Compositionality in a Parallel Architecture for Language Processing

Compositionality has been a central concept in linguistics and philosophy for decades, and it is increasingly prominent in many other areas of cognitive science. Its status, however, remains contentious. Here, I reassess the nature and scope of the principle of compositionality (Partee, 1995) from the perspective of psycholinguistics and cognitive neuroscience. First, I review classic arguments for compositionality and conclude that they fail to establish compositionality as a property of human language. Next, I state a new competence argument, acknowledging the fact that any competent user of a language L can assign to most expressions in L at least one meaning which is a function only of the meanings of the expression's parts and of its syntactic structure. I then discuss selected results from cognitive neuroscience, indicating that the human brain possesses the processing capacities presupposed by the competence argument. Finally, I outline a language processing architecture consistent with the neuroscience results, where semantic representations may be generated by a syntax-driven stream and by an “asyntactic” processing stream, jointly or independently. Compositionality is viewed as a constraint on computation in the former stream only.     

What Compositionality Still Can Do

Proponents of deflationism about meaning often claim that the principle of compositionality, when properly understood, places no constraint whatsoever on the nature of lexical meaning. This deflationary thesis admits of both strong and weak readings. On the strong reading, the principle does not rule out any theory of lexical meaning either alone or in conjunction with other independently plausible semantic assumptions. On the weak reading, the principle alone does not rule out any such theory. I argue that, though weak deflationism about compositionality has better initial prospects than strong, neither version of the thesis is credible. In particular, weak deflationism cannot be maintained in conformity with the deflationist commitment to explaining facts about phrase meanings in compositional terms, that is, by appeal to facts about the lexicon.     

组合原则

组合原则是逻辑语义学的基本原则,表现为函项的思想和句法与语义的对应。就构造逻辑系统而言,组合原则是一种方法论。作为延伸到自然语言形式语义学的产物,组合原则起到一种核心灵魂的作用。组合原则在理论层面和应用领域皆获得一些有价值的结果。自然语言违反组合原则的实例主要表现为歧义现象、语用因素和句法与语义不对应的情况,这正是调整组合原则的适用条件从而促进形式语义学发展的契机。     

Processing idiomatic expressions: effects of semantic compositionality

Three experiments tested the main claims of the idiom decomposition hypothesis: People have clear intuitions on the semantic compositionality of idiomatic expressions, which determines the syntactic behavior of these expressions and how they are recognized. Experiment 1 showed that intuitions are clear only for a very restricted number of expressions, but for the majority of idioms, they are not consistent across speakers. Experiment 2 failed to support the claim that semantic compositionality influences the syntactic flexibility of idioms. Finally, Experiment 3 showed that idioms are more quickly recognized than their literal counterparts, regardless of compositionality and syntactic flexibility. All of the findings were at odds with the tenets of the idiom decomposition hypothesis. The theoretical implications of the results with respect to idiom processing and the notion of compositionality are discussed.     

The Myth of Reverse Compositionality

In the context of debates about what form a theory of meaning should take, it is sometimes claimed that one cannot understand an intersective modifier-head construction (e.g., ‘pet fish’) without understanding its lexical parts. Neo-Russellians like Fodor and Lepore contend that non-denotationalist theories of meaning, such as prototype theory and theory theory, cannot explain why this is so, because they cannot provide for the ‘reverse compositional’ character of meaning. I argue that reverse compositionality is a red herring in these debates. I begin by setting out some positive arguments for reverse compositionality and showing that they fail. Then I show that the principle of reverse compositionality has two big strikes against it. First, it is incompatible with all theories of meaning on the market, including the denotationalism favored by neo-Russellians. Second, it explains nothing that is not already explained by its venerable predecessor, the principle of (forward) compositionality.     

Why Compositionality Won't Go Away: Reflections on Horwich's 'Deflationary' Theory

Paul Horwich argues for a 'deflationary'account of compositionality, according to which, '. . . the compositionality of meaning imposes no constraint at all on how the meaning properties of words are constituted'. We have arrived at a tentative diagnosis, which is that Horwich fails to enforce several distinctions that turn out to be crucial. For example, sometimes he puts his main conclusion in the way we just quoted but sometimes, even on the following page, he puts it like this: 'understanding one of oneÕs own complex expressions (non-idiomatically) is, by definition, nothing over and above understanding its parts and knowing how they are combined'. We propose, in what follows, to consider how Horwich's deflationary account of compositionality fares if the distinction between theories of meaning and theories of understanding is properly attended to. Here's how we think it all turns out:
– Horwich is right to claim that compositionality is neutral with respect to the metaphysics of understanding expressions when 'understanding' refers to a (merely) dispositional state; but not when it refers to an occurrent state.
– Horwich is right strictu dictu to claim that compositionality is neutral with respect to the character of lexical meanings, but only strictu dictu.
– Compositionality taken, together with other constraints that semantic theories are required to satisfy, reduces the options in the theory of lexical meaning to a bare minimum.
The first part of the paper is about compositionality and understanding, the second part is about compositionality and lexical meaning.     

Communication and Strong Compositionality

Ordinary semantic compositionality (meaning of whole determined from meanings of parts plus composition) can serve to explain how a hearer manages to assign an appropriate meaning to a new sentence. But it does not serve to explain how the speaker manages to find an appropriate sentence for expressing a new thought. For this we would need a principle of inverse compositionality, by which the expression of a complex content is determined by the expressions of it parts and the mode of composition. But this presupposes that contents have constituent structure, and this cannot be taken for granted. However, it can be proved that if a certain principle of substitutivity is valid for a particular language, then the meanings expressed by its sentences can justifiably be treated as structured. In its simplest form, this principle says that if in a complex expression a constituent is replaced by another constituent with a different meaning, the new complexhas a meaning different from the original. This principle is again inversely related to the normal compositional principle of substitutivity. The combination of ordinary and inverse compositionality is here called strong compositionality. The proof is carried out in the algebraic framework developed by Wilfrid Hodges and Dag Westerståhl.     

Historical Semantic Chaining and Efficient Communication: The Case of Container Names

Semantic categories in the world's languages often reflect a historical process of chaining: A name for one referent is extended to a conceptually related referent, and from there on to other referents, producing a chain of exemplars that all bear the same name. The beginning and end points of such a chain might in principle be rather dissimilar. There is also evidence supporting a contrasting picture: Languages tend to support efficient, informative communication, often through semantic categories in which all exemplars are similar. Here, we explore this tension through computational analyses of existing cross‐language naming and sorting data from the domain of household containers. We find (a) formal evidence for historical semantic chaining, and (b) evidence that systems of categories in this domain nonetheless support near‐optimally efficient communication. Our results demonstrate that semantic chaining is compatible with efficient communication, and they suggest that chaining may be constrained by the functional need for efficient communication.     

Supervaluation-Style Truth Without Supervaluations

Kripke’s theory of truth is arguably the most influential approach to self-referential truth and the semantic paradoxes. The use of a partial evaluation scheme is crucial to the theory and the most prominent schemes that are adopted are the strong Kleene and the supervaluation scheme. The strong Kleene scheme is attractive because it ensures the compositionality of the notion of truth. But under the strong Kleene scheme classical tautologies do not, in general, turn out to be true and, as a consequence, classical reasoning is no longer admissible once the notion of truth is involved. The supervaluation scheme adheres to classical reasoning but violates compositionality. Moreover, it turns Kripke’s theory into a rather complicated affair: to check whether a sentence is true we have to look at all admissible precisification of the interpretation of the truth predicate we are presented with. One consequence of this complicated evaluation condition is that under the supervaluation scheme a more proof-theoretic characterization of Kripke’s theory becomes inherently difficult, if not impossible. In this paper we explore the middle ground between the strong Kleene and the supervaluation scheme and provide an evaluation scheme that adheres to classical reasoning but retains many of the attractive features of the strong Kleene scheme. We supplement our semantic investigation with a novel axiomatic theory of truth that matches the semantic theory we have put forth.     

Recurrence

Standard compositionality is the doctrine that the semantic content of a compound expression is a function of the semantic contents of the contentful component expressions. In 1954 Hilary Putnam proposed that standard compositionality be replaced by a stricter version according to which even sentences that are synonymously isomorphic (in the sense of Alonzo Church) are not strictly synonymous unless they have the same logical form. On Putnam??s proposal, the semantic content of a compound expression is a function of: (i) the contentful component expressions; and (ii) the expression??s logical form. Kit Fine recently expanded and modified Putnam??s idea into a sweeping theory in philosophy of language and philosophy of mind. The present paper is a detailed critique of Fine??s ??semantic relationism.?? Fine??s notion of coordination is explained in terms of the familiar pragmatic phenomenon of recognition. A serious error in Fine??s formal disproof of standard Millianism is exposed. It is demonstrated furthermore that Church??s original criticism of Putnam??s proposal can be extended to Fine??s semantic relationism. Finally, it is also demonstrated that the positive position Fine proffers to supplant standard Millianism is in fact exactly equivalent to standard Millianism, so that Fine??s overall position not only does not displace standard Millianism but is in fact inconsistent.     

A study of metrics of distance and correlation between ranked lists for compositionality detection

Compositionality in language refers to how much the meaning of some phrase can be decomposed into the meaning of its constituents and the way these constituents are combined. Based on the premise that substitution by synonyms is meaning-preserving, compositionality can be approximated as the semantic similarity between a phrase and a version of that phrase where words have been replaced by their synonyms. Different ways of representing such phrases exist (e.g., vectors (Kiela and Clark, 2013) or language models (Lioma, Simonsen, Larsen, and Hansen, 2015)), and the choice of representation affects the measurement of semantic similarity.We propose a new compositionality detection method that represents phrases as ranked lists of term weights. Our method approximates the semantic similarity between two ranked list representations using a range of well-known distance and correlation metrics. In contrast to most state-of-the-art approaches in compositionality detection, our method is completely unsupervised. Experiments with a publicly available dataset of 1048 human-annotated phrases shows that, compared to strong supervised baselines, our approach provides superior measurement of compositionality using any of the distance and correlation metrics considered.     

Compositionality and sandbag semantics

It is a common view that radical contextualism about linguistic meaning is incompatible with a compositional explanation of linguistic comprehension. Recently, some philosophers of language have proposed theories of ‘pragmatic’ compositionality challenging this assumption. This paper takes a close look at a prominent proposal of this kind due to François Recanati. The objective is to give a plausible formulation of the view. The major results are threefold. First, a basic distinction that contextualists make between mandatory and optional pragmatic processes needs to be revised. Second, the pragmatic theory can with stand a Davidsonian objection only by rejecting the importance of a distinction between primitive and non-primitive semantic items. Thirdly, however, the theory is now open to a worry about how it should be understood: either the theory consists in a very broad functionalist generalization about communication, which makes it explanatorily inert, or it boils down to a highly particularist view about linguistic meaning. Finally, I argue that Recanati’s notion of ‘occasion meaning’ is problematic and suggest replacing it with the notion of speaker meaning, which is explanatorily more basic.     

Semantic Compositionality and Truth-Conditional Content

It is widely held that hearers grasp an utterance's truth conditions by assigning contents to the linguistic expressions used, and combining these contents according to semantic composition rules. To preserve compositionality of truth-conditional content while accounting for context-sensitivity that is not traceable to overt linguistic form, semanticists posit covert linguistic structure. The strongest justification for this approach is the allegedly unconstrained nature of the alternative, whereby a process of 'free pragmatic enrichment' supplies constituents of content that are not traceable to (overt or covert) encoded meaning. This paper argues that free enrichment is tightly constrained by purely pragmatic factors, thus undermining the motivation for semantic compositionality.     

Against the identification of assertoric content with compositional value

This essay investigates whether the things we say are identical to the things our sentences mean. It is argued that these theoretical notions should be distinguished, since assertoric content does not respect the compositionality principle. As a paradigmatic example, Kaplan??s formal language LD is shown to exemplify a failure of compositionality. It is demonstrated that by respecting the theoretical distinction between the objects of assertion and compositional values certain conflicts between compositionality and contextualism are avoided. This includes the conflict between eternalism and the semantics of tense, the embedding problems for contextualism about epistemic modals and taste claims, and the conflict between direct reference and the semantics of bound pronouns (and monstrous operators). After presenting the theoretical picture which distinguishes assertoric content from compositional semantic value, some objections to the picture are addressed. In so doing, the objection from King (Philos Perspect 17(1):195?C246, 2003) stemming from apparent complications with the interaction of temporal expressions and attitude reports is assessed and shown to be non-threatening.     

Representante-representativo,représentant-représentation,ideational representative: Which one is a Freudian concept? On the translation of Vorstellungsrepräsentanz in Spanish,French and English1

The Freudian expression Vorstellungsrepräsentanz ( Freud, 1915b, 1915c ), which is rendered in the Standard Edition as ideational representative, is commonly translated in Spanish as representante-representativo and in French as représentant-représentation, among other renderings. An interdisciplinary conceptual inquiry, which applies linguistic semantics to the evaluation of the available Spanish and French renderings, concludes that this compound expression should be translated in these languages as representante ideativo and représentant idéatif, respectively, renderings which happen to correspond to Strachey’s translation into English in the SE. In contrast to most Spanish and French translations, this proposal conforms to the semantic principle of compositionality. On the one hand, it provides a suitable translation of the two parts of the compound. Thus it renders Vorstellung as idea, with the classical meaning of image or mental representation, which can be traced back to Hume’s empiricist philosophy, and it renders Repräsentanz as representative, with the meaning of delegate. On the other hand, its linguistic form preserves the attributive meaning relationship which exists between both concepts in the original German expression. Against the background of these semantic considerations, a theoretical question concerning Freudian metapsychology is discussed: the drive has a psychic representative, but is there a (mental) representation of the drive?     

The evidence for relativism

The aim of this paper is to examine the kind of evidence that might be adduced in support of relativist semantics of a kind that have recently been proposed for predicates of personal taste, for epistemic modals, for knowledge attributions and for other cases. I shall concentrate on the case of taste predicates, but what I have to say is easily transposed to the other cases just mentioned. I shall begin by considering in general the question of what kind of evidence can be offered in favour of some semantic theory or framework of semantic theorizing. In other words, I shall begin with the difficult question of the empirical significance of semantic theorizing. In Sect. 2, I outline a relativist semantic theory, and in Sect. 3, I review four types of evidence that might be offered in favour of a relativistic framework. I show that the evidence is not conclusive because a sophisticated form of contextualism (or indexical relativism) can stand up to the evidence. However, the evidence can be taken to support the view that either relativism or the sophisticated form of contextualism is correct.     

What differentiates declarative and procedural memories: reply to Cohen, Poldrack, and Eichenbaum (1997)

CPE claim that procedural and declarative representations differ on two important dimensions: flexibility and compositionality. I have proposed that the apparent flexibility of a memory depends entirely on the transfer conditions. Any retest is, in some sense, a test of flexibility, because something has changed since the original encoding episodic. I have argued that if one changes something that does not provide support to memory performance, the memory will appear flexible, and resistant to changes in the environment. If one changes the very thing that the representation codes, the memory will appear inflexible and easily disrupted by changes in the environment. This principle is equally true for procedural and declarative memory. CPE contend that procedural representations lack compositionality. An ideal test of this claim would examine the representation of a task that is widely agreed to be procedural (e.g. that has been demonstrated to be learned normally by amnesic patients, and in the absence of awareness by neurologically intact subjects). Such experiments appear not to have been conducted, and the fact is that many tasks that are widely agreed to be procedural probably are not compositional. They appear to be, as CPE contend, biases in a processing system; it is hard to imagine how repetition priming could be compositional. Nevertheless, this is not true of all procedural memories. There is a good deal of evidence that motor behaviour is organised hierarchically and has compositionality. There is every reason to think that most if not all motor behaviour is procedural; motor behaviour might be driven by goals that are declarative, but the low-level operations that actually manipulate effectors are closed to consciousness, do not depend on the medial temporal lobe or diencephalon, and would therefore be classified as procedural. CPE framed their theory of differences between procedural and declarative memory systems as an account of the deficit in amnesic patients. They therefore predict that the learning of amnesic patients should not show flexibility or compositionality. There is already at least one study showing learning in amnesic patients that is as flexible as that of control participants (Knowlton & Squire, 1996). There are not, to my knowledge, data on whether the motor skill learning of amnesic patients shows compositionality, but one might expect that it would, given that it does in neurologically intact participants, and given that motor skill learning appears unimpaired in amnesic patients. Thus, the conception of declarative and procedural memory provided by CPE may not provide a complete account of amnesic performance. The anatomic distinction between procedural and declarative memory systems appears quite strong, and there is therefore reason to believe that there are accompanying computational differences. There does not, however, appear to be sufficient evidence to support those differences proposed by CPE.     

Truth in the Tractatus

My paper takes issue both with the standard view that the Tractatus contains a correspondence theory and with recent suggestions that it features a deflationary or semantic theory. Standard correspondence interpretations are mistaken, because they treat the isomorphism between a sentence and what it depicts as a sufficient condition of truth rather than of sense. The semantic/deflationary interpretation ignores passages that suggest some kind of correspondence theory. The official theory of truth in the Tractatus is an obtainment theory – a sentence is true iff the state of affairs it depicts obtains. This theory differs from deflationary theories in that it involves an ontology of states of affairs/facts; and it can be transformed into a type of correspondence theory: a sentence is true iff it corresponds to, i.e. depicts an obtaining state of affairs (fact). Admittedly, unlike correspondence theories as commonly portrayed, this account does not involve a genuinely truth-making relation. It features a relation of correspondence, yet it is that of depicting, between a meaningful sentence and its sense – a possible state of affairs. What makes for truth is not that relation, but the obtaining of the depicted state of affairs. This does not disqualify the Tractatus from holding a correspondence theory, however, since the correspondence theories of Moore and Russell are committed to a similar position. Alternatively, the obtainment theory can be seen as a synthesis of correspondence, semantic and deflationary approaches. It does justice to the idea that what is true depends solely on what is the case, and it combines a semantic explanation of the relation between a sentence and what it says with a deflationary account of the agreement between what the sentence says and what obtains or is the case if it is true     

Why Metaphors have no Meaning: Considering Metaphoric Meaning in Davidson

Abstract

Since the publication of Donald Davidson’s essay “What Metaphors Mean” (1978) – in which he famously asserts that metaphor has no meaning – the views expressed in it have mostly met with criticism: prominently from Mary Hesse and Max Black. This article attempts to explain Davidson’s surprise-move regarding metaphor by relating it to elements in the rest of his work in semantics, such as the principle of compositionality, radical interpretation and the principle of charity. I conclude that Davidson’s views on metaphor are not only consistent with his semantic theory generally, but that his semantics also depend on these insights. Eventually, the debate regarding Davidson’s views on metaphor should be conducted on the level of his views on the nature of semantics, the relationship between language and the world and the possibility that there is something like conceptual schemes.

Sedert die publikasie van Donald Davidson se opstel What Metaphors Mean (1978) – waarin hy die berugte stelling maak dat metafoor geen betekenis het nie – is sy sieninge meestal begroet met kritiek, ook van prominente figure soos Mary Hesse en Max Black. Hierdie artikel poog om’n verduideliking te vind vir Davidson se verrassende skuif aangaande metafoor, deur sy sieninge hieroor te kontekstualiseer teen die agtergrond van elemente uit die res van sy werk in semantiek, soos die beginsel van komposisionaliteit, radikale interpretasie en die beginsel van rasionele akkomodasie (charity). Ek kom tot die gevolgtrekking dat Davidson se sieninge aangaande metafoor nie slegs naatloos aansluit by sy algemene sieninge aangaande semantiek nie, maar dat die res van sy semantiese teorie ook athang van sy sieninge aangaande metafoor. Uiteindelik behoort die debat rakende Davidson se sieninge aangaande metafoor gevoer te word op die vlak van die aard van semantiek, die verhouding tussen taal en die werldikheid en die moontlike bestaan van konseptuele skemas.