (Graph 5) that I obtained through the mosaic() function from the package vcd in R, and in an association plot (Graph 6) that was built through the assocs() function found in the same R library. The mosaic plot and the association plot confirm that the only period of time where there’s a significant discrepancy from the expected data are the years between 1946 and 1967, when we assist to a strong increase of the negative uses of “guerra” and an even more important decrease in the frequency of the positively connoted uses of the word. On the other hand, the second period has the data that deviates the least from the expected values, as visually shown in the association plot. Graph 5 - Mosaic plot of negative and positive uses of “guerra” across three time periods. The color of shading corresponds to the sign of a residual, and the intensity shows its relative importance. Graph 6 - Negative and positive uses of “guerra” across three time periods: association plot of residuals. The color of shading and its orientation upor downward corresponds to the sign of a residual, and the intensity of shading shows its relative importance. 3) Discussion of results and Conclusions Concerning the first hypothesis, I came to the conclusion that, despite my assumptions, during the ventennio fascista the connoted use of the word “guerra” was not sensitively more frequent than the neutral use of the same word. Nevertheless, this does not necessarily invalidate Gentile’s theory regarding the polarization of discourse about war in newspapers. A deeper reflection on this inquiry reveals that analyzing the stance of a single word cannot provide a sufficiently precise understanding of such a complex social phenomenon. Additionally, the increase in positively or negatively connoted uses of “guerra” does not imply a corresponding decrease in its neutral use. In fact, returning to the annotation guidelines (Table 2), one can observe that some neutral uses are part of phrasemes, fixed expressions, or constructs indicating temporality, such as “corrispondente di guerra”, “dopo guerra”, or “navi da guerra”. This means that the neutral and non-neutral uses of “guerra” are not completely interchangeable. In fact, the writer’s choice to use one instead of the other also depends on the context and the function of the term. In other words, as there is no perfect overlap between the two usages, they shouldn’t be statistically compared without having first filtered out the neutral variants of “guerra” that are not perfectly interchangeable with the positive and negative uses. As for the second hypothesis, our study confirmed that, between 1946 and 1967, on the Italian daily newspapers the negative use of the word “guerra” noticeably increased. As a result, the initial assumptions regarding the widespread fear of nuclear war and the influence of cultural movements