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.

This paper proposes an approach to capture the pragmatic context needed to infer irony in tweets. We aim to test the validity of two main hypotheses: (1) the presence of negations, as an internal propriety of an utterance, can help to detect the disparity between the literal and the intended meaning of an utterance, (2) a tweet containing an asserted fact of the form N ot(P 1 ) is ironic if and only if one can assess the absurdity of P 1 . Our first results are encouraging and show that deriving a pragmatic contextual model is feasible. MotivationIrony is a complex linguistic phenomenon widely studied in philosophy and linguistics (Grice et al., 1975;Sperber and Wilson, 1981;Utsumi, 1996). Despite theories differ on how to define irony, they all commonly agree that it involves an incongruity between the literal meaning of an utterance and what is expected about the speaker and/or the environment. 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 covers these devices focusing for the first time on the automatic detection of irony in French tweets.According to (Grice et al., 1975;Searle, 1979;Attardo, 2000), the search for a non-literal meaning starts when the hearer realizes that the speaker's utterance is context-inappropriate, that is an utterance fails to make sense against the context. For example, the tweet: "Congratulation #lesbleus for your great match!" is ironic if the French soccer team has lost the match. An analysis of a corpus of French tweets shows that there are two ways to infer such a context: (a) rely exclusively on the lexical clues internal to the utterance, or (b) combine these clues with an additional pragmatic context external to the utterance. In (a), the speaker intentionally creates an explicit juxtaposition of incompatible actions or words that can either have opposite polarities, or can be semantically unrelated, as in "TheVoice is more important than Fukushima tonight". Explicit opposition can also arise from an explicit positive/negative contrast between a subjective proposition and a situation that describes an undesirable activity or state. For instance, in " I love when my phone turns the volume down automatically" the writer assumes that every one expects its cell phone to ring loud enough to be heard. In (b), irony is due to an implicit opposition between a lexicalized proposition P describing an event or state and a pragmatic context external to the utterance in which P is false or is not likely to happen. In other words, the writer asserts or affirms P while he intends to convey P such that P = N ot(P ) or P = P . The irony occurs because the writer believes that his audience can detect the disparity between P and P on the basis of contextual knowledge or common background shared with the writer. For example, in "#Hollande is really a good diplomat #Algeria.", the writer critics the foreign policy of the French pre...

Hate Speech and harassment are widespread in online communication, due to users' freedom and anonymity and the lack of regulation provided by social media platforms. Hate speech is topically focused (misogyny, sexism, racism, xenophobia, homophobia, etc.), and each specific manifestation of hate speech targets different vulnerable groups based on characteristics such as gender (misogyny, sexism), ethnicity, race, religion (xenophobia, racism, Islamophobia), sexual orientation (homophobia), and so on. Most automatic hate speech detection approaches cast the problem into a binary classification task without addressing either the topical focus or the target-oriented nature of hate speech. In this paper, we propose to tackle, for the first time, hate speech detection from a multi-target perspective. We leverage manually annotated datasets, to investigate the problem of transferring knowledge from different datasets with different topical focuses and targets. Our contribution is threefold: (1) we explore the ability of hate speech detection models to capture common properties from topic-generic datasets and transfer this knowledge to recognize specific manifestations of hate speech; (2) we experiment with the development of models to detect both topics (racism, xenophobia, sexism, misogyny) and hate speech targets, going beyond standard binary classification, to investigate how to detect hate speech at a finer level of granularity and how to transfer knowledge across different topics and targets; and (3) we study the impact of affective knowledge encoded in sentic computing resources (SenticNet, EmoSenticNet) and in semantically structured hate lexicons (HurtLex) in determining specific manifestations of hate speech. We experimented with different neural models including multitask approaches. Our study shows that: (1) training a model on a combination of several (training sets from several) topic-specific datasets is more effective than training a model on a topic-generic dataset; (2) the multi-task approach outperforms a single-task model when detecting both the hatefulness of a tweet and its topical focus in the context of a multi-label classification approach; and (3) the models incorporating EmoSenticNet emotions, the first level emotions of SenticNet, a blend of SenticNet and EmoSenticNet emotions or affective features based on Hurtlex, obtained the best results. Our results demonstrate that multi-target hate speech detection from existing datasets is feasible, which is a first step towards hate speech detection for a specific topic/target when dedicated annotated data are missing. Moreover, we prove that domain-independent affective knowledge, injected into our models, helps finer-grained hate speech detection.