Underspecification theory

Figurative polysemy, in which a word's original meaning is extended into domains for which it did not originally apply, is a pervasive property of human language. Previous research using cross-linguistic (English/Spanish) data has shown that an important set of similarities and differences in patterns of figurative verb polysemy between counterpart verbs in two languages can be explained by detailed analysis of the event structures of each verb (McNally and Spalek, 2022). Here we treat a complementary case involving counterpart verbs in two languages which share the same basic event structures but differ in details of conceptual (or 'root') content, most importantly in the semantic restrictions on their participants. We show how, in this sort of case, the verbs will describe the same types of situations and be amenable to similar figurative extensions as long as their respective semantic restrictions can be met, but not otherwise. Teasing apart contrasts due to variation in event structural vs. conceptual content can shed light both on debates about the relation between root and grammatical content, as well as on the challenges of establishing lexical equivalences for verbs, be it for purposes of translation or for cross-linguistic comparison and the creation of lexical resources.

This paper analyzes a case of Italian word-formation, in which the semantics of the derived words appears to contain mutually exclusive ambiguities. Italian productively derives verbs of removal from nouns. These verbs have the general semantic form A removes X from Y. However, there are two subtypes that differ in whether the nominal base of the derived verb is taken to be the FIGURE or the GROUND: scremare 'to skim'; (crema 'cream') is a FIGURE-verb, and stanare 'to make come out of the burrow' (tana 'burrow') is a GROUND-verb. Traditional approaches are at a loss to give a uniform account for the semantics of derivational processes of these two kinds. In this paper, a uniform analysis is proposed. It is based upon the model of lexical semantics known as two-level semantics. Two-level semantics makes a distinction between a layer of meaning which is defined by grammar, and a level of interpretation, which is based upon conceptual knowledge. We propose that the derivation of Italian denominal verbs of removal of both types starts from a single underspecified representation, which is then specified at the conceptual level; depending on the concept type of the base, the denominal verb is either a FIGURE-verb or a GROUND-verb. In general, this paper argues that the two-level approach to semantics can be fruitfully improved by combining it with the notion of underspecification. And on a more general level, it is an example of how language specific semantics may be embedded in cognitive structure. (2) GROUND-verbs stanare 'to make come out of the burrow'; cf. tana 'burrow' snidare 'to drive out' cf. nido 'nest' scarcerare 'to release (from prison)' cf. carcere 'prison' scassare 'to remove from a box' cf. cassa 'box; case' sganciare 'to unhook; to unfasten' cf. gancio 'hook' sbarcare 'to unload; to disembark' cf. barca 'boat' Both types of DVRs are causative verbs of removal of an object X from another object Y. We call the movable object FIGURE, and the unmovable one GROUND. Thus, DVRs can be assigned the general semantic structure A removes figure X from ground Y. The two subtypes differ in the way the two locative slots FIGURE and GROUND are filled by the nominal base. In FIGURE-verbs, the nominal base fills the FIGURE-slot (cf. scremare 'to remove the cream from something') and in GROUND-verbs, it fills the GROUND-slot (cf. stanare 'to remove something from the burrow'). The question is whether we can account for both types of DVRs with one derivational process, or whether we have to assume two processes. In this paper, we argue that one underspecified derivational process can account for the two subtypes. The difference between the subtypes is only spelled out at the conceptual level. This is confirmed by the observation that the concept type of the base determines to which class a DVR belongs. The paper is organized as follows: In the first part (section 1), we discuss some general assumptions which underlie this study. We give a descriptive account of Italian DVRs (section 2) and propose a rule which generates them with an underspecified semantics (sections 3 and 4). We then discuss the problem of how this underspecification is resolved (section 5) and how nouns must be represented in the lexicon in order to make the resolution possible (section 6).

Recent research has shown that toddlers' lexical representations are phonologically detailed, quantitatively much like those of adults. Studies in this article explore whether toddlers' and adults' lexical representations are qualitatively similar. Psycholinguistic claims (Lahiri & Marslen-Wilson, 1991; Lahiri & Reetz, 2002, 2010) based on underspecification (Kiparsky, 1982 et seq.) predict asymmetrical judgments in lexical processing tasks; these have been supported in some psycholinguistic research showing that participants are more sensitive to noncoronal-to-coronal (pop → top) than to coronal-to-noncoronal (top → pop) changes or mispronunciations. Three experiments using on-line visual world procedures showed that 19-month-olds and adults displayed sensitivities to both noncoronal-to-coronal and coronal-to-noncoronal mispronunciations of familiar words. No hints of any asymmetries were observed for either age group. There thus appears to be considerable developmental continuity in the nature of early and mature lexical representations. Discrepancies between the current findings and those of previous studies appear to be due to methodological differences that cast doubt on the validity of claims of psycholinguistic support for underspecification.

Sgaw Karen is a Tibeto-Burman language with both lexical tone and a major-minor syllable divide. Unlike major syllables, minor syllables are structurally reduced and do not bear overt phonological tone. This acoustic study investigates the fundamental frequency of these minor syllables, summarizing surface pitch patterns and providing evidence for a phonetic target. The first acoustic look at minor syllables in Karenic languages, a phonetically understudied language family, 58 unique minor syllable sequences of each possible length (~500 tokens) are elicited from one native speaker from the Karen State, Myanmar (Burma). Analysis of minor syllable sequences adjacent to lexical tones shows (1) a low pitch target around 185 Hz to which minor syllables asymptotically approximate when given sufficient time, and (2) this target is not explicable by contextual interpolation. Results address the general problem of tonal underspecification and interpret findings in the context of a minor syllable typology. Acoustic surface patterns of the Sgaw Karen minor syllable point to phonetic pitch targets of gradient strengths. Findings provide a foundation for detailed tonal/ intonational analysis of Sgaw Karen, and demonstrates a framework for acoustic documentation of underspecified tonal units in understudied languages of Southeast Asia in general.

Knowledge about semantic associations between words is effective to disambiguate word senses. The aim of this paper is to investigate the role and the relevance of telic information from SIMPLE in the disambiguation of basic action types of Italian HOLD verbs (prendere, ‘to take’, raccogliere, ‘to pick up’, pigliare ‘to grab’ etc.). We propose an experiment to compare the results obtained with telic information from SIMPLE with basic co-occurrence information extracted from corpora (most salient verbs modifying nouns) classified in terms of general semantic classes to avoid data sparseness.

Transparent segments have been a well known challenge for accounts of patterns of long distance agreement, such as vowel and consonant harmony. Two standard ways to account for transparency are autosegmental feature spreading with underspecification (e.g. Kiparsky 1981; Steriade 1987) and Agreement by Correspondence (ABC; Walker 2000; Walker & Rose 2004; Hansson 2001). Both, however, fail to derive the multiple instances of transparency encountered in Tsilhqut'i­n (Cook 1993; 2013). Here, non-retracted dorsals act both as transparent and as opaque to the process of vowel retraction, depending on which side of the trigger, a retracted sibilant, they are located. On the other hand, both retracted and non-retracted dorsals are transparent in sibilant harmony, in which sibilants are forced to agree in retraction. I propose a superset approach that combines feature spreading and underspecification with ABC: All dorsals are transparent in sibilant harmony, because they are outside the...

It has traditionally been assumed that the meaning of some or all phrasal idioms is noncompositional. However, I will argue here that the aspectual meaning of idioms is completely systematic: there are no special aspectual restrictions on idioms, and moreover, the aspectual properties of an idiom are compositional, combining the aspectual properties of its syntactic constituents in the usual way. I will show that this observation supports the theory of Distributed Morphology (Halle & Marantz 1994). 1 Aspectual Classes of Idioms First, it is worth noting that all aspectual classes contain idiomatic VPs. In what follows, I will assume the familiar Vendlerian classes (states, activities, achievements, and accomplishments), identified by an array of tests from the literature (see Vendler 1967, Dowty 1979, Mittwoch 1991, among many others). Subclasses of achievements and accomplishments will also be distinguished. However, the aspectual parallelism between idiomatic and non-idiomatic VPs is independent of this classification.

It has traditionally been assumed that the meaning of some or all phrasal idioms is noncompositional. However, I will argue here that the aspectual meaning of idioms is completely systematic: there are no special aspectual restrictions on idioms, and moreover, the aspectual properties of an idiom are compositional, combining the aspectual properties of its syntactic constituents in the usual way. I will show that this observation supports the theory of Distributed Morphology (Halle & Marantz 1994). 1 Aspectual Classes of Idioms First, it is worth noting that all aspectual classes contain idiomatic VPs. In what follows, I will assume the familiar Vendlerian classes (states, activities, achievements, and accomplishments), identified by an array of tests from the literature (see Vendler 1967, Dowty 1979, Mittwoch 1991, among many others). Subclasses of achievements and accomplishments will also be distinguished. However, the aspectual parallelism between idiomatic and non-idiomatic VPs is independent of this classification.

The count/mass distinction is a widely discussed topic across languages and linguistic theories have covered a great part of peculiarities which appear in relation to this phe- nomenon. Abstract nouns have often been left out of consideration, possibly due to the fact that their reference is abstract and the application of some relevant features of count or mass reference, such as cumulativity, divisiveness or atomicity, does not seem to be possible. This paper presents a thorough study of lexical features of a subset of abstract nouns and their distribution in COCA which suggests that countability in abstract nouns has to be determined relative to their semantic category. Focusing on eventuality denoting nominals which comprise a substantial part of abstract nouns, I argue that these nouns resemble concrete nouns in that the countability distinctions are expressed in surprising similarity. I find that the core feature underlying the distinction between abstract count and abstract m...

In this study, we propose a discriminative training algorithm to jointly minimize mispronunciation detection errors (i.e., false rejections and false acceptances) and diagnosis errors (i.e., correctly pinpointing mispronunciations but incorrectly stating how they are wrong). An optimization procedure, similar to Minimum Word Error (MWE) discriminative training, is developed to refine the ML-trained HMMs. The errors to be minimized are obtained by comparing transcribed training utterances (including mispronunciations) with Extended Recognition Networks [3] which contain both canonical pronunciations and explicitly modeled mispronunciations. The ERN is compiled by handcrafted rules, or data-driven rules. Several conclusions can be drawn from the experiments: (1) data-driven rules are more effective than hand-crafted ones in capturing mispronunciations; (2) compared with the ML training baseline, discriminative training can reduce false rejections and diagnostic errors, though false acceptances increase slightly due to a small number of false-acceptance samples in the training set.

Die vorliegende Dissertation befasst sich mit der Zählbarkeit abstrakter Substantive, die in der Literatur zur Semantik von Substantiven kaum berücksichtigt wurden. Die Erforschung einer Teilmenge von Abstrakta anhand der Annotation von lexikalischen Merkmalen und einer gezielten Korpusstudie ermöglicht einen Einblick in das Sortenpotential dieser Nomina. Aus dieser Untersuchung lässt sich eine Reihe von Generalisierungen ableiten, anhand derer eine semantische Analyse der Zählbarkeit bei Ereignisnominalisierungen bereitgestellt wird. Das wesentliche Kriterium, das zählbare Ereignisse von nicht-zählbaren unterscheidet ist die Vagheit der Atome, die in Abhängigkeit zur Aktionsart des zugrundeliegenden Ereignisses bestimmt wird. Mit der Ausarbeitung der Zählbarkeit in Ereignissen stelle ich erste Anregungen für eine kompositionelle Analyse vor. Demzufolge trägt diese Ausarbeitung wesentlich zur Diskussion der Zählbarkeit und Ereignissemantik in der aktuellen Forschung bei.

We present a study of constructions of the type <light verb + noun> in European Portuguese. We analyse these constructions as complex predicates where both the light verb and the noun share an important role in the predication and we focus on the aspectual combinatory properties of the two elements of the complex predicate. We propose that light verbs inherit the feature specification of the corresponding main verbs, but that the light verbs are underspecified regarding (some of) the event structure features.

The main task that this thesis deals with is the provision of a comprehensive analysis covering a meaningful subset of English and developing a computational implementation that is able to show understanding of this language subset, in part via limited visualisation. There is a well accepted analogy that says that eventualities exist in time, in ways that are similar to how objects exist in space. This analogy is used as a framework to investigate in detail those activities that are the eventuality analogue of plural and mass objects—multiple instances of an activity, or continuous occurrence of an activity respectively. These are called extended activities. We examine the ways in which natural language is used to describe these kinds of activities, and discuss ways in which the meanings of such language can be represented. We concentrate on language that is in the form of instructions, and discuss the special relationship between instructions and activities. Using the idea that som...

We take the opportunity of the publication of some of the papers of the ESSLLI workshop TYTLES (TYpe Theory and LExical Semantics, ESSLLI 2015, Barcelona) to provide an overview of the possibilities that type theory offers to model lexical semantics, especially the type-theoretical frameworks that properly model compositional semantics. origins of this issue: esslli workshop on type theory and lexical semantics 2015 The program of the ESSLLI 2015 workshop held in Barcelona 1 consisted of twelve selected talks. The corresponding extended abstracts, together with an introduction and a conclusion by the workshop organisers, are available on the web as Cooper and Retoré (2015); it includes: A. Introduction (slides), by Robin Cooper and Christian Retoré.

This paper describes an attempt to gain more insight into the mechanisms that underlie lexicalised systematic polysemy. This phenomenon is interpreted as systematic sense combinations that are valid for more than one word. The hierarchical structure of WordNet is exploited to create a working definition of systematic polysemy and extract polysemic patterns at a level of generalisation that allows the identification of fine-grained semantic relations between the senses of the words participating in the systematic polysemic pattern.

Which normally-transitive English verbs can omit their objects (I ate), and why? This paper explores three factors that have been suggested to facilitate object omission: (i) how strongly a verb selects its object (Resnik 1993); (ii) a verb's frequency (Goldberg 2005); (iii) the extent to which the verb is associated with a routine-a recognized, conventional series of actions within a community (Levin & Rapaport Hovav 2014; Martí 2015). To operationalize (iii), this paper leverages the assumption that a given verb may be more strongly associated with a routine in one community than another. Comparing writings across communities, this paper offers corpus and experimental evidence that verbs omit their objects more readily in the communities where they are more strongly associated with a routine. Object-omitting uses of verbs are analyzed, following other work, as intransitive aspectual activities describing an agent's routine actions; so the hearer's task is not to recover a missing object, but to recognize the routine described by the verb. More broadly, the paper explores how the meaning and syntactic potential of verbs are shaped by the practices of the people who use them.

This paper examines the acoustics of underspecification and vowel harmony (VH) in Turkish. In Turkish, vowels in suffixes that change according to VH rules are widely believed to be underspecified for rounding and/or backness. Underspecification has previously been thought to persist in the acoustic realization of underspecified segments. In the present study, it was hypothesized that underspecification in Turkish vowels persists in their acoustic realization even after specification due to incomplete specification and that harmonizing vowels assimilate to particular vowel categories rather than the trigger vowels themselves. An experiment was run comparing underspecified suffix vowels to their specified counterparts in roots. Of the vowels underspecified for rounding, unrounded high vowels had a significantly higher F2 and [i] had a significantly lower F2 compared to their specified counterparts at all three points of measurement. This could be a case of enhancement, where the feat...

This paper revisits Oroch vowel harmony previously analyzed in Tolskaya (2008) and propose a new analysis within the framework of modified contrastive specification (Dresher, 2009). In particular, I show that the vowel patterns in Oroch are better explained under a contrastive hierarchy [low] &gt; [coronal] &gt; [Retracted Tongue Root (RTR)] &gt; [labial] than under Tolskaya`s Stratal OT approach. The proposed hierarchy not just assigns proper feature specifications to each vowel in Oroch but also provides a simple, unified solution to the transparency and opacity in Oroch vowel harmony: The transparency of /i/ to RTR harmony is due to the lack of contrastive [-RTR] specification while the opacity of /i, ?/ is due to its contrastive height feature specification ([-low]). In addition, the putative simple vowel /ae/ receives a better treatment as a diphthong /ia/ and its peculiar behavior resembling /i/ is ascribed to its high front vowel portion /i/.

In this paper we present state-of-the-art results on the computational classification of semantic type coercion, accomplished using a novel geometric method which is both context-sensitive and generalisable. We show that this method improves accuracy on a SemEval dataset over previous work, and gives promising results on a new more challenging experimental setup involving the same data. In addition to a description of our distributional semantic methodology and the results obtained on an established dataset, we offer an overview of the linguistic phenomenon of coercion and an analysis of the geometric features by which our results are achieved.

We discuss an extension of the standard logical rules (functional application and abstraction) in Categorial Grammar (CG), in order to deal with some specific cases of polysemy. We borrow from Generative Lexicon theory which proposes the mechanism of {\em coercion}, next to a rich nominal lexical semantic structure called {\em qualia structure}. In a previous paper we introduced coercion into the framework of {\em sign-based} Categorial Grammar and investigated its impact on traditional Fregean compositionality. In this paper we will elaborate on this idea, mostly working towards the introduction of a new semantic dimension. Where in current versions of sign-based Categorial Grammar only two representations are derived: a prosodic one (form) and a logical one (modelling), here we introduce also a more detaled representation of the lexical semantics. This extra knowledge will serve to account for linguistic phenomena like {\em metonymy\/}.

Following Kennedy and Levin (2001), I will use the term predicate of gradual change to mean almost all accomplishments and all activities, excluding a class of cases describing abrupt transitions like those in (1), discussed in Rothstein (2004): One can separate work on the aspectual nature of gradual change into two broad classes, mereological and degree-based, exemplified in papers like Krifka (1989) and Hay et al. (1999). The degree-based work seeks to explain aspectual properties in terms of degree scales with the same structure as those of adjectives; the mereological work assumes a different, arguably more general, partial order structure, which is also taken to be play a role in the semantics of mass terms. This paper argues that both kinds of structure are necessary for verbs of gradual change and proposes a unified formulation, as well as a diagnostic for separating the cases. I will argue that for a class of problematic predicates lying in a boundary region, and for processes like deverbal adjective creation, a coercion operation mapping from one structure to the other is also necessary. A mereological account is most naturally motivated through the examples of incremental themes (Dowty 1991). The unifying property of these verbs was succinctly stated in Krifka (1989): If an incremental theme verb expresses a relation R and R holds in event e and Θ(e) is the incremental theme, then in any sub-event of e in which R holds a portion or part of Θ(e) stands in the relation R. Incremental theme predicates include eat, learn, write, mow, perform, and read. Degree achievements, in contrast, provide the point of departure for a degree based analysis (Hay et al. 1999, Zucchi 1998, Kennedy and McNally 1999, Kratzer 2000), because they offer a rich stock of cases in which telicity is determined with no reference at all to parts. For example, a suitcase closing (Filip 1999) does not progress by having more and more of the suitcase closed, but by having the suitcase more and more closed. Hay et al. (1999) (HKL) explicitly argue for extending a degree account to the class of incremental theme verbs, suggesting for example, that for a verb like eat, telicity is directly linked to the size or volume of the theme and Kennedy and Levin (2001) (KL) argue for extending it to all verbs of gradual change. They point out, correctly, that while a degree-based acount like theirs can be formulated to explain the aspectual properties Following KL, I shall assume that predicates of gradual change are quantizable properties of events. That is, I assume every degreeable verb α has a change function ∆ e for which it is sensible to write (2). Following HKL, I will refer to q as the difference argument. Mereological accounts may be unified with degree accounts by assuming change functions for them as well, with parts ordered by the part-whole relation as their range. Somewhat surprisingly, the HKL analysis is then completely compatible with a mereological account. The key to their analysis is the INCREASE operator, which requires a domain for which a notion of difference is coherent. I will show that what Krifka calls the Remainder Principle, often found among mereological axiomatizations (Krifka 1998), is sufficient to define differences: The Remainder Principle requires that for any two ordered elements, x < y, there exists a remainder r which does not overlap x and which can be summed with x to give y. In a word, r is the difference between x and y. For any partially ordered domain with differences, quantizing those difference arguments will allow the definition of INCREASE; the entire HKL account of telicity follows.

This paper describes an attempt to gain more insight into the mechanisms that underlie lexicalised systematic polysemy. This phenomenon is interpreted as systematic sense combinations that are valid for more than one word. The hierarchical structure of WordNet is exploited to create a working definition of systematic polysemy and extract polysemic patterns at a level of generalisation that allows the identification of fine-grained semantic relations between the senses of the words participating in the systematic polysemic pattern.

This paper analyzes a case of Italian word-formation, in which the semantics of the derived words appears to contain mutually exclusive ambiguities. Italian productively derives verbs of removal from nouns. These verbs have the general semantic form A removes X from Y. However, there are two subtypes that differ in whether the nominal base of the derived verb is taken to be the FIGURE or the GROUND: scremare 'to skim'; (crema 'cream') is a FIGURE-verb, and stanare 'to make come out of the burrow' (tana 'burrow') is a GROUND-verb. Traditional approaches are at a loss to give a uniform account for the semantics of derivational processes of these two kinds. In this paper, a uniform analysis is proposed. It is based upon the model of lexical semantics known as two-level semantics. Two-level semantics makes a distinction between a layer of meaning which is defined by grammar, and a level of interpretation, which is based upon conceptual knowledge. We propose that the derivation of Italian denominal verbs of removal of both types starts from a single underspecified representation, which is then specified at the conceptual level; depending on the concept type of the base, the denominal verb is either a FIGURE-verb or a GROUND-verb. In general, this paper argues that the two-level approach to semantics can be fruitfully improved by combining it with the notion of underspecification. And on a more general level, it is an example of how language specific semantics may be embedded in cognitive structure. (2) GROUND-verbs stanare 'to make come out of the burrow'; cf. tana 'burrow' snidare 'to drive out' cf. nido 'nest' scarcerare 'to release (from prison)' cf. carcere 'prison' scassare 'to remove from a box' cf. cassa 'box; case' sganciare 'to unhook; to unfasten' cf. gancio 'hook' sbarcare 'to unload; to disembark' cf. barca 'boat' Both types of DVRs are causative verbs of removal of an object X from another object Y. We call the movable object FIGURE, and the unmovable one GROUND. Thus, DVRs can be assigned the general semantic structure A removes figure X from ground Y. The two subtypes differ in the way the two locative slots FIGURE and GROUND are filled by the nominal base. In FIGURE-verbs, the nominal base fills the FIGURE-slot (cf. scremare 'to remove the cream from something') and in GROUND-verbs, it fills the GROUND-slot (cf. stanare 'to remove something from the burrow'). The question is whether we can account for both types of DVRs with one derivational process, or whether we have to assume two processes. In this paper, we argue that one underspecified derivational process can account for the two subtypes. The difference between the subtypes is only spelled out at the conceptual level. This is confirmed by the observation that the concept type of the base determines to which class a DVR belongs. The paper is organized as follows: In the first part (section 1), we discuss some general assumptions which underlie this study. We give a descriptive account of Italian DVRs (section 2) and propose a rule which generates them with an underspecified semantics (sections 3 and 4). We then discuss the problem of how this underspecification is resolved (section 5) and how nouns must be represented in the lexicon in order to make the resolution possible (section 6).

Computational semantics has long relied upon the Montague correspondance between syntax and semantics, which is not by itself well suited for the computing of some phenomena, such as logical polysemy, addressed by recent advances in lexical semantics. Our aim is to integrate the results of lexical semantics studies such as the Generative Lexicon Theory in a straightforward way with the existing computing of logical forms, in order to form a Montagovian framework for lexical semantics. In addition, we will outline a way to integrate other kinds of semantic information.

In this paper, I explore the relation between methods of lexical representation involving decomposition and the theory of types as used in linguistics and programming semantics. I identify two major approaches to lexical decomposition in grammar, what I call parametric and predicative strategies. I demonstrate how expressions formed with one technique can be translated into expressions of the other. I then discuss argument selection within a type theoretic approach to semantics, and show how the predicative approach to decomposition can be modeled within a type theory with richer selectional mechanisms. In particular, I show how classic Generative Lexicon representations and operations can be viewed in terms of types and selection. * I would like to thank Nicholas Asher, with whom I have been developing the Type Composition Logic adopted here as the type theoretic interpretation of GL. I would also like to thank Ray Jackendoff, Jose Castano, Roser Sauri, Patrick Hanks, and Chungmin Lee for useful critical remarks. All errors and misrepresentations are, of course, my own.

In this paper, we describe the Argument Selection and Coercion task, currently in development for the SemEval-2 evaluation exercise scheduled for 2010. This task involves characterizing the type of compositional operation that exists between a predicate and the arguments it selects. Specifically, the goal is to identify whether the type that a verb selects is satisfied directly by the argument, or whether the argument must change type to satisfy the verb typing. We discuss the problem in detail and describe the data preparation for the task.

As a human-specific trait, language offers a unique window on human cognition. Grammatical constraints on the ways we speak about events, for instance, have long been thought to reveal the representational formats that our minds impose on the ways that we think about events. In recent research, verbs that name events of death have stood out as key counterexamples to standard theories of the grammatical constraints on possible verbs. The special status of these thanatological verbs raises two important questions: why, given the vast number of verbs in any language, is it that verbs of death hold this special status, and what do they tell us about the restrictions on the representational format for possible verbs? This paper reexamines the evidence coming from verbs of death, confirming that they are counterexamples to standard theories, but that their behaviour suggests a more revealing constraint on our mental representations-that our minds impose strict restrictions on the format o...

Existe una similitud en las alternancias de la vocal temática de los verbos españoles de segunda y tercera conjugación que siempre ha llamado la atención a los gramáticos hispanistas. Pocos, sin embargo, se han dado a la tarea de explicar estas similitudes, las cuales, en general, se consideran formas idiosincrásicas de la lengua española. En este estudio el autor propone un análisis novedoso que da cuenta de las alternancias de las vocales temática y radical de los verbos españoles. Demuestra la existencia de un rasgo fonético flotante en la representación subyacente de los verbos de tercera conjugación que marca la diferencia entre éstos y los verbos de segunda conjugación. También se demuestra que una regla de esparcimiento (armonía vocal) del rasgo [+alto] es responsable de la existencia de la mayoría de los verbos de tercera conjugación que en realidad son verbos de segunda conjugación en su forma subyacente. Este análisis predice que no deben existir en español verbos con vocal temática media precedidos de una vocal radical alta.

Recent research has shown that toddlers' lexical representations are phonologically detailed, quantitatively much like those of adults. Studies in this article explore whether toddlers' and adults' lexical representations are qualitatively similar. Psycholinguistic claims (Lahiri & Marslen-Wilson, 1991; Lahiri & Reetz, 2002, 2010) based on underspecification (Kiparsky, 1982 et seq.) predict asymmetrical judgments in lexical processing tasks; these have been supported in some psycholinguistic research showing that participants are more sensitive to non-coronal-to-coronal (pop → top) than to coronal-to-noncoronal (top → pop) changes or mispronunciations. Three experiments using on-line visual world procedures showed that 19-month-olds and adults displayed sensitivities to both noncoronal-to-coronal and coronal-to-noncoronal mispronunciations of familiar words. No hints of any asymmetries were observed for either age group. There thus appears to be considerable developmental continuity in the nature of early and mature lexical representations. Discrepancies between the current findings and those of previous studies appear to be due to methodological differences that cast doubt on the validity of claims of psycholinguistic support for lexical underspecification.

This paper discusses the strategies used for the anticausative alternation and the constraints on their distribution in two early Italian vernaculars, Old Florentine and Old Neapolitan, focusing on the emergence of aspectual notions such as telicity in determining variability in the occurrence of the reflexive morpheme si/se, the main/ only strategy for anticausatives with some inherently telic and punctual verbs, depending on the vernacular. It is also shown that in the early varieties the reflexive in the anticausative alternation mainly signals the suppression of the Actor, only gradually coming to mark the presence of a terminal point in the meaning of the verb/predicate.

Although English grammar encodes a number of semantic contrasts with tense and aspect marking, these semantics are currently ignored by Abstract Meaning Representation (AMR) annotations. This paper extends sentence-level AMR to include a coarse-grained treatment of tense and aspect semantics. The proposed framework augments the representation of finite predications to include a four-way temporal distinction (event time before, up to, at, or after speech time) and several aspectual distinctions (including static vs. dynamic, habitual vs. episodic, and telic vs. atelic). This will enable AMR to be used for NLP tasks and applications that require sophisticated reasoning about time and event structure.

Bondu-so (Dogon; Mali) vowel harmony exhibits both typologi-cally and theoretically interesting properties. The language's vocalic system displays surface patterns that implicate a ten-vowel system with an underlying [ATR] contrast at three vowel heights that is not immediately apparent given only mid vowels maintain an [ATR] contrast on the surface. The current paper presents previously unaccounted for data that show alternations associated with Bondu-so vowel harmony correlate not only with the [ATR] specification of a given root vowel, but also with properties of the root-final consonant. We appeal to a combination of featural and prosodic licensing to analyze these outcomes and do so in a modified version of the Parallel Structures Model of feature geometry. The PSM framework has been employed in studies of consonant assimilation and consonant-vowel interaction, but to our knowledge, the current paper is the first to extend it specifically to the analysis of vowel harmony.

[this version is the authors' accepted manuscript; please seek permission before citing]

In this paper we present state-of-the-art results on the computational classification of semantic type coercion, accomplished using a novel geometric method which is both context-sensitive and generalisable. We show that this method improves accuracy on a SemEval dataset over previous work, and gives promising results on a new more challenging experimental setup involving the same data. In addition to a description of our distributional semantic methodology and the results obtained on an established dataset, we offer an overview of the linguistic phenomenon of coercion and an analysis of the geometric features by which our results are achieved.

This paper investigates the meaning adaptability of change of state (CoS) verbs. It argues that both coercion and underspecification are necessary mechanisms in order to properly account for the semantic adaptability observable for CoS verbs in combination with their complements. This type of meaning adaptability has received little formal attention to date, although some recent work has already led the way on this topic (Spalek 2014, Lukassek and Spalek 2016, Asher et al. 2017). Our paper is part of a cross-linguistic case study of einfrieren and congelar. We model the meaning adaptability of this test case within Type Composition Logic (TCL) (Asher 2011). We build on Asher's coercion mechanism and introduce an additional mechanism for underspecification that exploits the fine-grained type system in TCL.

In this article we explore a methodology based on unsupervised learning that aims at automatically classifying the arguments of a verb, a classification crucially important in the process of sense isolation in lexicographic projects. More concretely, we strive to exploit the lexical semantic information obtained by combining a clustering algorithm with Word Embeddings and argue that the combination of both represents a fertile methodological ground for automatically classifying the selectional behaviour of a predicate. From the perspective of practical lexicography, our methodology should help to significantly shorten the time invested in isolating context-related senses of a predicate, in our case a verb. In particular, we illustrate, based on a case study of two Spanish verbs, how our methodology can provide a lexicographer with a general picture of a verb’s behaviour in a corpus and argue that the clusters we provide automatically encode insightful generalizations that facilitate the process of decision taking of a lexicographer about which senses are most common for a particular predicate, according to the different types of contexts it is most frequently used in.

As a deadjectival verb, English clean should be a clear-cut result verb, yet it often shows hallmarks of a manner verb. This paper investigates this dual behavior in light of manner/result complementarity: the proposal that verbs lexicalize either manner or result meaning components, but not both. We demonstrate that once lexicalized and contextually determined meaning components are distinguished, clean conforms to manner/result complementarity. It can be a result verb, not entailing a particular manner. However, given its strong association with cleaning routines, some uses simply lexicalize manner. Crucially, in manner uses the result drops out, consistent with manner/result complementarity. The manner-only and result-only uses of clean, then, instantiate related senses, each conforming to manner/result complementarity.

Change of state verbs and verbs of directed motion are shown to share interpretive properties and argument realization properties which are claimed to follow from the fact that they both lexically encode, or lexicalize, as scale. The semantic core of each verb in both classes is a stative scalar attribute, and the events in the denotation of the verb involve a change in the value of that scalar attribute. There is a similar typology of scalar attributes which form the core of both classes. In contrast, incremental theme verbs, which have also been claimed to be associated with a scale, do not show the same interpretive and argument realization properties. This is claimed to follow from the fact that the scale associated with these verbs is introduced by the direct object and not lexicalized by the verb.

This paper addresses the question of whether the four-way Vendler classification is appropriate for verbs or VPs. It suggests that the Vendler classification is not appropriate as a classification of verbs, and offers a different classification for the elements of lexicalized meaning which determine the aspectual potential of verbs. It highlights the importance of a class of verbs with a lexically encoded scale, illustrating that a large class of scalar verbs cannot be classified once and for all either as activities, accomplishments or achievements, though the lexically encoded scale accounts for most of the aspectual behavior of these verbs. The aspectual and syntactic significance of a lexicalized scale is explored. The class of verbs lexicalizing a scale is shown not to be the same as the class of verbs selecting an incremental theme. This is justified both on semantic grounds and on syntactic grounds. There appears to be more justification for recognizing the four-way Vendler classification at the VP level, though it is demonstrated that accomplishments do not have a uniform internal temporal structure, predominantly because of the variety of sources of incremental structure.

We investigate the English verbs climb and cut, cited as counterexamples to manner/result complementarity: the proposal that verbs lexicalize either manner or result meaning components, but not both. Once their lexicalized meaning is identified and distinguished from contextually determined elements of meaning, cut and climb conform to manner/result complementarity. We show that cut is basically a result verb, with a prototypical manner often inferred. However, as it lexicalizes a result prototypically brought about in a certain manner, some uses simply lexicalize this manner. Crucially, in manner uses, the result component drops out, consistent with manner/result complementarity. In contrast, climb is essentially a manner verb. Once its lexicalized manner is accurately identified and distinguished from meaning contributed by context, the upward direction associated with many uses can be shown to arise from inference. However, climb has some restricted uses which lexicalize a result. Importantly, on these uses, the manner component is lost. With both verbs, then, the manner-only and result-only uses instantiate different, though related, senses of the relevant verb, with each sense conforming to manner/result complementarity.

In this paper we present a novel argument against strict locality in vowel harmony: a vowel's feature may have a double identity, active in one process and neutral in another. Such is the behavior of [back] in Votic [i]. It is invisible to harmony, while simultaneously triggering an assimilation process. We argue that no feature-sharing account of this phenomenon is plausible, including the relatively powerful extension of Span Theory that permits vowels in a harmonic span to remain unasso-ciated (and unharmonized) with the span's head. We offer instead an account based on the Agreement-By-Correspondence approach to long-distance assimilation.

The question of how much meaning can be encoded in a single word is a central issue in natural language semantics. In this paper, I test the theory that verbs display manner-result complementarity: that verbs either describe an action performed by an agent, or a change undergone by a patient, but not both (Rappaport-Hovav & Levin 2010). Given data from an acceptability judgment experiment as well as a corpus study, I propose that instrumental verbs such as slice and chop are counter-examples to complementarity. These results suggest that verbal meaning is packaged into a single verbal root (see Beavers & Koontz-Garboden 2012), but that this verbal root may be associated with more than one part of an event structure.

After a quick overview of the field of study known as “Lexical Semantics”, where we advocate the need of accessing additional information besides syntax and Montague- style semantics at the lexical level in order to complete the full analysis of an utterance, we summarize the current formulations of a well-known theory of that field. We then propose and justify our own model of the Generative Lexicon Theory, based upon a variation of classical compositional semantics, and outline its formalization. Additionally, we discuss the theoretical place of informational, knowledge-related data supposed to exist within the lexicon as well as within discourse and other linguistic constructs.

Computational semantics has long relied upon the Montague correspondance between syntax and semantics, which is not by itself well suited for the computing of some phenomena, such as logical polysemy, addressed by recent advances in lexical semantics. Our aim is to integrate the results of lexical semantics studies such as the Generative Lexicon Theory in a straightforward way with the existing computing of logical forms, in order to form a Montagovian framework for lexical semantics. In addition, we will outline a way to integrate other kinds of semantic information.