Semantic Compositionality

This dissertation is dedicated to the philosophical analysis of variables. This task presents considerable challenges. We believe that any satisfactory characterization of variables essentially involves semantic considerations which eventually treat variables as meaningful constituents of languages. Hence, we believe that variables are not only syntactic category but also a semantic category. The scope of our investigation here is restricted. Specifically, we focus solely on firstorder languages. We do not explore other formal languages or variable expressions in natural languages. We believe this restriction is theoretically benign. First-order languages are taken to be canonical formal languages, and variable expressions that occur in first-order languages are paradigmatic for variable symbols. We point out that from the syntactic perspective, there is no unique syntactic category that variables of first-order languages belong to. Firstly, there are free occurrences of individual variables in a formula. These are regarded as first-order terms, together-but not exclusively-with individual constants. Secondly, there are bound occurrences of variables in a formula that are part of the mechanism of quantification. Bound occurrences of variables are not supposed to be interpreted independently of the quantifiers which govern these occurrences of variables. But we notice that this is only partially true relative to standard Tarskian objectual semantics. In this semantics, formulas involving a bound occurrence of a variable are evaluated via recursive decomposition that involves the semantic interpretation of a free occurrence of a variable. Finally, there are occurrences of predicate symbols and also individual constants. In the case of the former and the latter, these are also interpretable as variables, though model-theoretically their interpretation is fixed relative to a particular structure in which formulas are evaluated. Nonetheless, their interpretation may vary from one structure to another. Finally first-order formulas themselves may be treated as variables. Admittedly, their interpretation may vary from one structure to another. In this dissertation we do not pay attention to proof-theory of first-order languages. In proof-theoretic approaches to first-order logic, the aforementioned symbols are not interpreted at all. Rather, they are either regarded as schematic expressions. As we shall see schematic expressions are semantically interpretable as another species of being a variable. Now the chief question is: how to give a semantic characterization of variables? Preliminary, one could say that the semantic value of variables is indeterminate. However, these characterization is open to interpretation. Indeterminacy should not be understood as

Este artigo tem como objetivo destacar a diferença fundamental entre dois princípios essenciais no logicismo de Frege. À primeira vista, o princípio da composicionalidade parece ter ocupado o lugar do princípio do contexto no projeto de Frege. No entanto, nossa intenção é demonstrar que esses dois princípios residem em domínios distintos da lógica e da filosofia, apesar
de estarem conectados a um mesmo projeto. Para atingir esse objetivo, inicialmente, realizaremos uma análise das obras "Die Grundlagen der Arithmetik" e "Über Sinn und Bedeutung". Em seguida, exploraremos a proposta de formalização do princípio da composicionalidade apresentada por Lehmann. Por fim, investigaremos se Frege, de fato, abandonou o princípio do contexto e discutiremos a conciliação entre esses dois  princípios.
Palavras-chave: Frege. Princípio do contexto. Princípio da composicionalidade.

This  paper  addresses  the  concept  of  semantic  compositionality  applied  to  complex  words  (especially  derived)  and  suggests  linguistic  indicators  to  classify  morphological  structures  in  terms  of  compositionality.  While  doing  so,  it  discusses  the  relationship  between  semantic  compositionality,  morphological  (formal)  transparency,  special  meaning,  and  lexicalization.  Although  semantic  compositionality  is  a  concept  increasingly evoked in syntax-based morphological analysis, there is no orderly proposal for  indicators  of  this  property  for  linguistic  expressions.  Based  on  data  from  Brazilian  Portuguese, and in order to guide the classification of empirical data on morphologically complex  words,  possible  indicators  obtained  through  linguistic  tests  (i.e.,  formal  transparency, formation of paraphrases and use of synonyms for roots and affixes, adverbial or nominal modification, maintenance and loss of (morpho)phonological properties) and psycholinguistic  events  (i.e,  word  association,  lexical  decision,  slips  of  the  tongue  and  facetious creations) are listed and discussed.

Аннотация. В статье представлен инструментарий для формализации синтаксических теорий в рамках минималистской программы Н. Хомского (Minimalist Program) в генеративной лингвистике. Он представляет собой набор формальных конструкций в языке Agda, позволяющих определить деревья вывода, синтаксические объекты и основные понятия минималистской теории-операцию соединения (Merge), признаки, копии, цепи, cкомандование и др. В то же время, в формализме опущены операция согласования (Agree) и теория фаз. Ключевые слова: минималистская программа, генеративная лингвистика, Agda.

A recent ''third wave'' of neural network (NN) approaches now delivers state-ofthe-art performance in many machine learning tasks, spanning speech recognition, computer vision, and natural language processing. Because these modern NNs often comprise multiple interconnected layers, work in this area is often referred to as deep learning. Recent years have witnessed an explosive growth of research into NN-based approaches to information retrieval (IR). A significant body of work has now been created. In this paper, Kezban Dilek Onal and Ye Zhang contributed equally. Maarten de Rijke and Matthew Lease contributed equally.

While the field of heritage language (HL) bilingualism has grown substantially in recent years, no studies have considered heritage Hebrew speakers dominant in American English. Expanding HL studies to new language pairs is crucial to understand the generalizability of prior findings across diverse linguistic contexts. In the current study, we assess 40 adult participants (16 M, 24 F) and present an overview of their lexical abilities, as derived from a quantitative and qualitative analysis of performance on the Multilingual Naming Test (MINT) and a narrative elicitation task. We consider target accuracy, types of non-target responses, and cross-linguistic influence in the form of code-switching and calquing. Participants’ non-target responses indicated a strong grasp of Hebrew root-pattern word formation and creativity in the face of lexical gaps. Code-switching and calquing patterns in the narratives suggest that the dominant English is the clear framing language, from which speak...

This paper presents and analyzes lexical and syntactic evidence from heritage Russian as spoken by bilinguals dominant in American English. The data come from the Russian Learner Corpus, a new resource of spoken and written materials produced by heritage re-learners and L2 learners of Russian. The paper focuses on lexical phrase violations, which we divide further into transfer-based structures and novel creations, showing that the latter are used by heritage speakers, but generally not freely available to L2 learners. In constructing innovative expressions, heritage speakers follow general principles of compositionality. As a result, novel constructions are more semantically transparent than their correlates in the baseline or dominant language. We argue that such semantically transparent, compositional patterns are based on structures that are universally available across languages. However, L2 speakers resort to these universal strategies for creating novel phrases much less ofte...

While the field of heritage language (HL) bilingualism has grown substantially in recent years, no studies have considered heritage Hebrew speakers dominant in American English. Expanding HL studies to new language pairs is crucial to understand the generalizability of prior findings across diverse linguistic contexts. In the current study, we assess 40 adult participants (16 M, 24 F) and present an overview of their lexical abilities, as derived from a quantitative and qualitative analysis of performance on the Multilingual Naming Test (MINT) and a narrative elicitation task. We consider target accuracy, types of non-target responses, and cross-linguistic influence in the form of code-switching and calquing. Participants’ non-target responses indicated a strong grasp of Hebrew root-pattern word formation and creativity in the face of lexical gaps. Code-switching and calquing patterns in the narratives suggest that the dominant English is the clear framing language, from which speak...

Several scholars from the latter school of thought have proposed that there was a two-word stage in the course of language evolution, in which utterances could not combine more than two words. These models agree that the putative two-word stage did not exhibit syntax. However, they disagree on whether or not there existed rules for inferring the semantic relationship between the two words expressing a compositional proposition. Focusing on semantically transitive events, I combine in the present paper language evolution models with previous empirical studies in linguistics to argue that the two-word stage was indeed governed by rules for inferring the compositional meaning of the utterance, in that (1) words were either associated with fixed (“predetermined”) semantic roles (i.e., agent, patient, predicate) or (2) there was a fixed order of semantic roles and the same words could be assigned different semantic roles in different utterances. Given the proposed existence of rules for ...

Central to modern semantics is Frege's distinction between force and sense. According to that distinction the content of an illocutionary act-a self-standing utterance of a sentence S-can be divided into two components. One is the proposition P that S's linguistic meaning and context associates with it. The other is S's illocutionary force. Illocutionary forces are usually analysed as communicative intentions with which S is uttered. If the force is assertoric, then the speaker intends that her audience believe P, or some such condition. If the force is commissive-that characteristic of orders and requests-the speaker intends that the audience see to it that it is true. And so on. The force/sense distinction-designate it Force/Sense-is associated with another thesis. This is the truth-conditional embedding principle: TC-Embedding: If a sentence S is embedded in a compound sentence (…S…), then the speech act performed with S is a propositional act: an act that involves uttering S with a propositional (true/false-assessable) content and no other content. Force/Sense implies TC-Embedding, given plausible background assumptions and the fact that forces don't seem to embed. Despite their venerated status, we dispute both Force/Sense and TC-Embedding. We think linguistic phenomena across the board show both are wrong. However, in this paper, we restrict ourselves to figurative speech and irony in particular. We argue that irony embeds in compound sentences. This fact seriously challenges TC-Embedding and thus Force/Sense. We examine a range of responses, but the overall picture is a negative one. Theorists need to go back to the drawing board about the nature of illocutionary acts. 1. Consider irony. Suppose Max has a very low opinion of George's talents.

This paper presents and analyzes lexical and syntactic evidence from heritage Russian as spoken by bilinguals dominant in American English. The data come from the Russian Learner Corpus, a new resource of spoken and written materials produced by heritage re-learners and L2 learners of Russian. The paper focuses on lexical phrase violations, which we divide further into transfer-based structures and novel creations, showing that the latter are used by heritage speakers, but generally not freely available to L2 learners. In constructing innovative expressions, heritage speakers follow general principles of compositionality. As a result, novel constructions are more semantically transparent than their correlates in the baseline or dominant language. We argue that such semantically transparent, compositional patterns are based on structures that are universally available across languages. However, L2 speakers resort to these universal strategies for creating novel phrases much less ofte...

This paper presents and analyzes lexical and syntactic evidence from heritage Russian as spoken by bilinguals dominant in American English. The data come from the Russian Learner Corpus, a new resource of spoken and written materials produced by heritage re-learners and L2 learners of Russian. The paper focuses on lexical phrase violations, which we divide further into transfer-based structures and novel creations, showing that the latter are used by heritage speakers, but generally not freely available to L2 learners. In constructing innovative expressions, heritage speakers follow general principles of compositionality. As a result, novel constructions are more semantically transparent than their correlates in the baseline or dominant language. We argue that such semantically transparent, compositional patterns are based on structures that are universally available across languages. However, L2 speakers resort to these universal strategies for creating novel phrases much less ofte...

Uncertain information is ubiquitous in the Semantic Web, due to methods used for collecting data and to the inherently distributed nature of the data sources. It is thus very important to develop probabilistic Description Logics (DLs) so that the uncertainty is directly represented and managed at the language level. The DISPONTE semantics for probabilistic DLs applies the distribution semantics of probabilistic logic programming to DLs. In DISPONTE, axioms are labeled with numeric parameters representing their probability. These are often difficult to specify or to tune for a human. On the other hand, data is usually available that can be leveraged for setting the parameters. In this paper, we present EDGE that learns the parameters of DLs following the DISPONTE semantics. EDGE is an EM algorithm in which the required expectations are computed directly on the binary decision diagrams that are built for inference. Experiments on two datasets show that EDGE achieves higher areas under...

This paper presents and analyzes lexical and syntactic evidence from heritage Russian as spoken by bilinguals dominant in American English. The data come from the Russian Learner Corpus, a new resource of spoken and written materials produced by heritage re-learners and L2 learners of Russian. The paper focuses on lexical deviations from baseline Russian at a single-and multi-word level, which we divide further into transfer-based structures and novel creations, showing that the latter are used by heritage speakers, but generally not freely available to L2 learners. In constructing innovative expressions, heritage speakers follow general principles of compositionality. As a result, such innovative expressions are more semantically transparent than their correlates in the baseline or dominant language. We contend that semantically transparent, compositional patterns are based on structures that are universally available across languages. However, L2 speakers resort to these universal strategies for creating novel phrases much less often than heritage speakers. In their linguistic creativity, heritage speakers' utterances parallel those of L1 child learners rather than L2 speakers.

Резюме: Задача статьи - дать новое определение явления семантической композициональности, которое раскрывало бы его коммуникативную и когнитивную стороны. Сначала рассматривается так называемое «математическое» определение композициональности Ходжеса, обсуждается вопрос о его «пустоте». В качестве средства преодоления этой пустоты вводится процедурная трактовка значения, в рамках которой значение становится абстрактным алгоритмом (теория Московакиса). Однако этого не достаточно, поскольку возникает интенсиональная логика с бесконечной интенсиональной иерархией, что никак не соответствует лингвистической практике. Я вижу выход в «приземлении» этой иерархии в ситуации произнесения с по-мощью использования её параметров, а также процедуры их вербализации. В первом случае значение ока-зывается локальным, во втором-глобальным, при этом последнее значение относится к выражению, полученному как ситуационное расширение исходного. Далее я анализирую трансформации ситуаций произнесения в результате речевых и неречевых действий агентов, и, следуя предложению Ротшильда и Ялджина, ставлю на первое место понятие «истинности дискурса», понимая его как успешность коммуникации. Это позволяет определить понятия локальной композициональности и некомпозициональности. В силу локальности значения для каждой ситуации выделяются фокус и периферия. Вклад в значение элементов периферии минимален и не влияет на успешность коммуникации, что делает эти элементы локально некомпози-циональными. Связь между локальной некомпозициональностью и контекстной зависимостью состоит в том, что вербализация и умолчание значения контекстных параметров создают или устраняют некомпозициональные компоненты выражений. Осуществление такого рода преобразований есть условие распознавание рациональным агентом композициональности и некомпозициональности. Мой общий вывод состоит в том, что композициональность и некомпозициональность всегда присутствуют одновременно, поскольку ситуации произнесения в общем случае предполагают дифференциацию фокуса и периферии, граница которых при этом подвижна. Тотальная композициональность есть экстремальный случай, в котором такой дифференциации не происходит. Ключевые слова: композициональность, значение, контекст, семантический аглоритм, фокус коммуникации, локальное значение.

Abstract: The task of the article is to generate a new definition of the phenomenon of semantic compositionality that would reveal its communicative and cognitive aspects. First, the so-called "mathematical" definition of Hodges is considered, and the question of its "vacuity" is discussed. As a means of overcoming this vacuity, a procedural in-terpretation of meaning is introduced, within which the meaning becomes an abstract algorithm (the theory of Mo-schovakis). However, this is not enough, since intensional logic arises with an infinite intensional hierarchy, which in no way corresponds to linguistic practice. I see the way out in the "landing" of this hierarchy in the situation of speaking by using its parameters, as well as the procedure for their verbalization. In the first case, the meaning is local, in the second – global, while the latter meaning relates to the expression obtained as a situational extension of the original. Next, I analyze the transformation of the situations as a result of speech acts and of non-verbal acts of agents, and following the proposal of Rotshild and Yalcin, I put first the notion of "truth of discourse", understand-ing it as the success of communication. This allows me to define the concepts of local compositionality and non-compositionality. Because of the locality of the meaning, focus and periphery are highlighted for each situation. The contribution to the meaning of peripheral elements is minimal and does not affect the success of communication, which makes these elements locally non-compositional. The relationship between local non-compositionality and contextual dependency is that the verbalization and defaulting of the meaning of contextual parameters create or eliminate non-compositional components of expressions. Realization of such transformations is a condition of rec-ognition of compositionality and non-compositionality. My general conclusion is that compositionality and non-compositionality are always present simultaneously, since the situations of speaking in the general case assume dif-ferentiation of focus and periphery, the boundary of which is mobile at the same time. Total compositionality is an extreme case in which there is no such differentiation.

We propose a nonmonotonic Description Logic of typicality able to account for the phenomenon of combining prototypical concepts, an open problem in the fields of AI and cognitive modelling. Our logic extends the logic of typicality ALC + T R , based on the notion of rational closure, by inclusions p :: T(C) D (" we have probability p that typical Cs are Ds "), coming from the distributed semantics of probabilistic Description Logics. Additionally, it embeds a set of cognitive heuristics for concept combination. We show that the complexity of reasoning in our logic is EXPTIME-complete as in ALC.

It's no secret that different of the subfields in cognitive science dispute what the correct solution is to various problems that they each investigate in their separate ways. Sometimes this is due to differing antecedent ideas about what is the appropriate way to investigate the phenomenon, other times it is due to differing antecedent ideas about what principles an adequate solution should embody, and still other times it is due to differing antecedent ideas concerning what the dispute is about.. . as for example when they use the same terminology for different phenomena. This paper is an investigation into these differing antecedent ideas in the realm of meaning and compositionality as they play out in linguistics, cognitive psychology , and philosophy of language. The focus is on the notions of subjective meaning and objective meaning, and a main conclusion is that there needs be a " two-factor semantic theory " to accommodate the overall goals of both sides. Some steps are made towards that end, but various previous attempts are argued to have missed the point. In the end it will be shown that, clearly, more work needs to be done along a number of specified dimensions, especially on the subjective side of the dispute.

During the last decades, many cognitive architectures (CAs) have been realized adopting different assumptions about the organization and the representation of their knowledge level. Some of them (e.g. SOAR [35]) adopt a classical symbolic approach, some (e.g. LEABRA[48]) are based on a purely con-nectionist model, while others (e.g. CLARION [59]) adopt a hybrid approach combining connectionist and symbolic representational levels. Additionally, some attempts (e.g. biSOAR) trying to extend the representational capacities of CAs by integrating diagrammatical representations and reasoning are also available [34]. In this paper we propose a reflection on the role that Conceptual Spaces, a framework developed by Peter Gärdenfors [24] more than fifteen years ago, can play in the current development of the Knowledge Level in Cognitive Systems and Architectures. In particular, we claim that Conceptual Spaces offer a lingua franca that allows to unify and generalize many aspects of the symbolic, sub-symbolic and diagrammatic approaches (by overcoming some of their typical problems) and to integrate them on a common ground. In doing so we extend and detail some of the arguments explored by Gärdenfors [23] for defending the need of a conceptual, intermediate, representation level between the symbolic and the sub-symbolic one. In particular we focus on the advantages offered by Conceptual Spaces (w.r.t. symbolic and sub-symbolic approaches) in dealing with the problem of compositionality of representations based on typicality traits. Additionally, we argue that Conceptual Spaces could offer a unifying framework for interpreting many kinds of diagrammatic and analogical representations. As a consequence, their adoption could also favor the integration of diagrammatical representation and reasoning in CAs.

This paper examines a class of indefinite pronouns in Bengali (“k-words”) with two aims—first to bring in empirical, descriptive semantic data not previously examined, and secondly to motivate a two-dimensional analysis of polarity-sensitivity that depends both on structural licensing and pragmatic preconditions, which in turn are sensitive to both a domain of alternatives and the extent to which that domain is exhaustified. In Bengali, a k-word like kono (“some”/“any”) is morphologically composed of an interrogative (e.g. kon) along with the particle -o (Chatterji 1926; Thompson 2010).  The k-word—with Negative Polarity (NPI) and Free Choice (FCI) readings—has been analysed as indetermintes triggering alternatives (Ramchand 1997) similar to Japanese (Kratzer and Shimoyama 2002). In previous work (Ullah 2005) I argue that Bengali k-words divide into emphatics and non-emphatics and that emphatics are subject to licensing in anti-veridical contexts (cf. Giannakidou 1997, 1998). In this paper I identify an intonational diagnostic that teases apart emphatics from non-emphatics and then show that non-emphatics obtain a referentially vague (RV) reading of the type called “epistemic indefinites” e.g. algún in Spanish (Alonso-Ovalle 2006). Next, examining interpretations in the scope of modals, I show non-emphatic k-words systematically contrast with, on one side, non-polarity-sensitive indefinites and, on the other side, FCIs. Giannakidou and Quer (2013) argue that certain FCI and RV indefinites in Spanish, Catalan, and Greek are best explained as requiring preconditions that trigger individual-alternatives but differ critically in that FC requires exhaustification of the domain while the RV requires only partial exhaustification. My paper argues that Giannakidou and Quer-type analysis better explains the Bengali k-words in all three readings—NPI, FCI, and RV. This approach supports converging work that treats polarity not as a monolithic class but as a cluster of phenomena emerging from the mix of grammar, semantics, and pragmatics.

ABSTRACT. The Principle of Semantic Compositionality (sometimes
called 'Frege's Principle') is the principle that the meaning of
a (syntactically complex) whole is a function only of the meanings
of its (syntactic) parts together with the manner in which these parts
were combined. This principle has been extremely influential
throughout the history of formal semantics; it has had a tremendous
impact upon modern linguistics ever since Montague Grammars
became known; and it has more recently shown up as a guiding
principle for a certain direction in cognitive science.
Despite the fact that The Principle is vague or underspecified at
a number of points - such as what meaning is, what counts as a part,
what counts as a syntactic complex, what counts as combination -
this has not stopped some people from viewing The Principle as
obviously true, true almost by definition. And it has not stopped other
people from viewing The Principle as false, almost pernicious in its
effect. And some of these latter theorists think that it is an empirically
false principle while others think of it as a methodologically
wrong-headed way to proceed.
In fact, there are approximately 318 arguments against The
Principle which can be found in the literature, whereas there are only
three (or maybe four) arguments proposed in favor of The Principle.
This paper will adjudicate among these arguments. And at the end
it will suggest some other way to look at what proponents of
compositionality really want.

Frege’s distinction between force and sense is a central pillar of modern thinking about meaning. This is the idea that a self-standing utterance of a sentence S can be divided into two components. One is the proposition P that S’s linguistic meaning and context associates with it. The other is S’s illocutionary force. The force/sense distinction is associated with another thesis, the embedding principle, that implies that the only content that embeds in compound sentences is propositional content. We argue that both the Force/Sense distinction and the principle of embedding are seriously challenged by figurative language, and irony in particular. We conclude that theorists need to go back to the drawing board about the nature of illocutionary acts.