Iberica 13 Ibérica 40 (2020): 217-244 ISSN: 1139-7241 / e-ISSN: 2340-2784 Abstract The present work carries out a discursive analysis of two 2019 institutional reports on the world’s economy and their Spanish translations, and endeavours to determine the role that emotion plays in the way each text is lexically verbalized. A prior lexical analysis is carried out to obtain a 60-keyword list used to create a lexical taxonomy which may allow comparison of the proportion of –emotion-free– technolects and that of emotion-laden items conceptualizing the economic situation with a positive or negative polarity in either text. Subsequently, automatic processing is conducted with Lingmotif (moreno-Ortiz, 2017), a Sentiment Analysis software to establish the positive and/or negative polarity and the degree of emotional intensity in the texts. The analysis reveals how there exist rhetorical-linguistic differences in the way emotion is deployed between the reports in line with the goals of each institution, and that these differences are mirrored in their Spanish translations. Keywords: economic discourse, emotion, Sentiment Analysis, economic lexicon, English, Spanish. Resumen La emoción para pronosticar una recesión: análisis bilingüe léxico y de sentimiento de los informes económicos mundiales de la ONU y del FMI para el año 2019 Este trabajo lleva a cabo un análisis discursivo de dos informes institucionales de 2019 sobre la economía mundial y sus traducciones al español, y se propone Emotion to forecast a recession: A bilingual lexical and sentiment analysis of the un and imf world economy reports for 2019 María Ángeles Orts Llopis Universidad de Murcia (Spain) mageorts@um.es 217 Ibérica 40 (2020): 217-244 mAríA ÁnGElES OrTS llOPiS determinar el papel que desempeña la emoción en la forma en que cada texto está verbalizado léxicamente. Se ha realizado previamente un análisis del léxico que ha permitido obtener una lista de 60 palabras clave utilizadas para crear una taxonomía léxica capaz de permitir la comparación de la proporción de tecnicismos libres de emoción y la de elementos cargados de emoción que conceptualizan la situación económica con una polaridad positiva o negativa en cada texto. Posteriormente, se ha realizado un procesamiento automático con Lingmotif (moreno-Ortiz, 2017), un software de análisis de sentimiento que establece la polaridad positiva y/o negativa y el grado de intensidad emocional de los textos. Dicho análisis revela la existencia de diferencias retórico- lingüísticas en la forma en que se emplea la emoción en cada uno de los informes que están relacionadas con los objetivos comunicativos de cada institución y que quedan reflejadas en sus traducciones al español. Palabras clave: discurso económico, emoción, análisis de sentimiento, léxico económico, inglés, español. 1. Introduction The world economy is, nowadays, a topic of global concern, particularly since the summer of 2008, the date marking the onset of the Global Systemic Crisis. During the years that have followed, economy has evolved from being a subject only for connoisseurs to becoming a topic of general interest, as the evolution of global finances is warily monitored by experts and followed by laypeople through traditional media and the social networks. Since the outbreak of the crisis, though, the factors that make economy a matter of general interest, and even of anxiety, seem to have become more sophisticated. Political and social events –such as the advent of the Trump administration in the uSA, the Euro crisis and Brexit in Europe, the changing role of emerging economies and the increasing meteorological risks due to climatic change in the world at large– have an influence on global markets and their performance, and have been identified as “market sentiment indicators”, these defined as “a combination of all investors’ views, opinions and feelings”1. Consequently, trade policy controversies, financial instabilities and high debt levels constitute the focus of interest of both specialists and laypeople. in tune with this concern, as an aftermath of the previous crisis and at the threshold of a new 2020 mayhem, think tanks such as the imf and the un endeavour to warn the world at large, but mainly policymakers, of the potential dangers that loom in the future by means of recurrent reports on 218 the economic prospects for the years and months to come. A corpus of such reports constitutes the materials under scrutiny in the present work. if emotions in language is a flourishing field in areas such as Appraisal Theory and Pragmatics (foolen, 2012; Schwarz-friesel, 2015; mackenzie & Alba-Juez, 2019, just as a few examples), with the exception of some studies by social scientists and anthropologists (Berezin, 2005, 2009, among others) it is by no means usual to associate emotion and economy. in fact, economics has traditionally been regarded as a rational, regular science, and the antithesis of emotion and sentiment (Bandelj, 2009). And even if the technicalisation of the economic profession is ever increasing (Biel & Sosoni, 2017), which favours the coinage of new terms and the opacity of its discourse, the language of economics and finance is, nonetheless, quite close to human sentiment, plagued as it is with metaphors indicating a range of emotions such as fear and trust, where fear is connected to market anxiety and panic, and trust is connected to market confidence (Charteris-Black, 2004; Charteris-Black & Ennis, 2001; Herrera Soler & White, 2012). Hence, the present analysis is based upon the hypothesis that sentiment in economic language is expressed with items of emotion-laden value –words that indicate a process of assignment of sentiment on the part of the addresser of the text– which, in economics, take the shape of “discourse metaphors”. Zinken defines discourse metaphors as “verbal expressions containing a construction that evokes an analogy negotiated in the discourse community” (2007: 445). Along these lines, for musolff and Zinken (2009) and Zinken et al. (2008), metaphor is first a mode of argumentation, framing arguments and highlighting discursive developments. Hence, discourse metaphors shall be understood here as hackneyed metaphorical projections prototypical of some genres which present a lexicalized form with a relatively fixed meaning (Garrido, 2013). it is, then, our hypothesis that the type of genre under analysis, the institutional economic report, is not only colonised by emotion-free technolects, the terms of art of the profession, but also by emotion-laden items, liable to generate the basic sentiments depicted in Plutchik’s model (1980) –joy-sadness, anger-fear, trust-distrust and surprise-anticipation. These emotion-laden words and expressions have the potential to turn rational or, rather, emotion-free, concepts like the human body, objects, substances, as well as structure abstract concepts, into emotion-triggers. A plethora of studies within the realm of linguistics applied to professional languages are explanatory of diverse aspects of metaphoricity, and some EmOTiOn TO fOrECAST A rECESSiOn: A BilinGuAl lExiCAl AnD SEnTimEnT AnAlySiS Ibérica 40 (2020): 217-244 219 approach the subject of economic metaphors with heuristic explanations, endeavouring to explain how abstract conceptualizations of economy, growth, market, market movements, etc. are made more comprehensible through images (Silaški & Đurović, 2010; Skorczynska, 2010). But, conversely, other authors (Kövecses, 2003, 2010; Charteris-Black, 2014) demonstrate that, in economic language, metaphorical images are linked to the rhetorical concept of pathos, that is to say, the heightening of emotional impact. Specifically, a series of studies demonstrate how the economy is often portrayed as a human being who suffers maladies, describing economic problems as objects subject to breakdowns or erosion (Orts & rojo, 2010; Schäffner, 2012; Arrese, 2015, just to mention some). in the course of our analysis we will see how “source concepts are organized into mininarratives or ‘scenarios’ that dominate the discourse manifestations of source domains”, as defined and developed by mussolf (2006: 23). We, then, seek to explicitly measure the expression of emotions in economic language with the systematic study of emotion-laden items and discourse metaphors, aiming to capture how economic problems and their solutions are described, where such lexical elements map the economic scenario with a positive or negative polarity. in a second stage, this study also aims to deploy Sentiment Analysis software in the attempt to capture the receiver of the message’s attention with a mixture of technical data and emotional messages. Sentiment Analysis, or SA, constitutes the germ of extensive literature destined to the description of different software designs for computational linguistics and automatic analyses of texts (Taboada, 2016). in the area of economics, SA has provided new possibilities to unveil some of the complexities of financial markets, and as one of the instruments provided by Ai (Artificial intelligence), researchers have deployed its algorithms to analyse the trends of stock markets, aiding companies and investors in the fields of decision-making and risk management (Devitt & Ahmad, 2007; ruiz-martínez et al., 2012; uhr et al., 2014, among others). nevertheless, to our knowledge, in Applied linguistics and, specifically, in Genre Analysis, this type of software has not been used to provide (as this study endeavours to evidence) interesting connections between words and the emotional subtext of a professional genre such as the institutional report. Hence, we are poised to demonstrate how the combination of quantitative and qualitative linguistic analyses is very useful to shed some light on how it is that sentiment on similar information contents (a financial account on the state of the world’s mAríA ÁnGElES OrTS llOPiS Ibérica 40 (2020): 217-244220 economy in 2019) is expressed differently, according to the role and scope of the organization behind it. 2. Materials and methods The role that textual typology can play is very relevant when assessing the presence of potentially emotion-laden lexical items in economic discourse. As loma-Osorio explains (2004), economic texts addressed both to expert and non-expert receivers perform informative and persuasive functions, constituting a channel to disseminate the technolects that strictly belong to the monosemic, expert-to-expert scope of communication, as well as innovative coinages of new discourse metaphors and neologisms. Specifically, the institutional financial reports currently submitted to analysis constitute a means to spread economic information while describing the global economic panorama emotionally enough to exert some kind of response from the readers (mainly policymakers and economic actors). Our parallel corpus consists of two 2019 reports on the state of the world economy in their original English versions and their Spanish translations. The first group of subcorpora is made up of two versions of the “World Economic Situation and Prospects 2019”, issued by the un in the course of the abovementioned year. The English version (henceforth, WESP) and its Spanish translation (hereafter, WESPES) are, indeed, a product of the united nations Department of Economic and Social Affairs, working together with a group of un Commissions specialized in the scrutiny of economic analysis and policy in several parts of the world. WESP is defined by the institution as “a flagship publication on the state of the world economy”2, monitoring macroeconomic trends and strategies for a large number of economies worldwide. As its name suggests, apart from economic issues, the un Department of Economic and Social Affairs not only aims to offer hard and fast economic information, but it monitors social and environmental dimensions of sustainability, seeking the implementation of the 2023 Agenda for Sustainable Development. The second group of subcorpora is offered by the imf. The imf’s World Economic Outlook update for January (henceforth WEO in its English original)3 is one of the reports that the organization issues yearly, whose official translation into Spanish, as offered by the institution, will be referred to as WEOES. EmOTiOn TO fOrECAST A rECESSiOn: A BilinGuAl lExiCAl AnD SEnTimEnT AnAlySiS Ibérica 40 (2020): 217-244 221 Contrary to WESP, WEO is mainly concerned with economic affairs worldwide, since the imf’s primary purpose is the vigilance of the international monetary system, supervising issues in the macroeconomic and financial scope that have an influence on the stability of global and financial aspects. in this sense, and in tune with the role of the imf as a watchdog of international economy, the report will presumably concentrate upon macroeconomic affairs and emphasise technical information. Table 1 summarizes the lexical structure of the different subcorpora after an examination with the Lingmotif software (moreno-Ortiz, 2017): in the table we can see how the Spanish subcorpora are longer than their English originals. This confirms what was stated by Vázquez Ayora (1977: 336) and Berg Selisson (2012: 120), that Spanish translations from English are systematically longer than their originals for a number of reasons, such as the fact that many one-word English constructions are translated into phrases of two or three words; English, in contrast, is a much more economical language, which translators usually tend to amplify when rendering their Spanish versions. Hence, WESPES consists of 3496 tokens, from 1057 types, structured in 106 sentences which contain 1600 lexical items and 1614 grammatical items functionally speaking, and, formally speaking, 3170 single words and 292 multiwords. On the other hand, WEOES (the imf report in Spanish) is an even larger subcorpus, with 4385 words or tokens and 1203 forms or types. it should be noted that lexical richness is, comparatively, greater in the un subcorpora than in the imf ones, where there is a higher number of words repeated frequently. This can be seen in the 30.23% token/type ratio in WESPES and the 27.43% in WEOES. The number of sentences is lower in WESPES, which again indicates a higher sentence length in the imf subcorpus. The study, as we anticipated in our previous section, consists of two phases: the first is purely lexical and carries out automatic processing by means of mAríA ÁnGElES OrTS llOPiS Ibérica 40 (2020): 217-244222 T C T Corpus Sentences Tokens Types T/T Ratio Items Lex. Words Func. Words Single words Multi words WESP 106 2623 915 34.88% 2524 1600 924 2389 92 WEO 137 3454 1044 30.22% 3375 2122 1253 3340 115 WESPES 106 3496 1057 30.23% 2524 1600 1614 3170 292 WEOES 137 4385 1203 27.43% 3375 2122 1970 4092 293 Table 1. Lexical structure of WESPES and WEOES, according to Lingmotif (Moreno-Ortiz, 2017). I Antconc 3.5.8. (Anthony, 2019), a software for the extraction of data. Antconc 3.5.8. is a text analysis program that allows lists of words, their frequencies and corpus keywords to be collected from textual corpora, allowing comparisons between frequencies, placements, n-grams (ergo, more or less fixed verbal sequences) and large-scale concordances. The objective of this phase was to obtain a lexical classification of the four subcorpora with a list of keywords indicating the relevance and frequency of a) emotion- free economic technolects or b) emotion-laden discourse metaphors disclosing positive or negative feelings in the description of the global economic scenario. lexical units were marked as metaphorical when the semantic tension created between the contextual meaning in which the unit is used and its basic meaning (as defined by our source of reference, in this case unTErm, the united nations Terminology Database4), can be explained in terms of cross-domain mapping (Steen, 2009: 36-37). for example, the word ‘burden’ generally refers to a serious or difficult obstacle. in economy, the term points to a second, metaphorical meaning: the responsibility to pay back an excessive amount of money. likewise, the word ‘hike’ is “a long walk in the countryside”, whereas its metaphorical meaning points to “a sudden increase in the amount or level of something”. To obtain the keyword list, an ad-hoc reference corpus was built for each language, consisting of previous versions of the un and imf reports combined provided for 2017 and 2018, containing 179,450 tokens (in the English reference corpus) and 216,970 tokens (in the Spanish reference corpus). The usage of a similar reference corpus is in line with Garofalo’s work (2017) and was aimed at underlining the differences in content and lexical choices between past editions of the reports and the ones subject to scrutiny, highlighting how singularly global market conditions are depicted by each of the think tanks in the present economic scenario. in a second phase, SA processing was used through Lingmotif v.1.0 software (moreno Ortiz, 2017). in this case, the purpose was to study the sentiment polarity (positive, negative or neutral) and the intensity of feeling (neutral to extremely intense) in each of the texts. using the results of the first phase, a complementary plugin or lexicon (one in English, one in Spanish) was added to the software, consisting of a list of emotion-laden terms, mostly discourse metaphors, as the emotion-laden items in the corpus. With the aid of the plugin, the texts were processed again to observe the differences in emotional involvement of the different segments within each text. EmOTiOn TO fOrECAST A rECESSiOn: A BilinGuAl lExiCAl AnD SEnTimEnT AnAlySiS Ibérica 40 (2020): 217-244 223 According to the hypotheses here formulated, it is to be expected that there may be, in different measures in each subcorpus, a) both emotion-free technolects (the specialised terminology of the field) and emotion-laden items (terms with a sentiment value), and also b) different polarities of feeling in both texts, and different intensities of emotion, probably in tune with the social purpose that each institution wishes to accomplish. 3. Results 3.1. Phase one. The lexical study The method followed for the use of this software rendering a description of the lexical map in the imf and un subcorpora in each language entailed the elaboration of a list of stopwords, or those with grammatical, but without lexical meaning. A stopword list was made in English and in Spanish to filter each subcorpus in turn, so as to eliminate the elements void of lexicality. On the other hand, so as to further restrict the framework of analysis, a list of lemmas was also incorporated in either language to cluster all the morphological variants of the different types. Table 3 presents the results obtained by word keyness for either English subcorpus. Due to the length of the corpus overall and the space restrictions of this work, only the first 60 keywords were chosen, many of which will be discussed below. Table 2 shows, then, each term in order of keyness and their relative frequency, since the subcorpora were different in size. mAríA ÁnGElES OrTS llOPiS Ibérica 40 (2020): 217-244224 EmOTiOn TO fOrECAST A rECESSiOn: A BilinGuAl lExiCAl AnD SEnTimEnT AnAlySiS Ibérica 40 (2020): 217-244 225 WESP Order Keyness Relative frequency (f%) Term 1 1,60 341.059 develop 2 1,98 245.751 economy 3 0,72 83.235 sustainable 4 0,80 80.591 risk 5 0,34 70.420 prospect 6 0,34 54.745 poverty 7 0,99 53.122 global 8 0,76 45.092 trade 9 0,91 37.623 policy 10 0,30 37.529 multilateral 11 0,19 36.561 dispute 12 1,45 36.180 growth 13 0,23 36.084 agenda 14 0,15 35.914 escalate 15 0,15 35.914 tension 16 0,30 32.948 inequality 17 0,95 32.087 country 18 0,15 30.932 condition 19 0,27 30.772 longer 20 0,19 28.750 diversification 21 0,15 28.321 fundamental 22 0,34 27.026 infrastructure 23 0,11 26.935 eradicating 24 0,11 26.935 poses 25 0,11 26.935 steep 26 0,69 26.405 financial 27 0,15 24.914 rural 28 0,23 24.388 vulnerability 29 0,15 23.661 vital 30 0,23 21.129 environment 31 0,15 20.818 inclusive 32 0,27 20.376 social 33 0,30 20.359 levels 34 0,11 20.250 resource 35 0,42 20.195 debt 36 0,19 19.514 require 37 0,15 19.388 reach 38 0,38 19.013 high 39 0,19 18.772 progress 40 0,30 18.742 rise 41 0,30 18.742 significant 42 0,11 18.685 institution 43 0,30 18.592 large 44 0,08 17.957 contagion 45 0,08 17.957 delivering 46 0,08 17.957 disorderly 47 0,08 17.957 fail 48 0,08 17.957 matched 49 0,08 17.957 obligation 50 0,08 17.957 power 51 0,08 17.957 redressing 52 0,08 17.957 responsive 53 0,08 17.957 retaliation 54 0,08 17.957 transparency 55 0,15 17.657 target 56 0,11 17.465 concern 57 0,11 17.465 fragility 58 0,11 17.465 heightened 59 0,11 17.465 severe 60 1,60 17.465 tariff Table 2. Keywords in WESP and WEO. I WEO Order Keyness Relative frequency (f%) Term 1 1,30 191.648 economy 2 0,41 122.378 tension 3 0,43 111.664 tariff 4 0,35 97.870 condition 5 0,29 66.395 develop 6 0,32 65.935 sentiment 7 0,75 64.557 trade 8 0,26 62.725 concern 9 0,49 55.339 risk 10 0,26 52.639 tighten 11 1,36 50.617 growth 12 0,55 47.090 price 13 0,72 44.538 global 14 0,14 43.706 escalate 15 0,17 43.501 prospect 16 0,78 42.117 policy 17 0,69 41.323 financial 18 0,29 41.113 weak 19 0,23 39.785 projection 20 0,14 38.325 brexit 21 0,23 35.112 fall 22 0,14 33.198 barrier 23 0,38 30.152 oil 24 0,26 28.371 uncertainty 25 0,14 27.584 spreads 26 0,17 27.073 sovereign 27 0,69 26.360 market 28 0,09 26.224 dispute 29 0,09 26.224 shutdown 30 0,12 25.481 aversion 31 0,12 25.481 resolve 32 0,38 23.909 projected 33 0,29 23.272 forecast 34 0,32 22.501 domestic 35 0,09 21.751 softening 36 0,12 21.658 elevated 37 0,12 21.658 regulatory 38 0,14 21.061 unchanged 39 0,35 20.227 expected 40 0,52 19.982 rate 41 0,09 19.545 emission 42 0,12 19.146 notably 43 0,35 18.409 lower 44 0,09 17.982 burden 45 0,14 17.655 revisions 46 0,06 17.483 passthrough 47 0,06 17.483 shadow 48 0,06 17.483 spark 49 0,06 17.483 truce 50 0,14 17.309 slower 51 0,12 17.276 industrial 52 0,09 16.765 contraction 53 0,09 16.765 hikes 54 0,12 16.746 momentum 55 0,14 16.049 slow 56 0,09 15.766 weighed 57 0,12 15.356 announced 58 0,23 15.312 activity 59 0,14 14.953 slowdown 60 1,30 14.947 range in the table some of the words have been highlighted in bold (those indicating economic technolects, or terms with a standardized, specific usage in economics, according to unTErm), red (the emotion-laden items with a metaphorical value), green (those lexical items labelled ‘social terms’, expressing or verbalizing social involvement) and blue, marking intensifiers or softeners. if we take either subcorpus at hand, we can see that among the first 60 keywords in WESP only 4 are technolects (‘economy’, ‘trade’, ‘financial’ and ‘tariff ’). WEO, in contrast, duplicates that number, with 8 terms of art: ‘price’, ‘economy’, ‘tariff ’, ‘trade’, ‘financial’, ‘spread’, ‘market’ and ‘sovereign’. Words of a social character were found in WESP, such as ‘sustainable’, ‘environment’ and ‘social’, which an Antconc 3.5.8. search rendered as having a positive polarity (as in ‘sustainable growth/development/finance’ or ‘social protection/goals/services’). in WEO, no term with a clear social orientation was found, which seems to be symptomatic of greater narrowness of topic. Once the technolects had been identified, we proceeded to analyse the sentiment items in the list. The selection was undertaken searching for their collocations and concordances in text with the aid of Antconc 3.5.8. (Anthony, 2019) to spot the potentially emotional lexical elements in context. finally, Excel allowed a new numerical compilation of the different semantic values of positive and negative evaluation, this time taking into account their grammatical function. Our Antconc search unveiled 30 items in WESP: 9 positive terms (‘develop’, ‘multilateral’, ‘growth’, ‘diversification’, ‘eradicating’, ‘resource’, ‘progress’ and ‘transparency’) and 14 negative (‘risk’, ‘poverty’, ‘dispute’, ‘escalate’, ‘tension’, ‘inequality’, ‘vulnerability’, ‘debt’, ‘contagion’, ‘disorderly’, ‘fail’, ‘retaliation’, ‘concern’ and ‘fragility’), to which emotion-laden value intensifiers and softeners used to express positive (2) and negative (5) polarities were added. Specifically, the intensifiers ‘steep’, ‘severe’, ‘high’, ‘significant’, ‘large’ and ‘heightened’ were found in negative contexts (as in ‘steep reductions/number of disputes’, ‘severe disruptions’, ‘significant vulnerabilities/slowdown/damage’, ‘large imbalances/risk/ repercussions’ and ‘heightened stress/volatility/tensions’), whereas ‘fundamental’ and ‘vital’ were found always in a positive context (as in ‘fundamental shift/transformation’, or in ‘vital economic transformation’). As far as WEO is concerned, 28 emotion-laden items were uncovered, and the difference in polarity is even more abrupt (23 negative, against 5 positive items): 18 negative nouns were found (‘tension’, ‘concern’, ‘risk’, ‘tighten’, ‘escalate’, ‘weak’, ‘Brexit’, ‘fall’, ‘barrier’, ‘uncertainty’, ‘dispute’, ‘shutdown’, mAríA ÁnGElES OrTS llOPiS Ibérica 40 (2020): 217-244226 ‘aversion’, ‘burden’, ‘shadow’, ‘contraction’, ‘slowdown’ and ‘hike’), plus 5 intensifiers with a negative value (‘elevated’, ‘unchanged’, ‘slower/ slow’ and ‘softening’, as in ‘elevated debt burdens/geopolitical risks’, ‘unchanged yields’, ‘slow global growth/demand’ or ‘softening momentum’), and, finally, 5 positive sentiment words (‘growth’, ‘resolve’, ‘truce’, ‘momentum’ and ‘sentiment’). Table 3 shows these results more explicitly: As we suggested at the beginning of the study, our search for concordances and collocations led to the conclusion that positive and negative items seem to be constellated as discourse metaphors triggering positive or negative responses. Hence, of the negative words found in WESP, ‘dispute’, ‘retaliation’, ‘risk’, ‘escalate’, ‘tension’ and ‘disorderly’ correspond to the framing of economy as a conflict where economic problems (‘poverty’, ‘inequality’) are enemies, as in: (1) Eradicating poverty by 2030 will require both double-digit growth in Africa and steep reductions in income inequality. EmOTiOn TO fOrECAST A rECESSiOn: A BilinGuAl lExiCAl AnD SEnTimEnT AnAlySiS Ibérica 40 (2020): 217-244 227 Lexical taxonomy in WESP Lexical taxonomy in WEO Technolects (4) Social words (3) Technolects (8) Social words (0) ECONOMY, TRADE, FINANCIAL, TARIFF SUSTAINABLE, ENVIRONMENT, SOCIAL ECONOMY, TARIFF, TRADE, FINANCIAL, SPREAD, MARKET, SOVEREIGN ––– Negative sentiment words (19) Negative sentiment words (13) Emotion-laden items Intensifiers and softeners Emotion-laden items Intensifiers and softeners POVERTY, INEQUALITY, TENSION, DISPUTE, RETALIATION, RISK, ESCALATE, DISORDERLY VULNERABILITY, CONTAGION, FRAGILITY, FAIL, DEBT, CONCERN STEEP SEVERE HIGH SIGNIFICANT LARGE HEIGHTENED TENSION, CONCERN, RISK TIGHTEN, ESCALATE, WEAK, BREXIT, FALL, BARRIER, UNCERTAINTY, DISPUTE, SHUTDOWN, AVERSION, BURDEN, SHADOW, CONTRACTION SLOWDOWN, HIKE ELEVATED UNCHANGED SLOWER/ SLOW SOFTENING Positive sentiment words (11) Positive sentiment words (5) Emotion-laden items Intensifiers and softeners Emotion-laden items Intensifiers and softeners DEVELOP, MULTILATERAL, GROWTH, DIVERSIFICATION, RESOURCE, PROGRESS RESPONSIVE, TRANSPARENCY FUNDAMENTAL VITAL GROWTH, RESOLVE, TRUCE, MOMENTUM, SENTIMENT ––– Table 3. Lexical taxonomies in WESP and WEO. A in WEO, in contrast, the array of negative words pictures different metaphorical scenarios. Economy as a conflict is also present, with words like ‘tension’, ‘escalate’, ‘risk’ and ‘dispute’, most also existent in WESP, but the economy is mainly conceptualized as a journey full of obstacles (as with words like ‘Brexit’, ‘barrier’, ‘fall’, ‘burden’, ‘slowdown’ and ‘hike’ –always in reference to interest rates and/or tariffs, and with a negative value), sometimes a problem zone where global economies are humans that suffer (‘concern’, ‘uncertainty’, ‘aversion’, ‘weak’). See, for instance: (2) multiple factors weigh on the region’s outlook, including weak oil output growth, which offsets an expected pickup in non-oil activity. Problems are also seen as objects that break (as in ‘tighten’, ‘shutdown’ and ‘contraction’). Even a word like ‘development’, potentially positive, is used in negative contexts, as in: (3) But these developments occurred against a backdrop of weakening financial market sentiment, trade policy uncertainty, and concerns about China’s outlook. regarding positive scenarios, in WESP there is a conceptualization of economy as a living organism (‘develop’, ‘growth’, ‘responsive’), or as a journey (‘progress’) where economic systems are seen as objects or buildings (‘multilateral’, ‘diversification’, ‘transparency’) whose design can be improved, as in: (4) This also requires progress towards a more inclusive, flexible and responsive multilateral system. in WEO the human organism conceptualization seems to be stronger (‘growth’, ‘resolve’, ‘sentiment’) while framing economy as a conflict scenario (‘truce’) or a journey (‘momentum’) show some, if scarcer, examples. Table 4 below exhibits the findings for the Spanish subcorpora, WESPES and WEOES. mAríA ÁnGElES OrTS llOPiS Ibérica 40 (2020): 217-244228 Table 4. Keywords in WESPES and WEOES. EmOTiOn TO fOrECAST A rECESSiOn: A BilinGuAl lExiCAl AnD SEnTimEnT AnAlySiS Ibérica 40 (2020): 217-244 229 T WESPES Order Keyness Relative frequency (f%) Term 1 1,60 210.707 economía 2 1,14 96.249 desarrollo 3 0,51 90.774 financiero 4 0,83 88.567 política 5 0,57 86.720 riesgo 6 0,60 72.461 comercio 7 0,86 69.733 país 8 0,31 65.206 perspectiva 9 0,20 64.517 tensión 10 0,26 62.880 pobreza 11 0,92 58.381 mundial 12 0,46 57.238 gran 13 0,20 52.360 básico 14 1,14 44.240 crecimiento 15 0,49 41.760 sostenible 16 0,23 39.992 desigualdad 17 0,17 38.785 agenda 18 0,11 36.867 desequilibrio 19 0,14 35.559 condición 20 0,20 35.230 multilateral 21 0,14 33.798 diversificación 22 0,29 33.041 clima 23 0,29 32.800 mundo 24 0,11 31.883 financiación 25 0,49 29.971 nivel 26 0,17 29.839 objetivos 27 0,11 29.269 escalada 28 0,17 29.239 servicio 29 0,09 27.650 arancel 30 0,09 27.650 controversia 31 0,09 27.650 obstáculo 32 0,11 27.366 graves 33 0,14 27.189 eficaz 34 0,11 25.857 formulación 35 0,11 24.602 amenaza 36 0,11 24.602 logro 37 0,26 24.266 podría 38 0,23 23.949 infraestructura 39 0,09 23.172 erradicar 40 0,17 22.474 ritmo 41 0,14 21.048 deben 42 0,11 20.999 tecnología 43 0,26 20.973 medida 44 0,09 20.960 percepción 45 0,09 20.960 produjo 46 0,11 20.316 fundamental 47 0,17 19.972 social 48 0,09 19.393 apoyar 49 0,09 19.393 correcciones 50 0,09 19.393 desencadenar 51 0,09 19.393 riqueza 52 0,14 18.573 vulnerabilidad 53 0,20 18.505 reducir 54 0,06 18.433 afectaría 55 0,06 18.433 imperativo 56 0,06 18.433 obligaciones 57 0,06 18.433 proporcionar 58 0,06 18.433 receptivo 59 0,06 18.433 rurales 60 1,60 18.433 transparencia WEOES Order Keyness Relative frequency (f%) Term 1 1,09 156.933 economía 2 0,34 134.603 tensión 3 0,55 133.463 comercial 4 0,68 126.398 financiero 5 0,27 107.682 arancel 6 0,34 107.239 emergente 7 0,30 101.281 condición 8 0,25 87.591 actitud 9 0,32 84.536 perspectiva 10 0,57 78.832 mercado 11 0,43 73.910 precio 12 0,39 65.055 riesgo 13 0,27 64.896 desaceleración 14 0,57 64.518 política 15 1,07 54.429 crecimiento 16 0,16 53.325 preocupación 17 0,23 52.802 factor 18 0,21 51.289 prevé 19 0,41 47.974 informe 20 0,14 44.889 moderación 21 0,11 44.868 aversión 22 0,21 43.908 proyecta 23 0,14 42.452 restrictivas 24 0,16 42.268 proyección 25 0,25 40.825 previsto 26 0,25 40.141 petróleo 27 0,21 39.305 materia 28 0,14 37.341 soberanos 29 0,62 36.893 mundial 30 0,11 36.537 brexit 31 0,11 36.537 ímpetu 32 0,14 36.055 expectativa 33 0,18 32.063 pronóstico 34 0,14 29.723 endurecimiento 35 0,14 29.053 deterioro 36 0,09 28.301 cierre 37 0,09 28.301 próximos 38 0,21 28.180 incertidumbre 39 0,25 26.981 medida 40 0,07 26.921 disputa 41 0,09 26.401 vehículos 42 0,11 25.242 barreras 43 0,11 25.242 nuevos 44 0,09 24.894 levemente 45 0,14 24.360 revisiones 46 0,09 23.642 subida 47 0,25 22.945 baja 48 0,07 22.444 modere 49 0,07 22.444 títulos 50 0,14 21.101 ritmo 51 0,16 20.564 banco 52 0,07 20.236 obedece 53 0,11 19.491 diferenciales 54 0,09 19.369 finales 55 0,11 19.112 semestre 56 0,07 18.671 arancelarias 57 0,07 18.671 flojo 58 0,07 18.671 sufrido 59 0,09 18.171 pasado 60 1,09 17.947 tregua Among the 60 items on the list, WESPES shows 5 technical terms (economía, financiero, comercio, financiación, arancel), as opposed to 11 in WEOES (economía, comercial, financiero, arancel, mercado, precio, soberanos, títulos, diferenciales, banco and arancelarias). The Spanish version of the imf report shows, then, more technical terms than any other subcorpus. in contrast, three words of a social character (sostenible, clima, social) were found in the un subcorpus, but as it happened with WEO, no word of this kind is in WEOES’s list of most frequent words. regarding sentiment items, WESPES shows 24: 11 positive words and intensifiers, 13 negative (12 emotion-laden nouns and verbs –riesgo, tensión, pobreza, desigualdad, desequilibrio, escalada, controversia, obstáculo, amenaza, desencadenar, vulnerabilidad, obligación– plus an intensifier used only in negative contexts: grande). WEOES, in turn, shows 22 emotion-laden words: 14 negative terms, plus 3 negative intensifiers, and like its English version, 5 positive terms, no positive intensifiers. We can also see these taxonomies in Table 5 below: mAríA ÁnGElES OrTS llOPiS Ibérica 40 (2020): 217-244230 Lexical taxonomy in WESPES Lexical taxonomy in WEOES Technolects (5) Social words (3) Technolects (11) Social words (0) ECONOMÍA, FINANCIERO, COMERCIO, FINANCIACIÓN, ARANCEL SOSTENIBLE, CLIMA, SOCIAL ECONOMÍA, COMERCIAL, FINANCIERO, ARANCEL, MERCADO, PRECIO, SOBERANOS, TÍTULOS, DIFERENCIALES, BANCO, ARANCELARIAS. ––– Negative sentiment words (13) Negative sentiment words (17) Emotion-laden items Intensifiers and softeners Emotion-laden items Intensifiers and softeners RIESGO, TENSIÓN, POBREZA, DESIGUALDAD, DESEQUILIBRIO, ESCALADA, CONTROVERSIA, OBSTÁCULO, AMENAZA, DESENCADENAR, VULNERABILIDAD, OBLIGACIÓN GRANDE TENSIÓN, RIESGO, DESACELERACIÓN, PREOCUPACIÓN, AVERSIÓN, BREXIT, ENDURECIMIENTO, DETERIORO, CIERRE, INCERTIDUMBRE, DISPUTA, BARRERAS, SUBIDA FLOJO, LEVEMENTE, (A LA) BAJA Positive sentiment words (11) Positive sentiment words (5) Emotion-laden items Intensifiers and softeners Emotion-laden items Intensifiers and softeners DESARROLLO, CRECIMIENTO, MULTILATERAL, DIVERSIFICACIÓN, EFICAZ, LOGRO, APOYAR, RECEPTIVO, TRANSPARENCIA FUNDAMENTAL, IMPERATIVO CRECIMIENTO, MODERACIÓN, ÍMPETU, TREGUA, ACTITUD ––– Table 5. Lexical taxonomies in WESPES and WEOES. A As regards discourse metaphors, most of the words in WESPES belong to the war/conflict conceptualization, since they are deployed to identify threats, to present the un as a battling/healing force and to establish itself as a defender against negative elements of the global economy, with words like controversia (‘controversy’), amenaza (‘threat’), riesgo (‘risk’), escalada (‘escalation’), tension (‘tension’); as in WESP, economic problems are enemies, as with pobreza (‘poverty’), desigualdad (‘inequality’), desequilibrio (‘imbalance’), vulnerabilidad (‘vulnerability’) and desencadenar (‘unchain’). See, for example, 5: (5) Estos d e s e qui l i br io s al e j a n a ún más el logro de los objetivos de erradicar la pobreza y crear empleos decentes para todos. (“These imbalances push the targets of eradicating poverty and creating decent jobs for all further from reach”, in the English version). Other conceptualizations present positive words in a scenario where economy is a living/human organism with words like desarrollo (‘development’), crecimiento (‘growth’), logro (‘achievement’) and economic systems and structures are objects and buildings to be repaired, as in multilateral, diversificación (‘diversification’), apoyar (‘support’), transparencia (‘transparency’): (6) La di v e rs i f ic a c i ón re f o rz a rá la resiliencia y, en muchos casos, es también una necesidad ambiental. (“Diversification will strengthen resilience, and in many cases is also an environmental necessity”). in WEOES economy as a conflict is also present, with words like tensión, riesgo, disputa, restrictive; also, economic problems are objects that experiment breakdowns, with endurecimiento (‘tightening’), deterioro (‘deterioration’) and cierre (‘shutdown’). let us see an example: (7) [..] otros factores que agudizan el r ie s g o a la baja son la incertidumbre acerca del programa de políticas de los nuevos gobiernos, un c i e rr e prolongado del gobierno federal en Estados Unidos, y te n s io ne s geopolíticas en Oriente Medio y Asia oriental. (“[..] other factors adding downside risk include uncertainty about the policy agenda of new administrations, a protracted uS federal government shutdown, as well as geopolitical tensions in the middle East and East Asia”). The economy as a journey is, as in its English version, present in sentences containing words such as desaceleración (‘slowdown’), barreras (‘barriers’), EmOTiOn TO fOrECAST A rECESSiOn: A BilinGuAl lExiCAl AnD SEnTimEnT AnAlySiS Ibérica 40 (2020): 217-244 231 Brexit, and subida, translated from ‘high’ or ‘higher,’ frequently used in the context of interest rates and tariffs: (8) A pesar del estímulo fiscal que neutraliza en parte el impacto de la subida de los aranceles de Estados unidos, la economía de China se desacelerará. (“Despite fiscal stimulus that offsets some of the impact of higher uS tariffs, China’s economy will slow down”). lastly, country economies and markets in WEOES show a human element, both in negative and in positive sentiment words –preocupación (concern), aversión (‘aversion’), incertidumbre (‘uncertainty’), actitud (‘sentiment’), crecimiento (‘growth’) and moderación (‘moderation’). Our final example shows this: (9) [..] si la ac t it ud del mercado se recupera, una mayor confianza y condiciones financieras más favorables podrían reforzarse mutuamente y elevar el c re c i mie n to por encima del pronóstico de base. (“if [..] market sentiment recovers, then improved confidence and easier financial conditions could reinforce each other to lift growth above the baseline forecast”). 3.2. Phase two. Deploying Sentiment Analysis As Garofalo (2017: 57) points out, Sentiment Analysis endows discourse analysis with useful empirical data for the researcher, offering computational and objectifiable keys to strengthen the merely qualitative study of observable phenomena. As noted above, this paper uses the Lingmotif software (moreno- Ortiz, 2017) to find out the polarity and intensity of feeling in the two sets of reports under study. Specifically, the difference between Lingmotif –developed by the Tecnolengua group of the university of málaga– and other opinion mining methods lies in the fact that the system admits the processing of extensive texts, and not just mini-texts in the style of tweets. Additionally, it includes an internal lexicon with words and phrases with emotional weight which are detected in the process of analysis. Lingmotif evaluates the text under analysis on a scale from 0 to 100 from “extremely negative” to “extremely positive”, based on a semantic sentiment orientation in the text (Text Sentiment Score, or TSS). it also offers a rating scale of sentiment intensity (Text Sentiment intensity, or TSi), calculated as an average between emotion- ridden and emotion-free ones, assigning ratings on a scale from -5 to +5, where 0 is an expression of neutrality. it is important to note at this point that, on the one hand, the software not only detects simple units, but also localizes syntagmatic units, so that, for example, the positive term ‘buoyant’ becomes mAríA ÁnGElES OrTS llOPiS Ibérica 40 (2020): 217-244232 negative when added the modifier ‘less’ and the negative term ‘corruption’ acquires a positive value when the verbal form ‘reducing’ is added (both in WESP). This is possible due to the existence of context rules (of inversion, intensification and attenuation) in the system, to accommodate possible sentiment modifiers (moreno-Ortiz, 2017: 133). These context rules (CVS or ‘contextual valence shifters’) are Lingmotif’s mechanism to detect words or phrases that may appear in the immediate vicinity of the identified value unit, modifying its valence and increasing, neutralizing or giving it an opposite value to that originally given (fernández-Cruz & Baixauli-Pérez, 2018: 45). figures 1 to 4 show the overall results provided by the software: EmOTiOn TO fOrECAST A rECESSiOn: A BilinGuAl lExiCAl AnD SEnTimEnT AnAlySiS Ibérica 40 (2020): 217-244 233 MARÍA ÁNGELES ORTS LLOPIS 0 is an expression of neutrality. It is important to note at this point that, on the one hand, the software not only detects simple units, but also localizes syntagmatic units, so that, for example, the positive term ‘buoyant’ becomes negative when added the modifier ‘less’ and the negative term ‘corruption’ acquires a positive value when the verbal form ‘reducing’ is added (both in WESP). This is possible due to the existence of context rules (of inversion, intensification and attenuation) in the system, to accommodate possible sentiment modifiers (Moreno-Ortiz, 2017: 133). These context rules (CVS or ‘contextual valence shifters’) are Lingmotif’s mechanism to detect words or phrases that may appear in the immediate vicinity of the identified value unit, modifying its valence and increasing, neutralizing or giving it an opposite value to that originally given (Fernández- Cruz & Baixauli-Pérez, 2018: 45). Figures 1 to 4 show the overall results provided by the software: TSS TSI CVS segments Positive items Negative items Neutral items Positive Score Negative Score 45 97 43 154 175 1271 291 360 Figure 1. WESP Sentiment Score and Intensity values. TSS TSI CVS segments Positive items Negative items Neutral items Positive Score Negative Score 45 91 37 167 197 1443 302 389 Figure 2. WESPES Sentiment Score and Intensity values. Initially, the results seem to, somehow, contradict our initial diagnosis as to the – comparatively– more negative character of the IMF reports, compared to the UN ones. From Figures 1 and 2 above, it is clear that the values of the UN report in the English original and its translation seem similar, with identical TSSs, weighing the scores of positive, negative and neutral lexical items. These two figures show how they are both initially, the results seem to, somehow, contradict our initial diagnosis as to the –comparatively– more negative character of the imf reports, compared to the un ones. from figures 1 and 2 above, it is clear that the values of the un report in the English original and its translation seem similar, with identical TSSs, weighing the scores of positive, negative and neutral lexical items. These two figures show how they are both slightly (and identically) negative and somewhat different in their TSis: 97 for WESP and 91 for WESPES (both extremely intense). mAríA ÁnGElES OrTS llOPiS Ibérica 40 (2020): 217-244234 EMOTION TO FORECAST A RECESSION: A BILINGUAL LEXICAL AND SENTIMENT ANALYSIS slightly (and identically) negative and somewhat different in their TSIs: 97 for WESP and 91 for WESPES (both extremely intense). TSS TSI CVS segments Positive items Negative items Neutral items Positive Score Negative Score 42 74 45 123 187 1813 231 375 Figure 3. WEO Sentiment Score and Intensity values. TSS TSI CVS segments Positive items Negative items Neutral items Positive Score Negative Score 41 71 39 134 216 2001 229 415 Figure 4. WEOES Sentiment Score and Intensity values. In turn, the results for the IMF reports show that both sentiment score and intensity are marginally weaker in the Spanish translation but that, still, both WEO and WEOES are slightly negative and fairly intense. If, on the other hand, we take the two sets of reports, the UN’s and the IMF’s, an important difference in intensity –they are evaluated as ‘extremely’ and ‘fairly’ intense, respectively– is observable between them. TSI is calculated as the proportion of sentiment (positive or negative) vs neutral scores. Besides, lower intensity values in the IMF reports as compared to the UN’s ones are directly related to two variables, length and lexical richness. As far as length is concerned, since intensity is calculated in the software as an absolute measure and categorized by text length, the longer the text (in this case, WEO and WEOES), the more improbable for its words to be emotion-laden than in the shorter ones (WESP and WESPES). Lexical richness is also relevant: WESP was by far the most lexically rich text, which accounts for its higher intensity, and, at the other side of the scale, WEOES was the longest and least lexically rich text, and also the text with the lowest intensity. TSS, on the other hand, is calculated weighing positive and negative items in proportion in turn, the results for the imf reports show that both sentiment score and intensity are marginally weaker in the Spanish translation but that, still, both WEO and WEOES are slightly negative and fairly intense. if, on the other hand, we take the two sets of reports, the un’s and the imf’s, an important difference in intensity –they are evaluated as ‘extremely’ and ‘fairly’ intense, respectively– is observable between them. TSi is calculated as the proportion of sentiment (positive or negative) vs neutral scores. Besides, lower intensity values in the imf reports as compared to the un’s ones are directly related to two variables, length and lexical richness. As far as length is concerned, since intensity is calculated in the software as an absolute measure and categorized by text length, the longer the text (in this case, WEO and WEOES), the more improbable for its words to be emotion-laden than in the shorter ones (WESP and WESPES). lexical richness is also relevant: WESP was by far the most lexically rich text, which accounts for its higher intensity, and, at the other side of the scale, WEOES was the longest and least lexically rich text, and also the text with the lowest intensity. TSS, on the other hand, is calculated weighing positive and negative items in proportion to neutral scores. in the light of our lexical reports, we predicted that WEO and WEOES would be more negative than WESP and WESPES, but in fact the software shows that the contrary is true. However, the lesser negative polarity in the imf’s reports may be explained because of the substantially lower number of neutral items in WESP and WESPES as compared to WEO and WEOES. However, Lingmotif’s findings regarding the number of positive and negative items somehow confirm our results in the lexical part of the study. Even if the four texts seem to be equally negative, and disregarding the presence of neutral items, it is verifiable that WEO and WEOES do have more negative items and higher negative scores (187 and 216, respectively) than WESP and WESPES (175 and 197, respectively). Subsequently, and upon the basis of the lexical study in the previous subsection, a plugin or reference dictionary was prepared for each subcorpus to refine the results of the analysis and provide value units linking to the conceptual framework of each of the reports. Because sentiment is very closely connected to topic, this additional wordlist was prepared to complement the central lexicon Lingmotif provides and takes into account the sentiment in our taxonomy. in the procedure followed, the words in context, and their collocations and n-grams were scrutinized to generate a set of value units selected for their representativeness and were assigned a negative or positive value in accordance with their metaphorical meaning in the text. The wildcard All was used to match all parts of speech: EmOTiOn TO fOrECAST A rECESSiOn: A BilinGuAl lExiCAl AnD SEnTimEnT AnAlySiS Ibérica 40 (2020): 217-244 235 Table 7 shows the plugins made for the respective subcorpora in English and in Spanish, with a difference in amount (49 items in the English plugin, 42 mAríA ÁnGElES OrTS llOPiS Ibérica 40 (2020): 217-244236 EMOTION TO FORECAST A RECESSION: A BILINGUAL LEXICAL AND SENTIMENT ANALYSIS ,ALL,-5 ,ALL,-3 ,ALL,-5 ,ALL,2 ,ALL,-5 ,ALL,-3 ,ALL,-5 ,ALL,-5 ,ALL,-3 ,ALL,-5 ,ALL,-5 ,ALL,-5 ,ALL,-5 ,ALL,-5 ,ALL,3 ,ALL,-3 ,ALL,-5 ,ALL,3 ,ALL,-5 ,ALL,-5 ,ALL,3 ,ALL,3 ,ALL,-3 ,ALL,-2 ,ALL,-5 ,ALL,-3 ,ALL,-5 ,ALL,-5 ,ALL,-5 ,ALL,-3 ,ALL,3 ,ALL,-5 ,ALL,3 ,ALL,3 ,ALL,-3 ,ALL,3 ,ALL,-3 ,ALL,-5 ,ALL,-5 ,ALL,-3 ,ALL,-3 ,ALL,-3 ,ALL,5 ,ALL,3 ,ALL,3 ,ALL,-3 ,ALL,-5 <ímpetu>,ALL,-5 ,ALL,5 ,ALL,-5 ,ALL,2 ,ALL,-3 ,ALL,3 ,ALL,5 ,ALL,3 ,ALL,3 ,ALL,-5 ,ALL,3 ,ALL,-5 ,ALL,-3 ,ALL,2 ,ALL,-5 ,ALL,-3 ,ALL,-5 ,ALL,-5 ,ALL,-3 ,ALL,-3 ,ALL,-5 ,ALL,-5 ,ALL,3 ,ALL,-3 ,ALL,-5 ,ALL,-5 ,ALL,5 ,ALL,-3 ,ALL,-3 ,ALL,5 ,ALL,-5 ,ALL,-3 ,ALL,5 ,ALL,-5 ,ALL,5 ,ALL,-5 ,ALL,-3 ,ALL,5 ,ALL,5 ,ALL,-5 ,ALL,-3 ,ALL,3 ,ALL,-5 ,ALL,-3 Table 7. Ad-hoc plugin for English and Spanish. Table 7 shows the plugins made for the respective subcorpora in English and in Spanish, with a difference in amount (49 items in the English plugin, 42 in the Spanish one) reflecting concomitances with the taxonomy above. The keywords were selected manually in the segments processed by Lingmotif, taking note of those that, according to our results, were outside the valuation assigned by default by the internal dictionary of the application, or whose valuation was newly adjusted to concur with our conceptual study. Once the keywords had been selected, they were assigned a negative or positive valence, depending on their context in the text, in a scale from -5 to +5. Thus, for example, words that in context were found to very negatively depict the economy as a conflict scenario (‘dispute’, ‘risk’, ‘escalate’ and ‘tension’; disputa, riesgo, tensión), or as a journey (‘Brexit’, ‘barrier’, ‘fall’, ‘burden’, ‘slowdown’; Brexit, barrera, obstáculo, desaceleración) were assigned a -5 valence, whereas intensifiers with a negative value (‘severe’, ‘heightened’) were all found to contain less negativity and were assigned a -3 valence. Most positive words conceptualizing economy as a living organism (‘develop’, ‘growth’, ‘responsive’; desarrollo, crecimiento, receptivo), and framing economic in the Spanish one) reflecting concomitances with the taxonomy above. The keywords were selected manually in the segments processed by Lingmotif, taking note of those that, according to our results, were outside the valuation assigned by default by the internal dictionary of the application, or whose valuation was newly adjusted to concur with our conceptual study. Once the keywords had been selected, they were assigned a negative or positive valence, depending on their context in the text, in a scale from -5 to +5. Thus, for example, words that in context were found to very negatively depict the economy as a conflict scenario (‘dispute’, ‘risk’, ‘escalate’ and ‘tension’; disputa, riesgo, tensión), or as a journey (‘Brexit’, ‘barrier’, ‘fall’, ‘burden’, ‘slowdown’; Brexit, barrera, obstáculo, desaceleración) were assigned a -5 valence, whereas intensifiers with a negative value (‘severe’, ‘heightened’) were all found to contain less negativity and were assigned a -3 valence. most positive words conceptualizing economy as a living organism (‘develop’, ‘growth’, ‘responsive’; desarrollo, crecimiento, receptivo), and framing economic systems as objects or buildings (‘multilateral’, ‘diversification’; multilateral, diversificación) were assigned a 3 valence, since they were found to have a moderate (not very intense) sentiment weight in context. likewise, intensifiers with a positive value (almost absent in the Spanish subcorpora and with scarce examples in the English ones, like ‘vital’ or ‘fundamental’) were assigned a 3 valence. Only words with more emotional weight in context, like ‘transparency’, ‘momentum’, ‘truce’ or ‘sustainable’ (transparencia, , ímpetu, tregua, sostenible) were assigned a 5 valence. figure 5 shows the results of the analysis performed by Lingmotif: One of the most striking findings is the fact that the four reports now present the maximum TSi –100, all of them “extremely intense”– score. This is undoubtedly due to the fact that the number of sentiment items has risen in the scrutiny of the reports overall, the number of neutral items having, consequently, decreased. Sentiment polarities have also changed, predictably EmOTiOn TO fOrECAST A rECESSiOn: A BilinGuAl lExiCAl AnD SEnTimEnT AnAlySiS Ibérica 40 (2020): 217-244 237 MARÍA ÁNGELES ORTS LLOPIS systems as objects or buildings (‘multilateral’, ‘diversification’; multilateral, diversificación) were assigned a 3 valence, since they were found to have a moderate (not very intense) sentiment weight in context. Likewise, intensifiers with a positive value (almost absent in the Spanish subcorpora and with scarce examples in the English ones, like ‘vital’ or ‘fundamental’) were assigned a 3 valence. Only words with more emotional weight in context, like ‘transparency’, ‘momentum’, ‘truce’ or ‘sustainable’ (transparencia, , ímpetu, tregua, sostenible) were assigned a 5 valence. Figure 5 shows the results of the analysis performed by Lingmotif: Corpus TSS TSI Sent. Items Pos. Items Neg. Items Neut. Items WESP 48 100 415 223 192 1170 WEO 37 100 420 190 230 1689 WESPES 44 100 428 218 210 1372 WEOES 34 100 440 195 245 1904 Figure 5. New values assigned the 4 subcorpora with ad hoc plugin. One of the most striking findings is the fact that the four reports now present the maximum TSI –100, all of them “extremely intense”– score. This is undoubtedly due to the fact that the number of sentiment items has risen in the scrutiny of the reports overall, the number of neutral items having, consequently, decreased. Sentiment polarities have also changed, predictably going down in the two reports by the IMF (from 42 to 37 in the English original and from 41 to 34 in the Spanish one, both sets turning from slightly negative to fairly negative in sentiment polarity) and increasing in the case of the UN’s reports (from 45 to 48 sentiment polarity in WESP and from 45 to 44 in WESPES). As remarked upon above, these changes are in line with the incorporation of our findings in the previous section. In the first place, adding up all the emotion-laden items as sentiment factors implies that intensity increases, since new sentiment words were integrated into the plugin and some others were granted more intensity than previously, as is the case of all the items assigned a -5 or 5 valence. Secondly, the deployment of our new taxonomy also points to differences in sentiment scores. In fact, it is the reason for the larger amount of negative items in the IMF reports (187 before, to 230 after the plugin in WEO, 216 to 245 in WEOES) and the greater number of positive items in the UN ones (154 to 223 in WESP, 167 to 218 in WESPES). Hence, WEO, as the most adverse of reports in our corpus, show fragments that Lingmotif evaluates as having an extremely negative score, where items that had not been picked up by the software as negative, now are: (10) China’s economy slowed in 2018 mainly due to financial regulatory tightening to rein in shadow banking activity and off-budget local government investment, and as a result of the widening trade dispute with the United States, which intensified the slowdown toward the end of the year. And in its Spanish version, in WEOES: (11) La economía china se desaceleró en 2018 debido principalmente al endurecimiento de las regulaciones financieras con el fin de frenar la actividad de la banca paralela y excluir del presupuesto la inversión de los gobiernos locales, y también como resultado de la creciente disputa comercial con Estados Unidos, que agravó la desaceleración hacia finales del año. As we can see in the Spanish version, the verbal form agravó constitutes a more explicit, negative version than its referent in the original, ‘intensified’. Additionally, an increase of positive polarity is observable in WESP and WESPES: (12) Nevertheless, this must leave space for countries to adopt effective tax policies that enhance domestic public finance for sustainable development. going down in the two reports by the imf (from 42 to 37 in the English original and from 41 to 34 in the Spanish one, both sets turning from slightly negative to fairly negative in sentiment polarity) and increasing in the case of the un’s reports (from 45 to 48 sentiment polarity in WESP and from 45 to 44 in WESPES). As remarked upon above, these changes are in line with the incorporation of our findings in the previous section. in the first place, adding up all the emotion-laden items as sentiment factors implies that intensity increases, since new sentiment words were integrated into the plugin and some others were granted more intensity than previously, as is the case of all the items assigned a -5 or 5 valence. Secondly, the deployment of our new taxonomy also points to differences in sentiment scores. in fact, it is the reason for the larger amount of negative items in the imf reports (187 before, to 230 after the plugin in WEO, 216 to 245 in WEOES) and the greater number of positive items in the un ones (154 to 223 in WESP, 167 to 218 in WESPES). Hence, WEO, as the most adverse of reports in our corpus, show fragments that Lingmotif evaluates as having an extremely negative score, where items that had not been picked up by the software as negative, now are: (10) China’s economy slowed in 2018 mainly due to financial regulatory tightening to rein in shadow banking activity and off-budget local government investment, and as a result of the widening trade dispute with the united States, which intensified the slowdown toward the end of the year. And in its Spanish version, in WEOES: (11) La economía china se d e s a c e l e r ó en 2018 debido principalmente al e nd ure c imi e nt o de las regulaciones financieras con el fin de f r e na r la actividad de la banca paralela y excluir del presupuesto la inversión de los gobiernos locales, y también como resultado de la creciente d is p ut a comercial con Estados Unidos, que ag r av ó la d e sa c e l e r ac i ón hacia finales del año. As we can see in the Spanish version, the verbal form agravó constitutes a more explicit, negative version than its referent in the original, ‘intensified’. Additionally, an increase of positive polarity is observable in WESP and WESPES: (12) nevertheless, this must leave space for countries to adopt effective tax policies that enhance domestic public finance for sustainable development. mAríA ÁnGElES OrTS llOPiS Ibérica 40 (2020): 217-244238 (13) Sin embargo, esto debe dejar espacio para que los países adopten políticas fiscales e f ic a c e s que me j o re n las finanzas públicas nacionales en pos del de s a rr ol l o s os te n ib l e. All in all, the integration of the plugin into the process of analysis with Lingmotif has shown how connected sentiment is to the way in which information is conceptualized by the sender of the message. This conceptualization is made, in economic discourse, through the deployment of an array of discourse metaphors that transmit sentiment and attitude to heighten emotional impact. Even if differences in intensity and polarity before and after the application of the plugin are not abysmal, the results are really relevant, since they confirm that there exist interesting asymmetries between texts, aligned with the intentions of their issuers. 4. Discussion and conclusions Our study has endeavoured to demonstrate how the combination of qualitative and quantitative linguistic analyses can offer interesting data as to the way in which institutions issue very similar financial information in different ways, according to their aims and scope. The reports under scrutiny show that the 2008 crisis is not over, and seems to lengthen towards a new economic mayhem in 2020; but the un and the imf do not communicate this information in the same way. from the onset of this study we hypothesized that the lexicon of the reports under analysis would combine technical terms belonging to the area of economics, together with emotion-laden items with a metaphorical value depicting sentiment, but with uneven combinations, in tune with the institution’s role and goals. in fact, our initial lexical analysis showed how the imf reports seemed to be more technical and scarcer in emotion-laden items and discourse metaphors than the un’s. One could infer from such results that the imf (a body which is solely involved in macroeconomic analyses and recommendations) is more bent on technical issues and less involved emotionally. nonetheless, our processing of the texts through Lingmotif has unveiled that the reports issued by this institution are, in fact, acutely intense, and even more negative in character than the un’s. This appears to support the discourse metaphor constellations uncovered initially, which suggested that the imf resorts to a wider range of conceptualizations (the economy is a battlefield, but also a sick living organism, a vehicle whose failure to get to a destination is mapped as EmOTiOn TO fOrECAST A rECESSiOn: A BilinGuAl lExiCAl AnD SEnTimEnT AnAlySiS Ibérica 40 (2020): 217-244 239 wreckage– and its problems, objects in need of repairing. We illustrated all of this above, but the following extract also summarizes the institution’s dire warnings: (14) risks to global growth tilt to the downside. An escalation of trade tensions beyond those already incorporated in the forecast remains a key source of risk to the outlook. financial conditions have already tightened since the fall. A range of triggers beyond escalating trade tensions could spark a further deterioration in risk sentiment with adverse growth implications, especially given the high levels of public and private debt. On the other hand, in the first batch of results, the un reports actually showed more sentiment words overall, exuding a social inspiration that is seemingly scant in the other group of reports. This, and the higher number of positive items, showed a wider-ranging scope and a more positive character in this report than in the imf’s. Eventually, processing WESP and WESPES with the SA software confirmed that the un reports contain negative items, but these are actually fewer than the positive ones. in tune with such processing, we saw how the discourse metaphors in the un reports tended to stick more narrowly to the framing of the economic situation as an arena of conflict, but also used very positive imagery with words such as ‘growth’, ‘achievement’, ‘transparency’ and ‘multilateral’, which point to an incessant fight of economies towards improvement and wellbeing: (15) Strengthening multilateralism is, therefore, central to advancing sustainable development across the globe. (16) Trade in services also contributes to inclusiveness, resilience, and diversification. This conceptualization could respond to the formulation that a weak economy is something that countries have to fight against to protect their citizens, but that there might be a series of possible way-outs for the solution to problems, a sense of shared objectives in the social sphere to overcome them, in the picture portrayed by the un’s reports. in conclusion, the present study has tried to demonstrate that the discourse of economy is not exempt from emotion, nor are the official reports on the situation of the world finances –by institutions as prominent as the un and the imf– written in a detached way. On the contrary, international think tanks mAríA ÁnGElES OrTS llOPiS Ibérica 40 (2020): 217-244240 emit their diagnoses on the state of the global economy deploying emotional triggers with the aim of promoting international response, mainly from economic policymakers and operators. This study has carried out a scrutiny of the emotion-laden items and discourse metaphors with positive or negative values that these institutions apply when issuing their annual reports directed to warn the world about the symptoms of the crisis that they seem to predict. nevertheless, there is a lack of linearity in sentiment polarity and intensity between reports. As we stated above, our second objective was to detect which of the reports –the imf’s or the un’s– carried more emotional weight, and whether such weight would be maintained in their translated versions. On the one hand, this study has proved that the way in which the global economic scenario is pictured is different, in line with the sociological or ideological purposes that each institution is set to attain. On the other hand, it has unveiled slight rhetorical-linguistic differences between the economic discourse of the two languages. Overall, there are differences of verbalization in the English originals and their Spanish translations, since even if the versions transmit analogous meanings overall, they are embodiments of the different ways to conceive economic discourse in either language, where English has conciseness as one of its main characteristics and Spanish tends to be more verbose and categorical, mainly in the use of adverbials. more importantly, the polarized expression of positivity or negativity is embedded in the way that each think tank describes the economic difficulties that the world has ahead of it, and the role that such an institution plays in the global scenario. Hence, probably our results, if restricted in size and scope to these reports, may be ‘softer’, in the case of the un, ‘harsher’, regarding the imf, in line with the way in which each institution depicts itself: the un being broader in its scope, highlighting social problems and environmental issues that affect world economies from the humanitarian standpoint; the imf’s bleak discourse, on the other hand, showing the institution’s role as a watchdog in charge of monitoring changes in the economy, as it issues regulatory advice to governments in the face of a new economic havoc that seems to loom in the world’s horizon. Article history: Received 31 July 2019 Received in revised form 23 January 2020 Accepted 17 February 2020 EmOTiOn TO fOrECAST A rECESSiOn: A BilinGuAl lExiCAl AnD SEnTimEnT AnAlySiS Ibérica 40 (2020): 217-244 241 References mAríA ÁnGElES OrTS llOPiS Ibérica 40 (2020): 217-244242 Anthony, L. (2019). AntConc (Version 3.5.8.). Tokyo: Waseda University. URL: [17/01/2020]. Arrese, A. (2015). “Las metáforas de la crisis del euro en la prensa española”. Communication and Society 28(2): 19-38. Bandelj, N. (2009). “Emotions in economic action and interaction”. Theory and Society 38: 347-366. Berezin, M. (2005). “Emotions and the Economy” in N.J. Smelser & R. Swedberg (eds.), Handbook of Economic Sociology, 109-127. New York/Princeton: Russell Sage Foundation/ Princeton University Press. Berezin, M. (2009). “Exploring emotions and the economy: new contributions from sociological theory”. Theory and Society 38: 335-346. Berk-Seligson, S. (2012). The Bilingual Courtroom: Court Interpreters in the Judicial Process. Chicago: University of Chicago Press. Biel, Ł. & V. Sosoni (2017). “The translation of economics and the economics of translation”. Perspectives 25(3): 351-361. Charteris-Black, J. & T. Ennis (2001). “A comparative study of metaphor in Spanish and English financial reporting”. English for Specific Purposes 20: 249-266. Charteris-Black, J. (2004). Corpus Approaches to Critical Metaphor Analysis. Basingstoke: Palgrave Macmillan. Charteris-Black, J. (2014). Analysing Political Speeches: Rhetoric, Discourse and Metaphor. Basingstoke: Palgrave Macmillan. Devitt, A. & K. Ahmad (2007). “Sentiment polarity identification in financial news: A cohesion-based approach” in Proceedings of the 45th Annual Meeting of the Association of Computational Linguistics, 984-991. Prague: Association for Computational Linguistics. Fernández-Cruz, J. & C. Baixauli-Pérez (2018). “Análisis de sentimiento con Lingmotif. Aproximación metodológica al análisis de intervenciones en series temporales de la valencia” in La simulación en ingeniería, 43-51. Esmeraldas, Ecuador: PUCESE. Foolen, A. (2012). “The relevance of emotion for language and linguistics.” in A. Foolen, U. Lüdtke, T.P. Racine & J. Zlatev. (eds.), Moving Ourselves, Moving Others. Motion and Emotion in Intersubjectivity, Consciousness and Language, 349-368. Amsterdam: John Benjamins. Garofalo, G. (2017). La insoportable levedad del acusar. El lenguaje de la valoración en las querellas de la Fiscalía Superior de Cataluña contra Carme Forcadell i Lluís. Milan: FrancoAngeli. Garrido, J. (2013). “Motion metaphors in discourse construction” in F. Gonzálvez-García, M.S. Peña Cervel & L. Pérez Hernández (eds.), Metaphor and Metonymy Revisited beyond Contemporary Theory of Metaphor: Recent Developments and Applications, 109-131. Amsterdam: John Benjamins. Herrera-Soler, H. & M. White (eds.) (2012). Metaphor and Mills: Figurative Language in Business and Economics. Berlin/Boston: De Gruyter Mouton. Kövecses, Z. (2003). Metaphor and Emotion: Language, Culture, and Body in Human Feeling. Cambridge: CUP. Kövecses, Z. (2010). Metaphor: A Practical Introduction (2nd ed.). Oxford: OUP. Loma-Osorio, M. (2004). Estructura y función del texto económico: fundamentos de una léxico- gramática del discurso económico en español y en inglés. PhD Dissertation, Universidad Complutense de Madrid. Mackenzie, J.L. & L. Alba-Juez (eds.) (2019). Emotion in Discourse. Amsterdam: John Benjamins. Moreno-Ortiz, A. (2017). “Lingmotif: A User- focused Sentiment Analysis Tool”. Procesamiento del Lenguaje Natural 58: 133-140. Musolff, A. (2006). “Metaphor scenarios in public discourse”. Metaphor and Symbol 21(1): 23-38. Musolff, A. & J. Zinken (eds.) (2009). Metaphor and Discourse. Basingstoke: Palgrave MacMillan. Plutchik, R. (1980). Emotion: A Psychoevolutionary Synthesis. New York: Harper and Row. Rojo, A. & M.Á. Orts (2010). “Metaphorical pattern analysis in financial texts: Framing the crisis in positive or negative metaphorical terms”. Journal of Pragmatics 42(12): 3300-3313. Ruiz-Martínez, J.M., R. Valencia-García & F. García-Sánchez (2012). “Semantic-based sentiment analysis in financial news” in Proceedings of the 1st International Workshop on Finance and Economics on the Semantic Web, 38- 51. Mª Ángeles Orts Llopis is Professor at the Department of Translation and interpreting at the university of murcia, where she currently lectures on legal and Economic Translation. She has specialized in legal, economic and business English and has published extensively in such areas. NOTES 1 According to ifOrEx, “market sentiment is the combined attitude of investors toward a specific market or instrument”, as in . 2 WESP is available at . 3 WEO is available at . 4 unTErm is available at . EmOTiOn TO fOrECAST A rECESSiOn: A BilinGuAl lExiCAl AnD SEnTimEnT AnAlySiS Ibérica 40 (2020): 217-244 243 Schäffner, C. (2012). “Finding space under the umbrella: the Euro crisis, metaphors, and translation”. Journal of Specialized Translation 17(b): 250-270. Schwarz-Friesel, M. (2015). “Language and emotion. The cognitive linguistic perspective” in U.M. Lüdtke (ed.), Emotion in Language, 157-173. Amsterdam: John Benjamins. Silaški, N. & T. Đurović (2010). “The conceptualization of the global financial crisis via the economy is a person metaphor –a contrastive study of English and Serbian”. Facta Universitatis, Series: Linguistics and Literature 8(2): 129-139. Skorczynska, H. (2010). “A corpus-based evaluation of metaphors in a business English textbook”. English for Specific Purposes 29: 30-42. Steen, G. (2009). “Three kinds of metaphor in discourse: A linguistic taxonomy” in A. Musolff & J. Zinken (eds.), Metaphor and Discourse, 25-39. Basingstoke: Palgrave Macmillan. Taboada, M. (2016). “Sentiment analysis: An overview from Linguistics”. Annual Review of Linguistics 2(1): 325-347. Vázquez Ayora, G. (1977). Introducción a la Traductología. Washington: Georgetown University Press. Zinken, J. (2007). “Discourse metaphors: The link between figurative language and habitual analogies”. Cognitive Linguistics 18(3): 445-466. Zinken, J., I. Hellsten & B. Nerlich (2008). “Discourse metaphors” in R. Dirven, R. Frank, T. Ziemke & J. Zlatev (eds.), Body, Language, and Mind. Vol. 2: Sociocultural Situatedness, 363-385. Berlin: Mouton de Gruyter.