Making sense of word senses: the comprehension of polysemy depends on sense overlap

Studies of polysemy are few in number and are contradictory. Some have found differences between polysemy and homonymy (L. Frazier & K. Rayner, 1990), and others have found similarities (D. K. Klein & G. Murphy, 2001). The authors investigated this issue using the methods of D. K. Klein and G. Murphy (2001), in whose study participants judged whether ambiguous words embedded in word pairs (e.g., tasty chicken) made sense as a function of a cooperating, conflicting, or neutral context. The ambiguous words were independently rated as having low, moderate, or highly overlapping senses to approximate a continuum from homonymy to metonymic polysemy. The effects of meaning dominance were examined. Words with highly overlapping meanings (e.g., metonymy) showed reduced effects of context and dominance compared with words with moderately or low overlapping meanings (e.g., metaphorical polysemy and homonymy). These results suggest that the comprehension of ambiguous words is mediated by the semantic overlap of alternative senses/meanings.     

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.     

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.     

The power of the written word: Comparing word meaning inferencing from auditory and written modalities throughout grade school

Previous research indicates that word learning from auditory contexts may be more effective than written context at least through fourth grade. However, no study has examined contextual differences in word learning in older school-aged children when reading abilities are more developed. Here we examined developmental differences in children’s ability to deduce the meanings of unknown words from the surrounding linguistic context in the auditory and written modalities and sought to identify the most important predictors of success in each modality. A total of 89 children aged 8–15 years were randomly assigned to either read or listen to a narrative that included eight novel words, with five exposures to each novel word. They then completed three posttests to assess word meaning inferencing. Children across all ages performed better in the written modality. Vocabulary was the only significant predictor of success on the word inferencing task. Results indicate support for written stimuli as the most effective modality for novel word meaning deduction. Our findings suggest that the presence of orthographic information facilitates novel word learning even for early, less proficient readers.     

A probabilistic computational model of cross-situational word learning

Words are the essence of communication: They are the building blocks of any language. Learning the meaning of words is thus one of the most important aspects of language acquisition: Children must first learn words before they can combine them into complex utterances. Many theories have been developed to explain the impressive efficiency of young children in acquiring the vocabulary of their language, as well as the developmental patterns observed in the course of lexical acquisition. A major source of disagreement among the different theories is whether children are equipped with special mechanisms and biases for word learning, or their general cognitive abilities are adequate for the task. We present a novel computational model of early word learning to shed light on the mechanisms that might be at work in this process. The model learns word meanings as probabilistic associations between words and semantic elements, using an incremental and probabilistic learning mechanism, and drawing only on general cognitive abilities. The results presented here demonstrate that much about word meanings can be learned from naturally occurring child-directed utterances (paired with meaning representations), without using any special biases or constraints, and without any explicit developmental changes in the underlying learning mechanism. Furthermore, our model provides explanations for the occasionally contradictory child experimental data, and offers predictions for the behavior of young word learners in novel situations.     

Learning new vocabulary during childhood: Effects of semantic training on lexical consolidation and integration

Research suggests that word learning is an extended process, with offline consolidation crucial for the strengthening of new lexical representations and their integration with existing lexical knowledge (as measured by engagement in lexical competition). This supports a dual memory systems account, in which new information is initially sparsely encoded separately from existing knowledge and integrated with long-term memory over time. However, previous studies of this type exploited unnatural learning contexts, involving fictitious words in the absence of word meaning. In this study, 5- to 9-year-old children learned real science words (e.g., hippocampus) with or without semantic information. Children in both groups were slower to detect pauses in familiar competitor words (e.g., hippopotamus) relative to control words 24 h after training but not immediately, confirming that offline consolidation is required before new words are integrated with the lexicon and engage in lexical competition. Children recalled more new words 24 h after training than immediately (with similar improvements shown for the recall and recognition of new word meanings); however, children who were exposed to the meanings during training showed further improvements in recall after 1 week and outperformed children who were not exposed to meanings. These findings support the dual memory systems account of vocabulary acquisition and suggest that the association of a new phonological form with semantic information is critical for the development of stable lexical representations.     

Meaning construction and integration in children with hydrocephalus

Text comprehension processes were investigated in children with hydrocephalus, a neurodevelopmental disorder associated with good word decoding, but deficient reading comprehension. In Experiment 1, hydrocephalus and control groups were similar in processes related to activating word meanings and using context to enhance meaning. The hydrocephalus group was poorer at suppressing contextually irrelevant meanings. In Experiment 2, the hydrocephalus group had difficulty integrating information from an earlier read sentence to understand a new sentence as textual distance between the two propositions increased, suggesting difficulty in reactivation processes related to comprehension. Results are discussed in relation to cognitive and neurocognitive models of comprehension.     

Constraints Children Place on Word Meanings

This paper views lexical acquisition as a problem of induction: Children must figure out the meaning of a given term, given the large number of possible meanings any term could have. If children had to consider, evaluate, and rule out an unlimited number of hypotheses about each word in order to figure out its meaning, learning word meanings would be hopeless. Children must, therefore, be limited in the kinds of hypotheses they consider as possible word meanings. This paper considers three possible constraints on word meanings: (1) The whole object assumption which leads children to interpret novel terms as labels for objects—not parts, substances, or other properties of objects: (2) The taxonomic assumption which leads children to consider labels as referring to objects of like kind, rather than to objects that are thematically related: and (3) The mutual exclusivity assumption which leads children to expect each object to have only one label. Some of the evidence for these constraints is reviewed.     

Capacities underlying word learning

Children are strikingly good at learning the meanings of words. Current controversy focuses on the relative importance of different capacities in this learning process including principles of association, low-level attentional mechanisms, special word learning constraints, syntactic cues and theory of mind. We argue that children succeed at word learning because they possess certain conceptual biases about the external world, the ability to infer the referential intentions of others and an appreciation of syntactic cues to word meaning. Support for this view comes from studies exploring the phenomena of fast mapping, the whole object bias, the acquisition of names for entities belonging to different ontological kinds and the effect of lexical contrast. Word learning is not the result of a general associative learning process, nor does it involve specialized constraints. The ability to learn the meanings of words depends on a number of capacities, some of which are specific to language and unique to humans, others of which are potentially shared with other species.     

Poor reading comprehension despite fast word decoding in children with hydrocephalus

Children with hydrocephalus decode words better than they understand what they read. We tested whether children with hydrocephalus (from myelomeningocele or aqueduct stenosis) (1) decode words slowly, (2) use decoding processes similar to those of neurologically intact peers, and (3) comprehend poorly to the extent that they are slow decoders. We compared speed of word decoding in 33 children with hydrocephalus and 33 controls matched on a pairwise basis for age, grade, and word decoding accuracy. The children with hydrocephalus were as fast as controls in reading words, but, unlike controls, they did not demonstrate an effect of spelling-sound regularity. Further, decoding speed did not contribute to reading comprehension beyond word decoding accuracy. The reading comprehension deficits of good decoders with hydrocephalus are not related to early-stage processing deficits in word recognition speed. Likely origins of comprehension failure in this group are discussed.     

Just a talking book? Word learning from watching baby videos

This study examined the relationship between viewing an infant DVD and expressive and receptive language outcomes. Children between 12 and 15 months were randomly assigned to view Baby Wordsworth, a DVD highlighting words around the house marketed for children beginning at 12 months of age. Viewings took place in home settings over 6 weeks. After every 2 weeks and five exposures to the DVD, children were assessed on expressive and receptive communication measures. Results indicated there was no increased growth on either outcome for children who had viewed the DVD as compared to children in the control group, even after multiple exposures. After controlling for age, gender, cognitive developmental level, income, and parent education, the most significant predictor of vocabulary comprehension and production scores was the amount of time children were read to.     

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.     

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.     

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.     

Eye movements,word familiarity,and vocabulary acquisition

Two experiments examined the effects of word familiarity on word recognition and text comprehension during silent reading. Readers' eye movements were monitored as they read sentences containing words that varied in familiarity as assessed by printed estimates of word frequency, subjective ratings of familiarity, and a multiple‐choice test of meaning knowledge. Effects of word frequency were unaffected by differences in subjective familiarity rating for high frequency words. Differential effects of familiarity rating were observed in low frequency conditions. In addition, processing time on high and low frequency words did not differ when familiarity was held constant for moderately familiar words. Readers spent more initial processing time on novel words than familiar words. Performance on a vocabulary test administered after the reading session demonstrated that readers successfully acquired and retained new word meanings. Finally, reanalysis of word processing time as a function of vocabulary test performance demonstrated a systematic relationship between online processing patterns and memory for novel word meaning.     

Context effects in sentence comprehension: A study of the subjective lexicon

These studies explore the role of context in determining what information about the meanings of words is activated in memory at the time a word is encountered in a sentence. Using a color-naming paradigm, it was shown that both meanings of a word that has two distinct meanings are activated in memory at the time the word is heard in a sentence. This activation occurs even when there is sufficient contextual information to indicate which meaning was intended by the speaker. These results support the hypothesis that there exists in memory an isolable subjective lexicon. They suggest that context which is effective in disambiguating lexical ambiguities in the language has its effect only at a relatively late stage in the cognitive processing involved in language comprehension.     

Children's Use of Generics in Inductive Inferences

Online comprehension of naturally spoken and perceptually degraded words was assessed in 95 children ages 12 to 31 months. The time course of word recognition was measured by monitoring eye movements as children looked at pictures while listening to familiar target words presented in unaltered, time-compressed, and low-pass-filtered forms. Success in word recognition varied with age and level of vocabulary development, and with the perceptual integrity of the word. Recognition was best overall for unaltered words, lower for time-compressed words, and significantly lower in low-pass-filtered words. Reaction times were fastest in compressed, followed by unaltered and filtered words. Results showed that children were able to recognize familiar words in challenging conditions and that productive vocabulary size was more sensitive than chronological age as a predictor of children's accuracy and speed in word recognition.     

Dynamics of Word Comprehension in Infancy: Developments in Timing, Accuracy, and Resistance to Acoustic Degradation

Online comprehension of naturally spoken and perceptually degraded words was assessed in 95 children ages 12 to 31 months. The time course of word recognition was measured by monitoring eye movements as children looked at pictures while listening to familiar target words presented in unaltered, time-compressed, and low-pass-filtered forms. Success in word recognition varied with age and level of vocabulary development, and with the perceptual integrity of the word. Recognition was best overall for unaltered words, lower for time-compressed words, and significantly lower in low-pass-filtered words. Reaction times were fastest in compressed, followed by unaltered and filtered words. Results showed that children were able to recognize familiar words in challenging conditions and that productive vocabulary size was more sensitive than chronological age as a predictor of children's accuracy and speed in word recognition.     

Idiom comprehension deficits in relation to corpus callosum agenesis and hypoplasia in children with spina bifida meningomyelocele

Idioms are phrases with figurative meanings that are not directly derived from the literal meanings of the words in the phrase. Idiom comprehension varies with: literality, whether the idiom is literally plausible; compositionality, whether individual words contribute to a figurative meaning; and contextual bias. We studied idiom comprehension in children with spina bifida meningomyelocele (SBM), a neurodevelopmental disorder associated with problems in discourse comprehension and agenesis and hypoplasia of the corpus callosum. Compared to age peers, children with SBM understood decomposable idioms (which are processed more like literal language) but not non-decomposable idioms (which require contextual analyses for acquisition). The impairment in non-decomposable idioms was related to congenital agenesis of the corpus callosum, which suggests that the consequences of impaired interhemispheric communication, whether congenital or acquired in adulthood, are borne more by configurational than by compositional language.     

Making Sense of Compound Nouns: A Study of Word Relatedness in Telugu

Lexical development is typically viewed as elaboration, differentiation, and integration of semantic codes—codes that signify the meanings embodied in the words. Our earlier work based on 50 noun + noun (NN) compounds in Telugu has shown that children in the age group 8–14 years exhibit clear-cut developmental trends in producing and segmenting the NN compound nouns and generating words that are related in meaning to the target compounds. The database for the present study is drawn from our earlier work, and it consists of 1800 words reported to be related in meaning to the 50 target compound nouns by 36 children (12 III grade children, 12 VI grade children, and 12 IX grade children) and 600 words produced by 12 adults. A thorough analysis of the individual word associations generated by the subjects revealed that children tended to generate: (1) compounds with the same head word as the target word but with a new modifier word, (2) novel compounds that have phonetic/phonological association with the target words, most of which are actually nonwords in the language; and (3) new single-stem nouns and new compounds that are considerably fewer in number than those produced by adult subjects. Some of the theoretical and pedagogical implications of the differences in performance of children vs. adult subjects in the encoding of word meanings in an experimental context are discussed in this paper.