Probing Lexical Ambiguity: Word Vectors Encode Number and Relatedness of Senses

Lexical ambiguity—the phenomenon of a single word having multiple, distinguishable senses—is pervasive in language. Both the degree of ambiguity of a word (roughly, its number of senses) and the relatedness of those senses have been found to have widespread effects on language acquisition and processing. Recently, distributional approaches to semantics, in which a word's meaning is determined by its contexts, have led to successful research quantifying the degree of ambiguity, but these measures have not distinguished between the ambiguity of words with multiple related senses versus multiple unrelated meanings. In this work, we present the first assessment of whether distributional meaning representations can capture the ambiguity structure of a word, including both the number and relatedness of senses. On a very large sample of English words, we find that some, but not all, distributional semantic representations that we test exhibit detectable differences between sets of monosemes (unambiguous words; N = 964), polysemes (with multiple related senses; N = 4,096), and homonyms (with multiple unrelated senses; N = 355). Our findings begin to answer open questions from earlier work regarding whether distributional semantic representations of words, which successfully capture various semantic relationships, also reflect fine-grained aspects of meaning structure that influence human behavior. Our findings emphasize the importance of measuring whether proposed lexical representations capture such distinctions: In addition to standard benchmarks that test the similarity structure of distributional semantic models, we need to also consider whether they have cognitively plausible ambiguity structure.     

Shifting senses in lexical semantic development

Most words are associated with multiple senses. A DVD can be round (when describing a disc), and a DVD can be an hour long (when describing a movie), and in each case DVD means something different. The possible senses of a word are often predictable, and also constrained, as words cannot take just any meaning: for example, although a movie can be an hour long, it cannot sensibly be described as round (unlike a DVD). Learning the scope and limits of word meaning is vital for the comprehension of natural language, but poses a potentially difficult learnability problem for children. By testing what senses children are willing to assign to a variety of words, we demonstrate that, in comprehension, the problem is solved using a productive learning strategy. Children are perfectly capable of assigning different senses to a word; indeed they are essentially adult-like at assigning licensed meanings. But difficulties arise in determining which senses are assignable: children systematically overestimate the possible senses of a word, allowing meanings that adults rule unlicensed (e.g., taking round movie to refer to a disc). By contrast, this strategy does not extend to production, in which children use licensed, but not unlicensed, senses. Children’s productive comprehension strategy suggests an early emerging facility for using context in sense resolution (a difficult task for natural language processing algorithms), but leaves an intriguing question as to the mechanisms children use to learn a restricted, adult-like set of senses.     

Polysemy and the subjective lexicon: Semantic relatedness and the salience of intraword senses

Subjective perceptions of the senses of polysemous English words are collected in questionnaire studies and the effects of variability in semantic distances among these senses are examined in an experiment. In the first of two questionnaire studies, native speakers produce meanings for 175 polysemous words; from their responses, the most frequently produced meaning for each word is identified as its dominant sense. In a second questionnaire, independent subjects rate the semantic relatedness between the dominant meaning and the other senses generated for each word in the first study. Relatedness measures vary, raising the possibility that polysemous words vary in terms of the salience of their different senses in different contexts. This is confirmed in an experiment showing that salience ratings are influenced by the interacting factors of sentential context, extent of relatedness of the senses, and the dominance status of the senses.This research was supported by a grant from the Special Research Fund, University of Western Australia, to Kevin Durkin.     

Context and homograph meaning resolution.

We describe an activation-based model of word recognition and apply it to the process of resolving the meaning of homographs presented in context. The interpretation of homographs was assessed by asking participants to decide whether a target word was related to the meaning of a sentence containing a homograph. These relatedness decisions varied systematically with the relative frequency of the homograph meanings, delay, and the nature of the sentence context. In the model, it was assumed that orthographic and contextual information combine additively to determine the activation of word meanings, and that the probability of a "related" response is determined by the activation level of the related meaning. The model accurately accounts for all observed effects, as well as their interaction. We conclude that the core process of lexical ambiguity resolution may be quite simple.     

On Tichy’s determiners and Zalta’s abstract objects

It is not a common practice to postulate meaning entities treated as objects of some kind. The paper demonstrates two ways of introducing meaning-objects in two logics of natural language, Tichy’s Transparent Intensional Logic and Zalta’s Intensional Logic of Abstract Objects. Tichy’s theory belongs to the Fregean line of thinking, with what he calls ‘constructions’ as Fregean senses, and ‘determiners’ as object-like meaning entities constructed by the senses. Zalta’s theory belongs to Meinongian logics and he postulates a rich realm of abstract Meinongian objects to play the role of meanings. The paper analyses the mechanisms of reference in both conceptions and it offers a comparison of the mediating meaning-objects and the framework designed to expose this mediation in both theories. An attempt is made to expose how the treatment of the meaning entities depends upon the theory of meaning which is assumed.     

Cross-language influences: translation status affects intraword sense relatedness

Many words have more than one meaning, and these meanings vary in their degree of relatedness. In the present experiment, we examined whether this degree of relatedness is influenced by whether or not the two meanings share a translation in a bilingual’s other language. Native English speakers with Spanish as a second language (i.e., English-Spanish bilinguals) and native Spanish speakers with English as a second language (i.e., Spanish-English bilinguals) were presented with pairs of phrases instantiating different senses of ambiguous English words (e.g., dinner date–expiration date) and were asked to decide whether the two senses were related in meaning. Critically, for some pairs of phrases, a single Spanish translation encompassed both meanings of the ambiguous word (joint-translation condition; e.g., mercado in Spanish refers to both a flea market and the housing market), but for others, each sense corresponded to a different Spanish translation (split-translation condition; e.g., cita in Spanish refers to a dinner date, but fecha refers to an expiration date). The proportions of “yes” (related) responses revealed that, relative to monolingual English speakers, Spanish–English bilinguals consider joint-translation senses to be less related than split-translation senses. These findings exemplify semantic cross-language influences from a first to a second language and reveal the semantic structure of the bilingual lexicon.     

Processing ambiguous verbs: evidence from eye movements

In 2 eye-tracking experiments, participants read verbs that had 2 (unrelated) meanings or 2 (related) senses in contexts that disambiguated before or after the verb, to the dominant or subordinate interpretation. A 3rd experiment used unambiguous verbs. The results indicated that the language processor used information about context in the early stages of resolving meaning ambiguities but only during integration for sense ambiguities. Effects of preference were delayed for both types of verbs. The results contrast with findings concerning the processing of nouns (e.g., K. Rayner & S. A. Duffy, 1986). For meaning ambiguities, the authors argue that delays in resolution allow both meanings to reach a high level of activation, thus reducing effects of frequency. For sense ambiguities, the authors argue that the processor does not access multiple senses but activates one underspecified meaning and uses context to home in on the appropriate sense.     

THE DEVELOPMENT OF NONMETRIC SPACE ANALYSIS: A LETTER TO PROFESSOR JOHN ROSS

A series of recent studies (Cliff, 1968, 1969; Cliff, Bradley, and Girard, 1970) describe a mathematical model for the cognitive processes involved in an individual's endorsements (or nonendorsements) of personality inventory items. In this model, the inventory items are assumed to occupy positions in a multidimensional space of subjective meaning, and the probability of endorsement is assumed to be some mathematical funotion of these positions.     

Syntax,Semantics, and Intentional Aspects

Abstract

It is widely assumed that the meaning of at least some types of expressions involves more than their reference to objects, and hence that there may be co-referential expressions which differ in meaning. It is also widely assumed that “syntax does not suffice for semantics”, i.e. that we cannot account for the fact that expressions have semantic properties in purely syntactical or computational terms. The main goal of the paper is to argue against a third related assumption, namely that what is responsible for a difference in meaning between co-referential expressions is the computational difference in the cognitive functioning of the expressions. “Intentional aspects” of expressions—those features which their meanings involve in addition to reference—cannot be syntacticized, since they are individuated not in terms of any cognitive feature, but rather in terms of those properties of the referents through which the expressions refer to them, and cognitive features cannot determine such properties in exactly the same sense as they cannot determine reference.     

Predication

In Latent Semantic Analysis (LSA) the meaning of a word is represented as a vector in a high-dimensional semantic space. Different meanings of a word or different senses of a word are not distinguished. Instead, word senses are appropriately modified as the word is used in different contexts. In N-VP sentences, the precise meaning of the verb phrase depends on the noun it is combined with. An algorithm is described to adjust the meaning of a predicate as it is applied to different arguments. In forming a sentence meaning, not all features of a predicate are combined with the features of the argument, but only those that are appropriate to the argument. Hence, a different "sense" of a predicate emerges every time it is used in a different context. This predication algorithm is explored in the context of four different semantic problems: metaphor interpretation, causal inferences, similarity judgments, and homonym disambiguation.     

Ineffability,Signification and the Meaning of Life

Abstract

There is an apparent tension between two familiar platitudes about the meaning of life: (i) that ‘meaning’ in this context means ‘value’, and (ii) that such meaning might be ineffable. I suggest a way of trying to bring these two claims together by focusing on an ideal of a meaningful life that fuses both the axiological and semantic senses of ‘significant’. This in turn allows for the possibility that the full significance of a life might be ineffable not because its axiological significance is ineffable, but because its semantic significance is ineffable in virtue of the signification relation itself being unsignifiable. I then explore to what degree this claim about signification can be adequately defended.     

Inferentialism and the Normativity of Meaning

There may be various reasons for claiming that meaning is normative, and additionally, very different senses attached to the claim. However, all such claims have faced fierce resistance from those philosophers who insist that meaning is not normative in any nontrivial sense of the word. In this paper I sketch one particular approach to meaning claiming its normativity and defend it against the anti-normativist critique: namely the approach of Brandomian inferentialism. However, my defense is not restricted to inferentialism in any narrow sense for it encompasses a much broader spectrum of approaches to meaning, connected with the Wittgensteinian and especially Sellarsian view of language as an essentially rule-governed enterprise; and indeed I refrain from claiming that the version of inferentialism I present here is in every detail the version developed by Brandom.     

Aging and the Search for Meaning

SUMMARY

The natural science model is not adequate alone to deal with some of the profound problems of gerontology because it does not deal well with meaning and values issues. This paper develops the topic “Aging and the Search for Meaning” using phenomenological and hermeneutical methods. It is assumed that finding a meaning in life is a primary drive across the years. Such a quest requires a hermeneutic, i.e. a means by which life may be symbolized or given meaning and direction. Using Viktor Frankl's meaning theory, the paper introduces an approach which describes the process by which the meaning of aging can be interpreted.     

Semantic diversity: A measure of semantic ambiguity based on variability in the contextual usage of words

Semantic ambiguity is typically measured by summing the number of senses or dictionary definitions that a word has. Such measures are somewhat subjective and may not adequately capture the full extent of variation in word meaning, particularly for polysemous words that can be used in many different ways, with subtle shifts in meaning. Here, we describe an alternative, computationally derived measure of ambiguity based on the proposal that the meanings of words vary continuously as a function of their contexts. On this view, words that appear in a wide range of contexts on diverse topics are more variable in meaning than those that appear in a restricted set of similar contexts. To quantify this variation, we performed latent semantic analysis on a large text corpus to estimate the semantic similarities of different linguistic contexts. From these estimates, we calculated the degree to which the different contexts associated with a given word vary in their meanings. We term this quantity a word’s semantic diversity (SemD). We suggest that this approach provides an objective way of quantifying the subtle, context-dependent variations in word meaning that are often present in language. We demonstrate that SemD is correlated with other measures of ambiguity and contextual variability, as well as with frequency and imageability. We also show that SemD is a strong predictor of performance in semantic judgments in healthy individuals and in patients with semantic deficits, accounting for unique variance beyond that of other predictors. SemD values for over 30,000 English words are provided as supplementary materials.     

Word Senses as Clusters of Meaning Modulations: A Computational Model of Polysemy

Most words in natural languages are polysemous; that is, they have related but different meanings in different contexts. This one-to-many mapping of form to meaning presents a challenge to understanding how word meanings are learned, represented, and processed. Previous work has focused on solutions in which multiple static semantic representations are linked to a single word form, which fails to capture important generalizations about how polysemous words are used; in particular, the graded nature of polysemous senses, and the flexibility and regularity of polysemy use. We provide a novel view of how polysemous words are represented and processed, focusing on how meaning is modulated by context. Our theory is implemented within a recurrent neural network that learns distributional information through exposure to a large and representative corpus of English. Clusters of meaning emerge from how the model processes individual word forms. In keeping with distributional theories of semantics, we suggest word meanings are generalized from contexts of different word tokens, with polysemy emerging as multiple clusters of contextually modulated meanings. We validate our results against a human-annotated corpus of polysemy focusing on the gradedness, flexibility, and regularity of polysemous sense individuation, as well as behavioral findings of offline sense relatedness ratings and online sentence processing. The results provide novel insights into how polysemy emerges from contextual processing of word meaning from both a theoretical and computational point of view.     

Evolution of word meanings through metaphorical mapping: Systematicity over the past millennium

One way that languages are able to communicate a potentially infinite set of ideas through a finite lexicon is by compressing emerging meanings into words, such that over time, individual words come to express multiple, related senses of meaning. We propose that overarching communicative and cognitive pressures have created systematic directionality in how new metaphorical senses have developed from existing word senses over the history of English. Given a large set of pairs of semantic domains, we used computational models to test which domains have been more commonly the starting points (source domains) and which the ending points (target domains) of metaphorical mappings over the past millennium. We found that a compact set of variables, including externality, embodiment, and valence, explain directionality in the majority of about 5000 metaphorical mappings recorded over the past 1100 years. These results provide the first large-scale historical evidence that metaphorical mapping is systematic, and driven by measurable communicative and cognitive principles.     

生命意义感对大学新生日常烦心事和心理适应的影响

为验证日常烦心事对个体心理适应的作用过程,并探讨生命意义感是否在烦心事知觉与心理适应之间起调节作用。研究采用纵向追踪方式对153名大学新生进行新近生活经历问卷、压力知觉量表、生命意义感量表和心理适应的测量。结果显示压力知觉在日常烦心事与心理适应之间起完全中介作用;拥有意义在烦心事知觉与心理适应之间起调节作用,但追求意义的调节作用不显著。这说明日常烦心事会通过压力知觉影响个体心理适应,拥有意义对个体心理适应具有调节作用。     

Modelling the effects of semantic ambiguity in word recognition

Most words in English are ambiguous between different interpretations; words can mean different things in different contexts. We investigate the implications of different types of semantic ambiguity for connectionist models of word recognition. We present a model in which there is competition to activate distributed semantic representations. The model performs well on the task of retrieving the different meanings of ambiguous words, and is able to simulate data reported by Rodd, Gaskell, and Marslen-Wilson [J. Mem. Lang. 46 (2002) 245] on how semantic ambiguity affects lexical decision performance. In particular, the network shows a disadvantage for words with multiple unrelated meanings (e.g., bark) that coexists with a benefit for words with multiple related word senses (e.g., twist). The ambiguity disadvantage arises because of interference between the different meanings, while the sense benefit arises because of differences in the structure of the attractor basins formed during learning. Words with few senses develop deep, narrow attractor basins, while words with many senses develop shallow, broad basins. We conclude that the mental representations of word meanings can be modelled as stable states within a high-dimensional semantic space, and that variations in the meanings of words shape the landscape of this space.     

The meaningful life in Japan and the United States: Levels and correlates of meaning in life

Culture supplies people with the provisions to derive meaning from life. However, no research has examined cultural variation in the two principal dimensions of meaning in life, presence of meaning and search for meaning. The present investigation adapted theories of self-concept and cognitive styles to develop a dialectical model of meaning in life, which predicted cultural differences in the tendency to experience search for meaning as opposed to, or harmonious with, presence of meaning. Using data from American (N = 1183) and Japanese (N = 982) young adults, mean levels and correlates of presence of meaning and search for meaning were examined. As predicted, Americans reported greater presence of meaning; Japanese reported greater search for meaning. In accordance with the model, search for meaning was negatively related to presence of meaning and well-being in the United States (opposed) and positively related to these variables in Japan (harmonious). Thus, the search for meaning appears to be influenced by culture, and search for meaning appears to moderate cultural influences on presence of meaning.     

Meaning making coping,making sense,and post-traumatic growth following the 9/11 terrorist attacks

According to the meaning making model, traumatic events disrupt global meaning systems, and meaning making coping helps restore congruency between global meaning and appraisals of traumatic events. We examined the contextual and coping predictors of two specific meanings made: having made sense and perceiving post-traumatic growth in a nationally representative sample of 1004 adults approximately six weeks after the September 11th terrorist attacks in the United States. Although the two meanings made were positively correlated, they had very different predictors: having made sense was primarily predicted by some aspects of meaning focused coping as well as other typically adaptive coping (e.g., active coping), but perceived growth was predicted by nearly all types of coping. Further, having made sense was related to less distress, but perceived growth was related to more distress. These results suggest that having made sense is a product of meaning focused and active coping and appears adaptive, while post-traumatic growth reflects a mix of positive functioning and continued distress and coping efforts.