Parsing to Learn

Learning a language by parameter setting is almost certainly less onerous than composing a grammar from scratch. But recent computational modeling of how parameters are set has shown that it is not at all the simple mechanical process sometimes imagined. Sentences must be parsed to discover the properties that select between parameter values. But the sentences that drive learning cannot be parsed with the learner's current grammar. And there is not much point in parsing them with just one new grammar. They must apparently be parsed with all possible grammars, in order to find out which one is most successful at licensing the language. The research task is to reconcile this with the fact that the human sentence parsing mechanism, even in adults, has only very limited parallel parsing capacity. I have proposed that all possible grammars can be folded into one, if parameter values are fragments of sentential tree structures that the parser can make use of where necessary to assign a structure to an input sentence. However, the problem of capacity limitations remains. The combined grammar will afford multiple analyses for some sentences, too many to be computed on-line. I propose that the parser computes only one analysis per sentence but can detect ambiguity, and that the learner makes use of unambiguous input only. This provides secure information but relatively little of it, particularly at early stages of learning where few grammars have been excluded and ambiguity is rife. I consider three solutions: improving the parser's ability to extract unambiguous information from partially ambiguous sentences, assuming default parameter values to temporarily eliminate ambiguity, reconfiguring the parameters so that some are subordinate to others and do not present themselves to the learner until the others have been set. A more radical alternative is to give up the quest for error-free learning and permit parameters to be set without regard for whether the parser may have overlooked an alternative analysis of the sentence. If it can be assumed that the human parser keeps a running tally of the parameter values it has accessed, then the learner would do nothing other than parse sentences for comprehension, as adults do. The most useful parameter values would become more and more easily accessed; the noncontributors would drop out of the running. There would be no learning mechanism at all, over and above the parser. But how accurate this system would be remains to be established.     

Prediction and substantiation: A new approach to natural language processing

This paper describes a new approach to natural language processing which results in a very robust and efficient system. The approach taken is to integrate the parser with the rest of the system. This enables the parser to benefit from predictions that the rest of the system makes in the course of its processing. These predictions can be invaluable as guides to the parser in such difficult problem areas as resolving referents and selecting meanings of ambiguous words. A program, called FRUMP for Fast Reading Understanding and Memory Program, employs this approach to parsing. FRUMP skims articles rather than reading them for detail. The program works on the relatively unconstrained domain of news articles. It routinely understands stories it has never before seen. The program's success is largely due to its radically different approach to parsing.     

Memory requirements and local ambiguities of parsing strategies

We present a method for calculating lower bounds on the space required and local ambiguities entailed by parsing strategies. A fast, compact natural language parser must implement a strategy with low space requirements and few local ambiguities. It is also widely assumed in the psycholinguistics literature that extremely limited short-term space is available to the human parser, and that sentences containing center-embedded constructions are incomprehensible because processing them requires more space than is available. However, we show that the parsing strategies most psycholinguists assume require less space for processing center-embedded constructions than for processing other perfectly comprehensible constructions. We present alternative strategies for which center-embedded constructions do require more space than other constructions.     

Early language and the development of children's reading skills

There is considerable agreement that vocabulary plays a central role in reading acquisition, but there is less agreement about whether this association is direct or indirect through phonological and print-related knowledge. Longitudinal data from 77 African-American children were analyzed to examine the relationship between language skills, phonological knowledge and print processing skills at pre-kindergarten and kindergarten with reading at pre-kindergarten through second grade. Analyses indicated that home and child care experiences were related to reading indirectly through language and that language and phonological knowledge were both directly related to acquisition of reading skill. This study of African-American children and previous studies of lower- and middle-income children indicate both language and phonological skills play an important role in children becoming successful readers, and that experiences at home and in child care during in early childhood play a role in the acquisition of reading through their enhancement of early language and phonological skills.     

Asynchronous language acquisition in developmental dysphasia

A longitudinal study was conducted to document and compare evolution of children with linguistic acquisition impairment. To determine whether development of the analytic mechanisms underlying linguistic processing occured in similar fashion, two children with mixed developmental dysphasia were assessed from 4 to 5:6 years of age with psycholinguistic tests at 6-months interval. Spontaneous speech and language production (consonant repertory in initial word position, MLU, and lexical diversity) were investigated in a standardized symbolic play context. The phonologic and lexico-morphologic evolution analyses revealed a marked improvement in motor control of phonology and in the application of morphosyntaxic rules in child 1, whereas child 2 was still impaired in phonology and morphosyntax. The singular developmental changes in spontaneous speech results indicate dynamic relationships between various language production facets and variability in the kind of deficit and lexical automation presented by these children. These contrasts in the evolution of language production profiles between child 1 and child 2 also underline the importance of longitudinal studies in the analysis of the atypical linguistic processing paths used by children with developmental dysphasia.     

The kindergarten-path effect: studying on-line sentence processing in young children

A great deal of psycholinguistic research has focused on the question of how adults interpret language in real time. This work has revealed a complex and interactive language processing system capable of rapidly coordinating linguistic properties of the message with information from the context or situation (e.g. Altmann & Steedman, 1988; Britt, 1994; Tanenhaus, Spivey-Knowlton, Eberhard & Sedivy, 1995; Trueswell & Tanenhaus, 1991). In the study of language acquisition, however, surprisingly little is known about how children process language in real time and whether they coordinate multiple sources of information during interpretation. The lack of child research is due in part to the fact that most existing techniques for studying language processing have relied upon the skill of reading, an ability that young children do not have or are only beginning to acquire. We present here results from a new method for studying children's moment-by-moment language processing abilities, in which a head-mounted eye-tracking system was used to monitor eye movements as participants responded to spoken instructions. The results revealed systematic differences in how children and adults process spoken language: Five Year Olds did not take into account relevant discourse/pragmatic principles when resolving temporary syntactic ambiguities, and showed little or no ability to revise initial parsing commitments. Adults showed sensitivity to these discourse constraints at the earliest possible stages of processing, and were capable of revising incorrect parsing commitments. Implications for current models of sentence processing are discussed.     

A probabilistic model of semantic plausibility in sentence processing

Experimental research shows that human sentence processing uses information from different levels of linguistic analysis, for example, lexical and syntactic preferences as well as semantic plausibility. Existing computational models of human sentence processing, however, have focused primarily on lexico-syntactic factors. Those models that do account for semantic plausibility effects lack a general model of human plausibility intuitions at the sentence level. Within a probabilistic framework, we propose a wide-coverage model that both assigns thematic roles to verb–argument pairs and determines a preferred interpretation by evaluating the plausibility of the resulting ( verb , role , argument ) triples. The model is trained on a corpus of role-annotated language data. We also present a transparent integration of the semantic model with an incremental probabilistic parser. We demonstrate that both the semantic plausibility model and the combined syntax/semantics model predict judgment and reading time data from the experimental literature.     

Disfluencies, language comprehension, and Tree Adjoining Grammars

Disfluencies include editing terms such as uh and um as well as repeats and revisions. Little is known about how disfluencies are processed, and there has been next to no research focused on the way that disfluencies affect structure-building operations during comprehension. We review major findings from both computational linguistics and psycholinguistics, and then we summarize the results of our own work which centers on how the parser behaves when it encounters a disfluency. We describe some new research showing that information associated with misarticulated verbs lingers, and which adds to the large body of data on the critical influence of verb argument structures on sentence comprehension. The paper also presents a model of disfluency processing. The parser uses a Tree Adjoining Grammar to build phrase structure. In this approach, filled and unfilled pauses affect the timing of Substitution operations. Repairs and corrections are handled by a mechanism we term “Overlay,” which allows the parser to overwrite an undesired tree with the appropriate, correct tree. This model of disfluency processing highlights the need for the parser to sometimes coordinate the mechanisms that perform garden-path reanalysis with those that do disfluency repair. The research program as a whole demonstrates that it is possible to study disfluencies systematically and to learn how the parser handles filler material and mistakes. It also showcases the power of Tree Adjoining Grammars, a formalism developed by Aravind Joshi which has yielded results in many different areas of linguistics and cognitive science.     

Speaking a tone language enhances musical pitch perception in 3–5‐year‐olds

Young children learn multiple cognitive skills concurrently (e.g., language and music). Evidence is limited as to whether and how learning in one domain affects that in another during early development. Here we assessed whether exposure to a tone language benefits musical pitch processing among 3–5‐year‐old children. More specifically, we compared the pitch perception of Chinese children who spoke a tone language (i.e., Mandarin) with English‐speaking American children. We found that Mandarin‐speaking children were more advanced at pitch processing than English‐speaking children but both groups performed similarly on a control music task (timbre discrimination). The findings support the Pitch Generalization Hypothesis that tone languages drive attention to pitch in nonlinguistic contexts, and suggest that language learning benefits aspects of music perception in early development. A video abstract of this article can be viewed at: https://youtu.be/UY0kpGpPNA0     

The Role of Language Skill in Child Psychopathology: Implications for Intervention in the Early Years

In this narrative review, we suggest that children’s language skill should be targeted in clinical interventions for children with emotional and behavioral difficulties in the preschool years. We propose that language skill predicts childhood emotional and behavioral problems and this relationship may be mediated by children’s self-regulation and emotion understanding skills. In the first sections, we review recent high-quality longitudinal studies which together demonstrate that that children’s early language skill predicts: (1) emotional and behavioral problems, and this relationship is stronger than the reverse pattern; (2) self-regulation skill; this pattern may be stronger than the reverse pattern but moderated by child age. Findings also suggest that self-regulation skill mediates the relation between early language skill and children’s emotional and behavioral problems. There is insufficient evidence regarding the mediating role of emotion understanding. In subsequent sections, we review evidence demonstrating that: (1) particular kinds of developmentally targeted parent–child conversations play a vital role in the development of language skill, and (2) some current clinical interventions, directly or indirectly, have a beneficial impact on children’s vocabulary and narrative skills, but most approaches are ad hoc. Targeting language via parent–child conversation has the potential to improve the outcomes of current clinical interventions in the preschool years.     

Predictors of early vocabulary growth in children born preterm and full term: A study of processing speed and medical complications

Delays in expressive vocabulary may be harbingers of long-term language difficulties. In toddlers born full term (FT), individual differences in language processing speed are associated with variation in expressive vocabulary growth. Children born preterm (PT) are at increased risk for persistent language deficits. Here, we evaluate predictors of early vocabulary growth in PT toddlers in relation to two sources of variability: language processing speed and medical complications of prematurity. Vocabulary growth from 16 to 30 months (adjusted for degree of prematurity) was modeled longitudinally using parent reports in English-speaking FT (n = 63; ≥37 weeks, ≥2495 g) and PT (n = 69; ≤32 weeks, <1800 g) children, matched on sex and socioeconomic status. Children were tested in the “looking-while-listening task” at 18 months to derive a measure of language processing speed. Each PT child was assessed for number of medical complications (13 maximum), based on medical chart reviews. PT and FT children displayed similar vocabulary trajectories; however, birth group disparities began to emerge by 30 months. PT children were slower in language processing speed than FT children. Critically, language processing speed predicted expressive vocabulary size at 30 months; interactions with birth group were not significant (all p > .20). In PT children, faster language processing speed predicted stronger outcomes regardless of number of medical complications; slower processing speed and more medical complications predicted poorer outcomes. Faster processing speed reflected favorable neuropsychological processes associated with faster expressive vocabulary growth that overrode the impact of medical complications on language outcomes in PT children.     

Language Acquisition is Language Change

According to the theory of Universal Grammar, the primary linguistic data guides children through an innately specified space of hypotheses. On this view, similarities between child-English and adult-German are as unsurprising as similarities between cousins who have never met. By contrast, experience-based approaches to language acquisition contend that child language matches the input, with nonadult forms being simply less articulated versions of the forms produced by adults. This paper reports several studies that provide support for the theory of Universal grammar, and resist explanation on experience-based accounts. Two studies investigate English-speaking children’s productions, and a third examines the interpretation of sentences by Japanese speaking children. When considered against the input children are exposed to, the findings of these and other studies are consistent with the continuity hypothesis, which supposes that child language can differ from the language spoken by adults only in ways that adult languages can differ from each other.     

Theories of early language acquisition

What features of brain processing and neural development support linguistic development in young children? To what extent is the profile and timing of linguistic development in young children determined by a pre-ordained genetic programme? Does the environment play a crucial role in determining the patterns of change observed in children growing up? Recent experimental, neuroimaging and computational studies of developmental change in children promise to contribute to a deeper understanding of how the brain becomes wired up for language. In this review, the muttidisciplinary perspectives of cognitive neuroscience, experimental psycholinguistics and neural network modelling are brought to bear on four distinct areas in the study of language acquisition: early speech perception, word recognition, word learning and the acquisition of grammatical inflections. It is suggested that each area demonstrates how linguistic development can be driven by the interaction of general learning mechanisms, highly sensitive to particular statistical regularities in the input, with a richly structured environment which provides the necessary ingredients for the emergence of linguistic representations that support mature language processing. Similar epigenetic principles, guiding the emergence of linguistic structure, apply to all these domains, offering insights into phenomena ranging from the precocity of young infant's sensitivity to speech contrasts to the complexities of the problem facing the young child learning the arabic plural.     

When learners surpass their models: the acquisition of American Sign Language from inconsistent input

The present study examines the impact of highly inconsistent input on language acquisition. The American deaf community provides a unique opportunity to observe children exposed to nonnative language models as their only linguistic input. This research is a detailed case study of one child acquiring his native language in such circumstances. It asks whether this child is capable of organizing a natural language out of input data that are not representative of certain natural language principles. Simon is a deaf child whose deaf parents both learned American Sign Language (ASL) after age 15. Simon's only ASL input is provided by his late-learner parents. The study examines Simon's performance at age 7 on an ASL morphology task, compared with eight children who have native signing parents, and also compared with Simon's own parents. The results show that Simon's production of ASL substantially surpasses that of his parents. Simon's parents, like other late learners of ASL, perform below adult native signing criteria, with many inconsistencies and errors in their use of ASL morphology. In contrast, Simon's performance is much more regular, and in fact on most ASL morphemes is equal to that of children exposed to a native signing model. The results thus indicate that Simon is capable of acquiring a regular and orderly morphological rule system for which his input provides only highly inconsistent and noisy data. In addition, the results provide some insight into the mechanisms by which such learning may occur. Although the ASL situation is rare, it reveals clues that may contribute to our understanding of the human capacity for language learning.     

Language facilitates event memory in early childhood: Child comprehension,adult-provided linguistic support and delayed recall at 16 months

Adult-provided supportive language facilitates memory for the past in preverbal and verbal children. Work conducted with 18-month-olds indicates that children benefit from supportive adult language when tested after a 4-week delay but not when tested immediately after sequence demonstration; moreover, findings reveal that supportive language provided only at test may be more facilitative of recall after a delay relative to supportive language provided only at encoding. In the present study, we examined whether child language comprehension abilities moderated the extent to which preverbal children benefitted from supportive language provided at encoding and test. The findings indicated that child language comprehension and supportive language provided at encoding were unassociated with performance at baseline or immediate imitation; however, the moderating effect of child language comprehension on adult-provided supportive language at encoding and test was observed after a 1-week delay. Correlations revealed continuous associations between general comprehension scores and recall performance after the 1-week delay on sequences presented in the most supportive condition at encoding. Taken together, the presented findings reveal that the complex interplay between language and cognition is established in early childhood, with foundational relations emerging before children are capable of verbally reporting on the past.     

Null subjects: a problem for parameter-setting models of language acquisition

Some languages, like English, require overt surface subjects, while others, like Italian and Spanish, allow "null" subjects. How does the young child determine whether or not her language allows null subjects? Modern parameter-setting theory has proposed a solution, in which the child begins acquisition with the null subject parameter set for either the English-like value or the Italian-like value. Incoming data, or the absence thereof, force a resetting of the parameter if the original value was incorrect. This paper argues that the single-value solution cannot work, no matter which value is chosen as the initial one, because of inherent limitations in the child's parser, and because of the presence of misleading input. An alternative dual-value solution is proposed, in which the child begins acquisition with both values available, and uses theory-confirmation procedures to decide which value is best supported by the available data.     

Social Cognition and Externalizing Psychopathology: An Investigation of the Mediating Role of Language

The present study integrates findings from three lines of research on the association of social cognition and externalizing psychopathology, language and externalizing psychopathology, and social cognition and language functioning using Structural Equation Modeling (SEM). To date these associations have been examined in pairs. A sample of 354 clinic-referred children (aged 7 to14 years) recruited from a children’s mental health centre were tested on measures of language, social cognition, working memory, and child psychopathology. We compared a hypothesized model presenting language functioning as a mediator of the association between social cognition and externalizing psychopathology to a model presenting the independent contribution of language and social cognition to externalizing psychopathology. As hypothesized, we found that the mediation model fits the data better than the alternative model. Our findings have implications for developing and modifying intervention techniques for children with dual language and externalizing psychopathology. This research was supported by a grant from Health Canada (NHRDP, 6606-4835) to the third author.