Abstract:Expectation-based theories of language comprehension, in particular Surprisal Theory, go a long way in accounting for the behavioral correlates of word-by-word processing difficulty, such as reading times. An open question, however, is in which component(s) of the Event-Related brain Potential (ERP) signal Surprisal is reflected, and how these electrophysiological correlates relate to behavioral processing indices. Here, we address this question by instantiating an explicit neurocomputational model of incremen… Show more
“…The complementary nature of these meaning representations is underlined by recent advances in the neurocognition of language, where evidence suggests that lexical retrieval (the mapping of words onto lexical semantics) and semantic integration (the integration of word meaning into the unfolding representation of propositional meaning) are two distinct processes involved in word-by-word sentence processing (see [44,45] for explicit neurocognitive models). More specifically, this neurocognitive perspective on comprehension suggests that there is no compositionality at the lexical level; that is, word forms in context are mapped into word meaning representations, which are subsequently integrated into a compositional phrasal-/utterance-level meaning representation.…”
Section: Dfs and Distributional Semantics Offer Complementary Meaning Representationsmentioning
Natural language semantics has recently sought to combine the complementary strengths of formal and distributional approaches to meaning. More specifically, proposals have been put forward to augment formal semantic machinery with distributional meaning representations, thereby introducing the notion of semantic similarity into formal semantics, or to define distributional systems that aim to incorporate formal notions such as entailment and compositionality. However, given the fundamentally different 'representational currency' underlying formal and distributional approaches-models of the world versus linguistic co-occurrence-their unification has proven extremely difficult.Here, we define a Distributional Formal Semantics that integrates distributionality into a formal semantic system on the level of formal models. This approach offers probabilistic, distributed meaning representations that are also inherently compositional, and that naturally capture fundamental semantic notions such as quantification and entailment. Furthermore, we show how the probabilistic nature of these representations allows for probabilistic inference, and how the information-theoretic notion of "information" (measured in terms of Entropy and Surprisal) naturally follows from it. Finally, we illustrate how meaning representations can be derived incrementally from linguistic input using a recurrent neural network model, and how the resultant incremental semantic construction procedure intuitively captures key semantic phenomena, including negation, presupposition, and anaphoricity.
“…The complementary nature of these meaning representations is underlined by recent advances in the neurocognition of language, where evidence suggests that lexical retrieval (the mapping of words onto lexical semantics) and semantic integration (the integration of word meaning into the unfolding representation of propositional meaning) are two distinct processes involved in word-by-word sentence processing (see [44,45] for explicit neurocognitive models). More specifically, this neurocognitive perspective on comprehension suggests that there is no compositionality at the lexical level; that is, word forms in context are mapped into word meaning representations, which are subsequently integrated into a compositional phrasal-/utterance-level meaning representation.…”
Section: Dfs and Distributional Semantics Offer Complementary Meaning Representationsmentioning
Natural language semantics has recently sought to combine the complementary strengths of formal and distributional approaches to meaning. More specifically, proposals have been put forward to augment formal semantic machinery with distributional meaning representations, thereby introducing the notion of semantic similarity into formal semantics, or to define distributional systems that aim to incorporate formal notions such as entailment and compositionality. However, given the fundamentally different 'representational currency' underlying formal and distributional approaches-models of the world versus linguistic co-occurrence-their unification has proven extremely difficult.Here, we define a Distributional Formal Semantics that integrates distributionality into a formal semantic system on the level of formal models. This approach offers probabilistic, distributed meaning representations that are also inherently compositional, and that naturally capture fundamental semantic notions such as quantification and entailment. Furthermore, we show how the probabilistic nature of these representations allows for probabilistic inference, and how the information-theoretic notion of "information" (measured in terms of Entropy and Surprisal) naturally follows from it. Finally, we illustrate how meaning representations can be derived incrementally from linguistic input using a recurrent neural network model, and how the resultant incremental semantic construction procedure intuitively captures key semantic phenomena, including negation, presupposition, and anaphoricity.
“…For example, when the model processes "a boy plays soccer", it does not only recover the explicit literal propositional content, but it also constructs a more complete situation model, in which a boy is likely to be playing outside, on a field, with a ball, etc. In this way, it differs from other connectionist models of language processing, that typically employ simpler meaning representations, such as case-roles (e.g., [26][27][28][29]). Crucially, Frank et al [23]'s model generalizes to sentences and meaning representations that it has not seen during training, exhibiting different levels of semantic systematicity.…”
Decades of studies trying to define the extent to which artificial neural networks can exhibit systematicity suggest that systematicity can be achieved by connectionist models but not by default. Here we present a novel connectionist model of sentence production that employs rich situation model representations originally proposed for modeling systematicity in comprehension. The high performance of our model demonstrates that such representations are also well suited to model language production. Furthermore, the model can produce multiple novel sentences for previously unseen situations, including in a different voice (actives vs. passive) and with words in new syntactic roles, thus demonstrating semantic and syntactic generalization and arguably systematicity. Our results provide yet further evidence that such connectionist approaches can achieve systematicity, in production as well as comprehension. We propose our positive results to be a consequence of the regularities of the microworld from which the semantic representations are derived, which provides a sufficient structure from which the neural network can interpret novel inputs.
“…This Surprisal measure is influenced by both linguistic experience, as well as knowledge about the world [ 5 ]. As Brouwer et al [ 48 ] point out, this view of the P600 as reflecting comprehension-centric Surprisal follows from the RI Theory. Just as syntactic models determine the likelihood of alternative analyses based on linguistic experience, the RI model recovers interpretations that reflect the distributional characteristics of the utterances it is exposed to [ 48 ].…”
Section: Introductionmentioning
confidence: 99%
“…In fact, the decomposition of language comprehension into retrieval and integration is made even more explicit in the computational instantiation of RI theory. In this model, retrieval is instantiated by the function which maps an incoming orthographic/acoustic word form onto a representation of word meaning , while taking the unfolding utterance context —the utterance meaning constructed prior to the current word—into account [ 48 ]. The output of this function serves as input to the function which serves to integrate the retrieved word meaning into the unfolding utterance context , to produce an updated utterance meaning .…”
Section: Introductionmentioning
confidence: 99%
“…The most recent instantiation of RI theory thus predicts the P600 component of the ERP signal, which indexes the amount of effort involved in updating the unfolding utterance meaning representation with the retrieved meaning of an incoming word, to be the locus that is specifically sensitive to expectancy/Surprisal effects [ 48 ] and insensitive to association effects. That is, integration effort is assumed to increase to the extent that the utterance meaning representation resulting from integrating this word meaning is semantically, pragmatically, or structurally unexpected, given the utterance meaning representation prior to integration.…”
Expectation-based theories of language processing, such as Surprisal theory, are supported by evidence of anticipation effects in both behavioural and neurophysiological measures. Online measures of language processing, however, are known to be influenced by factors such as lexical association that are distinct from—but often confounded with—expectancy. An open question therefore is whether a specific locus of expectancy related effects can be established in neural and behavioral processing correlates. We address this question in an event-related potential experiment and a self-paced reading experiment that independently cross expectancy and lexical association in a context manipulation design. We find that event-related potentials reveal that the N400 is sensitive to both expectancy and lexical association, while the P600 is modulated only by expectancy. Reading times, in turn, reveal effects of both association and expectancy in the first spillover region, followed by effects of expectancy alone in the second spillover region. These findings are consistent with the Retrieval-Integration account of language comprehension, according to which lexical retrieval (N400) is facilitated for words that are both expected and associated, whereas integration difficulty (P600) will be greater for unexpected words alone. Further, an exploratory analysis suggests that the P600 is not merely sensitive to expectancy violations, but rather, that there is a continuous relation. Taken together, these results suggest that the P600, like reading times, may reflect a meaning-centric notion of Surprisal in language comprehension.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.