The relation between language and thought has been central in many disciplines including philosop... more The relation between language and thought has been central in many disciplines including philosophy, linguistics, psychology, anthropology, to mention just some. For the Ancient Greek thinkers, logos refers to both thought (specifically, ability of humans to think logically) and language, therefore a symmetry is projected in the relation of the two: the ability to think needs language to express thought. Moreover, for thinkers like Aristotle this relation is universal: thoughts do not vary according to language, but language is the universal vehicle for representing thought. For Plato, likewise, as lucidly expressed in Cratylus, objects have their essence (ousia) and the task of name-givers is to discover the essence and name accordingly. The Greek view is descriptive and representational: the thought and the world exist independent of language, and language serves to describe it. Language does not by itself create a reality, although in rhetoric and in sophistry the truth of sentences can be manipulated to exploit gaps in logic that can lead to flawed conclusions. In contrast to this descriptive view, the so-called Sapir-Whorf hypothesis stipulates a primacy of language over thought such that thought is determined by the language. This position-widely popular in (mostly non-linguistic) circles, and epitomized in Wittgenstein's famous saying The limits of my language means the limits of my worldis one of relativism and language supremacy: in its strongest form, the position says that the language spoken by a community (a) fully determines the thought frame within which the community understands the world, and therefore (b) places limitations on how linguistic agents perceive and construct the world. The linguistic supremacy position views language as a world creating and restricting device, in contrast to the Greek view where the function of language is merely to represent and describe the world. In this paper, we will approach the question of language, thought and reality by studying how grammar and the lexicon encode our relation to the world. We will address the fundamental categories of knowledge and belief, and focus on specific grammatical devices such as mood morphemes (subjunctive and indicative), attitude verbs of knowledge and belief, and expressions of possibility and necessity such as modal verbs (must, may, will, might). What is the function of these expressions? How much do they tell us about the nature of knowing and believing? Is a language like English, which lacks subjunctive and indicative, missing something crucial compared to languages such as Greek and Italian which have productive mood morphology? Is the absence of mood a deficiency for knowledge or belief? We will argue that the function of modal and propositional attitude verbs is to be truth manipulators: they express whether speakers are fully or partially committed to the truth of a proposition, and whether they do so based on objective (i.e. truth), or subjective criteria. In believing, commitment relies on subjectively veridical constructs of truth that commit speakers solipsistically, and 1
Recognizing speech acts (SA) is crucial for capturing meaning beyond what is said, making communi... more Recognizing speech acts (SA) is crucial for capturing meaning beyond what is said, making communicative intentions particularly relevant to identify urgent messages. This paper attempts to measure for the first time the impact of SA on urgency detection during crises, in tweets. We propose a new dataset annotated for both urgency and SA, and develop several deep learning architectures to inject SA into urgency detection while ensuring models generalisability. Our results show that taking speech acts into account in tweet analysis improves information type detection in an out-of-type configuration where models are evaluated in unseen event types during training. These results are encouraging and constitute a first step towards SA-aware disaster management in social media.
HAL (Le Centre pour la Communication Scientifique Directe), May 22, 2022
This paper aims at measuring transfer learning performances across different types of crises rela... more This paper aims at measuring transfer learning performances across different types of crises related to sudden or unexpected events (like earthquakes, terror attacks, explosions, technological incidents) that cannot be foreseen by emergency services and on the occurrence of which they have virtually no control. Although sudden crises are present in most existing crisis datasets, as far as we are aware, no one studied their impact on classifiers performances when evaluated in an out-of-type scenario in which models are tested on a particular type of crisis unseen during training. Our contribution is threefold: (1) A new dataset of about 3,800 French tweets related to four sudden events that occurred in France annotated for both relatedness (i.e., useful vs. not useful for emergency responders) and urgency (i.e., not useful vs. urgent vs. not urgent), (2) A set of monotask and multitask zero-shot learning experiments to transfer knowledge across events and types, and finally, (3) Experiments involving few-shot learning to measure the amount of sudden events instances needed during training to guarantee good performances. When compared to a cross-event setting, our preliminary results are encouraging and show that transfer from predictable ecological crisis to sudden events is feasible and constitutes a first step towards real-time crisis management systems from social media content.
Social media networks have become a space where users are free to relate their opinions and senti... more Social media networks have become a space where users are free to relate their opinions and sentiments which may lead to a large spreading of hatred or abusive messages which have to be moderated. This paper presents the first French corpus annotated for sexism detection composed of about 12,000 tweets. In a context of offensive content mediation on social media now regulated by European laws, we think that it is important to be able to detect automatically not only sexist content but also to identify if a message with a sexist content is really sexist (i.e. addressed to a woman or describing a woman or women in general) or is a story of sexism experienced by a woman. This point is the novelty of our annotation scheme. We also propose some preliminary results for sexism detection obtained with a deep learning approach. Our experiments show encouraging results.
HAL is a multi-disciplinary open access archive for the deposit and dissemination of scientific r... more HAL is a multi-disciplinary open access archive for the deposit and dissemination of scientific research documents, whether they are published or not. The documents may come from teaching and research institutions in France or abroad, or from public or private research centers. L’archive ouverte pluridisciplinaire HAL, est destinée au dépôt et à la diffusion de documents scientifiques de niveau recherche, publiés ou non, émanant des établissements d’enseignement et de recherche français ou étrangers, des laboratoires publics ou privés. Sentence meaning as argumentative dialogues Davide Catta, Alda Mari, Christian Retoré
Italian is a well-known exception to the cross-linguistic generalization according to which `beli... more Italian is a well-known exception to the cross-linguistic generalization according to which `belief' predicates are indicative selectors across languages. We newly propose that languages that select the subjunctive with epistemic predicates allow us to see a systematic polysemy between what we call an expressive-`belief' (featuring only a doxastic dimension) and an inquisitive-`belief' (featuring both a doxastic and an epistemic dimension conveying doxastic certainty (in the assertion) and epistemic uncertainty (in the presupposition)). We offer several previously unseen contrasts proving this distinction and offer a new analysis for mood choice cross-linguistically. We argue that the distinction between expressive and inquisitive attitudes is not an idiosyncrasy of non-factive epistemics. We provide novel data, showing that fictional predicates (dream, imagine) license the subjunctive. We explain the indicative/subjunctive alternation by again appealing to epistemic ...
Instruments are expressed in language by various means: prepositions, postpositions, affixes incl... more Instruments are expressed in language by various means: prepositions, postpositions, affixes including case marks, nonfinite verbs, etc. We consider here 12 languages from five families in order to be able to identify the different meaning components that structure instrumentality. This work has been made possible partly via the STIC-Asia cooperation framework.
HAL (Le Centre pour la Communication Scientifique Directe), Jun 21, 2022
Discovered by (Austin,1962) and extensively promoted by , speech acts (SA) have been the object o... more Discovered by (Austin,1962) and extensively promoted by , speech acts (SA) have been the object of extensive discussion in the philosophical and the linguistic literature, as well as in computational linguistics where the detection of SA have shown to be an important step in many down stream NLP applications. In this paper, we attempt to measure for the first time the role of SA on urgency detection in tweets, focusing on natural disasters. Indeed, SA are particularly relevant to identify intentions, desires, plans and preferences towards action, providing therefore actionable information that will help to set priorities for the human teams and decide appropriate rescue actions. To this end, we come up here with four main contributions: (1) A two-layer annotation scheme of SA both at the tweet and subtweet levels, (2) A new French dataset of 6,669 tweets annotated for both urgency and SA, (3) An in-depth analysis of the annotation campaign, highlighting the correlation between SA and urgency categories, and (4) A set of deep learning experiments to detect SA in a crisis corpus. Our results show that SA are correlated with urgency which is a first important step towards SA-aware NLP-based crisis management on social media.
International audienceSubjunctive/indicative are triggered by (non)veridicality in the assertion ... more International audienceSubjunctive/indicative are triggered by (non)veridicality in the assertion or presupposition. With emotives, the assertion contains a sentiment ordering source S that partitions the emotive domain E into p (positive extent) and ¬p (negative). Emotives are nonveridical in the assertion and veridical in the presupposition. Languages differently parametrize mood choice, which can be selected by the assertion or the presupposition
KNOW: A Journal on the Formation of Knowledge, 2021
The relation between language and thought has been central in many disciplines including philosop... more The relation between language and thought has been central in many disciplines including philosophy, linguistics, psychology, anthropology, to mention just some. For the Ancient Greek thinkers, logos refers to both thought (specifically, ability of humans to think logically) and language, therefore a symmetry is projected in the relation of the two: the ability to think needs language to express thought. Moreover, for thinkers like Aristotle this relation is universal: thoughts do not vary according to language, but language is the universal vehicle for representing thought. For Plato, likewise, as lucidly expressed in Cratylus, objects have their essence (ousia) and the task of name-givers is to discover the essence and name accordingly. The Greek view is descriptive and representational: the thought and the world exist independent of language, and language serves to describe it. Language does not by itself create a reality, although in rhetoric and in sophistry the truth of sentences can be manipulated to exploit gaps in logic that can lead to flawed conclusions. In contrast to this descriptive view, the so-called Sapir-Whorf hypothesis stipulates a primacy of language over thought such that thought is determined by the language. This position-widely popular in (mostly non-linguistic) circles, and epitomized in Wittgenstein's famous saying The limits of my language means the limits of my worldis one of relativism and language supremacy: in its strongest form, the position says that the language spoken by a community (a) fully determines the thought frame within which the community understands the world, and therefore (b) places limitations on how linguistic agents perceive and construct the world. The linguistic supremacy position views language as a world creating and restricting device, in contrast to the Greek view where the function of language is merely to represent and describe the world. In this paper, we will approach the question of language, thought and reality by studying how grammar and the lexicon encode our relation to the world. We will address the fundamental categories of knowledge and belief, and focus on specific grammatical devices such as mood morphemes (subjunctive and indicative), attitude verbs of knowledge and belief, and expressions of possibility and necessity such as modal verbs (must, may, will, might). What is the function of these expressions? How much do they tell us about the nature of knowing and believing? Is a language like English, which lacks subjunctive and indicative, missing something crucial compared to languages such as Greek and Italian which have productive mood morphology? Is the absence of mood a deficiency for knowledge or belief? We will argue that the function of modal and propositional attitude verbs is to be truth manipulators: they express whether speakers are fully or partially committed to the truth of a proposition, and whether they do so based on objective (i.e. truth), or subjective criteria. In believing, commitment relies on subjectively veridical constructs of truth that commit speakers solipsistically, and 1
In this paper we examine the epistemic and temporal uses of Italian and Greek future and argue th... more In this paper we examine the epistemic and temporal uses of Italian and Greek future and argue that, in both readings, future morphemes in these two languages are epistemic modal operators that assess indirect evidence at the utterance time. We show that the future reading arises when an overt adverb is used, or in virtue of a mechanism ensuring indirect access to the eventuality. By treating FUT as an epistemic modal across its available interpretations, our account differs from previous analyses that posit metaphysical/epistemic ambiguity for FUT operator.
The paper proposes a fresh look at emotive factives and explains variation in mood choice across ... more The paper proposes a fresh look at emotive factives and explains variation in mood choice across languages by distinguishing nonveridicality in the assertion (triggering the subjunctive) and veridicality in the presupposition (triggering the indicative). Proceeding of CLS 51, April 2015.
In this paper, we discuss the purely epistemic reading of the future morphemes in Greek and Itali... more In this paper, we discuss the purely epistemic reading of the future morphemes in Greek and Italian. In this reading, which we call ‘epistemic future’, the future morpheme behaves as equivalent to the universal epistemic modal must. Like must, epistemic future appears to have an evidential component: it cannot be used if the speaker knows that the prejacent proposition p is true. The judgment with the future is epistemically weaker than an unmodalized assertion, and relies on partial knowledge supporting p. We show that partial knowledge is not indirect knowledge. Our analysis renders epistemic futures and must nonveridical, therefore epistemically weaker than unmodalized positive assertions— and contrasts with von Fintel and Gillies 2010 who argue that must is ‘strong’. We explain the effect of strength by arguing that future and must are biased modals in that they partially support p in the set of best worlds. All modals that come with ordering sources are biased, therefore strong...
Uploads
Papers by Alda Mari