Reference to Abstract Objects in Discourse
Abstract
Preface. Introduction. 1. From Events to Propositions: a Tour of Abstract Entities, Eventualities and the Nominals that Denote them. 2. A Crash Course in DRT. 3. Attitudes and Attitude Descriptions. 4. The Semantic Representation for Sentential Nominals. 5. Problems for the Semantics of Nominals. 6. Anaphora and Abstract Entities. 7. A Theory of Discourse Structure for an Analysis of Abstract Entity Anaphora. 8. Applying the Theory of Discourse Structure to the Anaphoric Phenomena. 9. Applications of the Theory of Discourse Structure to Concept Anaphora and VP Ellipsis. 10. Model Theory for Abstract Entities and its Philosophical Implications. Conclusion. Bibliography. Index.
Chapters (11)
This book is about abstract objects and the ways we refer to them in natural language. Abstract objects are things like propositions, properties, states of affairs and facts. They have no spatio-temporal location, usually no causal efficacy, and are not perceived by the senses. They may be universals, like properties, and apply to many concrete objects or they may be particulars. Traditionally, abstract objects have been studied by metaphysicians, logicians and, in particular, ideal-language philosophers. These philosophers’ efforts to “regiment” and to systematize the realm of abstract objects have revealed the pitfalls and paradoxes that threaten naive conceptions of these entities, the conceptions exhibited in the ways we ordinarily speak and think about them. But those interested in natural language have also paid a price for this hegemony. The ideal-language philosopher’s interest in the natural language semantics of expressions denoting such abstract entities is often much like an inquisitor’s interest in the views of a heretic.
The main purpose of this chapter and its main contribution to the literature on natural language metaphysics is to construct
a typology of abstract entities, organized according to a spectrum of world immanence. Expanding on the work of Vendler, Gary, and others, I will look at the distributional data concerning what sort of abstract
object referring expressions go with which contexts. This will serve to divide eventualities from purely abstract objects
like propositions. But further refinements will indicate that the typology contains a spectrum of world immanence among abstract
objects-ranging from quite concrete objects like events to purely abstract objects of thought. I will distinguish not only
between eventualities and abstract entities, but also between facts, simple propositions, and projective propositions among
the more abstract entities. I will also adopt a typology of eventualities accepted by many linguists and philosophers.
The last chapter introduced a typology of abstract entities in natural language metaphysics. To investigate and analyze the
typology further, I need to give a detailed semantics of sentential and verbal nominals. In order to do this, I will introduce
the semantic framework, Discourse Representation Theory (DRT), within which the analysis will take place. That is the task
of the present chapter. Much of this chapter rehearses material and motivation in DRT that has already surfaced elsewhere.
All of this fragment, for instance, is covered in Kamp and Reyle (199+).1 But unlike the top-down analysis of Kamp and Reyle, the analysis and construction procedure of the DR-theoretic structures
presented here is bottom up. It describes a program that has been implemented in LISP and PROLOG.2 The principal novel contribution of this chapter is to give a semantics for the bottom up construction procedure that shows
the construction process is semantically coherent and gives the right results. This semantics yields insights into the possibility
of a compositional treatment of DRT, and it also lays the foundations for the semantic analysis of sentential nominals.
In the analysis of abstract entities, propositions requires careful consideration. This chapter presents the results of previous
research in DRT on attitudes and attitude reports that bears on the analysis of propositions in natural language metaphysics.
My analysis of abstract objects in subsequent chapters starts from the view of propositions as I have portrayed them in my
work on the attitude. This view differs from the usual approaches to propositions in the philosophical literature in several
ways. First, it implies that the identity conditions of propositions are radically context dependent. Because of this, traditional
puzzles about belief like those proposed by Frege and by Kripke (1979) receive a novel solution. Further, this context sensitivity
suggests that the traditional search for one criterion of identity for propositions in terms of sets of possible worlds as
in Montague (1968) or in terms of alphabetic variants of particular logical forms as in Fodor (1975), or in terms of structured
states of affairs as in Barwise and Perry (1983), Soames (1987), and Salmon (1986) is deeply misguided. Second, while the
analysis of attitudes and propositions requires postulating that mental states are structured and have contents, we need not
postulate propositions as independent, abstract entities themselves. Attitude talk does not commit us to Platonism, contrary
to what Bealer (1982) and many others have suggested. Third, DRT integrates and makes precise within a truth conditional framework
some of the ideas of conceptual role theories of meaning (Field 1972). DRT thus combines advantages of what are often seen
as incompatible approaches. Fourth, the extension of the DRS construction procedure to attitudes introduces concepts and concept
discourse referents into the semantics. The discussion of concepts will also figure in my discussion of abstract entities
and anaphora. Finally, the work on attitudes introduces machinery to develop a more sophisticated notion of update meaning
in the dynamic interpretation of attitude reports than what was presented in chapter two, or in Kamp (1981), Heim (1982),
and Groenendijk and Stokhof (1987). It is just this sort of machinery that will be needed in the development of an account
of discourse structure. The dynamic theory of attitude interpretation follows a dynamic theory of attitude formation.
In chapter 1, I introduced several tasks for any linguistically sensitive, semantic theory of abstract entities. One was to
provide a semantics of the variety of constructions that make up the class of sentential nominals, by which I mean all those
constructions which have an intuitive, semantic connection to an associated sentence. These include: derived nominals, a variety
of gerund phrases (of-ing gerunds, ACC- ing gerunds and POSS-ing gerunds), that clauses, for infinitival phrases, naked infinitive phrases and noun phrases involving a variety of common nouns that may combine with
that clauses or gerund phrases. The class of common nouns itself divides into three: nouns that are nominalizations of some prepositional
attitude verb, nouns that are nominalizations of other complemented verbs, and nouns that are not nominalizations of any verb.
This chapter continues the development of the DRS construction procedure for various types of nominals. I look at some problematic
nominals--the ACC-ing and POSS-ing nominals as well as the naked infinitive complements of perception reports. I will argue that all are instances of an IP
syntactic construction; this syntactic assumption, together with the assumptions already made in chapter 4 about the syntax-semantics
interface and the DRS construction procedure, will account for much of the semantic behavior of these peculiar nominals. In
the last two sections, I return to the difficulties for an account of nominals raised at the beginning of chapter 4.
The investigations of previous chapters suggest that subDRSs and predicative DRSs, as well as discourse referents, may serve
as terms of anaphoric relations. The use of DRSs as anaphoric antecedents has already pointed to the curious ability of pronouns
to refer anaphorically to a sum of what seem to be disparate, abstract objects. In this chapter, I discuss various types of
abstract entity anaphora within the DRT framework--in particular, anaphoric reference to events, facts, and propositions,
which are representative of eventuality and abstract object anaphora generally. I will indicate how various anaphoric phenomena
involving eventualities and abstract entities have a uniform analysis within the DRT framework. This will lay the groundwork
for the next three chapters, in which I examine the intricate effects of discourse structure on these anaphoric relations.
Even without bringing in discourse structure, however, there is much to say about abstract entity anaphora.
The last chapter introduced the DR-theoretic approach to abstract entity anaphora, on which various DR-theoretic structures
derived from the DRS construction algorithm or C-abstraction may furnish antecedents to discourse referents introduced by
pronouns. To handle the full range of abstract object anaphoric phenomena, however, one must be able to refer anaphorically
not only to DR-theoretic structures introduced by the construction procedure but also to DR- theoretic structures constructed
from bits of text that “naturally hang together.” These segments generate DR-theoretic structures that also serve as anaphoric
antecedents. The main task of this chapter is to develop a formal theory of discourse structure and discourse segmentation
suitable for the analysis of abstract object anaphora.
The last chapter developed SDRS theory, a theory of discourse structure using DR-theoretic structures for the purpose of imposing
constraints on anaphora. I elaborated several phases of SDRS construction: determining basic constituents, constraints on
attachment, rules for attachment, and rules for constituent revision. This chapter introduces the constraints on anaphora,
in particular abstract entity anaphora, and investigates the predictions of the theory. I will work out a variety of examples
of abstract entity anaphora and look at two other applications of SDRS theory: event anaphora and anaphoric relations between
abstract objects of different types.
In this chapter I apply the theory of discourse structure elaborated in the previous two chapters to concept anaphora. The semantics and discourse effects of parallelism and contrast as well as other rhetorical relations will play an important role in the analysis of concept anaphora. Discourse parallelism and contrast are natural extensions of syntactic intra-sentential constraints that many researchers have claimed guide anaphoric processes like VP ellipsis.1 The discourse constraints, however, are needed to handle examples of VP ellipsis that other theories cannot analyze.
Throughout this book, I have principally concerned myself with natural language semantics and the natural language metaphysics
underlying it So for the most part, I have concentrated on discussions of DRSs, the partial models whose task it is to reflect
to some extent natural language metaphysics in their discourse referent types. This chapter concentrates on themes of real
metaphysics, the model theoretic interpretation of the DRS construction procedure and its possible philosophical implications.
... This paper provides the first steps towards a formal analysis of the interaction between intentional structure and informational structure. Our framework for discourse structure analysis is sdrt (Asher 1993). The basic representational structures of that theory may be used to characterise cognitive states. ...
... It's unclear how Grosz and Sidner would represent this. sdrt (Asher, 1993) is in a good position to be integrated with a theory of cognitive states, because it uses the same basic structures (discourse representation structures or drss) that have been used in Discourse Representation Theory (drt) to represent different attitudes like beliefs and desires (Kamp 1981, Asher 1986, 1987, Kamp 1991, Asher and Singh, 1993. ...
... In sdrt (Asher, 1993), an nl text is represented by a segmented drs (sdrs), which is a pair of sets containing: the drss or sdrss representing respectively sentences and text segments, and discourse relations between them. Discourse relations, modelled after those proposed by Hobbs (1985), Polanyi (1985) and Thompson and Mann (1987), link together the constituents of an sdrs. ...
This paper is about the flow of inference between communicative intentions, discourse structure and the domain during discourse processing. We augment a theory of discourse interpretation with a theory of distinct mental attitudes and reasoning about them, in order to provide an account of how the attitudes interact with reasoning about discourse structure.
... Nous utilisons l'étiquette de construction indépendamment du cadre théorique des Grammaires de constructions, dans la mesure où notre intérêt dans la présente étude est de montrer que les relations discursives sont responsables de la modulation de la structure informationnelle au niveau de la phrase. Les relations rhétoriques appréhendent la fonction pragmatique d'un fragment de discours par rapport aux autres tout en rendant compte de la structure hiérarchique du texte (Mann et Thompson 1988, Asher 1993. Plus précisément, le rapport « partie-de » qui relie l'associé de jusqu'à à un ensemble contextuellement saillant est une instanciation parmi d'autres de la relation d'Élaboration (Mann et Thompson 1988, Asher 1993. ...
... Les relations rhétoriques appréhendent la fonction pragmatique d'un fragment de discours par rapport aux autres tout en rendant compte de la structure hiérarchique du texte (Mann et Thompson 1988, Asher 1993. Plus précisément, le rapport « partie-de » qui relie l'associé de jusqu'à à un ensemble contextuellement saillant est une instanciation parmi d'autres de la relation d'Élaboration (Mann et Thompson 1988, Asher 1993. Nous montrons que jusqu'à marque l'extension d'un ensemble clos, signalant la clôture épistémique de cet ensemble et que cette fonction se projette au niveau discursif dans la mesure où la construction mirative marque la clôture du topique (Van Kuppevelt 1995a), signalant en même temps la fin d'une unité discursive complexe. ...
... Nous avons montré que l'associé de jusqu'à est relié à un ensemble contextuellement saillant par une relation partie/tout en tant qu'il représente l'élément le plus inattendu de cet ensemble. Nous montrerons dans la présente section que le lien partie/tout joue un rôle dans l'établissement de la cohérence discursive à travers la relation rhétorique d'Élaboration (Mann et Thompson 1988, Asher 1993. Les relations rhétoriques appréhendent la fonction pragmatique d'un fragment par rapport à un autre en montrant que la cohérence discursive est gouvernée par le principe de la pertinence. ...
La construction [il n’est pas/il n’y a pas jusqu’à SN+ relative négative] subit une spécialisation fonctionnelle dans le marquage du focus étroit, déterminée par la composante mirative associée à jusqu’à, réanalysé comme particule scalaire-additive. Nous poursuivons une modélisation du discours qui met au jour une hiérarchie des topiques à l’intérieur d’une unité discursive complexe pour montrer que la construction mirative marque la clôture du topique, ce qui reflète le rôle de clôture d’un ensemble que joue l’associé de jusqu’à. La présence en position de focus de la construction mirative d’un constituant relié par un rapport partie/tout à un ensemble contextuellement saillant explique la relation asymétrique, de subordination, qu’elle établit en tant qu’unité discursive par rapport à l’unité discursive contenant le tout.
... Therefore, it is crucial for a novel task-based dialogue system to be able to analyze the rhetorical structure and topic structure of dialogue by discourse parsing and topic segmentation. The discourse parsing aims to identify the rhetorical structure and relationships between various discourse units of dialogue [4,2], such as QAP (Question Answer Pair) and Ack (Acknowledgment). Unlike dialogue act recognition (DAR), it not only expresses the pragmatics of utterance but also characterizes the object utterance on which it acts at a higher level. ...
... Based on the Rhetorical Structure Theory (RST) [25] for documents, the Segmented Discourse Representation Theory (SDRT) [4,2] is the popular rhetorical structure representation of dialogue. It uses the arc between various discourse units with a tree structure to model the rhetorical relations in functional terms. ...
... For the rhetorical case, our approach has three more right relations than the baseline(BART). Specifically, compared with baseline, our method achieves gold relations on [5,7], [0, 2], [2,3], [7,8]. All of these relations have arc distances of less than 3. ...
The advancement of large language models (LLMs) has propelled the development of dialogue systems. Unlike the popular ChatGPT-like assistant model, which only satisfies the user's preferences, task-oriented dialogue systems have also faced new requirements and challenges in the broader business field. They are expected to provide correct responses at each dialogue turn, at the same time, achieve the overall goal defined by the task. By understanding rhetorical structures and topic structures via topic segmentation and discourse parsing, a dialogue system may do a better planning to achieve both objectives. However, while both structures belong to discourse structure in linguistics, rhetorical structure and topic structure are mostly modeled separately or with one assisting the other in the prior work. The interaction between these two structures has not been considered for joint modeling and mutual learning. Furthermore, unsupervised learning techniques to achieve the above are not well explored. To fill this gap, we propose an unsupervised mutual learning framework of two structures leveraging the global and local connections between them. We extend the topic modeling between non-adjacent discourse units to ensure global structural relevance with rhetorical structures. We also incorporate rhetorical structures into the topic structure through a graph neural network model to ensure local coherence consistency. Finally, we utilize the similarity between the two fused structures for mutual learning. The experimental results demonstrate that our methods outperform all strong baselines on two dialogue rhetorical datasets (STAC and Molweni), as well as dialogue topic datasets (Doc2Dial and TIAGE).
... Por lo que respecta a la interacción de la partícula no con el dominio verbal, la naturaleza del objeto semántico que se obtiene cuando emerge la lectura marcada de eventualidad inhibida ha suscitado controversia entre los distintos autores. En el apartado 1 señalamos cómo algunos autores proponen que dicho objeto semántico se corresponde con un hecho o proposición (Asher 1993;Kamp y Reyle 1993), mientras que otros postulan que se trata de una eventualidad (Horn 1989;de Swart 1996;Cooper 1999 ponen de manifiesto que el objeto semántico denotado por el predicado en estos casos se corresponde con una eventualidad, y no con un hecho o proposición 10 . Entre estos diagnósticos se encuentran la compatibilidad con complementos temporales de duración y con modificadores de frecuencia. ...
... La primera prueba consiste en la combinación de <no + nominalización deverbal eventiva> con complementos temporales de duración. Este tipo de complementos pueden modificar a eventualidades (véase (31a)), pero no a hechos o proposiciones (véase (31b) 11 ) (Asher 1993;Przepiórkowski 1999;Higginbotham 2000). ...
... El estatus de los hechos dentro de la ontología semántica es controvertido. Por ejemplo, autores comoAsher (1993Asher ( , 2000 oKratzer (2002) consideran que los hechos son objetos semánticos distintos de las proposiciones, mientras que otros autores como Weiser (2008) equiparan los hechos a las proposiciones. Por su parte,Parsons (1990: cap. 3) distingue dos tipos de hechos -hechos proposicionales (propositional facts) y hechos materiales (material facts)-y sostiene que los primeros guardan determinadas semejanzas con las proposiciones. ...
Este artículo aborda la interacción de la negación con las nominalizaciones deverbales. Se examina qué tipos de nominalizaciones son compatibles con la negación y se pone el foco de estudio en las nominalizaciones eventivas. Se muestra cómo la interpretación que emerge con la construcción <no + nominalización deverbal eventiva> se corresponde, de forma paralela a lo que ocurre en el dominio verbal, con la lectura de eventualidad inhibida. Tras ello, se pone de manifiesto cómo la construcción denota una eventualidad que posee propiedades tanto estativas como eventivas.
... We demonstrate Llamipa's performance using three datasets annotated in the style of SDRT (Segmented Discourse Representation Theory Asher, 1993;Asher and Lascarides, 2003). Our central dataset is the Minecraft Structured Dialogue Corpus (MSDC; Thompson et al., 2024), a corpus of situated, chat-based, task-oriented dialogues, but we also show the generality of our approach by separately fine-tuning on STAC, a corpus of situated multiparty chats from an online game . ...
... There are two main theories that have investigated complete discourse structures for texts: Rhetorical Structure Theory (RST; Mann and Thompson, 1987) and SDRT (Asher, 1993;Asher and Lascarides, 2003). RST assigns relations between contiguous EDUs which are connected hierarchically to further EDUs in a treelike structure. ...
... We demonstrate LLaMIPa's performance using three datasets annotated in the style of SDRT (Segmented Discourse Representation Theory Asher, 1993;Asher and Lascarides, 2003). Our central dataset is the Minecraft Structured Dialogue Corpus (MSDC; Thompson et al., 2024), a corpus of situated, chat-based, task-oriented dialogues, but we show the generality of our approach by separately fine-tuning on STAC, a corpus of situated multi-party chats from an online game . ...
... There are two main theories that have investigated complete discourse structures for texts: Rhetorical Structure Theory (RST; Mann and Thompson, 1987) and SDRT (Asher, 1993;Asher and Lascarides, 2003). RST builds discourse trees over EDUs with the semantic relation labels on the head nodes dominating subtrees. ...
This paper provides the first discourse parsing experiments with a large language model (LLM) finetuned on corpora annotated in the style of SDRT (Asher, 1993; Asher and Lascarides, 2003). The result is a discourse parser, LLaMIPa (LLaMA Incremental Parser), which is able to more fully exploit discourse context, leading to substantial performance gains over approaches that use encoder-only models to provide local, context-sensitive representations of discourse units. Furthermore, it is able to process discourse data incrementally, which is essential for the eventual use of discourse information in downstream tasks.
... Rather, he merely notes that the running took place, and has hidden to avoid meeting the other subject. In the literature on complementation (Asher 1993;Hengeveld and Mackenzie 2008), this meaning is characterized as eventive (or state-of-affairs). The complementizer use of kak as in (3) is presumably related to its use as a manner question word in (1) and (2). ...
... Complementation studies define events (or states-of-affairs) as situations that (do not) occur in reality and are located in space and time (Asher 1993;Hengeveld and McKenzie 2008: 166;Vendler 1967a), e.g., Shooting occurred at a private residence, while propositions are defined as information units having a truth value (Asher 1993: 24-32), or events wrapped up in the mental frame of epistemic value and information source (Hengeveld and McKenzie 2008: 144), John said that he was shot at. The criteria of distinguishing between events and propositions have been widely discussed in the literature. ...
In Russian, the subordinator kak ‘how’ is both a manner question word and an eventive complementizer. The Russian linguistic tradition explains the colexification of the two functions in terms of a semantic shift from manner as characteristic of a situation to event description as a whole. Alternatively, a grammaticalization scenario from manner complements to event/propositional complements has been suggested: manner complements originally have a propositional frame, which is foregrounded concurrently with the loss of the manner meaning, giving rise to both eventive and propositional interpretations. This article is aimed at testing both hypotheses. We study several large Old Russian manuscripts, starting from the first available documents of the 11th century, and show that at the earliest documented period Old Russian kako/kakъ could be used in all types of complement clauses. It could introduce eventive, propositional, manner and irrealis purposive-like complements. Accordingly, the evolution of the subordinator kak in complementation involves a narrowing of its functional domain. We classify Old Russian texts based on the period and trace the gradual loss of particular functions during the centuries. Thus, we show that the Russian data supports the second grammaticalization scenario.
... 6 On the other hand, the decision of whether to attach new units in terms of subordination or coordination is made on the basis of computing the semantic content of the units (ibid.: (Polanyi 1988: 613) 617), a process which is not further specified but which involves inferences from world knowledge and linguistic knowledge. 7 Subordination in terms of structural attachment is most prominently dealt with in the framework of Segmented Discourse Representation Theory (SDRT; Asher 1993;Asher & Lascarides 2003;Lascarides & Asher 2007), a formal theory of the representation and interpretation of discourse that accounts for the hierarchical organization of discourse, relying on the concept of discourse relations between text spans. Just like in RST, each text segment in SDRT is connected to another via discourse relations with no overlap between text segments. ...
... Subordination is at the same time defined in semantic terms: "[ subordinates ] iff the main eventuality described in is a subsort of the main eventuality described in or the proposition associated with defeasibly implies that associated with . " (Asher 1993). Semantically, subordinated discourse units are assumed to provide a more detailed description of some aspect of the superordinated unit, thus introducing a "new level of detail in discourse" (Asher & Lascarides 2003: 8). ...
This handbook chapter addresses the question of whether the notion of embedding is relevant in discourse structure and if so, what constitutes embedding in discourse. We argue that discourse-structural subordination is a case in point, discussing this concept both from the viewpoint of different theories applying the notion of "rhetorical" or "discourse relations", as well as in goal-driven approaches to discourse structure. Furthermore, we provide an overview of how discourse-structural embedding is realized in language with an emphasis on the relationship between discourse-structural and syntactic subordination. Finally, we revisit the notion of discourse subordination from the viewpoint of the semantic and pragmatic processes that may be taken to lead to the establishment of a subordinating relation between discourse segments.
This is the peer-reviewed and accepted version of a chapter to appear in Anton Benz, Werner Frey, Hans-Martin Gärtner, Manfred Krifka, Mathias Schenner, Marzena Zygis (eds.): Handbook of Clausal Embedding. Berlin: Language Science Press.
... Discourse connectives are one of the most important aspects of discourse structure. They are lexicosyntactic elements that signal a pragmatic or semantic relation (contingency, expansion, contrast, etc.) between two discourse units such as verb phrases, clauses or sentences (Asher, 1993;Prasad et al., 2008). While the most well-known discourse connectives belong to syntactic classes such as coordinating and subordinating conjunctions (and, but, because), adverbs (however) or prepositional phrases (in sum), it is known that clitics can also function similarly as discourse connectives (König, 2002), and may convey additive, contrastive or concessive senses (Forker, 2016;Faller, 2020). ...
... Following the principles of the PDTB, we consider dA a DC when it links two segments that have an "abstract object" interpretation (propositions, eventualities, etc.) (Asher, 1993;Prasad et al., 2008). ...
... Caudal & Nicolas 2005), and discourse particles (cf. Abraham 1991) can lexicalize speech act functions, or second-order functions over speech act functions (i.e., they are modifiers of speech act functions) suggests that speech acts functions must have an event argument (and a degree argument too for some of them), and not just bear on abstract types of discourse referents à la Asher (1993). (See e.g., Danlos (2007) for an SDRT implementation of a number of speech act-level lexical items). ...
This paper aims at demonstrating the validity of a two-pronged hypothesis: (i) that the aspectual viewpoint content of the so-called ‘narrative imparfait’ (NIMPF) does not bear on the verb it marks (i.e., it does not combine with the event predicate denoted by said verb) but that, therefore, (ii) it must operate at a higher, discursive semantic level. To substantiate the above hypothesis, the paper first focuses on diachronic and synchronic evidence suggesting that the NIMPF does not contribute aspectual meaning at the sentence semantics-level – showing notably that it behaves like a ‘viewpoint neutral’ tense with respect to the verb it marks. The paper then discusses synchronic, discursive evidence supporting the view that the NIMPF actually indicates a partial, discourse-structurally incomplete, ‘ongoing’ narrative act. From these two facts, the paper concludes that NIMPF utterances refer to imperfectively viewed narrative speech act events, and constitute a separate speech act-level conventionalized reading of the imparfait, applying an imperfective viewpoint meaning to relational speech act functions, i.e., to rhetorical relations. It is argued that they should be endowed with a speech act event argument, and constitute an abstract type of event predicate which the viewpoint meaning of the NIMPF takes as its input.
... It is encoded in the LUD as in (1) (cf. Asher's elaboration relation (Asher, 1993)). In Japanese, the antecedent part can be syntactically determined, so far as the topic phrase is expressed with the topic marker. ...
In the German government (BMBF) funded project Verbmobil, a semantic formalism Language for Underspecified Discourse Representation Structures (LUD) is used which describes several DRSs and allows for underspecification. Dealing with Japanese poses challenging problems. In this paper, a treatment of multiple discourse relation constructions on the sentential level is shown, which are common in Japanese but cause a problem for the formalism,. The problem is to distinguish discourse relations which take the widest scope compared with other scope-taking elements on the one hand and to have them underspecified among each other on the other hand. We also state a semantic constraint on the resolution of multiple discourse relations which seems to prevail over the syntactic c-command constraint.
... Crucially, the literature has claimed that, although we can imagine of something that occurs in the real world which makes the external argument refrain from doing the corresponding affirmative eventuality, as well as of something that makes us expect in the first place that it should carry out this eventuality, these are pragmatic factors which are not encoded within the grammar of the sentences but provided by the extralinguistic world With all that, the literature has provided several grammatical contexts which coerce the inhibited eventuality reading. Among them we find the co-occurrence with durative modifiers introduced by durante 'for', which are compatible with predicates that describe an eventuality (32a) (Asher 1993;Higginbotham 2000). Because these modifiers measure the time during ...
This paper addresses the interaction of the negative particle "no" with Spanish zero event deverbal nominalizations. Firstly, it is shown that these nouns are compatible with "no". Secondly, following a series of grammatical tests, the interpretation(s) that arise with <"no" + zero event deverbal nominalization> are studied. To this respect, it is brought to light that these nouns do not pattern with verbal predicates, because unlike what happens with the latter, the former do not give rise to the inhibited eventuality reading. Thirdly, it is proposed that this interpretation is not available with zero event deverbal nominalizations because negation blocks the lexicalisation via Phrasal Spellout of the noun’s syntactic configuration in this case.
... Deriving DRSs from sentences compositionally is a nontrivial challenge. Efforts towards this goal include λ-DRT (Muskens, 1994;Kohlhase et al., 1996Kohlhase et al., , 1998, Compositional DRT (Muskens, 1996), and bottom-up DRT (Asher, 1993). All of these approaches use lambda calculus to compositionally combine partial meaning representations, which is intractable in broad-coverage semantic parsing (see e.g. the discussion by Artzi et al. (2015)). ...
... As we aim to demonstrate in this paper, discourse structure can help to clean up datasets for training and thereby improve training. MSDC Thompson et al. (2024b) provide full discourse annotations for the MDC, known as the Minecraft Structured Dialogue Corpus (MSDC), using the discourse theory and annotation principles of SDRT (Asher, 1993;Asher and Lascarides, 2003) extended to a multimodal environment, in which both nonlinguistic actions and discourse moves can enter into semantic relations like Elaboration, Correction, and Narration (Hunter et al., 2018;Asher et al., 2020). They follow annotation practices given for the STAC corpus (Asher et al., 2016). ...
... SDRT (Segmented Discourse Representation Theory) is used to fulfill the analyses, in which discourse structures are modeled into graphs made up of two types of semantic discourse units (DUs): elementary discourse units and complex discourse units [6,7,8,9]. An EDU is the most basic element as it often contains only 1 eventuality state or event [10]. ...
While many researchers have paid attention to Rabindranath Tagore's poems, few have analyzed his short stories, still less on a specific story, like The Renunciation. The fiction critiques the caste system of Hinduism and portrays a unique female character. This paper proposes that structural and coherence analysis can assist in a better appreciation of the story. The disruption of the timeline emphasizes the importance of a series of events occurred in the past. Coherence relations within them are established and analyzed so that the caste system's role as the origin of perpetuating tragedy among Hindus is explicitly revealed. The blank in the timeline along with the nightingale which appears throughout the novel indicates the plight of women in Hindu society. Analyses in this paper provide argumentation for views proposed in previous studies, namely Tagore's firm stance against the caste system and his empathy for suffering women, from a new angle. It also demonstrates the feasibility of linguistic theories appliance to the field of literary criticism.
... Some schemes accept them as markables such as the Gum corpus (Zeldes, 2017), but we do not annotate them because they are not discourse entities themselves but properties so they cannot pass our referentiality criterion. Abstract Entities: We left out reference to abstract objects like propositions, state-of-affairs, and other sort of such entities discussed by (Asher, 1993), as their inclusion immensely complicates the annotation task when handled along with conceptually simpler type of referents we aimed to capture in the present study. (See Zeyrek et al. (2010) for abstract object annotation in Turkish Discourse Bank (Zeyrek et al., 2013)). ...
... because, therefore), Temporal (e.g. and then, subsequently), and other types of relations. 2 In simple terms, connectives are words or short phrases that explicitly express a discourse relation between propositions (usually sentences, see Asher (1993) and its notion of "abstract objects" for further information). Connectives form a syntactically heterogeneous group (e.g., conjunctions and adverbs) and can be semantically ambiguous, as they may exhibit a discourse reading or a sentential reading. ...
Machine-readable inventories of connectives that provide information on multiple levels are a useful resource for automated discourse parsing, machine translation, text summarization and argumentation mining, etc. Despite Chinese being one of the world’s most widely spoken languages and having a wealth of annotated corpora, such a lexicon for Chinese still remains absent. In contrast, lexicons for many other languages have long been established. In this paper, we present 226 Chinese discourse connectives, augmented with morphological variations, syntactic (part-of-speech) and semantic (PDBT3.0 sense inventory) information, usage examples and English translations. The resulting lexicon, Chinese-DiMLex, is made publicly available in XML format, and is included in connective-lex.info, a platform specifically designed for human-friendly browsing of connective lexicons across languages. We describe the creation process of the lexicon, and discuss several Chinese-specific considerations and issues arising and discussed in the process. By demonstrating the process, we hope not only to contribute to research and educational purposes, but also to inspire researchers to use our method as a reference for building lexicons for their (native) language(s).
... Although this is less common in semantics, it has been argued that there are constraints that play a role in the construction of discourse. An example of this is the Right Frontier Constraint (see Asher, 1993;Asher & Lascarides, 2003;Polanyi, 1997;Webber, 1988), which states that a discourse constituent must be attached on the right frontier of the ongoing discourse. Discourse accessibility could be seen as a comparable sort of constraint that helps to determine the well-formedness of a discourse. ...
The general view in syntactic literature is that binding constraints can make antecedents syntactically inaccessible. However, several studies showed that antecedents which are ruled out by syntactic binding constraints still influence online processing of anaphora in some stages, suggesting that a cue-based retrieval mechanism plays a role during anaphora resolution. As in the syntactic literature, in semantic accounts like Discourse Representation Theory (DRT), formal constraints are formulated in terms of accessibility of the antecedent. We explore the discourse inaccessibility postulated in DRT by looking at its role in pronoun resolution of inter-sentential anaphoric relations in four off-line and two eye-tracking experiments. The results of the eye-tracking experiments suggest that accessibility has an effect on pronoun resolution from early on. The study quantifies evidence of inaccessible antecedents affecting pronoun resolution and shows that almost all evidence points to the conclusion that discourse-inaccessible antecedents are ruled out for pronoun resolution in processing. The only potential counter-example to this claim is also detected, but remains only as anecdotal evidence even after combining data from both eye-tracking studies. The findings in the study show that accessibility plays a significant role in the processing of pronoun resolution in a way which is potentially challenging for the cue-based retrieval mechanism. The paper argues that discourse accessibility can help expand the theories of retrieval beyond the syntactic and sentence-level domain and provides a window into the study of interference in discourse.
... Nous détaillons plus bas les définitions formelles des éventualités du Type 1 et du Type 2. Avant de présenter et commenter ces définitions, quelques précisions doivent être apportées, en particulier au sujet des opérateurs servant à saisir les rapports spatio-temporels entre les sous-événements de changement de relation et de changement d'emplacement (dans cette section et à la section 4.2). Ces opérateurs ont déjà été utilisés dans la littérature (ex : Asher, 1993, Asher et al., 1995, Kamp & Reyle, 1993. Les relations ⊆t, ≡t et ⫗ indiquent respectivement l'inclusion temporelle, l'équivalence/identité temporelle et la précédence immédiate (ou « aboutement »). ...
À partir de recherches antérieures sur l’expression du déplacement strict en français (par des verbes tels que arriver, partir, se rendre et sortir, par exemple), cette étude fait un bilan onomasiologique des composants sémantiques permettant de construire des éventualités ou procès de cette nature. Sont d’abord recensés les composants partagés par l’ensemble des éventualités de déplacement strict du français et qui déterminent leur structure spatio-temporelle. Divers composants de sens additionnels, susceptibles de s’agréger à la structure spatio-temporelle ainsi élaborée, sont ensuite passés en revue. Outre cet inventaire descriptif, quelques éléments d’un système de représentation formelle et graphique des éventualités de déplacement strict sont également exposés.
... MSDC Thompson et al. (2024) provided full discourse annotations for the Minecraft corpus, known as the Minecraft Structured Dialogue Corpus (MSDC), using the discourse theory and annotation principles of SDRT (Asher, 1993;Asher and Lascarides, 2003) extended to a multimodal environment, in which both nonlinguistic actions and discourse moves can enter into semantic relations like Elaboration, Correction, and Narration (Hunter et al., 2018;Asher et al., 2020). They followed annotation practices given for the STAC corpus (Asher et al., 2016). ...
When engaging in collaborative tasks, humans efficiently exploit the semantic structure of a conversation to optimize verbal and nonverbal interactions. But in recent "language to code" or "language to action" models, this information is lacking. We show how incorporating the prior discourse and nonlinguistic context of a conversation situated in a nonlinguistic environment can improve the "language to action" component of such interactions. We fine tune an LLM to predict actions based on prior context; our model, NeBuLa, doubles the net-action F1 score over the baseline on this task of Jayannavar et al.(2020). We also investigate our model's ability to construct shapes and understand location descriptions using a synthetic dataset.
... we show that lt-na in Akan is used for marking such relations. Asher (1993), Asher (1998), Lascarides and Asher (1993) and Asher and Lascarides (2003) As Asher & Lascarides note, discourses (9a) and (9b) have the same tense forms and aspectual classes (aka 'Aktionsarten'), yet they seem to imply different coherence relations between their sentences. In (9a), the sentences are interpreted as relating a story in which a certain sequence of events is described; hence they are understood to be temporally ordered. ...
Grammatical accounts of na in Akan identify two different forms: nà with a low tone (LT-na) and ná with a high tone (HT-na). LT-na functions in two ways: as a focus marker or a conjunction, the latter of which can take a prefix and be realized as ɛna. While some scholars treat them as two different na’s, others point to a commonality between the two. HT-na has been analyzed as functioning as past and future tenses, and a logical connector. We argue that these three HT-na along with the two LT-na are subcategories of a super-category. We propose that the super-category is a non-tonal na (call it Root-na), with a common basic meaning which explains all five seemingly unrelated interpretations. Root-na links the na-clause with something in the common ground, i.e., to something that appeared in the previous context or is presupposed. It is spelled out as a LT-na or HT-na, depending on the kind of linking. LT-na marks discourse coherence relations such as focus and narrative-sequence, both of which are shown in the linguistics literature to be anaphoric. HT-na is an intensional marker which links times or possible worlds.
... However, although they coincide in allowing clitic resumption of the predicate by means of a neuter clitic ho/lo, they differ in that only Catalan allows clitic resumption of the predicate by means of the clitic NE. 9 What is the difference between these clitics (i.e., ho/lo and NE)? Both clitics can be said to denote abstract semantic objects (Asher 1993); however, in Catalan, ho allows a broad spectrum of potential antecedents, including propositions, properties and situations, while en can resume properties but not propositions . ...
The clitic pronoun ne and the functional element de introducing nominal constituents have many nominal and prepositional functions across Romance languages. In this article, we focus on the nominal functions, singling out three different bundles of semantic features that characterize both ne and de. They can denote properties of individual entities, properties of kinds, or predicate properties. The article shows that Catalan ne and de display the three types of denotation, while Italian ne and de only display the first one. This article further supports the hypothesis that the indefinite determiner de can be overt or silent, thereby unifying de-phrases (and the Italian partitive article) with bare nouns. The analysis of de as an indefinite determiner is then extended to adjectival de, which is claimed to mark concord features on adjectives in both Catalan and Italian.
... For instance, the coherence relation PARALLEL enforces a particular interpretation of the pronoun. On the SDRT account this follows, because two sentential structures can be related by PARALLEL only if their arguments match (Asher 1993). 9 Resolutions to individuals other than Stanley fail to instantiate a structure that could count as parallel to the one assigned to the fi rst sentence Phil poked Stanley, and the two utterance units can thus not be connected with the relation PARALLEL. ...
Stojnić (2021) argues that the content of linguistic utterances is determined by the rules of natural language grammar more stringently than what is generally assumed. She proposes specifically that coherence relations are encoded by the linguistic structures and determine what individuals count as most prominent, thereby serving as the referents of free (“demonstrative”) pronouns. In this paper, I take a close look at the empirical evidence from English and Serbian that she offers in support of this position. Considering these data points in connection with additional linguistic data (also from German and Japanese), I argue that there is no compelling evidence for the assumption that coherence relations directly determine the resolution of pronouns. Instead, grammatical restrictions imposed by different types of pronouns and tenses have a larger impact on the meaning conventionally expressed by complex utterances than what is generally assumed in the literature on coherence relations.
... Many of these strings are just non-meaningful sequences of words but the space also includes coherent and consistent strings that form meaningful texts and conversations. Building on formal theories of textual and conversational meaning (Kamp and Reyle 1993;Asher 1993;De Groote 2006;Asher and Pogodalla 2010), Asher, Paul, and Venant (2017) use this subset of coherent and consistent texts and conversations to define the semantics and strategic consequences of conversational moves in terms of possible continuations. ...
Transformer-based language models have been shown to be highly effective for several NLP tasks. In this article, we consider three transformer models, BERT, RoBERTa, and XLNet, in both small and large versions, and investigate how faithful their representations are with respect to the semantic content of texts. We formalize a notion of semantic faithfulness, in which the semantic content of a text should causally figure in a model’s inferences in question answering. We then test this notion by observing a model’s behavior on answering questions about a story after performing two novel semantic interventions—deletion intervention and negation intervention. While transformer models achieve high performance on standard question answering tasks, we show that they fail to be semantically faithful once we perform these interventions for a significant number of cases (∼ 50% for deletion intervention, and ∼ 20% drop in accuracy for negation intervention). We then propose an intervention-based training regime that can mitigate the undesirable effects for deletion intervention by a significant margin (from ∼ 50% to ∼ 6%). We analyze the inner-workings of the models to better understand the effectiveness of intervention-based training for deletion intervention. But we show that this training does not attenuate other aspects of semantic unfaithfulness such as the models’ inability to deal with negation intervention or to capture the predicate–argument structure of texts. We also test InstructGPT, via prompting, for its ability to handle the two interventions and to capture predicate–argument structure. While InstructGPT models do achieve very high performance on predicate–argument structure task, they fail to respond adequately to our deletion and negation interventions.
... [...] Recognizing coherence relations may thus be just one way of using very general principles for simplifying our view of the world. (Hobbs 1990) Hobbsian ideas have been made formally precise in Segmented Discourse Representation Theory (SDRT, Asher 1993, Asher and Lascarides 2003, Asher and Vieu 2005, Hunter et al. 2018, which aims to model what coherence relations mean, and how discourse structures are constructed. In particular, SDRT models discourse structure as a graph over semantic representations of pieces of discourse or discourse units (DUs), which come in two types: (i) elementary discourse units (EDUs), which are the atoms of a given discourse, and (ii) complex discourse units (CDUs), which are built out of EDUs and may include only two or three EDUs or correspond to several paragraphs or even multiple pages of text. ...
Through imaginative engagement readers of fiction become, to an extraordinary extent, the narrator’s ‘children’: they often submit themselves to the narrator’s authority without reserve. But precisely because of that, readers are deeply at a loss when their trust is betrayed. This underscores a core function of fiction, namely to evoke emotional response in the reader. In this paper, we hypothesize how a reader’s imaginative engagement can be subjected to narrative frustration due to processing or moral complexity. The types of narrative frustration we consider differ in terms of their sources, and their emotional and behavioral impacts on the reader. Here, we break down these frustrations into their component parts, in an effort to better characterize the different classes of frustrations. We propose that frustrations arise from different combinations of local uncertainty, moral clash and global uncertainty. These sources of frustration in turn explain the reader’s emotional response and their consequent reading behavior as they imaginatively engage with fiction.
... 3 Perdicoyianni-Paléologou (2014). 4 Voir par exempleLyons (1977),Asher (1993),Ginzburg et Sag (2000),Moltmann (2013;2022). ...
This chapter is concerned with the reference to abstract objects by means of demonstrative pronouns in Homer. The corpus is made of books Θ and I of the Ilias. This survey is supplemented by probing the rest of the Ilias and the Odyssey. Demonstrative pronouns mostly point at proposition-like objects (propositions, facts, possibilities, as opposed to semi-abstract objects like events and situations). Our main finding is that two situations must be teased apart: intrasentential cataphora to clauses (which is mainly made in the singular), on one hand, and, on the other hand, anaphora and deixis (which mainly make use of plural demonstratives). Moreover, we detected two temporal strata in the Homeric text for anaphora and deixis: In the ancient parts, abstract reference is in the plural, while the recent layers feature both plural and singular terms. In these passages, a relevant variable that triggers the appearance of the singular is the reference to facts, as will be in classical Greek. Crucially, it is precisely in these latter cases that the textual tradition is the shakiest.
Ce chapitre traite de la référence aux objets abstraits au moyen de pronoms démonstratifs chez Homère. Le corpus est constitué des chants Θ et I de l'Iliade. L'étude est complétée par des sondages dans le reste de l'Iliade et dans l'Odyssée. Les pronoms démonstratifs pointent surtout vers des objets propositionnels (propositions, faits, possibilités, par opposition avec les objets semi-abstraits comme les événements et les situations). Notre principal résultat est que deux situations doivent être distinguées : d'une part, la cataphore intraphrastique à des propositions (qui se fait surtout au singulier) et, d'autre part, l'anaphore et la déixis (qui se font surtout au moyen de démonstratifs pluriels). De plus, nous avons repéré deux couches temporelles dans le texte homérique pour l'anaphore et la déixis : dans les parties anciennes, la référence abstraite est au pluriel, tandis que dans les passages récents, on trouve des termes tant au pluriel qu'un singulier. Dans ces derniers, une variable pertinente pour l'emploi du singulier est la référence aux faits, comme ce sera le cas en grec classique. Et c'est précisément dans ces derniers cas que la tradition textuelle est la plus hésitante.
... We have S ϕ as a referent representing the set of facets that individuate Sydney. Such referents are known as structured referents, in DRT (Asher, 1993;Asher and Lascarides, 2003;Brasoveanu, 2007;Kamp et al., 2011: Ch. 3-4). These referents can represent (union) sets of referents or sets of conditions. ...
The goal of this paper is to offer a unified account of Place as a central theoretical notion across different disciplines. We show that while psychology, geography and other sciences have been converging to a unified view of this notion, linguistics still offers a fragmented perspective. Consequently, place names lack a full-fledged analysis that connects this category to the psychological concept of place. We propose to overcome this impasse by introducing a multi-modal Discourse Representation Theory (DRT) account of place as a conceptual construct and place concepts as specific instances of this construct. We show that current variants of DRT permit us to model place names and their senses, i.e., the meaning(s) that individuals associate with Sydney. We then model non-linguistic place concepts, i.e., the mental representation(s) that individuals can have of the city carrying this name. We present a model of the relation between linguistic meaning and conceptual content via the notion of anchoring relations applied to place. We pair this formal treatment with a morpho-syntactic account of place names building on current generative syntax treatments of proper names. Once we have a morpho-syntactic and semantic model of place names, we use a frame semantics treatment to account for lexical relations among place names. We test the overarching model on a set of recalcitrant problems afflicting current linguistic and multi-disciplinary treatments of place. These are the grammatical complexity and lexical content of place names, place concepts and their networks, and inter-subjective, communicative models of place in discourse. By solving these problems, our account integrates several frameworks (DRT, conceptual analysis, generative syntax, frame semantics) and connects several disciplines (linguistics, psychology, geographic information science, communication models) via a novel, multi-modal account of place. We conclude by discussing the theoretical and empirical import of these results.
... Various frameworks exist that can be used to annotate discourse relations, such as RST (Mann and Thompson, 1988) and SDRT (Asher, 1993). In this work, we focus on the annotation of implicit discourse relations, following the framework used to annotate the Penn Discourse Treebank 3.0 (PDTB, Webber et al., 2019). ...
Disagreement in natural language annotation has mostly been studied from a perspective of biases introduced by the annotators and the annotation frameworks. Here, we propose to analyze another source of bias—task design bias, which has a particularly strong impact on crowdsourced linguistic annotations where natural language is used to elicit the interpretation of lay annotators. For this purpose we look at implicit discourse relation annotation, a task that has repeatedly been shown to be difficult due to the relations’ ambiguity. We compare the annotations of 1,200 discourse relations obtained using two distinct annotation tasks and quantify the biases of both methods across four different domains. Both methods are natural language annotation tasks designed for crowdsourcing. We show that the task design can push annotators towards certain relations and that some discourse relation senses can be better elicited with one or the other annotation approach. We also conclude that this type of bias should be taken into account when training and testing models.
... Our idea is to apply truth conditional semantics to LLMs by representing models themselves as strings. Semanticists have used strings and continuation semantics (Reynolds, 1974) -in which the meaning of a string s is defined in terms of its possible continuations, the set of longer strings S that contain s-to investigate the meaning and strategic consequences of conversational moves (Asher et al., 2017), temporal expressions (Fernando, 2004), generalized quantifiers (Graf, 2019), and the "dynamic" formal semantics of (Kamp and Reyle, 1993;Asher, 1993)(De Groote, 2006Asher and Pogodalla, 2011). In our case, we will use strings to define models A s . ...
... 2) Coordinating RRs, such as Narration, Parallel and Contrast, exist between units of the same information packaging level. SDRT theory (Asher, 1993;Asher and Lascarides, 2003;Lascarides and Asher, 2008) adopt the distinction of two types of RRs and consider they govern the hierarchical structure of discourses. ...
... For this approach to work, we need a meaning representation frame-work where (i) the formalism is language-neutral, (ii) there is aligned data both in terms of meaninglanguage(s), but also multilingually across different languages, and (iii) there is enough expressivity to cover for a wide range of language phenomena. Discourse Representation Structure (DRS), which satisfies our requirements well, is the formal meaning representation proposed in Discourse Representation Theory (DRT, Kamp 1981;Asher 1993;Kamp and Reyle 1993;Kadmon 2001;Kamp et al. 2011;Geurts et al. 2020). It covers a large variety of linguistic phenomena, including anaphors, presuppositions, temporal expressions and multisentence discourses and captures the semantics of negation, modals and quantification. ...
... Events can also introduce new entities into the discourse or narrative, through the use of creation predicates (Asher, 1993;Badia and Saurí, 2000). This is pervasive in procedural text, where the goal is to describe a sequence of transformations to apply to multiple objects to build up a goal object. ...
... While most work on abstractive meeting summarization implicitly assumes that conversation is merely a linear sequence of utterances without semantic relations between them, example (1) underscores the importance of semantic relations for conversational understanding. Drawing on similar insights, certain recent approaches exploit independent theories of discourse structure, such as Rhetorical Structure Theory (RST; Mann and Thompson, 1987) and Segmented Discourse Representation Theory (SDRT; Asher, 1993;Lascarides and Asher, 2008), to improve summarization. Accounts like RST and SDRT maintain that in a coherent conversation, each (roughly) clause-level unit should be semantically related to some other part of the conversation via a discourse relation such as Question-Answer Pair (QAP), Acknowledgement (Ack), Explanation, Contrast, etc., to reflect its contribution to the larger discourse. ...
A system that could reliably identify and sum up the most important points of a conversation would be valuable in a wide variety of real-world contexts, from business meetings to medical consultations to customer service calls. Recent advances in deep learning, and especially the invention of encoder-decoder architectures, has significantly improved language generation systems, opening the door to improved forms of abstractive summarization—a form of summarization particularly well-suited for multi-party conversation. In this paper, we provide an overview of the challenges raised by the task of abstractive meeting summarization and of the data sets, models, and evaluation metrics that have been used to tackle the problems.
Semantic theories for natural language assume many different kinds of objects, including (among many others) individuals, properties, events, degrees, and kinds. Formal type-theoretic semantics tames this 'zoo' of objects by assuming only a small number of ontologically primitive categories and by obtaining the objects of all other categories through constructions out of these primitives. This Element surveys arguments for this reduction of semantic categories. It compares the ontological commitments of different such reductions and establishes relations between competing foundational semantic ontologies. In doing so, it yields insights into the requirements on minimal semantic ontologies for natural language and the challenges for semantic ontology engineering.
This Element gives an introduction to the emerging discipline of natural language ontology. Natural language ontology is an area at the interface of semantics, metaphysics, and philosophy of language that is concerned with which kinds of objects are assumed by our best semantic theories. The Element reviews different strategies for identifying a language's ontological commitments. It observes that, while languages share a large number of their ontological commitments (such as to individuals, properties, events, and kinds), they differ in other commitments (for example, to degrees). The Element closes by relating different language and theory-specific ontologies, and by pointing out the merits and challenges of identifying inter-category relations within a single ontology.
This paper deals with three interrelated topics, linguistic anaphora, multi-modal anaphora and the top-down broadcasting of information using gestural post-holds in multimodal dialogue. Initially, a new solution for definite, pronominal and pro-adverbial anaphora is given based on the idea that an existentially quantified general term may output a definite reference. This approach is extended to multimodal anaphora, where part or all of an anaphor’s meaning is contributed by some sequence of iconic or deictic gestures. Anaphora exploit the semantic potential of their antecedents, they work, as tradition has it, “bottom-up”. An inverse relation, more general than cataphora, and investigated here for the first time, is “broadcasting”, where information is freely distributed top down and input to receiving sites (ports). Anaphora are modelled with the same top-down mechanism and the same applies for coherence relations in dialogue which generally show an anaphora-like behaviour. “Broadcasting” can be used in the context of anaphors, for example, to provide their gestural meaning parts but also for a verb’s multi-modal arguments for referring to a location, a direction or an area. As to multi-modal data, broadcasting is shown to be frequently tied up with gestural post-holds, the holding of a gesture’s stroke information independently of semantically alignable speech. This leads to considering post-holds from a new perspective, stressing their speech-independent function and their relevance for indicating topic-continuity. We show that multi-modal anaphora and especially broadcasting cross single contributions and turns. The data which let us develop these perspectives come from the SaGA (Speech and Gesture Alignment) corpus, a set of route-description dialogues generated in a VR-setting incorporating marker-based eye-tracking facilities. The calculus used to model the anaphora and broadcasting dynamics is the concurrent λΨ-calculus, a recently developed two-tiered machinery using a Ψ-calculus for input-output, data transport and broadcasting. The data transported are in a typed λ-calculus format incorporating Neo-Davidsonian representations; these data can be linguistic, gestural only or multi-modal. Multi-modal informational chunks are modelled as communicating agents sending and receiving information via input-output-channels. They are introduced incrementally on an empirically motivated construction or gesture-plus-construction or gesture only basis. The λΨ-calculus is also used for the multi-modal fusion component unifying gestural and linguistic information; hence, the paper is also a contribution to multi-modal fusion of linguistic and gestural input. Finally, it is shown how the presented algorithm can capture multi-modal coherence relations or a multi-modal anaphora resolution based on PTT ideas.
Discourse analysis is a highly applicable area of natural language processing. In English and other languages, resources for discourse-based tasks are widely available. Thai, however, has hitherto lacked such resources. We present the Thai Discourse Treebank, the first, large Thai corpus annotated in the style of the Penn Discourse Treebank. The resulting corpus has over 10,000 sentences and 18,000 instances of connectives in 33 different relations. We release the corpus alongside our list of 148 potentially polysemous discourse connectives with a total of 340 form-sense pairs and their classification criteria to facilitate future research. We also develop models for connective identification and classification tasks. Our best models achieve an F1 of 0.96 in the identification task and 0.46 on the sense classification task. Our results serve as benchmarks for future models for Thai discourse tasks.
This paper addresses how the negative particle "no" interacts with Spanish deverbal nominalizations that denote an event. Firstly, it is pointed out that, when preceded by negation, these nominalizations only give rise to the inhibited eventuality reading, contrary to what happens with verbs, which give rise to both the inhibited eventuality reading and the negated eventuality reading. Secondly, it is shown that, when these nominalizations co-occur with "no", their lexical aspect is modified, as they share properties with events, but with states as well. Thirdly, a proposal of analysis for <"no" + event deverbal nominalization> is presented. We claim that both the interpretation and lexical aspect of this construction follow from the interaction of negation with the syntactic configuration of the nominalization and, specifically, with the projection responsible for the triggering of the eventuality.
Certain propositional anaphora, like the Polarity Particles and Polar Additives discussed in this paper, are sensitive to the polarity of their antecedent clause. The paper establishes that discourse polarity—the polarity of the antecedent clause for the purposes of licensing subsequent anaphora—is influenced by complex factors, some of them syntactic, semantic, and pragmatic in nature.
The hyperintensional dynamic framework presented here gives a formal foundation to a distinction of polarity for propositional discourse referents and captures some of the central generalizations. It uses discourse referents for hyperintensional propositions, providing a level of representation that connects information from the discourse context, the proposition’s semantic content, and the information about the polarity of the antecedent clause. Therefore, it constitutes a step towards an analysis capturing the heterogeneous factors influencing discourse polarity.
As the counterpart to kinds of entities or objects [5], event kinds are used to account for a variety of linguistic facts in the event domain. The modification of event kinds is usually considered to be restricted, and temporal modifiers are claimed to be hardly acceptable [18]. In this paper I focus on verbal gerunds in English, which have been analyzed as event kind descriptions [15], and demonstrate that they accept temporal modification but remain kind-referring. I extend the analysis of frequency adjectives [13] to interpret both temporal and frequency modifiers in verbal gerunds, showing that the modification of event kinds is less restricted than commonly assumed.
The paper presents a detailed corpus-based analysis of the German prospective stehen vor NP light verb construction. The starting point of the analysis is the claim that the construction is restricted to change-of-state nouns in the NP-internal position (Fleischhauer & Gamerschlag 2019, Fleischhauer et al. 2019). Based on corpus data, I demonstrate that although the construction shows a strong preference for such nouns, other semantic types of nouns (such as state nouns or process nouns) occur in the construction as well. I argue that process nouns in particular require contextual support to be licensed within the construction. In the paper, I present an analysis of the prospective light verb construction in terms of current relevance. This analysis accounts for the observed preference for change-of-state NP-internal nouns as well as for the need to provide contextual support for process nouns. The notion current relevance is frequently employed in the analysis of the perfect aspect; the current paper represents the first attempt to extend this notion to the prospective aspect.*
Individual Differences in Anaphora Resolution: Language and cognitive effects explores anaphora resolution from different perspectives, and investigates various aspects of the phenomenon, as contributions include research protocols that combine old and new experimental methodologies as well as theoretical and empirical approaches. A central theme across volume contributions are the multiple linguistic and extralinguistic factors that constrain anaphora resolution, its processing and acquisition by a variety of populations (children and adults, monolinguals, bilinguals and second language learners) as well as the mechanisms underlying anaphora resolution. Anaphora resolution constitutes an ideal environment to test the interaction between domain-general cognitive systems and domain-specific linguistic sub-routines, since variability in referential preferences is not related to binding constraints (an integral part of syntax per se) but is closely tied to processing (functional constraints) modulated by the integration of discourse-filtered information.
Yüklemler adsıl üyeler yanında tümcesel üyeler de alabilmektedir. Güncel anlambilim çalışmalarının genel çerçevesini oluşturan Doğruluk Koşullu Anlambilim kuramına göre, özel adlar bir varlığa gönderimde bulunmak suretiyle onu zihnimizde çağrıştıran e tipinde öğeler, tümceler ise doğruluk değeri alabilen <s,t> tipinde önermelerdir. Ad ve önermenin yanında, bir tümcede üye rolü üstlenebilecek öğeler NE sözcükleri de barındırabilmektedir. Soru tümceleri olarak adlandırılan bu öğeler ise, barındırdıkları gizil bildirme tümcesini doğru kılacak önermeler kümesini ifade ederler. Üye rolü üstlenebilme bakımından NE tümceciklerinin dağılımı bazı kısıtlamalara tabidir. Bu kısıtlamalar çoğunlukla ana tümcede bulunan yüklemin iç/dış üyesi üzerinde uyguladığı seçme kısıtlamaları ile izah edilmektedir ki bu kısıtlamalara, üyenin hangi ontolojik ulamdan olması gerektiği bilgisi de dahildir. Ancak, NE tümcelerinin dağılımı görünenden daha karmaşıktır. Sözgelimi iç üye olarak sadece önerme seçen inan- yükleminin aksine, Türkçede gör- ve bil- gibi bazı yüklemler iç üye olarak hem NE tümcesi hem de şey/kişi/yer gibi adsıl öğeleri kabul etmekte iken bu serbesti getir- gibi bazı yüklemler için geçerli değildir. İngilizce gibi dillerde ise yüklemler iç üye olarak hem NE tümcesi hem de adsıl bir üye alma bakımından Türkçeden daha serbest davranmaktadır. Doğal dil ontolojisindeki rolü bakımından NE tümcelerinin soru, şey/kişi/yer ifadelerin ise varlık kodladığını varsayarsak, hem Türkçe-İngilizce eşleşmezliğini hem de tümcesel/adsıl üye seçme bakımından Türkçe yüklemler arasında görülen bu farkı nasıl açıklayabiliriz? Bu çalışmada söz konusu fark, NE tümcelerinin yorumu bakımından İngilizce ile Türkçe arasındaki farka dayandırılmakta, bunların İngilizcede hem soru hem de varlıklara gönderim yapabilmesi mümkünken Türkçede sadece soru kodlayabileceği öne sürülmektedir. Bu nedenle, üye olarak bir varlık seçen getir- eylemi NE tümceciklerini reddedecektir. Öte yandan, şey/kişi/yer gibi adsıl üyeler barındırmaları nedeniyle varlıklara gönderimde bulunuyor gibi görünen bazı ifadelerinin, aslında alanyazında Gizli Soru adıyla bilinen yapıları örneklendirdiği ve bu nedenle de Türkçede NE tümcelerinin her zaman sorulara gönderimde bulunduğu tezine aykırılık teşkil etmediği gösterilecektir. Bu tarz yapılar, şey/kişi/yer gibi sözcüklerin (ana yükleme bağlı olarak) gizli soru kodlayabilmesi sayesinde mümkün olabilmektedir. Bu veriler ışığında, NE tümcelerinin varlığa gönderimde bulunduğu Serbest İlgi Tümceciklerinin Türkçede mümkün olmadığı sonucuna varılacaktır.
Anaphora resolution is a traditional task in the natural language processing community, defined as a cohesion phenomenon where one entity points back to a previous entity. Event discourse deixis (EDD) is a kind of more complex anaphora in which the anaphors refer to event descriptions such as sentences or clauses. Event discourse deixis resolution (EDDR) is able to help machines understand the richer linguistic and semantic information in the discourse. However, compared to anaphora resolution, EDDR has received relatively less research attention. In this work, we investigate the EDDR task by designing the corresponding dataset and model. First, we manually construct a high-quality Chinese corpus for EDDR, including 4,417 documents and 5,929 event chains that consist of event antecedents and anaphors. Second, we propose a deep neural network model for EDDR, which formulates the task into two subtasks, namely event anaphor recognition and event antecedent recognition. Our model is trained under the two subtasks jointly so that the EDDR task can be performed end-to-end. Besides our final model, we also build 7 pipeline and joint models as baselines to build comprehensive benchmarks for follow-up research. Experimental results on our EDDR dataset show that our model outperforms all the baselines and achieves about 53%, 44%, 53% and 63% F1s using standard anaphora resolution metrics such as CoNLL, MUC, B3 and Ceafe. The performances show that EDDR is a challenging task and worth researching in the future. Our dataset and model will be released to facilitate the follow-up research.
In discussing English articles, we refer to the distinction between countable nouns and uncountable nouns. A countable noun is expressed in the “a + noun” form (e.g., an apple) or the plural noun form (e.g., apples), whereas an uncountable noun in the “Ø + noun” form (e.g., apple). Distinguishing what is countable from what is not seems fairly straightforward, and the use of countable and uncountable nouns in English should be easy. L2 learners, however, consider that English articles are among the most difficult grammatical items. A cognitive linguistics, which is an alternative to the traditional way of presenting English articles, claims that the countable / uncountable distinction applies not to the word, but to the referent the word denotes. The perception‐based principle of countability and a more general principle of individuation will be discussed here to draw pedagogical implications.
One of the most interesting aspects of natural language is how texts cohere, which involves the pragmatic or semantic relations that hold between clauses (addition, cause-effect, conditional, similarity), referred to as discourse relations. A focus on the identification and classification of discourse relations appears as an imperative challenge to be resolved to support tasks such as text summarization, dialogue systems, and machine translation that need information above the clause level. Despite the recent interest in discourse relations in well-known languages such as English, data and experiments are still needed for typologically different and less-resourced languages. We report the most comprehensive investigation of shallow discourse parsing in Turkish, focusing on two main sub-tasks: identification of discourse relation realization types and the sense classification of explicit and implicit relations. The work is based on the approach of fine-tuning a pre-trained language model (BERT) as an encoder and classifying the encoded data with neural network-based classifiers. We firstly identify the discourse relation realization type that holds in a given text, if there is any. Then, we move on to the sense classification of the identified explicit and implicit relations. In addition to in-domain experiments on a held-out test set from the Turkish Discourse Bank (TDB 1.2), we also report the out-domain performance of our models in order to evaluate its generalization abilities, using the Turkish part of the TED Multilingual Discourse Bank. Finally, we explore the effect of multilingual data aggregation on the classification of relation realization type through a cross-lingual experiment. The results suggest that our models perform relatively well despite the limited size of the TDB 1.2 and that there are language-specific aspects of detecting the types of discourse relation realization. We believe that the findings are important both in providing insights regarding the performance of the modern language models in a typologically different language and in the low-resource scenario, given that the TDB 1.2 is 1/20th of the Penn Discourse TreeBank in terms of the number of total relations.
This essay presents a philosophical and computational theory of the representation of de re, de dicto, nested, and quasi-indexical belief reports expressed in natural language. The propositional Semantic Network Processing System (SNePS) is used for representing and reasoning about these reports. In particular, quasi-indicators (indexical expressions occurring in intentional contexts and representing uses of indicators by another speaker) pose problems far natural-language representation and reasoning systems, because—unlike pure indicators—they cannot be replaced by coreferential NPs without changing the meaning of the embedding sentence. Therefore, the referent of the quasi-indicator must be represented in such a way that no invalid coreferential claims are entailed. The importance of quasi-indicators is discussed, and it is shown that all four of the above categories of belief reports can be handled by a single representational technique using belief spaces containing intensional entities. Inference rules and belief-revision techniques for the system ore also examined.