Datasets:

WINGNUS
/

ACL-OCL

	{
	"paper_id": "E17-1025",
	"header": {
	"generated_with": "S2ORC 1.0.0",
	"date_generated": "2023-01-19T10:51:57.820504Z"
	},
	"title": "Exploring the Impact of Pragmatic Phenomena on Irony Detection in Tweets: A Multilingual Corpus Study",
	"authors": [
	{
	"first": "Jihen",
	"middle": [],
	"last": "Karoui",
	"suffix": "",
	"affiliation": {
	"laboratory": "",
	"institution": "Universit\u00e9 de Toulouse",
	"location": {
	"country": "France"
	}
	},
	"email": "[email protected]"
	},
	{
	"first": "Farah",
	"middle": [],
	"last": "Benamara",
	"suffix": "",
	"affiliation": {
	"laboratory": "",
	"institution": "Universit\u00e9 de Toulouse",
	"location": {
	"country": "France"
	}
	},
	"email": "[email protected]"
	},
	{
	"first": "V\u00e9ronique",
	"middle": [],
	"last": "Moriceau",
	"suffix": "",
	"affiliation": {
	"laboratory": "LIMSI",
	"institution": "Universit\u00e9 Paris-Saclay",
	"location": {
	"country": "France"
	}
	},
	"email": "[email protected]"
	},
	{
	"first": "Viviana",
	"middle": [],
	"last": "Patti",
	"suffix": "",
	"affiliation": {
	"laboratory": "",
	"institution": "University of Turin",
	"location": {
	"country": "Italy"
	}
	},
	"email": "[email protected]"
	},
	{
	"first": "Cristina",
	"middle": [],
	"last": "Bosco",
	"suffix": "",
	"affiliation": {
	"laboratory": "",
	"institution": "University of Turin",
	"location": {
	"country": "Italy"
	}
	},
	"email": "[email protected]"
	},
	{
	"first": "Nathalie",
	"middle": [],
	"last": "Aussenac-Gilles",
	"suffix": "",
	"affiliation": {
	"laboratory": "",
	"institution": "Universit\u00e9 de Toulouse",
	"location": {
	"country": "France"
	}
	},
	"email": ""
	}
	],
	"year": "",
	"venue": null,
	"identifiers": {},
	"abstract": "This paper provides a linguistic and pragmatic analysis of the phenomenon of irony in order to represent how Twitter's users exploit irony devices within their communication strategies for generating textual contents. We aim to measure the impact of a wide-range of pragmatic phenomena in the interpretation of irony, and to investigate how these phenomena interact with contexts local to the tweet. Informed by linguistic theories, we propose for the first time a multi-layered annotation schema for irony and its application to a corpus of French, English and Italian tweets.We detail each layer, explore their interactions, and discuss our results according to a qualitative and quantitative perspective.",
	"pdf_parse": {
	"paper_id": "E17-1025",
	"_pdf_hash": "",
	"abstract": [
	{
	"text": "This paper provides a linguistic and pragmatic analysis of the phenomenon of irony in order to represent how Twitter's users exploit irony devices within their communication strategies for generating textual contents. We aim to measure the impact of a wide-range of pragmatic phenomena in the interpretation of irony, and to investigate how these phenomena interact with contexts local to the tweet. Informed by linguistic theories, we propose for the first time a multi-layered annotation schema for irony and its application to a corpus of French, English and Italian tweets.We detail each layer, explore their interactions, and discuss our results according to a qualitative and quantitative perspective.",
	"cite_spans": [],
	"ref_spans": [],
	"eq_spans": [],
	"section": "Abstract",
	"sec_num": null
	}
	],
	"body_text": [
	{
	"text": "Irony is a complex linguistic phenomenon widely studied in philosophy and linguistics (Grice et al., 1975; Sperber and Wilson, 1981; Utsumi, 1996) . Glossing over differences across approaches, irony can be defined as an incongruity between the literal meaning of an utterance and its intended meaning. For many researchers, irony overlaps with a variety of other figurative devices such as satire, parody, and sarcasm (Clark and Gerrig, 1984; Gibbs, 2000) . In this paper, we use irony as an umbrella term that includes sarcasm, although some researchers make a distinction between them, considering that sarcasm tends to be more aggressive (Lee and Katz, 1998; Clift, 1999) .",
	"cite_spans": [
	{
	"start": 86,
	"end": 106,
	"text": "(Grice et al., 1975;",
	"ref_id": "BIBREF21"
	},
	{
	"start": 107,
	"end": 132,
	"text": "Sperber and Wilson, 1981;",
	"ref_id": "BIBREF39"
	},
	{
	"start": 133,
	"end": 146,
	"text": "Utsumi, 1996)",
	"ref_id": "BIBREF47"
	},
	{
	"start": 419,
	"end": 443,
	"text": "(Clark and Gerrig, 1984;",
	"ref_id": "BIBREF12"
	},
	{
	"start": 444,
	"end": 456,
	"text": "Gibbs, 2000)",
	"ref_id": "BIBREF19"
	},
	{
	"start": 642,
	"end": 662,
	"text": "(Lee and Katz, 1998;",
	"ref_id": "BIBREF27"
	},
	{
	"start": 663,
	"end": 675,
	"text": "Clift, 1999)",
	"ref_id": "BIBREF13"
	}
	],
	"ref_spans": [],
	"eq_spans": [],
	"section": "Introduction",
	"sec_num": "1"
	},
	{
	"text": "Different categories of irony have been studied in the linguistic literature such as hyperbole, exaggeration, repetition or change of register (see section 3 for a detailed description). These categories were mainly identified in literary texts (books, po-ems, etc.) , and as far as we know, no one explored them in the context of social media. The goal of the paper is thus four folds: (1) analyse if these categories are also valid in social media contents, focusing on tweets which are short messages (140 characters) where the context may not be explicitly represented; (2) examine whether these categories are linguistically marked; (3) test if there is a correlation between the categories and markers; and finally (4) see if different languages have a preference for different categories.",
	"cite_spans": [
	{
	"start": 245,
	"end": 266,
	"text": "(books, po-ems, etc.)",
	"ref_id": null
	}
	],
	"ref_spans": [],
	"eq_spans": [],
	"section": "Introduction",
	"sec_num": "1"
	},
	{
	"text": "This analysis can be exploited in a purpose of automatic irony detection, which is progressively gaining relevance within sentiment analysis (Maynard and Greenwood, 2014; Ghosh et al., 2015) . In particular, it will bring out the most discriminant pragmatic features that need to be taken into account for an accurate irony detection, therefore helping systems improve beyond standard approaches that still heavily rely on features gleaned from the utterance-internal context Gonzalez-Ibanez et al., 2011; Liebrecht et al., 2013; Buschmeier et al., 2014; .",
	"cite_spans": [
	{
	"start": 141,
	"end": 170,
	"text": "(Maynard and Greenwood, 2014;",
	"ref_id": "BIBREF30"
	},
	{
	"start": 171,
	"end": 190,
	"text": "Ghosh et al., 2015)",
	"ref_id": "BIBREF17"
	},
	{
	"start": 476,
	"end": 505,
	"text": "Gonzalez-Ibanez et al., 2011;",
	"ref_id": "BIBREF20"
	},
	{
	"start": 506,
	"end": 529,
	"text": "Liebrecht et al., 2013;",
	"ref_id": "BIBREF29"
	},
	{
	"start": 530,
	"end": 554,
	"text": "Buschmeier et al., 2014;",
	"ref_id": "BIBREF11"
	}
	],
	"ref_spans": [],
	"eq_spans": [],
	"section": "Introduction",
	"sec_num": "1"
	},
	{
	"text": "To this end, informed by well-established linguistic theories of irony, we propose for the first time:",
	"cite_spans": [],
	"ref_spans": [],
	"eq_spans": [],
	"section": "Introduction",
	"sec_num": "1"
	},
	{
	"text": "\u2022 A multi-layered annotation schema in order to measure the impact of a wide-range of pragmatic phenomena in the interpretation of irony, and to investigate how these phenomena interact with context local to the tweet. The schema includes three layers: (1) irony activation types according to a new perspective of how irony activation happens-explicit vs. implicit, (2) irony categories as defined in previous linguistic studies, and (3) irony markers.",
	"cite_spans": [],
	"ref_spans": [],
	"eq_spans": [],
	"section": "Introduction",
	"sec_num": "1"
	},
	{
	"text": "\u2022 A multilingual corpus annotated according to this schema. As the expression of irony is very dependent on culture, we chose, for this first study, three Indo-European languages whose speakers share quite the same cultural background: French, English and Italian. The corpus is freely available for research purposes and can be downloaded here http://github.com/ IronyAndTweets/.",
	"cite_spans": [],
	"ref_spans": [],
	"eq_spans": [],
	"section": "Introduction",
	"sec_num": "1"
	},
	{
	"text": "\u2022 A qualitative and quantitative study, focusing in particular on the interactions between irony activation types and markers, irony categories and markers, and the impact of external knowledge on irony detection. Our results demonstrate that implicit activation of irony is a major challenge for future systems.",
	"cite_spans": [],
	"ref_spans": [],
	"eq_spans": [],
	"section": "Introduction",
	"sec_num": "1"
	},
	{
	"text": "The paper is organised as follows. We first present our data. Sections 3 and 4 respectively detail the annotation scheme and the annotation procedure. Section 5 discusses the reliability study whereas Section 6 the quantitative results. In Section 7, we compare our scheme to already existing schemes for irony stressing the originality of our approach and the importance of the reported results for automatic irony detection. Finally we end the paper by showing how the annotated corpora are actually exploited in automatic irony detection shared tasks.",
	"cite_spans": [],
	"ref_spans": [],
	"eq_spans": [],
	"section": "Introduction",
	"sec_num": "1"
	},
	{
	"text": "The datasets used in this study are tweets about hot topics discussed in the media. Our intuition behind choosing such topics is that the pragmatic context needed to infer irony is more likely to be understood by annotators compared to tweets that relate personal content. We relied on three corpora in French, English and Italian, referred to as F, E and I respectively. Table 1 shows the distribution of ironic vs. non ironic tweets in the data. The selection of ironic vs non-ironic tweets has been based on partly different criteria for the three addressed languages in order to tackle their features. In English and French, users employ specific hashtags (#irony, #sarcasm, #sarcastic) to mark their intention to be ironic. These hashtags have been often used as gold labels to detect irony in a supervised learning setting. Although this approach cannot be generalized well since not all ironic tweets contain hashtags, it has however shown to be quite reliable as good inter-annotator agreements (kappa around 0.75) between annotators' irony label and the reference irony hashtags have been reported (Karoui et al., 2015) . Nevertheless, irony corpus construction through hashtag filtering is not always possible for all languages. For instance, both in Czech and Italian, Twitter users generally do not use the sarcasm (i.e. '#sarkasmus', in Czech; '#sarcasmo' in Italian) or irony ('#ironie' in Czech or '#ironia' in Italian) hashtag variants to mark their intention to be ironic, thus in such cases relying on simple self-tagging for collecting ironic samples is not an option (Pt\u00e1\u010dek et al., 2014; Bosco et al., 2013) . Similar considerations hold for Chinese (Tang and Chen, 2014) . For what concerns Italian, we observe that even if occasionally Italian tweeters do use creative hashtags to explicitly mark the presence of irony, no generic shared hashtags have been used for longtime which can be considered as firmly established indicators of irony like those used for English.",
	"cite_spans": [
	{
	"start": 1107,
	"end": 1128,
	"text": "(Karoui et al., 2015)",
	"ref_id": "BIBREF26"
	},
	{
	"start": 1587,
	"end": 1608,
	"text": "(Pt\u00e1\u010dek et al., 2014;",
	"ref_id": "BIBREF34"
	},
	{
	"start": 1609,
	"end": 1628,
	"text": "Bosco et al., 2013)",
	"ref_id": "BIBREF7"
	},
	{
	"start": 1671,
	"end": 1692,
	"text": "(Tang and Chen, 2014)",
	"ref_id": "BIBREF44"
	}
	],
	"ref_spans": [
	{
	"start": 372,
	"end": 379,
	"text": "Table 1",
	"ref_id": "TABREF1"
	}
	],
	"eq_spans": [],
	"section": "Data",
	"sec_num": "2"
	},
	{
	"text": "The corpora built for English and French are new datasets built using the Twitter API as follows. We first selected 9 topics (politics, sport, artists, locations, Arab Spring, environment, racism, health, social media) discussed in the French media from Spring 2014 until Autumn 2015 and in the American media from Spring 2014 until Spring 2016. For each topic, we selected a set of keywords with and without hashtag: politics (e.g. Hollande, Obama), sport (e.g. #Zlatan, #FI-FAworldcup), etc. Then, we selected ironic tweets containing the topic keywords and the French (English) ironic hashtags. Finally, we selected non ironic tweets that contained only the topic keywords without the ironic hashtag. We removed duplicates, retweets and tweets containing pictures which would need to be interpreted to understand the ironic content. For English, since we were interested in ironic tweets for our annotation purpose, we stopped collecting messages when the number of ironic tweets was sufficient; this ex-plains the fact that classes of Ironic and Not Ironic tweets in the English dataset are pretty balanced, i.e. the amount of ironic tweets is not very low compared with the amount of not ironic ones.",
	"cite_spans": [],
	"ref_spans": [],
	"eq_spans": [],
	"section": "Ironic",
	"sec_num": null
	},
	{
	"text": "Italian data are instead extracted from two existing annotated data: the Sentipolc corpus, releasead for the shared task on sentiment analysis and irony detection in Twitter at Evalita 2014 (Basile et al., 2014) , and TW-SPINO which extends the Spinoza section of the Senti-TUT corpus (Bosco et al., 2013) . The Sentipolc dataset is a collection of Italian tweets derived from two existing corpora Senti-TUT and TWITA (Basile and Nissim, 2013) . It includes Twitter data exploiting specific keywords and hashtags marking political topics. In Sentipolc, each tweet has an annotation label among five mutually exclusive labels: positive opinion, negative opinion, irony, both positive and negative, and objective. TW-SPINO instead is from the Twitter section of Spinoza 1 , a popular collective Italian blog that publishes posts with sharp satire on politics. Since there is a collective agreement about the fact that these posts include irony mostly about politics, they represent a natural way to extend the sampling of ironic expressions. Moreover, while Sentipolc collects tweets spontaneously posted by Italian Twitter users, Spinoza's posts are selected and revised by an editorial staff, which explicitly characterize the blog as satiric. Such difference will possibly have a reflection on the types and variety of irony we detect in the tweets.",
	"cite_spans": [
	{
	"start": 190,
	"end": 211,
	"text": "(Basile et al., 2014)",
	"ref_id": "BIBREF5"
	},
	{
	"start": 285,
	"end": 305,
	"text": "(Bosco et al., 2013)",
	"ref_id": "BIBREF7"
	},
	{
	"start": 418,
	"end": 443,
	"text": "(Basile and Nissim, 2013)",
	"ref_id": "BIBREF4"
	}
	],
	"ref_spans": [],
	"eq_spans": [],
	"section": "Ironic",
	"sec_num": null
	},
	{
	"text": "To define our annotation schema, we analyzed the different categories of irony studied in the linguistic literature. Several categories have been proposed, as shown in the first column of Table 2 . Since all these categories have been found in a specific genre (literary texts), the first step was to check their presence on a small subset of 150 ironic tweets from our corpus. Three observations resulted from this first step, regarding irony activation, irony categories, and irony markers.",
	"cite_spans": [],
	"ref_spans": [
	{
	"start": 188,
	"end": 195,
	"text": "Table 2",
	"ref_id": null
	}
	],
	"eq_spans": [],
	"section": "A multi-layered annotation schema for irony in social media",
	"sec_num": "3"
	},
	{
	"text": "We observed that incongruity in ironic tweets often consists of at least two propositions (or words) P 1 and P 2 which are in contradiction to each other 1 http://www.spinoza.it/ (i.e. P 2 = Contradiction(P 1 )). It is the presence of this contradiction that activates irony. This contradiction can be at a semantic, veracity or intention level. P 1 and P 2 can be both part of the internal context of an utterance (that is explicitly lexicalized), or one is present and the other one implied. We thus defined two types of irony activation: EXPLICIT and IMPLICIT.",
	"cite_spans": [],
	"ref_spans": [],
	"eq_spans": [],
	"section": "Irony activation",
	"sec_num": "3.1"
	},
	{
	"text": "In EXPLICIT activation, one needs to rely exclusively on the lexical clues internal to the utterance, like in (1) where there is a contrast between P 1 that contains no opinion word, and P 2 which refers to a situation which is commonly judged as being negative, but in a communicative context which is clearly unsuitable w.r.t. to the one expressed in P 1 .",
	"cite_spans": [],
	"ref_spans": [],
	"eq_spans": [],
	"section": "Irony activation",
	"sec_num": "3.1"
	},
	{
	"text": "(1) L'Italia [attende spiegazioni] P 1 da cos\u00ec tanti paesi che comincio a pensare che le nostre richieste [finiscano nello spam] P 2 . (Italy is [waiting for explanations] P 1 from so many countries that I suspect our requests are being [labeled as spam] P 2 .)",
	"cite_spans": [],
	"ref_spans": [],
	"eq_spans": [],
	"section": "Irony activation",
	"sec_num": "3.1"
	},
	{
	"text": "Example 2shows another example of explicit semantic contradiction between P 1 and P 2 .",
	"cite_spans": [],
	"ref_spans": [],
	"eq_spans": [],
	"section": "Irony activation",
	"sec_num": "3.1"
	},
	{
	"text": "(2) Ben non ! [Matraquer et crever des yeux] P 1 , [ce n'est pas violent et \u00e7a respecte les droits] P 2 !!! #ironie (Well, no ! [Clubbing and putting up eyes] P 1 , [it is not violent and it does respect human rights] P 2 !!! #irony)",
	"cite_spans": [],
	"ref_spans": [],
	"eq_spans": [],
	"section": "Irony activation",
	"sec_num": "3.1"
	},
	{
	"text": "On the other hand, IMPLICIT activation arises from a contradiction between a lexicalized proposition P 1 describing an event or state and a pragmatic context P 2 external to the utterance in which P 1 is false, not likely to happen or contrary to the writer's intention. The irony occurs because the writer believes that his audience can detect the disparity between P 1 and P 2 on the basis of contextual knowledge or common background shared with the writer. For example, in (3), the negated fact in P 1 helps to recognize that the tweet is ironic.",
	"cite_spans": [],
	"ref_spans": [],
	"eq_spans": [],
	"section": "Irony activation",
	"sec_num": "3.1"
	},
	{
	"text": "(3) La #NSA a mis sur\u00e9coutes un pays entier. Pas d'inqui\u00e8tude pour la #Belgique: [ce n'est pas un pays entier.] P 1 #ironie (The #NSA wiretapped a whole country. No worries for #Belgium: [it is not a whole country.] P 1 #irony) \u2212\u2192 P 2: Belgium is a country.",
	"cite_spans": [],
	"ref_spans": [],
	"eq_spans": [],
	"section": "Irony activation",
	"sec_num": "3.1"
	},
	{
	"text": "Our categories Usage Metaphor (Ritchie, 2005; Burgers, 2010) Analogy Both :",
	"cite_spans": [
	{
	"start": 30,
	"end": 45,
	"text": "(Ritchie, 2005;",
	"ref_id": "BIBREF36"
	},
	{
	"start": 46,
	"end": 60,
	"text": "Burgers, 2010)",
	"ref_id": "BIBREF10"
	}
	],
	"ref_spans": [],
	"eq_spans": [],
	"section": "State of the art irony categories",
	"sec_num": null
	},
	{
	"text": "Covers analogy, simile, and metaphor. Involves similarity between two things that have different ontological concepts or domains, on which a comparison may be based Hyperbole (Berntsen and Kennedy, 1996; Mercier-Leca, 2003; Didio, 2007) Hyperbole/ Exaggeration Both Make a strong impression or emphasize a point Exaggeration (Didio, 2007) Euphemism (Muecke, 1978; Seto, 1998) Euphemism Both Reduce the facts of an expression or an idea considered unpleasant in order to soften the reality Rhetorical question (Barbe, 1995; Berntsen and Kennedy, 1996) ",
	"cite_spans": [
	{
	"start": 175,
	"end": 203,
	"text": "(Berntsen and Kennedy, 1996;",
	"ref_id": "BIBREF6"
	},
	{
	"start": 204,
	"end": 223,
	"text": "Mercier-Leca, 2003;",
	"ref_id": null
	},
	{
	"start": 224,
	"end": 236,
	"text": "Didio, 2007)",
	"ref_id": "BIBREF16"
	},
	{
	"start": 325,
	"end": 338,
	"text": "(Didio, 2007)",
	"ref_id": "BIBREF16"
	},
	{
	"start": 349,
	"end": 363,
	"text": "(Muecke, 1978;",
	"ref_id": "BIBREF32"
	},
	{
	"start": 364,
	"end": 375,
	"text": "Seto, 1998)",
	"ref_id": "BIBREF37"
	},
	{
	"start": 386,
	"end": 390,
	"text": "Both",
	"ref_id": null
	},
	{
	"start": 509,
	"end": 522,
	"text": "(Barbe, 1995;",
	"ref_id": null
	},
	{
	"start": 523,
	"end": 550,
	"text": "Berntsen and Kennedy, 1996)",
	"ref_id": "BIBREF6"
	}
	],
	"ref_spans": [],
	"eq_spans": [],
	"section": "Metaphor and Comparison",
	"sec_num": null
	},
	{
	"text": "Ask a question in order to make a point rather than to elicit an answer (P1: asking a question to have an answer, P2: no intention to have an answer because it is already known) Context shift (Haiman, 2001; Leech, 2016) Context Shift Exp A sudden change of the topic/frame, use of exaggerated politeness in a situation where this is inappropriate, etc. False logic or misunderstanding (Didio, 2007) False assertion Imp A proposition, fact or an assertion fails to make sense against the reality Oxymoron (Gibbs, 1994; Mercier-Leca, 2003) Oxymoron/ paradox Exp Equivalent to \"False assertion\" except that the contradiction is explicit Paradox (Tayot, 1984; Barbe, 1995) Situational irony (Shelley, 2001; Niogret, 2004) Other Both Humor or situational irony (irony where the incongruity is not due to the use of words but to a non intentional contradiction between two facts or events) Surprise effect, repetition, quotation marks, emoticons, exclamation, capital letter, crossed-out text, special signs (Haiman, 2001; Burgers, 2010) ",
	"cite_spans": [
	{
	"start": 192,
	"end": 206,
	"text": "(Haiman, 2001;",
	"ref_id": "BIBREF22"
	},
	{
	"start": 207,
	"end": 219,
	"text": "Leech, 2016)",
	"ref_id": "BIBREF28"
	},
	{
	"start": 234,
	"end": 237,
	"text": "Exp",
	"ref_id": null
	},
	{
	"start": 385,
	"end": 398,
	"text": "(Didio, 2007)",
	"ref_id": "BIBREF16"
	},
	{
	"start": 504,
	"end": 517,
	"text": "(Gibbs, 1994;",
	"ref_id": "BIBREF18"
	},
	{
	"start": 518,
	"end": 537,
	"text": "Mercier-Leca, 2003)",
	"ref_id": null
	},
	{
	"start": 642,
	"end": 655,
	"text": "(Tayot, 1984;",
	"ref_id": "BIBREF45"
	},
	{
	"start": 656,
	"end": 668,
	"text": "Barbe, 1995)",
	"ref_id": null
	},
	{
	"start": 687,
	"end": 702,
	"text": "(Shelley, 2001;",
	"ref_id": "BIBREF38"
	},
	{
	"start": 703,
	"end": 717,
	"text": "Niogret, 2004)",
	"ref_id": "BIBREF33"
	},
	{
	"start": 724,
	"end": 728,
	"text": "Both",
	"ref_id": null
	},
	{
	"start": 1002,
	"end": 1016,
	"text": "(Haiman, 2001;",
	"ref_id": "BIBREF22"
	},
	{
	"start": 1017,
	"end": 1031,
	"text": "Burgers, 2010)",
	"ref_id": "BIBREF10"
	}
	],
	"ref_spans": [],
	"eq_spans": [],
	"section": "Rhetorical question Both",
	"sec_num": null
	},
	{
	"text": "Words, expressions or symbols used to make a statement ironic Table 2 : Irony categories in our annotation schema.",
	"cite_spans": [],
	"ref_spans": [
	{
	"start": 62,
	"end": 69,
	"text": "Table 2",
	"ref_id": null
	}
	],
	"eq_spans": [],
	"section": "Markers",
	"sec_num": null
	},
	{
	"text": "Note that inferring irony in both types of activation requires some pragmatic knowledge. However, in case of IMPLICIT, the activation of irony happens only if the reader knows the context. To help annotators identify irony activation type, we apply the following rule: if P 1 and P 2 can be found in the tweet, then EXPLICIT, otherwise IMPLICIT.",
	"cite_spans": [],
	"ref_spans": [],
	"eq_spans": [],
	"section": "Markers",
	"sec_num": null
	},
	{
	"text": "Both explicit and implicit activation types can be expressed in different ways which we call irony categories. After a thorough inspection of how categories have been defined in linguistic literature, some of them were grouped, like hyperbole and exaggeration, as we observed that it is very difficult to distinguish them in short messages. We also discarded others, since we considered them as markers rather than irony categories (see the last row in Table 2 ). We finally retain eight categories, as shown in Table 2 : Five are more likely to be found in both types of activation (marked Both) while three may occur exclusively in a specific type (marked Exp for explicit or Imp for implicit).",
	"cite_spans": [],
	"ref_spans": [
	{
	"start": 453,
	"end": 460,
	"text": "Table 2",
	"ref_id": null
	},
	{
	"start": 512,
	"end": 519,
	"text": "Table 2",
	"ref_id": null
	}
	],
	"eq_spans": [],
	"section": "Irony categories",
	"sec_num": "3.2"
	},
	{
	"text": "Categories are not mutually exclusive. Example (5) shows a case of implicit irony activation where the user uses a false assertion P 1 and two rhetorical questions.",
	"cite_spans": [],
	"ref_spans": [],
	"eq_spans": [],
	"section": "Irony categories",
	"sec_num": "3.2"
	},
	{
	"text": "(5) @infos140 @mediapart Serge Dassault ?",
	"cite_spans": [],
	"ref_spans": [],
	"eq_spans": [],
	"section": "Irony categories",
	"sec_num": "3.2"
	},
	{
	"text": "Corruption ? Non ! Il doit y avoir une erreur.",
	"cite_spans": [],
	"ref_spans": [],
	"eq_spans": [],
	"section": "Irony categories",
	"sec_num": "3.2"
	},
	{
	"text": "[C'est l'image m\u00eame de la probit\u00e9 en politique] P 1 #ironie. (@infos140 @mediapart Serge Dassault? Corruption? No ! There must be an error. [He is the perfect image of probity in politics] P 1 #irony) \u2212\u2192 P 2 : Serge Dassault is involved and has been sentenced in many court cases.",
	"cite_spans": [],
	"ref_spans": [],
	"eq_spans": [],
	"section": "Irony categories",
	"sec_num": "3.2"
	},
	{
	"text": "As shown in Table 2 , linguistic literature considers other forms of irony categories, such as surprise effect, repetition, etc. Having a computational perspective in mind, we preferred to clearly distinguish between categories of irony which are pragmatic devices of irony as defined in the previous section, and irony markers which are a set of tokens (words, symbols, propositions) that may activate irony on the basis of the linguistic content of the tweet only. This distinction is also motivated by the fact that markers can either be present in distinct irony categories, not present at all, or present in non ironic tweets as well. Eighteen markers have been selected for our study. Some of them have shown their effectiveness when used as surface features in irony detection such as punctuation marks, capital letters, reporting speech verbs, emoticons, interjections, negations, opinion and emotion words Gonzalez-Ibanez et al., 2011; Reyes et al., 2013; Karoui et al., 2015) . We investigate in addition novel markers (cf. Table 5) : discourse connectives as they usually mark oppositions, argumentation chains and consequences; named entities and personal pronouns, as we assume they can be an indicator of the topic discussed in the tweet (media topic vs. a more personal tweet); URLs as they give contextual information that may help the reader to detect irony; and finally false propositions. These last four markers might be good features for an automatic detection of implicit irony, for example by detecting that an external context is needed. For example, in (2) markers are negations (no, not), punctuation (!, !!!), opinion word (violent) whereas in (3) markers are named entities (NSA, Belgium), negation (no, not) and false proposition (it is not a whole country).",
	"cite_spans": [
	{
	"start": 915,
	"end": 944,
	"text": "Gonzalez-Ibanez et al., 2011;",
	"ref_id": "BIBREF20"
	},
	{
	"start": 945,
	"end": 964,
	"text": "Reyes et al., 2013;",
	"ref_id": "BIBREF35"
	},
	{
	"start": 965,
	"end": 985,
	"text": "Karoui et al., 2015)",
	"ref_id": "BIBREF26"
	}
	],
	"ref_spans": [
	{
	"start": 12,
	"end": 19,
	"text": "Table 2",
	"ref_id": null
	},
	{
	"start": 1034,
	"end": 1042,
	"text": "Table 5)",
	"ref_id": "TABREF7"
	}
	],
	"eq_spans": [],
	"section": "Irony markers",
	"sec_num": "3.3"
	},
	{
	"text": "For each tweet t, the annotation works as follows 2 :",
	"cite_spans": [],
	"ref_spans": [],
	"eq_spans": [],
	"section": "Annotation procedure",
	"sec_num": "4"
	},
	{
	"text": "(a) Classify t into Ironic/Not ironic. In case annotators do not understand the tweet because of cultural references or lack of background knowledge, t can be classified into the No decision class. Note that this third class concerns only French and English corpora since the Italian corpus already has annotations for irony (cf. Section 2).",
	"cite_spans": [],
	"ref_spans": [],
	"eq_spans": [],
	"section": "Annotation procedure",
	"sec_num": "4"
	},
	{
	"text": "(b) If t is ironic, define its activation type: Can P 1 and P 2 be found in the tweet? If yes then explicit, otherwise implicit. Then specify the pragmatic devices used to express irony by selecting one or several categories.",
	"cite_spans": [],
	"ref_spans": [],
	"eq_spans": [],
	"section": "Annotation procedure",
	"sec_num": "4"
	},
	{
	"text": "(c) Identify text spans within the tweet that correspond to a pre-defined list of linguistic markers. Markers are annotated whatever the class of t. This is very important for analyzing the correlation between ironic (vs. non ironic) readings and the presence (vs. absence) of these markers.",
	"cite_spans": [],
	"ref_spans": [],
	"eq_spans": [],
	"section": "Annotation procedure",
	"sec_num": "4"
	},
	{
	"text": "Linguistic markers were automatically identified relying on dedicated resources for each language (opinion and emotion lexicons, intensifiers, interjections, syntactic parsers for named entities, etc.).",
	"cite_spans": [],
	"ref_spans": [],
	"eq_spans": [],
	"section": "Annotation procedure",
	"sec_num": "4"
	},
	{
	"text": "In case of missing markers or erroneous annotations, automatic annotations were manually corrected. Also, to ensure that the annotations were consistent with the instructions given in the manual, common errors are automatically detected: ironic tweets without activation type or irony category, absence of markers, etc. Annotators were asked to correct their errors before continuing to annotate new tweets. In order to evaluate the stability of the schema regarding language variations, we considered first the French set with a total of 2,000 tweets. Such tweets have been randomly selected from the ones collected as described in Section 2. In order to be sure to have a significant amount of ironic samples, 80% of the total tweets to be manually annotated were selected from the ironic set (i.e. tweets explicitly marked with hashtags like #ironie and #sarcasme) 3 . Three French native speakers were involved. The annotation of the French corpus followed a three-step procedure where an intermediate analysis of agreement and disagreement between the annotators was carried out. Annotators were first trained on 100 tweets, then were asked to annotate separately 300 tweets (this step allows to compute inter-annotator agreements, cf. next section), to finally annotate 1,700 tweets. In the last step, a revised version of the schema was provided. The adjudicated annotations performed in the second step are part of the corpus.",
	"cite_spans": [],
	"ref_spans": [],
	"eq_spans": [],
	"section": "Annotation procedure",
	"sec_num": "4"
	},
	{
	"text": "Then we annotated the English and Italian sets in two steps. First, a training phase (100 tweets each) and then the effective annotation, with respectively 550 and 500 tweets. Four native speakers were involved: two for English and two for Italian. All annotators are skilled in linguistics, researchers and PhD students in computational linguistics.",
	"cite_spans": [],
	"ref_spans": [],
	"eq_spans": [],
	"section": "Annotation procedure",
	"sec_num": "4"
	},
	{
	"text": "We report on the reliability of the annotation schema on the French data. Among 300 tweets, annotators agreed on 255 tweets (174 ironic and 63 not ironic), among which 18 have been classified as No decision. We get a Cohen's Kappa of 0.69 for Ironic/Not ironic classification which is a very good score. When compared to gold standard labels, we also obtained a good Kappa measure (0.62), which shows that French irony hashtags are quite reliable. We also noticed that more than 90% of the tweets annotated as No decision due to the lack of external context, are in fact ironic according to gold labels. We however decided to keep them for the experiments.",
	"cite_spans": [],
	"ref_spans": [],
	"eq_spans": [],
	"section": "Qualitative results",
	"sec_num": "5"
	},
	{
	"text": "For EXPLICIT vs. IMPLICIT, agreement on activation type knowing the tweet ironic obtained a Kappa of 0.65. It was interesting to note that implicit activation is the majority (76.42%). We observed the same tendency in the other languages too (cf. next section). This is an important result that shows that annotators are able to identify which are the textual spans that activate the incongruity in ironic tweets, whether explicit or implicit, and we expect automatic systems to do as good as humans, at best.",
	"cite_spans": [],
	"ref_spans": [],
	"eq_spans": [],
	"section": "Qualitative results",
	"sec_num": "5"
	},
	{
	"text": "Finally, for irony category identification, since the same ironic tweet can belong to several irony categories, we computed agreements by counting, for each tweet, the number of common categories and then dividing by the total number of annotated categories. We obtained 0.56 which is moderate. This score reflects the complexity of the identification of pragmatic devices. When similar devices are grouped together (mainly hyperbole/exaggeration and euphemism, as they are used to make the intended meaning either stronger or weaker), the score increases to 0.60.",
	"cite_spans": [],
	"ref_spans": [],
	"eq_spans": [],
	"section": "Qualitative results",
	"sec_num": "5"
	},
	{
	"text": "The main aim of our corpus-based study is to verify if the different linguistic theories and definitions made on irony can be applied to social media, especially to tweets, and to study its portability to several languages. Besides standard frequencies, we provide the correlations between irony activation types and markers and between categories and markers in order to bring out features that could be used in a perspective of automatic irony detection. In each corpus, all the frequencies presented here are statistically significant from what would be expected by chance using the \u03c7 2 test (p < 0.05). Table 3 gives the total number of annotated tweets and the activation type for ironic tweets. We observe that most irony activation types in the French and English corpora are implicit with respectively 73.01% and 66.28% while in the Italian corpus, explicit activation is the majority. Notice that the fact the analysis of the Italian dataset results in a different tendency on this respect can be possibly related to the absence of user-genereted ironic hashtags, while user explicitlty mark the intention to be ironic (see Section 2). Table 4 gives the percentage of tweets belonging to each category of irony split according to explicit vs. implicit activation, when applicable. Higher frequencies are in bold font. We note that oxymoron/paradox is the most frequent category for explicit irony in French, English and Italian. Concerning implicit irony, false assertion and other are the most frequent categories in French and English (other is the most frequent one in English because a majority of implicit ironic tweets use situational irony, e.g. Libertarian Ron Paul condemns Bill Clinton for taking advantage of 20y/o but would not support any law to protect her. #Monica). In Italian, false assertion, analogy and other are the most frequent categories. As classes are not mutually exclusive, there are 64/38 tweets (resp. in French and English) that belong to more than one category for explicit contradiction. The most frequent combinations are oxymoron/rhetorical question and oxymoron/other for both English and French; oxymoron/hyperbole for French and oxymoron/analogy for English. Concerning implicit activation, there are 134/62 tweets (resp. in French and English) that belong to more than one category. The most frequent combinations are false assertion/other and false assertion/hyperbole for both English and French; and analogy/other for English 4 . Table 5 provides the percentage of tweets containing markers for ironic (explicit or implicit) and non ironic tweets (row in gray). In French, intensifiers, punctuation marks and interjections are more frequent in ironic tweets whereas quotations are more frequent in non ironic tweets. In English, discourse connectors, quotations, comparison words and reporting speech verbs are twice as Other shift question assertion F E I F E I F E I F E I F E I F E I F E I F E I Ex 12 17 21 1 6 19 1 1 5 8 2 9 10 15 10 66 81 28 ---21 6 7 Im 2 13 26 ---1 1 4 10 7 5 14 1 12 ---56 20 34 32 65 19 Table 4 : Categories in explicit (Ex) or implicit (Im) activation in French, English and Italian (in %).",
	"cite_spans": [],
	"ref_spans": [
	{
	"start": 607,
	"end": 614,
	"text": "Table 3",
	"ref_id": "TABREF3"
	},
	{
	"start": 1145,
	"end": 1152,
	"text": "Table 4",
	"ref_id": null
	},
	{
	"start": 2481,
	"end": 2488,
	"text": "Table 5",
	"ref_id": "TABREF7"
	},
	{
	"start": 2871,
	"end": 3143,
	"text": "Other shift question assertion F E I F E I F E I F E I F E I F E I F E I F E I Ex 12 17 21 1 6 19 1 1 5 8 2 9 10 15 10 66 81 28 ---21 6 7 Im 2 13 26 ---1 1 4 10 7 5 14 1 12 ---56 20 34 32 65 19 Table 4",
	"ref_id": "TABREF1"
	}
	],
	"eq_spans": [],
	"section": "Quantitative results",
	"sec_num": "6"
	},
	{
	"text": "frequent in ironic tweets as in non ironic tweets whereas is it the opposite for personal pronouns. Note that there is no English ironic tweet containing URL since they were all annotated as no decision because of a lack of knowledge from the annotators who did not understand the tweet and the Web page pointed by the URL. In Italian, most of markers are more frequent in ironic tweets, while some, like quotations and URL, are more frequent in non ironic tweets 5 . Our study of negation as an irony marker actually considers negation words like no and not as well as periphrastic forms of negation such as ne ... pas in French. We however excluded lexical negations such as unreliable, unhappy, etc. We will further refine our analysis by considering more words that introduce negation. Also, regarding personal pronouns, they are more common in French and English than in Italian. Italian being a pro-drop language can in part motivate the difference detected with respect to pronouns.",
	"cite_spans": [],
	"ref_spans": [],
	"eq_spans": [],
	"section": "Ironic",
	"sec_num": null
	},
	{
	"text": "Then, we investigated the correlation between irony markers and irony activation types (resp. between irony markers and irony categories). Our aim is to analyze to what extent these markers can be indicators for irony prediction. Using the Cramer's V test (Cohen, 1988) on the number of occurrences of each marker, we found a statistically significant (p < 0.05) large correlation between markers and ironic/not ironic class for French (V = 0.156, df = 14) and Italian (V = 0.31, df = 6); between medium and large for English (V = 0.132, df = 9). We also found a large correlation between markers and irony activation types for French (V = 0.196, df = 16) , between medium and large for Italian (V = 0.138, df = 5) and medium for English (V = 0.083, df = 12). 6 We also analyzed the correlations per marker (df =1). The markers which are the most corre-lated to ironic/non ironic class are: negations, interjections, named entities and URL for French (0.140 < V < 0.410); negations, discourse connectors and personal pronouns for English (0.120 < V < 0.170); and quotations, named entities and URL for Italian (0.310 < V < 0.416).",
	"cite_spans": [
	{
	"start": 256,
	"end": 269,
	"text": "(Cohen, 1988)",
	"ref_id": "BIBREF14"
	},
	{
	"start": 635,
	"end": 655,
	"text": "(V = 0.196, df = 16)",
	"ref_id": null
	},
	{
	"start": 760,
	"end": 761,
	"text": "6",
	"ref_id": null
	}
	],
	"ref_spans": [],
	"eq_spans": [],
	"section": "Ironic",
	"sec_num": null
	},
	{
	"text": "The markers which are the most correlated to explicit/implicit activation are: opposition markers, comparison words and false assertion for French (0.140 < V < 0.190); opposition markers and discourse connectors for English (0.110 < V < 0.120); and discourse connectors, punctuation and named entities for Italian (0.136 < V < 0.213). Note that even if opinion words are very frequent in ironic tweets, they are however not correlated with either irony/non irony classification or explicit/implicit activation (V < 0.06), as many non ironic tweets also contain sentiment words.",
	"cite_spans": [],
	"ref_spans": [],
	"eq_spans": [],
	"section": "Ironic",
	"sec_num": null
	},
	{
	"text": "Finally, when analyzing which markers are correlated to irony categories, the more discriminant markers are: intensifiers, punctuation, false assertion and opinion words for French (large Cramer's V); negations, discourse connectors and personal pronouns for English (medium Cramer's V); and punctuation, interjections and named entities for Italian (medium Cramer's V).",
	"cite_spans": [],
	"ref_spans": [],
	"eq_spans": [],
	"section": "Ironic",
	"sec_num": null
	},
	{
	"text": "Most state of the art approaches rely on automatically built social media data collections to detect irony using a variety of features gleaned from the utterance-internal context going from ngram models, stylistic, to dictionary-based features (Burfoot and Baldwin, 2009; Gonzalez-Ibanez et al., 2011; Liebrecht et al., 2013; Joshi et al., 2015; Hern\u00e1ndez Far\u00edas et al., 2015) . In addition to the above more lexical features, many authors point out the contribution of pragmatic features, such as the use of common vs. rare words or synonyms (Barbieri and Saggion, 2014) . Recent work explores other kinds of contextual information like author profiles, conversational threads, or querying external sources of information (Bamman and Smith, 2015; Wallace et al., 2015; Karoui et F E I F E I F E I F E I F E I F E I F E I F E I F E I F E I Ex 7 2 1 37 58 15 6 41 29 2 14 -22 9 2 51 30 14 8 0 -3 0 -0 2 3 6 21 3 Im 6 4 7 34 61 9 4 29 16 4 15 -19 12 0 51 28 5 54 18 -3 3 -0 2 6 6 21 6 NI 5 10 0 58 75 9 4 13 18 0 0 -11 9 0 28 30 17 0 0 -2 0 -1 6 3 1 10 9 18 4 3 8 -31 21 5 14 2 11 8 8 4 97 100 65 1 17 0 48 75 -33 0 10 Im 3 11 6 2 6 -31 24 3 12 0 13 2 12 3 91 97 43 1 14 0 41 74 -29 0 2 NI 4 14 4 3 3 -30 40 1 2 2 12 4 6 1 82 88 98 3 7 1 35 68 -42 0 44 ",
	"cite_spans": [
	{
	"start": 244,
	"end": 271,
	"text": "(Burfoot and Baldwin, 2009;",
	"ref_id": "BIBREF9"
	},
	{
	"start": 272,
	"end": 301,
	"text": "Gonzalez-Ibanez et al., 2011;",
	"ref_id": "BIBREF20"
	},
	{
	"start": 302,
	"end": 325,
	"text": "Liebrecht et al., 2013;",
	"ref_id": "BIBREF29"
	},
	{
	"start": 326,
	"end": 345,
	"text": "Joshi et al., 2015;",
	"ref_id": "BIBREF25"
	},
	{
	"start": 346,
	"end": 376,
	"text": "Hern\u00e1ndez Far\u00edas et al., 2015)",
	"ref_id": "BIBREF23"
	},
	{
	"start": 543,
	"end": 571,
	"text": "(Barbieri and Saggion, 2014)",
	"ref_id": "BIBREF2"
	},
	{
	"start": 723,
	"end": 747,
	"text": "(Bamman and Smith, 2015;",
	"ref_id": "BIBREF0"
	},
	{
	"start": 748,
	"end": 769,
	"text": "Wallace et al., 2015;",
	"ref_id": "BIBREF49"
	},
	{
	"start": 770,
	"end": 779,
	"text": "Karoui et",
	"ref_id": "BIBREF26"
	}
	],
	"ref_spans": [
	{
	"start": 780,
	"end": 1107,
	"text": "F E I F E I F E I F E I F E I F E I F E I F E I F E I F E I Ex 7 2 1 37 58 15 6 41 29 2 14 -22 9 2 51 30 14 8 0 -3 0 -0 2 3 6 21 3 Im 6 4 7 34 61 9 4 29 16 4 15 -19 12 0 51 28 5 54 18 -3 3 -0 2 6 6 21 6 NI 5 10 0 58 75 9 4 13 18 0 0 -11 9 0 28 30 17 0 0 -2 0 -1 6 3 1 10",
	"ref_id": "TABREF1"
	},
	{
	"start": 1108,
	"end": 1331,
	"text": "9 18 4 3 8 -31 21 5 14 2 11 8 8 4 97 100 65 1 17 0 48 75 -33 0 10 Im 3 11 6 2 6 -31 24 3 12 0 13 2 12 3 91 97 43 1 14 0 41 74 -29 0 2 NI 4 14 4 3 3 -30 40 1 2 2 12 4 6 1 82 88 98 3 7 1",
	"ref_id": "TABREF1"
	}
	],
	"eq_spans": [],
	"section": "Related work",
	"sec_num": "7"
	},
	{
	"text": "Compared to automatic irony detection, little efforts have been done on corpus-based linguistic study of irony. Most of these efforts focus on analyzing the impact of irony in feeling expressions and emotions, by manually annotating tweets at both sentiment polarity and irony levels. E.g. Van Hee et al. (2016) distinguish between ironic, possibly ironic, and non-ironic tweets in English and Dutch. For ironic statements, polarity change that causes irony was annotated to specify whether the change comes from an opposition explicitly marked by a contrast between a positive situation and a negative one, an hyperbole, or an understatement. Stranisci et al. (2016) recently extend the Italian Senti-TUT schema (cf. Section 2) to mark the aspects of the topic being discussed in the tweet, as well as the sentiment expressed towards each aspect. Bosco et al. (2016) propose a second extension with the annotation of French tweets using three labels: positive irony, negative irony, and metaphorical expression. Current state of the art corpus-based studies are mainly oriented to a sentiment analysis perspective on irony, focusing almost exclusively on cap-turing tweet's overall sentiment, explicit polarity change, or syntactic irony patterns. We argue in this paper that irony should instead be an object of study by its own by proposing a more linguistic perspective in order to provide a deeper inspection of what are the inferential mechanisms that activate irony, either explicit or implicit, and the correlations between irony types and irony markers. As far as we know, this is the first study that investigates the portability of a wide-range of pragmatic devices in the interpretation of irony to social media data from a multilingual perspective.",
	"cite_spans": [
	{
	"start": 290,
	"end": 311,
	"text": "Van Hee et al. (2016)",
	"ref_id": "BIBREF48"
	},
	{
	"start": 644,
	"end": 667,
	"text": "Stranisci et al. (2016)",
	"ref_id": null
	},
	{
	"start": 848,
	"end": 867,
	"text": "Bosco et al. (2016)",
	"ref_id": "BIBREF8"
	}
	],
	"ref_spans": [],
	"eq_spans": [],
	"section": "2015).",
	"sec_num": null
	},
	{
	"text": "The French and Italian parts of the annotated corpus have been respectively exploited as datasets for the first irony detection shared tasks DEFT@TALN2017 7 and for the SEN-TIPOLC@Evalita shared task on irony detection 8 in both 2014 and 2016 editions (Basile et al., 2014; Barbieri et al., 2016) . In particular, currently only the first layer of the annotation scheme has been exploited aiming at detecting if a given tweet is ironic or not. The French task is ongoing. For what concerns Italian, in Sentipolc the irony detection task is one three related but independent subtasks focusing on subjectivity, polarity and irony detection, respectively. All tweets of the campaign are, therefore, annotated by a multi-layered annotation scheme including tags for all the three dimensions and available on the Task's website. In 2016 SENTIPOLC has been the most participated EVALITA task with a total of 57 submitted runs from 13 different teams. Not surprisingly, results of the 12 systems evaluated for irony detection seem to suggest that the task appears truly challenging. However, organizers observe that its complexity does not depend (only) on the inner structure of irony, but on unbalanced data distribution in Sentipolc (1 out of 7 examples is ironic in the training set, as they reflect the distribution in a realistic scenario) and on the overall availability of a limited amount of examples (probably not sufficient to generalise over the structure of ironic tweets). The plan is to organize an irony detection dedicated task including a larger and more balanced dataset of ironic tweets in future campaigns. In this perspective, it will be also interesting to investigate if the finer-grained annotation layers for irony proposed here can have a role in the annotation scheme proposed for the new task data.",
	"cite_spans": [
	{
	"start": 252,
	"end": 273,
	"text": "(Basile et al., 2014;",
	"ref_id": "BIBREF5"
	},
	{
	"start": 274,
	"end": 296,
	"text": "Barbieri et al., 2016)",
	"ref_id": "BIBREF3"
	}
	],
	"ref_spans": [],
	"eq_spans": [],
	"section": "Exploiting the annotated corpus for automatic irony detection",
	"sec_num": "8"
	},
	{
	"text": "In this paper, we proposed a multi-layered annotation schema for irony in tweets and a multilingual corpus-based study for measuring the impact of pragmatic phenomena in the interpretation of irony. The results show that our schema is reliable for French and that it is portable to English and Italian, observing relatively the same tendencies in terms of irony categories and markers. We observed correlations between markers and ironic/non ironic classes, between markers and irony activation types (explicit or implicit) and between markers and irony categories. These observations are interesting in a perspective of pragmatically and linguistically informed automatic irony detection, since it brings out the most discriminant features. On this line, we plan to accomplish a validation of the schema based on the definition of an automatic classification model built upon such annotated features. Moreover, an interesting challenge could be to apply the annota-tion schema to a new language also less culturally close to those addressed in this work.",
	"cite_spans": [],
	"ref_spans": [],
	"eq_spans": [],
	"section": "Conclusion and future work",
	"sec_num": "9"
	},
	{
	"text": "Finally, another perspective is to investigate how the application of our schema can contribute to shed light on the issue of distinguishing between irony and sarcasm. This issue is challenging, and only recently addressed from computational linguistics. In particular, new data-driven arguments for a possible separation between irony and sarcasm emerged from recent work on Twitter data (Sulis et al., 2016) . It could be interesting to see the relation between the finer-grained and pragmatic phenomena related to irony investigated in the present study and the higher-level distinction between irony and sarcasm.",
	"cite_spans": [
	{
	"start": 389,
	"end": 409,
	"text": "(Sulis et al., 2016)",
	"ref_id": "BIBREF43"
	}
	],
	"ref_spans": [],
	"eq_spans": [],
	"section": "Conclusion and future work",
	"sec_num": "9"
	},
	{
	"text": "The annotation manual is available at: github.com/ IronyAndTweets/Scheme",
	"cite_spans": [],
	"ref_spans": [],
	"eq_spans": [],
	"section": "",
	"sec_num": null
	},
	{
	"text": "Notice that at this stage such hashtags have been removed, and manual annotation have been applied to 2,000 tweets for all the layers foreseen by our schema. In this way, the reliability of self-tagging has been confirmed, and it was possible to identify the presence of irony also in tweets where it was not explicitly marked by hashtags.",
	"cite_spans": [],
	"ref_spans": [],
	"eq_spans": [],
	"section": "",
	"sec_num": null
	},
	{
	"text": "For what concerns Italian, at the current stage, only the category considered prevalent for implicit/explicit irony activation was annotated .",
	"cite_spans": [],
	"ref_spans": [],
	"eq_spans": [],
	"section": "",
	"sec_num": null
	},
	{
	"text": "For Italian, only values for markers automatically identified reliably, without need of manual correction, are reported (e.g. emoticons, negations). Values for other markers are currently missing since they require a manual check, for instance the case of capital letters, because of the presence in the Italian corpus where all the letters are capital.6 For both settings, frequencies < 5 were removed.",
	"cite_spans": [],
	"ref_spans": [],
	"eq_spans": [],
	"section": "",
	"sec_num": null
	},
	{
	"text": "https://deft.limsi.fr/2017/ 8 http://di.unito.it/sentipolc16",
	"cite_spans": [],
	"ref_spans": [],
	"eq_spans": [],
	"section": "",
	"sec_num": null
	}
	],
	"back_matter": [],
	"bib_entries": {
	"BIBREF0": {
	"ref_id": "b0",
	"title": "Contextualized sarcasm detection on Twitter",
	"authors": [
	{
	"first": "David",
	"middle": [],
	"last": "Bamman",
	"suffix": ""
	},
	{
	"first": "Noah",
	"middle": [
	"A"
	],
	"last": "Smith",
	"suffix": ""
	}
	],
	"year": 2015,
	"venue": "Proceedings of the International Conference on Web and Social Media, ICWSM 2015",
	"volume": "",
	"issue": "",
	"pages": "574--577",
	"other_ids": {},
	"num": null,
	"urls": [],
	"raw_text": "David Bamman and Noah A. Smith. 2015. Contextu- alized sarcasm detection on Twitter. In Proceedings of the International Conference on Web and Social Media, ICWSM 2015, pages 574-577.",
	"links": null
	},
	"BIBREF2": {
	"ref_id": "b2",
	"title": "Modelling Irony in Twitter: Feature Analysis and Evaluation",
	"authors": [
	{
	"first": "Francesco",
	"middle": [],
	"last": "Barbieri",
	"suffix": ""
	},
	{
	"first": "Horacio",
	"middle": [],
	"last": "Saggion",
	"suffix": ""
	}
	],
	"year": 2014,
	"venue": "Proceedings of Language Resources and Evaluation Conference (LREC)",
	"volume": "",
	"issue": "",
	"pages": "4258--4264",
	"other_ids": {},
	"num": null,
	"urls": [],
	"raw_text": "Francesco Barbieri and Horacio Saggion. 2014. Mod- elling Irony in Twitter: Feature Analysis and Eval- uation. In Proceedings of Language Resources and Evaluation Conference (LREC), pages 4258-4264.",
	"links": null
	},
	"BIBREF3": {
	"ref_id": "b3",
	"title": "Overview of the Evalita 2016 SENTIment POLarity Classification Task",
	"authors": [
	{
	"first": "Francesco",
	"middle": [],
	"last": "Barbieri",
	"suffix": ""
	},
	{
	"first": "Valerio",
	"middle": [],
	"last": "Basile",
	"suffix": ""
	},
	{
	"first": "Danilo",
	"middle": [],
	"last": "Croce",
	"suffix": ""
	},
	{
	"first": "Malvina",
	"middle": [],
	"last": "Nissim",
	"suffix": ""
	},
	{
	"first": "Nicole",
	"middle": [],
	"last": "Novielli",
	"suffix": ""
	},
	{
	"first": "Viviana",
	"middle": [],
	"last": "Patti",
	"suffix": ""
	}
	],
	"year": 2016,
	"venue": "Proceedings of Third Italian Conference on Computational Linguistics (CLiC-it 2016) & Fifth Evaluation Campaign of Natural Language Processing and Speech Tools for Italian. Final Workshop (EVALITA 2016)",
	"volume": "1749",
	"issue": "",
	"pages": "",
	"other_ids": {},
	"num": null,
	"urls": [],
	"raw_text": "Francesco Barbieri, Valerio Basile, Danilo Croce, Malvina Nissim, Nicole Novielli, and Viviana Patti. 2016. Overview of the Evalita 2016 SENTIment POLarity Classification Task. In Proceedings of Third Italian Conference on Computational Linguis- tics (CLiC-it 2016) & Fifth Evaluation Campaign of Natural Language Processing and Speech Tools for Italian. Final Workshop (EVALITA 2016), Napoli, Italy, December 5-7, 2016., volume 1749 of CEUR Workshop Proceedings. CEUR-WS.org.",
	"links": null
	},
	"BIBREF4": {
	"ref_id": "b4",
	"title": "Sentiment analysis on italian tweets",
	"authors": [
	{
	"first": "Valerio",
	"middle": [],
	"last": "Basile",
	"suffix": ""
	},
	{
	"first": "Malvina",
	"middle": [],
	"last": "Nissim",
	"suffix": ""
	}
	],
	"year": 2013,
	"venue": "Proceedings of of WASSA 2013",
	"volume": "",
	"issue": "",
	"pages": "100--107",
	"other_ids": {},
	"num": null,
	"urls": [],
	"raw_text": "Valerio Basile and Malvina Nissim. 2013. Senti- ment analysis on italian tweets. In Proceedings of of WASSA 2013, pages 100-107.",
	"links": null
	},
	"BIBREF5": {
	"ref_id": "b5",
	"title": "Overview of the Evalita 2014 SENTIment POLarity Classification Task",
	"authors": [
	{
	"first": "Valerio",
	"middle": [],
	"last": "Basile",
	"suffix": ""
	},
	{
	"first": "Andrea",
	"middle": [],
	"last": "Bolioli",
	"suffix": ""
	},
	{
	"first": "Malvina",
	"middle": [],
	"last": "Nissim",
	"suffix": ""
	},
	{
	"first": "Viviana",
	"middle": [],
	"last": "Patti",
	"suffix": ""
	},
	{
	"first": "Paolo",
	"middle": [],
	"last": "Rosso",
	"suffix": ""
	}
	],
	"year": 2014,
	"venue": "Proc. of EVALITA 2014",
	"volume": "",
	"issue": "",
	"pages": "50--57",
	"other_ids": {},
	"num": null,
	"urls": [],
	"raw_text": "Valerio Basile, Andrea Bolioli, Malvina Nissim, Vi- viana Patti, and Paolo Rosso. 2014. Overview of the Evalita 2014 SENTIment POLarity Classifica- tion Task. In Proc. of EVALITA 2014, pages 50-57, Pisa, Italy. Pisa University Press.",
	"links": null
	},
	"BIBREF6": {
	"ref_id": "b6",
	"title": "Unresolved contradictions specifying attitudesin metaphor, irony, understatement and tautology",
	"authors": [
	{
	"first": "Dorthe",
	"middle": [],
	"last": "Berntsen",
	"suffix": ""
	},
	{
	"first": "John",
	"middle": [
	"M"
	],
	"last": "Kennedy",
	"suffix": ""
	}
	],
	"year": 1996,
	"venue": "",
	"volume": "24",
	"issue": "",
	"pages": "13--29",
	"other_ids": {},
	"num": null,
	"urls": [],
	"raw_text": "Dorthe Berntsen and John M. Kennedy. 1996. Unresolved contradictions specifying attitudesin metaphor, irony, understatement and tautology. Po- etics, 24(1):13-29.",
	"links": null
	},
	"BIBREF7": {
	"ref_id": "b7",
	"title": "Developing Corpora for Sentiment Analysis: The Case of Irony and Senti-TUT",
	"authors": [
	{
	"first": "Cristina",
	"middle": [],
	"last": "Bosco",
	"suffix": ""
	},
	{
	"first": "Viviana",
	"middle": [],
	"last": "Patti",
	"suffix": ""
	},
	{
	"first": "Andrea",
	"middle": [],
	"last": "Bolioli",
	"suffix": ""
	}
	],
	"year": 2013,
	"venue": "IEEE Intelligent Systems",
	"volume": "28",
	"issue": "2",
	"pages": "55--63",
	"other_ids": {},
	"num": null,
	"urls": [],
	"raw_text": "Cristina Bosco, Viviana Patti, and Andrea Bolioli. 2013. Developing Corpora for Sentiment Analysis: The Case of Irony and Senti-TUT. IEEE Intelligent Systems, 28(2):55-63, March.",
	"links": null
	},
	"BIBREF8": {
	"ref_id": "b8",
	"title": "Tweeting and Being Ironic in the Debate about a Political Reform: the French Annotated Corpus TWitter-MariagePourTous",
	"authors": [
	{
	"first": "Cristina",
	"middle": [],
	"last": "Bosco",
	"suffix": ""
	},
	{
	"first": "Mirko",
	"middle": [],
	"last": "Lai",
	"suffix": ""
	},
	{
	"first": "Viviana",
	"middle": [],
	"last": "Patti",
	"suffix": ""
	},
	{
	"first": "Daniela",
	"middle": [],
	"last": "Virone",
	"suffix": ""
	}
	],
	"year": 2016,
	"venue": "Proceedings of the Tenth International Conference on Language Resources and Evaluation",
	"volume": "",
	"issue": "",
	"pages": "",
	"other_ids": {},
	"num": null,
	"urls": [],
	"raw_text": "Cristina Bosco, Mirko Lai, Viviana Patti, and Daniela Virone. 2016. Tweeting and Being Ironic in the De- bate about a Political Reform: the French Annotated Corpus TWitter-MariagePourTous. In Proceedings of the Tenth International Conference on Language Resources and Evaluation (LREC 2016).",
	"links": null
	},
	"BIBREF9": {
	"ref_id": "b9",
	"title": "Automatic satire detection: Are you having a laugh?",
	"authors": [
	{
	"first": "Clint",
	"middle": [],
	"last": "Burfoot",
	"suffix": ""
	},
	{
	"first": "Timothy",
	"middle": [],
	"last": "Baldwin",
	"suffix": ""
	}
	],
	"year": 2009,
	"venue": "Proceedings of the ACL-IJCNLP 2009 Conference Short Papers",
	"volume": "",
	"issue": "",
	"pages": "161--164",
	"other_ids": {},
	"num": null,
	"urls": [],
	"raw_text": "Clint Burfoot and Timothy Baldwin. 2009. Automatic satire detection: Are you having a laugh? In Pro- ceedings of the ACL-IJCNLP 2009 Conference Short Papers, pages 161-164, Suntec, Singapore, August. Association for Computational Linguistics.",
	"links": null
	},
	"BIBREF10": {
	"ref_id": "b10",
	"title": "Verbal irony: Use and effects in written discourse",
	"authors": [
	{
	"first": "Christian",
	"middle": [],
	"last": "Burgers",
	"suffix": ""
	}
	],
	"year": 2010,
	"venue": "",
	"volume": "",
	"issue": "",
	"pages": "",
	"other_ids": {},
	"num": null,
	"urls": [],
	"raw_text": "Christian Burgers. 2010. Verbal irony: Use and effects in written discourse. Ph.D. thesis, Radboud Univer- siteit Nijmegen.",
	"links": null
	},
	"BIBREF11": {
	"ref_id": "b11",
	"title": "An impact analysis of features in a classification approach to irony detection in product reviews",
	"authors": [
	{
	"first": "Konstantin",
	"middle": [],
	"last": "Buschmeier",
	"suffix": ""
	},
	{
	"first": "Philipp",
	"middle": [],
	"last": "Cimiano",
	"suffix": ""
	},
	{
	"first": "Roman",
	"middle": [],
	"last": "Klinger",
	"suffix": ""
	}
	],
	"year": 2014,
	"venue": "Proceedings of the 5th Workshop on Computational Approaches to Subjectivity, Sentiment and Social Media Analysis",
	"volume": "",
	"issue": "",
	"pages": "42--49",
	"other_ids": {},
	"num": null,
	"urls": [],
	"raw_text": "Konstantin Buschmeier, Philipp Cimiano, and Roman Klinger. 2014. An impact analysis of features in a classification approach to irony detection in prod- uct reviews. In Proceedings of the 5th Workshop on Computational Approaches to Subjectivity, Senti- ment and Social Media Analysis, pages 42-49, Bal- timore, Maryland, June. ACL.",
	"links": null
	},
	"BIBREF12": {
	"ref_id": "b12",
	"title": "On the pretense theory of irony",
	"authors": [
	{
	"first": "H",
	"middle": [],
	"last": "Herbert",
	"suffix": ""
	},
	{
	"first": "Richard",
	"middle": [
	"J"
	],
	"last": "Clark",
	"suffix": ""
	},
	{
	"first": "",
	"middle": [],
	"last": "Gerrig",
	"suffix": ""
	}
	],
	"year": 1984,
	"venue": "Journal of Experimental Psychology: General",
	"volume": "113",
	"issue": "1",
	"pages": "121--126",
	"other_ids": {},
	"num": null,
	"urls": [],
	"raw_text": "Herbert H. Clark and Richard J. Gerrig. 1984. On the pretense theory of irony. Journal of Experimental Psychology: General, 113(1):121-126.",
	"links": null
	},
	"BIBREF13": {
	"ref_id": "b13",
	"title": "Irony in conversation",
	"authors": [
	{
	"first": "Rebecca",
	"middle": [],
	"last": "Clift",
	"suffix": ""
	}
	],
	"year": 1999,
	"venue": "Language in Society",
	"volume": "28",
	"issue": "",
	"pages": "523--553",
	"other_ids": {},
	"num": null,
	"urls": [],
	"raw_text": "Rebecca Clift. 1999. Irony in conversation. Language in Society, 28:523-553.",
	"links": null
	},
	"BIBREF14": {
	"ref_id": "b14",
	"title": "Statistical Power Analysis for the Behavioral Sciences Second Edition",
	"authors": [
	{
	"first": "Jacob",
	"middle": [],
	"last": "Cohen",
	"suffix": ""
	}
	],
	"year": 1988,
	"venue": "",
	"volume": "",
	"issue": "",
	"pages": "",
	"other_ids": {},
	"num": null,
	"urls": [],
	"raw_text": "Jacob Cohen. 1988. Statistical Power Analysis for the Behavioral Sciences Second Edition. Lawrence Erl- baum Associates.",
	"links": null
	},
	"BIBREF15": {
	"ref_id": "b15",
	"title": "Semi-Supervised Recognition of Sarcasm in Twitter and Amazon",
	"authors": [
	{
	"first": "Dmitry",
	"middle": [],
	"last": "Davidov",
	"suffix": ""
	},
	{
	"first": "Oren",
	"middle": [],
	"last": "Tsur",
	"suffix": ""
	},
	{
	"first": "Ari",
	"middle": [],
	"last": "Rappoport",
	"suffix": ""
	}
	],
	"year": 2010,
	"venue": "Proceedings of the Fourteenth Conference on Computational Natural Language Learning",
	"volume": "",
	"issue": "",
	"pages": "107--116",
	"other_ids": {},
	"num": null,
	"urls": [],
	"raw_text": "Dmitry Davidov, Oren Tsur, and Ari Rappoport. 2010. Semi-Supervised Recognition of Sarcasm in Twitter and Amazon. In Proceedings of the Fourteenth Con- ference on Computational Natural Language Learn- ing, pages 107-116, Uppsala, Sweden, July. Associ- ation for Computational Linguistics.",
	"links": null
	},
	"BIBREF16": {
	"ref_id": "b16",
	"title": "Une approche s\u00e9manticos\u00e9miotique de l'ironie",
	"authors": [
	{
	"first": "Lucie",
	"middle": [],
	"last": "Didio",
	"suffix": ""
	}
	],
	"year": 2007,
	"venue": "",
	"volume": "",
	"issue": "",
	"pages": "",
	"other_ids": {},
	"num": null,
	"urls": [],
	"raw_text": "Lucie Didio. 2007. Une approche s\u00e9mantico- s\u00e9miotique de l'ironie. Ph.D. thesis, Universit\u00e9 de Limoges.",
	"links": null
	},
	"BIBREF17": {
	"ref_id": "b17",
	"title": "Semeval-2015 task 11: Sentiment Analysis of Figurative Language in Twitter",
	"authors": [
	{
	"first": "Aniruddha",
	"middle": [],
	"last": "Ghosh",
	"suffix": ""
	},
	{
	"first": "Guofu",
	"middle": [],
	"last": "Li",
	"suffix": ""
	},
	{
	"first": "Tony",
	"middle": [],
	"last": "Veale",
	"suffix": ""
	},
	{
	"first": "Paolo",
	"middle": [],
	"last": "Rosso",
	"suffix": ""
	},
	{
	"first": "Ekaterina",
	"middle": [],
	"last": "Shutova",
	"suffix": ""
	},
	{
	"first": "John",
	"middle": [],
	"last": "Barnden",
	"suffix": ""
	},
	{
	"first": "Antonio",
	"middle": [],
	"last": "Reyes",
	"suffix": ""
	}
	],
	"year": 2015,
	"venue": "",
	"volume": "",
	"issue": "",
	"pages": "",
	"other_ids": {},
	"num": null,
	"urls": [],
	"raw_text": "Aniruddha Ghosh, Guofu Li, Tony Veale, Paolo Rosso, Ekaterina Shutova, John Barnden, and Antonio Reyes. 2015. Semeval-2015 task 11: Sentiment Analysis of Figurative Language in Twitter. In Pro- ceedings of SemEval 2015, Co-located with NAACL, page 470478. ACL.",
	"links": null
	},
	"BIBREF18": {
	"ref_id": "b18",
	"title": "The poetics of mind: Figurative thought, language, and understanding",
	"authors": [
	{
	"first": "Raymond",
	"middle": [
	"W"
	],
	"last": "Gibbs",
	"suffix": ""
	}
	],
	"year": 1994,
	"venue": "",
	"volume": "",
	"issue": "",
	"pages": "",
	"other_ids": {},
	"num": null,
	"urls": [],
	"raw_text": "Raymond W. Gibbs. 1994. The poetics of mind: Figu- rative thought, language, and understanding. Cam- bridge University Press.",
	"links": null
	},
	"BIBREF19": {
	"ref_id": "b19",
	"title": "Irony in talk among friends. Metaphor and symbol",
	"authors": [
	{
	"first": "Raymond",
	"middle": [
	"W"
	],
	"last": "Gibbs",
	"suffix": ""
	}
	],
	"year": 2000,
	"venue": "",
	"volume": "15",
	"issue": "",
	"pages": "5--27",
	"other_ids": {},
	"num": null,
	"urls": [],
	"raw_text": "Raymond W. Gibbs. 2000. Irony in talk among friends. Metaphor and symbol, 15(1-2):5-27.",
	"links": null
	},
	"BIBREF20": {
	"ref_id": "b20",
	"title": "Identifying sarcasm in Twitter: a closer look",
	"authors": [
	{
	"first": "Roberto",
	"middle": [],
	"last": "Gonzalez-Ibanez",
	"suffix": ""
	},
	{
	"first": "Smaranda",
	"middle": [],
	"last": "Muresan",
	"suffix": ""
	},
	{
	"first": "Nina",
	"middle": [],
	"last": "Wacholde",
	"suffix": ""
	}
	],
	"year": 2011,
	"venue": "Proceedings of the 49th Annual Meeting of the Association for Computational Linguistics: Human Language Technologies: short papers",
	"volume": "2",
	"issue": "",
	"pages": "581--586",
	"other_ids": {},
	"num": null,
	"urls": [],
	"raw_text": "Roberto Gonzalez-Ibanez, Smaranda Muresan, and Nina Wacholde. 2011. Identifying sarcasm in Twit- ter: a closer look. In Proceedings of the 49th An- nual Meeting of the Association for Computational Linguistics: Human Language Technologies: short papers-Volume 2, pages 581-586. Association for Computational Linguistics.",
	"links": null
	},
	"BIBREF21": {
	"ref_id": "b21",
	"title": "Syntax and semantics. Logic and conversation",
	"authors": [
	{
	"first": "Peter",
	"middle": [],
	"last": "Herbert Paul Grice",
	"suffix": ""
	},
	{
	"first": "Jerry",
	"middle": [
	"L"
	],
	"last": "Cole",
	"suffix": ""
	},
	{
	"first": "",
	"middle": [],
	"last": "Morgan",
	"suffix": ""
	}
	],
	"year": 1975,
	"venue": "",
	"volume": "3",
	"issue": "",
	"pages": "41--58",
	"other_ids": {},
	"num": null,
	"urls": [],
	"raw_text": "Herbert Paul Grice, Peter Cole, and Jerry L. Morgan. 1975. Syntax and semantics. Logic and conversa- tion, 3:41-58.",
	"links": null
	},
	"BIBREF22": {
	"ref_id": "b22",
	"title": "Talk is cheap: Sarcasm, alienation, and the evolution of language",
	"authors": [
	{
	"first": "John",
	"middle": [],
	"last": "Haiman",
	"suffix": ""
	}
	],
	"year": 2001,
	"venue": "",
	"volume": "",
	"issue": "",
	"pages": "",
	"other_ids": {},
	"num": null,
	"urls": [],
	"raw_text": "John Haiman. 2001. Talk is cheap: Sarcasm, alien- ation, and the evolution of language. Oxford Uni- versity Press, USA.",
	"links": null
	},
	"BIBREF23": {
	"ref_id": "b23",
	"title": "Valento: Sentiment analysis of figurative language tweets with irony and sarcasm",
	"authors": [
	{
	"first": "Delia Iraz\u00fa Hern\u00e1ndez",
	"middle": [],
	"last": "Far\u00edas",
	"suffix": ""
	},
	{
	"first": "Emilio",
	"middle": [],
	"last": "Sulis",
	"suffix": ""
	},
	{
	"first": "Viviana",
	"middle": [],
	"last": "Patti",
	"suffix": ""
	},
	{
	"first": "Giancarlo",
	"middle": [],
	"last": "Ruffo",
	"suffix": ""
	},
	{
	"first": "Cristina",
	"middle": [],
	"last": "Bosco",
	"suffix": ""
	}
	],
	"year": 2015,
	"venue": "Proceedings of the 9th International Workshop on Semantic Evaluation",
	"volume": "",
	"issue": "",
	"pages": "694--698",
	"other_ids": {},
	"num": null,
	"urls": [],
	"raw_text": "Delia Iraz\u00fa Hern\u00e1ndez Far\u00edas, Emilio Sulis, Viviana Patti, Giancarlo Ruffo, and Cristina Bosco. 2015. Valento: Sentiment analysis of figurative language tweets with irony and sarcasm. In Proceedings of the 9th International Workshop on Semantic Evalu- ation (SemEval 2015), pages 694-698, Denver, Col- orado, June. ACL.",
	"links": null
	},
	"BIBREF24": {
	"ref_id": "b24",
	"title": "Irony Detection in Twitter: The Role of Affective Content",
	"authors": [
	{
	"first": "Delia Iraz\u00fa Hern\u00e1ndez",
	"middle": [],
	"last": "Far\u00edas",
	"suffix": ""
	},
	{
	"first": "Viviana",
	"middle": [],
	"last": "Patti",
	"suffix": ""
	},
	{
	"first": "Paolo",
	"middle": [],
	"last": "Rosso",
	"suffix": ""
	}
	],
	"year": 2016,
	"venue": "ACM Transactions on Internet Technologies",
	"volume": "16",
	"issue": "3",
	"pages": "",
	"other_ids": {},
	"num": null,
	"urls": [],
	"raw_text": "Delia Iraz\u00fa Hern\u00e1ndez Far\u00edas, Viviana Patti, and Paolo Rosso. 2016. Irony Detection in Twitter: The Role of Affective Content. ACM Transactions on Internet Technologies, 16(3):19:1-19:24.",
	"links": null
	},
	"BIBREF25": {
	"ref_id": "b25",
	"title": "Harnessing context incongruity for sarcasm detection",
	"authors": [
	{
	"first": "Aditya",
	"middle": [],
	"last": "Joshi",
	"suffix": ""
	},
	{
	"first": "Vinita",
	"middle": [],
	"last": "Sharma",
	"suffix": ""
	},
	{
	"first": "Pushpak",
	"middle": [],
	"last": "Bhattacharyya",
	"suffix": ""
	}
	],
	"year": 2015,
	"venue": "Proceedings of the 53rd Annual Meeting of the Association for Computational Linguistics and the 7th International Joint Conference on Natural Language Processing",
	"volume": "",
	"issue": "",
	"pages": "757--762",
	"other_ids": {},
	"num": null,
	"urls": [],
	"raw_text": "Aditya Joshi, Vinita Sharma, and Pushpak Bhat- tacharyya. 2015. Harnessing context incongruity for sarcasm detection. In Proceedings of the 53rd Annual Meeting of the Association for Computa- tional Linguistics and the 7th International Joint Conference on Natural Language Processing (Vol- ume 2: Short Papers), pages 757-762, Beijing, China, July. ACL.",
	"links": null
	},
	"BIBREF26": {
	"ref_id": "b26",
	"title": "Towards a contextual pragmatic model to detect irony in tweets",
	"authors": [
	{
	"first": "Jihen",
	"middle": [],
	"last": "Karoui",
	"suffix": ""
	},
	{
	"first": "Farah",
	"middle": [],
	"last": "Benamara",
	"suffix": ""
	},
	{
	"first": "V\u00e9ronique",
	"middle": [],
	"last": "Moriceau",
	"suffix": ""
	},
	{
	"first": "Nathalie",
	"middle": [],
	"last": "Aussenac-Gilles",
	"suffix": ""
	},
	{
	"first": "Lamia",
	"middle": [],
	"last": "Hadrich-Belguith",
	"suffix": ""
	}
	],
	"year": 2015,
	"venue": "Proceedings of the 53rd Annual Meeting of the Association for Computational Linguistics and the 7th International Joint Conference on Natural Language Processing",
	"volume": "2",
	"issue": "",
	"pages": "644--650",
	"other_ids": {},
	"num": null,
	"urls": [],
	"raw_text": "Jihen Karoui, Farah Benamara, V\u00e9ronique Moriceau, Nathalie Aussenac-Gilles, and Lamia Hadrich- Belguith. 2015. Towards a contextual pragmatic model to detect irony in tweets. In Proceedings of the 53rd Annual Meeting of the Association for Computational Linguistics and the 7th International Joint Conference on Natural Language Processing (Volume 2: Short Papers), pages 644-650, Beijing, China, July. ACL.",
	"links": null
	},
	"BIBREF27": {
	"ref_id": "b27",
	"title": "The differential role of ridicule in sarcasm and irony",
	"authors": [
	{
	"first": "J",
	"middle": [],
	"last": "Christopher",
	"suffix": ""
	},
	{
	"first": "Albert",
	"middle": [
	"N"
	],
	"last": "Lee",
	"suffix": ""
	},
	{
	"first": "",
	"middle": [],
	"last": "Katz",
	"suffix": ""
	}
	],
	"year": 1998,
	"venue": "Metaphor and Symbol",
	"volume": "13",
	"issue": "1",
	"pages": "1--15",
	"other_ids": {},
	"num": null,
	"urls": [],
	"raw_text": "Christopher J. Lee and Albert N. Katz. 1998. The differential role of ridicule in sarcasm and irony. Metaphor and Symbol, 13(1):1-15.",
	"links": null
	},
	"BIBREF28": {
	"ref_id": "b28",
	"title": "Principles of pragmatics",
	"authors": [
	{
	"first": "Geoffrey",
	"middle": [
	"N"
	],
	"last": "Leech",
	"suffix": ""
	}
	],
	"year": 2016,
	"venue": "",
	"volume": "",
	"issue": "",
	"pages": "",
	"other_ids": {},
	"num": null,
	"urls": [],
	"raw_text": "Geoffrey N. Leech. 2016. Principles of pragmatics. Routledge.",
	"links": null
	},
	"BIBREF29": {
	"ref_id": "b29",
	"title": "The perfect solution for detecting sarcasm in tweets #not",
	"authors": [
	{
	"first": "Christine",
	"middle": [],
	"last": "Liebrecht",
	"suffix": ""
	},
	{
	"first": "Florian",
	"middle": [],
	"last": "Kunneman",
	"suffix": ""
	},
	{
	"first": "Antal",
	"middle": [],
	"last": "Van Den",
	"suffix": ""
	},
	{
	"first": "",
	"middle": [],
	"last": "Bosch",
	"suffix": ""
	}
	],
	"year": 2013,
	"venue": "Proceedings of the 4th Workshop on Computational Approaches to Subjectivity, Sentiment and Social Media Analysis",
	"volume": "",
	"issue": "",
	"pages": "29--37",
	"other_ids": {},
	"num": null,
	"urls": [],
	"raw_text": "Christine Liebrecht, Florian Kunneman, and Antal Van den Bosch. 2013. The perfect solution for de- tecting sarcasm in tweets #not. In Proceedings of the 4th Workshop on Computational Approaches to Subjectivity, Sentiment and Social Media Analysis, pages 29-37, Atlanta, Georgia, June. Association for Computational Linguistics.",
	"links": null
	},
	"BIBREF30": {
	"ref_id": "b30",
	"title": "Proceedings of the Ninth International Conference on Language Resources and Evaluation (LREC'14)",
	"authors": [
	{
	"first": "Diana",
	"middle": [],
	"last": "Maynard",
	"suffix": ""
	},
	{
	"first": "Mark",
	"middle": [],
	"last": "Greenwood",
	"suffix": ""
	}
	],
	"year": 2014,
	"venue": "",
	"volume": "",
	"issue": "",
	"pages": "4238--4243",
	"other_ids": {},
	"num": null,
	"urls": [],
	"raw_text": "Diana Maynard and Mark Greenwood. 2014. Who cares about Sarcastic Tweets? Investigating the Im- pact of Sarcasm on Sentiment Analysis. In Pro- ceedings of the Ninth International Conference on Language Resources and Evaluation (LREC'14), pages 4238-4243, Reykjavik, Iceland, May. Euro- pean Language Resources Association (ELRA).",
	"links": null
	},
	"BIBREF32": {
	"ref_id": "b32",
	"title": "Irony markers",
	"authors": [
	{
	"first": "C",
	"middle": [],
	"last": "Douglas",
	"suffix": ""
	},
	{
	"first": "",
	"middle": [],
	"last": "Muecke",
	"suffix": ""
	}
	],
	"year": 1978,
	"venue": "Poetics",
	"volume": "7",
	"issue": "4",
	"pages": "363--375",
	"other_ids": {},
	"num": null,
	"urls": [],
	"raw_text": "Douglas C. Muecke. 1978. Irony markers. Poetics, 7(4):363-375.",
	"links": null
	},
	"BIBREF33": {
	"ref_id": "b33",
	"title": "Les figures de l'ironie dans A la recherche du temps perdu de Marcel Proust",
	"authors": [
	{
	"first": "Philippe",
	"middle": [],
	"last": "Niogret",
	"suffix": ""
	}
	],
	"year": 2004,
	"venue": "Editions L'Harmattan",
	"volume": "",
	"issue": "",
	"pages": "",
	"other_ids": {},
	"num": null,
	"urls": [],
	"raw_text": "Philippe Niogret. 2004. Les figures de l'ironie dans A la recherche du temps perdu de Marcel Proust. Edi- tions L'Harmattan.",
	"links": null
	},
	"BIBREF34": {
	"ref_id": "b34",
	"title": "Sarcasm Detection on Czech and English Twitter",
	"authors": [
	{
	"first": "Tom\u00e1\u0161",
	"middle": [],
	"last": "Pt\u00e1\u010dek",
	"suffix": ""
	},
	{
	"first": "Ivan",
	"middle": [],
	"last": "Habernal",
	"suffix": ""
	},
	{
	"first": "Jun",
	"middle": [],
	"last": "Hong",
	"suffix": ""
	}
	],
	"year": 2014,
	"venue": "Proceedings of COLING 2014, the 25th International Conference on Computational Linguistics",
	"volume": "",
	"issue": "",
	"pages": "213--223",
	"other_ids": {},
	"num": null,
	"urls": [],
	"raw_text": "Tom\u00e1\u0161 Pt\u00e1\u010dek, Ivan Habernal, and Jun Hong. 2014. Sarcasm Detection on Czech and English Twitter. In Proceedings of COLING 2014, the 25th Inter- national Conference on Computational Linguistics, pages 213-223, Dublin, Ireland, August. Dublin City University and ACL.",
	"links": null
	},
	"BIBREF35": {
	"ref_id": "b35",
	"title": "A multidimensional approach for detecting irony in twitter. Language resources and evaluation",
	"authors": [
	{
	"first": "Antonio",
	"middle": [],
	"last": "Reyes",
	"suffix": ""
	},
	{
	"first": "Paolo",
	"middle": [],
	"last": "Rosso",
	"suffix": ""
	},
	{
	"first": "Tony",
	"middle": [],
	"last": "Veale",
	"suffix": ""
	}
	],
	"year": 2013,
	"venue": "",
	"volume": "47",
	"issue": "",
	"pages": "239--268",
	"other_ids": {},
	"num": null,
	"urls": [],
	"raw_text": "Antonio Reyes, Paolo Rosso, and Tony Veale. 2013. A multidimensional approach for detecting irony in twitter. Language resources and evaluation, 47(1):239-268.",
	"links": null
	},
	"BIBREF36": {
	"ref_id": "b36",
	"title": "Frame-shifting in humor and irony",
	"authors": [
	{
	"first": "David",
	"middle": [],
	"last": "Ritchie",
	"suffix": ""
	}
	],
	"year": 2005,
	"venue": "Metaphor and Symbol",
	"volume": "20",
	"issue": "4",
	"pages": "275--294",
	"other_ids": {},
	"num": null,
	"urls": [],
	"raw_text": "David Ritchie. 2005. Frame-shifting in humor and irony. Metaphor and Symbol, 20(4):275-294.",
	"links": null
	},
	"BIBREF37": {
	"ref_id": "b37",
	"title": "On non-echoic irony",
	"authors": [
	{
	"first": "Ken-Ichi",
	"middle": [],
	"last": "Seto",
	"suffix": ""
	}
	],
	"year": 1998,
	"venue": "Relevance Theory: Applications and Implications",
	"volume": "37",
	"issue": "",
	"pages": "",
	"other_ids": {},
	"num": null,
	"urls": [],
	"raw_text": "Ken-ichi Seto. 1998. On non-echoic irony. Relevance Theory: Applications and Implications, 37:239.",
	"links": null
	},
	"BIBREF38": {
	"ref_id": "b38",
	"title": "The bicoherence theory of situational irony",
	"authors": [
	{
	"first": "Cameron",
	"middle": [],
	"last": "Shelley",
	"suffix": ""
	}
	],
	"year": 2001,
	"venue": "Cognitive Science",
	"volume": "25",
	"issue": "5",
	"pages": "775--818",
	"other_ids": {},
	"num": null,
	"urls": [],
	"raw_text": "Cameron Shelley. 2001. The bicoherence theory of situational irony. Cognitive Science, 25(5):775-818.",
	"links": null
	},
	"BIBREF39": {
	"ref_id": "b39",
	"title": "Irony and the use-mention distinction",
	"authors": [
	{
	"first": "Dan",
	"middle": [],
	"last": "Sperber",
	"suffix": ""
	},
	{
	"first": "Deirdre",
	"middle": [],
	"last": "Wilson",
	"suffix": ""
	}
	],
	"year": 1981,
	"venue": "",
	"volume": "49",
	"issue": "",
	"pages": "295--318",
	"other_ids": {},
	"num": null,
	"urls": [],
	"raw_text": "Dan Sperber and Deirdre Wilson. 1981. Irony and the use-mention distinction. Radical pragmatics, 49:295-318.",
	"links": null
	},
	"BIBREF42": {
	"ref_id": "b42",
	"title": "Annotating sentiment and irony in the online italian political debate on #labuonascuola",
	"authors": [],
	"year": null,
	"venue": "Proceedings of LREC 2016",
	"volume": "",
	"issue": "",
	"pages": "2892--2899",
	"other_ids": {},
	"num": null,
	"urls": [],
	"raw_text": "Annotating sentiment and irony in the online italian political debate on #labuonascuola. In Proceedings of LREC 2016, pages 2892-2899. ELRA.",
	"links": null
	},
	"BIBREF43": {
	"ref_id": "b43",
	"title": "Figurative messages and affect in Twitter: Differences between #irony, #sarcasm and #not. Knowledge-Based Systems",
	"authors": [
	{
	"first": "Emilio",
	"middle": [],
	"last": "Sulis",
	"suffix": ""
	},
	{
	"first": "D",
	"middle": [],
	"last": "Iraz\u00fa Hern\u00e1ndez Far\u00edas",
	"suffix": ""
	},
	{
	"first": "Paolo",
	"middle": [],
	"last": "Rosso",
	"suffix": ""
	},
	{
	"first": "Viviana",
	"middle": [],
	"last": "Patti",
	"suffix": ""
	},
	{
	"first": "Giancarlo",
	"middle": [],
	"last": "Ruffo",
	"suffix": ""
	}
	],
	"year": 2016,
	"venue": "",
	"volume": "108",
	"issue": "",
	"pages": "132--143",
	"other_ids": {},
	"num": null,
	"urls": [],
	"raw_text": "Emilio Sulis, D. Iraz\u00fa Hern\u00e1ndez Far\u00edas, Paolo Rosso, Viviana Patti, and Giancarlo Ruffo. 2016. Figura- tive messages and affect in Twitter: Differences be- tween #irony, #sarcasm and #not. Knowledge-Based Systems, 108:132 -143. New Avenues in Knowl- edge Bases for Natural Language Processing.",
	"links": null
	},
	"BIBREF44": {
	"ref_id": "b44",
	"title": "Chinese irony corpus construction and ironic structure analysis",
	"authors": [
	{
	"first": "Yijie",
	"middle": [],
	"last": "Tang",
	"suffix": ""
	},
	{
	"first": "Hsinhsi",
	"middle": [],
	"last": "Chen",
	"suffix": ""
	}
	],
	"year": 2014,
	"venue": "Proceedings of COLING 2014, the 25th International Conference on Computational Linguistics: Technical Papers",
	"volume": "",
	"issue": "",
	"pages": "1269--1278",
	"other_ids": {},
	"num": null,
	"urls": [],
	"raw_text": "Yijie Tang and HsinHsi Chen. 2014. Chinese irony corpus construction and ironic structure analysis. In Proceedings of COLING 2014, the 25th Inter- national Conference on Computational Linguistics: Technical Papers, pages 1269-1278.",
	"links": null
	},
	"BIBREF45": {
	"ref_id": "b45",
	"title": "L'ironie",
	"authors": [
	{
	"first": "Claudine",
	"middle": [],
	"last": "Tayot",
	"suffix": ""
	}
	],
	"year": 1984,
	"venue": "",
	"volume": "",
	"issue": "",
	"pages": "",
	"other_ids": {},
	"num": null,
	"urls": [],
	"raw_text": "Claudine Tayot. 1984. L'ironie. Ph.D. thesis, Claude Bernard University (Lyon).",
	"links": null
	},
	"BIBREF46": {
	"ref_id": "b46",
	"title": "Icwsm-a great catchy name: Semi-supervised recognition of sarcastic sentences in online product reviews",
	"authors": [
	{
	"first": "Oren",
	"middle": [],
	"last": "Tsur",
	"suffix": ""
	},
	{
	"first": "Dmitry",
	"middle": [],
	"last": "Davidov",
	"suffix": ""
	},
	{
	"first": "Ari",
	"middle": [],
	"last": "Rappoport",
	"suffix": ""
	}
	],
	"year": 2010,
	"venue": "ICWSM",
	"volume": "",
	"issue": "",
	"pages": "",
	"other_ids": {},
	"num": null,
	"urls": [],
	"raw_text": "Oren Tsur, Dmitry Davidov, and Ari Rappoport. 2010. Icwsm-a great catchy name: Semi-supervised recog- nition of sarcastic sentences in online product re- views. In ICWSM.",
	"links": null
	},
	"BIBREF47": {
	"ref_id": "b47",
	"title": "A unified theory of irony and its computational formalization",
	"authors": [
	{
	"first": "Akira",
	"middle": [],
	"last": "Utsumi",
	"suffix": ""
	}
	],
	"year": 1996,
	"venue": "Proceedings of COLING, the 16th conference on Computational Linguistics",
	"volume": "2",
	"issue": "",
	"pages": "962--967",
	"other_ids": {},
	"num": null,
	"urls": [],
	"raw_text": "Akira Utsumi. 1996. A unified theory of irony and its computational formalization. In Proceedings of COLING, the 16th conference on Computational Linguistics-Volume 2, pages 962-967. Association for Computational Linguistics.",
	"links": null
	},
	"BIBREF48": {
	"ref_id": "b48",
	"title": "Exploring the Realization of Irony in Twitter Data",
	"authors": [
	{
	"first": "Cynthia",
	"middle": [],
	"last": "Van Hee",
	"suffix": ""
	},
	{
	"first": "Els",
	"middle": [],
	"last": "Lefever",
	"suffix": ""
	},
	{
	"first": "V\u00e9ronique",
	"middle": [],
	"last": "Hoste",
	"suffix": ""
	}
	],
	"year": 2016,
	"venue": "Proceedings of LREC. European Language Resources Association (ELRA)",
	"volume": "",
	"issue": "",
	"pages": "",
	"other_ids": {},
	"num": null,
	"urls": [],
	"raw_text": "Cynthia Van Hee, Els Lefever, and V\u00e9ronique Hoste. 2016. Exploring the Realization of Irony in Twitter Data. In Proceedings of LREC. European Language Resources Association (ELRA).",
	"links": null
	},
	"BIBREF49": {
	"ref_id": "b49",
	"title": "Sparse, contextually informed models for irony detection: Exploiting user communities, entities and sentiment",
	"authors": [
	{
	"first": "Byron",
	"middle": [
	"C"
	],
	"last": "Wallace",
	"suffix": ""
	},
	{
	"first": "Eugene",
	"middle": [],
	"last": "Do Kook Choe",
	"suffix": ""
	},
	{
	"first": "",
	"middle": [],
	"last": "Charniak",
	"suffix": ""
	}
	],
	"year": 2015,
	"venue": "Proceedings of the 53rd Annual Meeting of the Association for Computational Linguistics and the 7th International Joint Conference on Natural Language Processing of the Asian Federation of Natural Language Processing",
	"volume": "",
	"issue": "",
	"pages": "1035--1044",
	"other_ids": {},
	"num": null,
	"urls": [],
	"raw_text": "Byron C. Wallace, Do Kook Choe, and Eugene Char- niak. 2015. Sparse, contextually informed mod- els for irony detection: Exploiting user communi- ties, entities and sentiment. In Proceedings of the 53rd Annual Meeting of the Association for Compu- tational Linguistics and the 7th International Joint Conference on Natural Language Processing of the Asian Federation of Natural Language Processing, ACL-IJCNLP 2015, pages 1035-1044. ACL.",
	"links": null
	}
	},
	"ref_entries": {
	"TABREF1": {
	"html": null,
	"content": "<table/>",
	"type_str": "table",
	"text": "Distribution of tweets in each corpus.",
	"num": null
	},
	"TABREF3": {
	"html": null,
	"content": "<table><tr><td>: Number of tweets in annotated corpora in</td></tr><tr><td>French (F), English (E) and Italian (I).</td></tr></table>",
	"type_str": "table",
	"text": "",
	"num": null
	},
	"TABREF5": {
	"html": null,
	"content": "<table><tr><td>Emoticon</td><td>Negation</td><td>Discourse</td><td>Humour #</td><td>Intensifier</td><td>Punctuation</td><td>False prop.</td><td>Surprise</td><td>Modality</td><td>Quotation</td></tr></table>",
	"type_str": "table",
	"text": "al.,",
	"num": null
	},
	"TABREF7": {
	"html": null,
	"content": "<table><tr><td/><td/><td colspan=\"2\">Negation</td><td colspan=\"3\">Discourse</td><td colspan=\"3\">Humour #</td><td colspan=\"3\">Intensifier</td><td colspan=\"3\">Punctuation</td><td colspan=\"3\">False prop.</td><td/><td>Modality</td><td>Quotation</td></tr><tr><td/><td colspan=\"2\">F E</td><td>I</td><td colspan=\"3\">F E I</td><td colspan=\"3\">F E I</td><td colspan=\"3\">F E I</td><td>F</td><td>E</td><td>I</td><td colspan=\"3\">F E I</td><td colspan=\"2\">F E I</td><td>F E I</td></tr><tr><td>Analogy</td><td colspan=\"2\">46 56</td><td>2</td><td>6</td><td colspan=\"2\">29 8</td><td>6</td><td colspan=\"2\">15 -</td><td colspan=\"3\">21 10 0</td><td>49</td><td>24</td><td>2</td><td colspan=\"2\">13 8</td><td>-</td><td>0</td><td>3</td><td>2</td><td>0</td><td>24 1</td></tr><tr><td>Context sh.</td><td colspan=\"3\">40 100 3</td><td>0</td><td colspan=\"2\">11 3</td><td>0</td><td>0</td><td>-</td><td>0</td><td>0</td><td>1</td><td>60</td><td>44</td><td>1</td><td>0</td><td>0</td><td>-</td><td>0</td><td>11 0</td><td>0</td><td>44 0</td></tr><tr><td>Euphemism</td><td colspan=\"2\">50 67</td><td>1</td><td>6</td><td>0</td><td>2</td><td>0</td><td>0</td><td>-</td><td colspan=\"3\">50 33 0</td><td>72</td><td>0</td><td>1</td><td colspan=\"2\">44 0</td><td>-</td><td>0</td><td>33 0</td><td>0</td><td>0</td><td>1</td></tr><tr><td>Hyperbole</td><td colspan=\"2\">25 42</td><td>1</td><td>5</td><td colspan=\"2\">25 2</td><td>3</td><td>8</td><td>-</td><td colspan=\"3\">57 38 0</td><td>56</td><td>21</td><td>2</td><td colspan=\"3\">53 46 -</td><td>0</td><td>0</td><td>0</td><td>8</td><td>4</td><td>0</td></tr><tr><td>Rhet. ques.</td><td colspan=\"2\">43 70</td><td>2</td><td>2</td><td colspan=\"2\">36 3</td><td>2</td><td colspan=\"2\">17 -</td><td colspan=\"2\">17 9</td><td>0</td><td>93</td><td>86</td><td>1</td><td>9</td><td>3</td><td>-</td><td>0</td><td>3</td><td>0</td><td>7</td><td>23 1</td></tr><tr><td>Oxymoron</td><td colspan=\"2\">35 59</td><td>3</td><td>4</td><td colspan=\"2\">43 6</td><td>0</td><td colspan=\"2\">14 -</td><td colspan=\"3\">21 10 1</td><td>49</td><td>26</td><td>2</td><td colspan=\"2\">11 0</td><td>-</td><td>0</td><td>2</td><td>1</td><td>5</td><td>20 0</td></tr><tr><td>False asser.</td><td colspan=\"2\">18 57</td><td>1</td><td>4</td><td colspan=\"2\">25 3</td><td>3</td><td>7</td><td>-</td><td colspan=\"3\">10 16 0</td><td>29</td><td>14</td><td>2</td><td colspan=\"3\">95 89 -</td><td>0</td><td>0</td><td>0</td><td>4</td><td>16 1</td></tr><tr><td>Other</td><td colspan=\"2\">26 62</td><td>2</td><td>5</td><td colspan=\"2\">31 3</td><td>5</td><td colspan=\"2\">18 -</td><td colspan=\"3\">15 11 0</td><td>45</td><td>20</td><td>2</td><td colspan=\"2\">11 3</td><td>-</td><td>0</td><td>2</td><td>1</td><td>8</td><td>25 1</td></tr><tr><td/><td colspan=\"3\">Opposition</td><td colspan=\"3\">Pers. pro.</td><td colspan=\"3\">Interjection</td><td colspan=\"3\">Comparison</td><td colspan=\"3\">Named E.</td><td colspan=\"3\">Report verb</td><td/><td>Opinion</td><td>URL</td></tr><tr><td/><td colspan=\"2\">F E</td><td>I</td><td colspan=\"3\">F E I</td><td colspan=\"3\">F E I</td><td colspan=\"3\">F E I</td><td>F</td><td>E</td><td>I</td><td colspan=\"3\">F E I</td><td colspan=\"2\">F E I</td><td>F E I</td></tr><tr><td>Analogy</td><td>6</td><td>11</td><td>2</td><td colspan=\"3\">38 19 2</td><td>6</td><td>0</td><td>3</td><td colspan=\"3\">43 42 3</td><td colspan=\"4\">100 100 17 2</td><td colspan=\"2\">16 0</td><td colspan=\"2\">41 68 -</td><td>13 0</td><td>1</td></tr><tr><td>Context sh.</td><td>0</td><td>11</td><td>1</td><td colspan=\"3\">40 33 1</td><td colspan=\"2\">20 0</td><td>2</td><td colspan=\"2\">20 6</td><td>0</td><td>80</td><td colspan=\"2\">100 8</td><td>0</td><td colspan=\"2\">22 0</td><td colspan=\"2\">60 68 -</td><td>0</td><td>0</td><td>1</td></tr><tr><td>Euphemism</td><td>0</td><td>0</td><td>0</td><td colspan=\"2\">22 0</td><td>0</td><td>6</td><td>0</td><td>1</td><td>0</td><td>0</td><td>0</td><td>94</td><td colspan=\"2\">100 2</td><td>0</td><td colspan=\"2\">33 0</td><td colspan=\"2\">56 67 -</td><td>22 0</td><td>1</td></tr><tr><td>Hyperbole</td><td>2</td><td>4</td><td>0</td><td colspan=\"3\">29 33 1</td><td colspan=\"2\">18 0</td><td>2</td><td>0</td><td>8</td><td>0</td><td>88</td><td>88</td><td>6</td><td>3</td><td colspan=\"2\">13 0</td><td colspan=\"2\">84 88 -</td><td>21 0</td><td>1</td></tr><tr><td>Rhet. ques.</td><td>3</td><td>15</td><td>1</td><td colspan=\"3\">31 27 0</td><td colspan=\"2\">13 2</td><td>1</td><td>2</td><td>5</td><td>0</td><td>90</td><td>97</td><td>9</td><td>1</td><td colspan=\"2\">17 0</td><td colspan=\"2\">45 73 -</td><td>25 0</td><td>1</td></tr><tr><td>Oxymoron</td><td colspan=\"2\">12 19</td><td>1</td><td colspan=\"3\">32 21 0</td><td colspan=\"2\">15 3</td><td>2</td><td>2</td><td>6</td><td>0</td><td>99</td><td colspan=\"3\">100 10 1</td><td colspan=\"2\">19 0</td><td colspan=\"2\">55 75 -</td><td>11 0</td><td>2</td></tr><tr><td>False asser.</td><td>3</td><td>4</td><td>1</td><td colspan=\"3\">31 36 1</td><td colspan=\"2\">13 0</td><td>1</td><td>2</td><td colspan=\"2\">13 1</td><td>90</td><td>93</td><td>8</td><td>1</td><td colspan=\"2\">13 0</td><td colspan=\"2\">45 79 -</td><td>25 0</td><td>0</td></tr><tr><td>Other</td><td>2</td><td>11</td><td>2</td><td colspan=\"3\">29 22 0</td><td colspan=\"2\">10 0</td><td>2</td><td>1</td><td colspan=\"2\">10 0</td><td>91</td><td>98</td><td>6</td><td>1</td><td colspan=\"2\">16 0</td><td colspan=\"2\">32 74 -</td><td>30 0</td><td>1</td></tr></table>",
	"type_str": "table",
	"text": "Markers in ironic (Exp or Imp) and non ironic (NI) tweets in French, English and Italian (in %). Markers with an * have not been studied in irony literature.",
	"num": null
	},
	"TABREF8": {
	"html": null,
	"content": "<table/>",
	"type_str": "table",
	"text": "Percentage of tweets in each ironic category containing markers in French, English and Italian.",
	"num": null
	}
	}
	}
	}