No full-text available
To read the full-text of this research,
you can request a copy directly from the author.
Abstract
This study uses self-organizing feature maps to model the acquisition of lexical and grammatical aspect. Previous research has identified a strong association between lexical aspect and grammatical aspect in child language, on the basis of which some researchers proposed innate semantic categories (Bickerton, 1984) or prelinguistic semantic space (Slobin, 1985). Our simulations indicate that this association can be modeled by self-organization and Hebbian learning principles in a feature-map model, without making particular assumptions about the structure of innate knowledge. In line with results from Li (1999), our study further attests to the utility of self-organizing neural networks in the study of language acquisition.
... Turkish, French, Italian, Polish, Greek, English) and were then encompassed to L2 acquisition research studies in English and Spanish (Bardovi-Harlig, 2000;Shirai, 1991;. Therefore, it could be said that the AH concentrates on the relationship between form and meaning and predicts the high relationships between lexical aspect and tense-aspect morphology usage since most theories distinguish between two types of aspects: lexical and grammatical (Li, 2000). It is assumed that lexical aspect traits influence the distribution of grammatical aspect morphemes in the early stages of development. ...
... They have been assigned to several aspectual classes based on the sort of event they signify. These lexical categories can be defined as follows (Haznedar, 2007;Li, 2000;): ...
... One of the aspects is the grammatical aspect, which relates to unique perspectives on certain circumstances. (Li, 2000). Johnson and Fay (2006, p.421) define the grammatical aspect as "a system for classifying utterances according to the perspective or viewpoint that conveys to the listener". ...
... Based on this theory, children and second language learners acquire language by starting with prototype of each category which share several characteristics with other members of that category (Gabriele, Martohardjono, & McClure, 2005). Most theories of tense and aspect introduce two sorts of aspect: lexical and grammatical (Li, 2000). As Johnson and Fay (2006) said several studies have shown a cooccurrence between lexical and grammatical aspects (e.g. ...
... One of the aspects mentioned in AH is grammatical aspect which refers to especial viewpoints of particular situations (Li, 2000). It is "a system for classifying utterances according to the perspective or viewpoint that convey to the listener" (Comrie, 1976& Smith, 1997, cited in Johnson & Fey, 2006. ...
... It is "a system for classifying utterances according to the perspective or viewpoint that convey to the listener" (Comrie, 1976& Smith, 1997, cited in Johnson & Fey, 2006. Therefore, whether a sentence presents an ongoing or completed action is a matter of grammatical aspect (Li, 2000). In English, bounded or perfective aspect focuses on the activity from outside which has a beginning and an ending, while unbounded or imperfective aspect looks at the event from the inside with no specific beginning or ending which is encoded with 'be' as the auxiliary verb and -ing which follows the main verb. ...
This study aims to investigate the claims of Aspect Hypothesis which holds that in the past tense, perfective verbs
precede imperfective verbs, having a different pattern of acquisition based on their lexical aspect, i.e., learners initiate using
perfective past marks from the prototypical verbs which contain the achievement and accomplishment lexical aspects, while they
start using imperfective past marks from the verbs whose lexical aspects are stative and activity. In this study, ten Iranian (from
Mashhad) Farsi speaking learners of English were given a film to watch and retell the story in impersonal narrative in past tense.
The verbs which they applied were classified into 4 lexical aspect categories with different semantic features like telicity,
durativity, and dynamicity; these categories include: state, achievement, accomplishment and activity. The results suggest that
atelic verbs like state and activity verbs were easier for the students to produce in past perfect and the atelic durative verbsactivities-
were produced with higher accuracy in past imperfective.
... It is a learning model in the sense that lexical representations of both languages can emerge from the statistical learning of the input speech. This property is similar to that of the SRN, but is based on our network in explicitly modeling lexical co-occurrences in the acquisition of word meanings (Li, 1999Li, , 2000 Farkaš & Li, 2001, in press). On the other hand, our model also has some of the representational features of BIMOLA: lexical forms are encoded by articulatory features of the phonemes of words (see also Li & MacWhinney, 2001). ...
... The design characteristics of the SOMBIP model are based on our self-organizing neural network model of language acquisition by young children (Li, 1999Li, , 2000 Farkaš & Li, 2001, in press). In recent years, self-organizing neural networks have become increasingly important for cognitive and perceptual studies (Hinton & Senjowski, 1999). ...
... One can also see a category of verbs indicating perceptual/mental activities in Chinese (marked as " Verbs-p " , including teng 'listen', tai 'look', gin 'see', seong 'think'). These effects of categorical emergence from statistical learning match up with results from our previous monolingual simulations (Li, 1999Li, , 2000 Farkaš & Li, in press). ...
Current connectionist models of bilingual language processing have been largely restricted to localist stationary models. To fully capture the dynamics of bilingual processing, we present SOMBIP, a self-organizing model of bilingual processing that has learning characteristics. SOMBIP consists of two interconnected self-organizing neural networks, coupled with a recurrent neural network that computes lexical co-occurrence constraints. Simulations with our model indicate that (1) the model can account for distinct patterns of the bilingual lexicon without the use of language nodes or language tags, (2) it can develop meaningful lexicalsemantic categories through self-organizing processes, (3) it can account for a variety of priming and interference effects based on associative pathways between phonology and semantics in the lexicon, and (4) it can explain lexical representation in bilinguals with different levels of proficiency and working memory capacity. These capabilities of our model are due to its design characteristics in that (a) it combines localist and distributed properties of processing, (b) it combines representation and learning, and (c) it combines lexicon and sentences in bilingual processing. Thus, SOMBIP serves as a new model of bilingual processing and provides a new perspective on connectionist bilingualism. It has the potential of explaining a wide variety of empirical and theoretical issues in bilingual research.
... A possible source of learners' knowledge of the telic vs. atelic contrast is the distributional bias in the target input. In the emergentist approach to the AH (Li 2000(Li , 2002, it is stressed that the early associations between aspectual morphemes and lexical categories are above all a consequence of learners' implicit capacity to analyze and record the probability of the co-occurrence of forms and meanings in the input they are exposed to. As noted, perfective morphemes are most often associated with telic verbs in the input. ...
The Aspect Hypothesis (AH) claims that L2 beginners use the perfective morpheme first with telic predicates (e.g., ‘arrive’, ‘build the house’) and only later with atelic ones (e.g., ‘know’, ‘work’). In contrast, the Lexical Underspecification Hypothesis (LUH) claims that beginners cannot represent the lexical aspect of L2 predicates (hence the telic vs. atelic distinction), because this distinction is a separate component of verb meaning. To investigate whether L2 learners distinguish between telic and atelic predicates, this study compares the responses from 299 L2 Italian learners (with different L1 backgrounds) and responses from 91 native speakers (NS) to the “ for / in + time span” adverbial test (Dowty 1979). The analysis shows that native speakers and L2 learners’ responses to the adverbial test diverge significantly, with learners’ proficiency and – to a lesser extent – L1 modulating their ratings. The results suggest that native speakers and beginning-intermediate L2 learners might not represent telicity alike, either because L2 aspectual competence is still developing or because beginning learners rely on the semantic representations of their L1. These findings support the predictions of the LUH and suggest caution when trying to assess learners’ aspectual representations.
... The discovery of nouns and verbs is enhanced to even higher levels when other assumptions (such as learning about co-occurrence relations among nonadjacent segments or using function words as markers of phrases to be analyzed) are added (Mintz 2005). Similarly, co-occurrences among words have been shown to also yield subcategories of nouns (animates versus inanimates) as well as subcategories of verbs (transitive versus intransitives, Elman 1990; Li 2000; Burgess and Lund 2000; See also Rogers and McClelland 2004). ...
Children learn how to learn language, and they get better as they go along. This article presents an overview of research inspired by a dynamic systems view of language learning that shows it to be a self-organizing process in which children create the units they need from the regularities present in the environment in which they are situated.
... Left-justified representations place emphasis on phonological similarities at the beginnings of words in linguistic processing (see Marslen-Wilson, 1987; Treiman & Zukowski, 1996 ), and this type of representations would be particularly appropriate, for example, for models that simulate the acquisition of prefixes (e.g., Li & MacWhinney, 1996 ), in which the initial contrasts of phonemes are at stake. Right-justified representations, on the other hand, are more appropriate for models that emphasize the phonological similarities at the ends of words in language acquisition (see Slobin, 1985), such as models of the acquisition of tense-aspect suffixes (Joanisse & Seidenberg, 1999; Li, 2000; Li & Shirai, 2000; MacWhinney, 1993; MacWhinney & Leinbach, 1991). In a right-justified representation, the examples inFig- ure 3 would thus look like those inFigure 4. Comparing this figure withFigure 3, we can clearly see the effects of left-versus right-justification on how phonological similarities are represented and captured (differently) in these vectors. ...
Much of the power of neural network modeling for language use and acquisition derives from a reliance on statistical regularities implicit in the phonological properties of words. Researchers have devised several methods for representing the phonology of words, but these methods are often either unable to represent realistically sized lexicons or inadequate in the ways they represent individual words. In this paper, we present a new phonological pattern generator (PatPho) that allows connectionist modelers to derive accurate phonological representations of the English lexicon. PatPho not only generates phonological patterns that can scale up to realistically sized lexicons, but also accurately and parsimoniously captures the similarity structures of the phonology of monosyllabic and multisyllabic words.
... Previous work by Li (2003), MacWhinney (2001a), Miikkulainen (1993Miikkulainen ( , 1997, and Ritter and Kohonen (1989) has shown that self-organizing neural networks, especially SOMs, are particularly suitable as models of the human lexicon. In our earlier work we used SOM to simulate language acquisition in various tasks: Li (1999Li ( , 2000 simulated the acquisition of lexical categories along with morphological acquisition (prefixes un-and dis-and suffixes -ing and -ed; see Li, 2003 for a summary); Farkas and Li (2001) modeled lexical category representation in an artificial corpus and a natural speech corpus based on parental input from the CHILDES database (MacWhinney, 2000); Farkas and Li (2002a) used growing nodes in SOM on the basis of the increasing vocabulary sizes during learning; Farkas and Li (2002b) modeled word confusion in production as a function of word frequency, word density, and rate of vocabulary increase; and Li and Farkas (2002) modeled lexical development in bilingual children. In all cases, the simulated patterns captured the development of basic linguistic categories from the statistical characteristics of the input data. ...
In this paper we present a self-organizing neural network model of early lexical development called DevLex. The network consists of two self-organizing maps (a growing semantic map and a growing phonological map) that are connected via associative links trained by Hebbian learning. The model captures a number of important phenomena that occur in early lexical acquisition by children, as it allows for the representation of a dynamically changing linguistic environment in language learning. In our simulations, DevLex develops topographically organized representations for linguistic categories over time, models lexical confusion as a function of word density and semantic similarity, and shows age-of-acquisition effects in the course of learning a growing lexicon. These results match up with patterns from empirical research on lexical development, and have significant implications for models of language acquisition based on self-organizing neural networks.
Deriving representations of meaning has been a long-standing problem in cognitive psychology and psycholinguistics. The lack of a m odel for representing semantic and grammatical knowledge has been a handicap in attempting to model the effects of semantic constraints in hum an syntactic processing. A computational model of high-dim ensional context space, the Hyperspace A nalogue to Language (H AL), is presented with a series of simulations modelling a variety of human empirical results. HAL learns its representations from the unsupervised processing of 300 million words of conversational text. W e propose that HAL's high-dim ensional context space can be used to (1) provide a basic categorisation of semantic and grammatical concepts, (2) model certain aspects of morphological ambiguity in verbs, and (3) provide an account of semantic context effects in syntactic processing. W e propose that the distributed and contextually derived representations that HAL acquires provide a basis for the subconceptual knowledge that can be used in accounting for a diverse set of cognitive phenomena.
Language research thrives on data collected from spontaneous interactions in naturally occurring situations. However, the process of collecting, transcribing, and analyzing naturalistic data can be extremely time-consuming and often unreliable. This book describes three basic tools for language analysis of transcript data by computer that have been developed in the context of the "Child Language Data Exchange System (CHILDES)" project. These are: the "CHAT" transcription and coding format, the "CLAN" package of analysis programs, and the "CHILDES" database. These tools have brought about significant changes in the way research is conducted in the child language field. They are being used with great success by researchers working with second language learning, adult conversational interactions, sociological content analyses, and language recovery in aphasia, as well as by students of child language development. The tools are widely applicable, although this book concentrates on their use in the child language field, believing that researchers from other areas can make the necessary analogies to their own topics. This thoroughly revised 2nd edition includes documentation on a dozen new computer programs that have been added to the basic system for transcript analysis. The most important of these new programs is the "CHILDES" Text Editor (CED) which can be used for a wide variety of purposes, including editing non-Roman orthographies, systematically adding codes to transcripts, checking the files for correct use of "CHAT," and linking the files to digitized audio and videotape. In addition to information on the new computer programs, the manual documents changed the shape of the "CHILDES/BIB" system--given a major update in 1994--which now uses a new computer database system. The documentation for the "CHILDES" transcript database has been updated to include new information on old corpora and information on more than a dozen new corpora from many different languages. Finally, the system of "CHAT" notations for file transcript have been clarified to emphasize the ways in which the codes are used by particular "CLAN" programs. The new edition concludes with a discussion of new directions in transcript analysis and links between the "CHILDES" database and other developments in multimedia computing and global networking. It also includes complete references organized by research topic area for the more than 300 published articles that have made use of the "CHILDES" database and/or the "CLAN" programs. LEA also distributes the "CLAN" programs and the complete "CHILDES" Database--including corpora from several languages and discourse situations--described in "The CHILDES Project." Be sure to choose the correct platform (IBM or Macintosh) for the "CLAN" programs; the "CHILDES" Database CD-ROM runs on both platforms.
It is hypothesized that creole languages are largely invented by children and show fundamental similarities, which derive from a biological program for language. The structures of Hawaiian Pidgin and Hawaiian Creole are contrasted, and evidence is provided to show that the latter derived from the former in a single generation. A realistic model of the processes of Creole formation shows how several specific historical and demographic factors interacted to restrict, in varying degrees, the access of pidgin speakers to the dominant language, and hence the nature of input to the children of those speakers. It is shown that the resulting similarities of Creole languages derive from a single grammar with a restricted list of categories and operations. However, grammars of individual Creoles will differ from this grammar to a varying extent: The degree of difference will correlate very closely with the quantity of dominant-language input, which in turn is controlled by extralinguistic factors. Alternative explanations of the above phenomena are surveyed, in particular, substratum theory and monogenesis: Both are found inadequate to account for the facts. Primary acquisition is examined in light of the general hypothesis, and it is suggested that the bioprogram provides a skeletal model of language which the child can then readily convert into the target language. Cases of systematic error and precocious learning provide indirect support for the hypothesis. Some conjectures are made concerning the evolutionary origins of the bioprogram and what study of Creoles and related topics might reveal about language origins.
Recent work in language acquisition has shown how linguistic form emerges from the operation of self-organizing systems. The emergentist framework emphasizes ways in which the formal structures of language emerge from the interaction of social patterns, patterns implicit in the input, and pressures arising from general aspects of the cognitive system. Emergentist models have been developed to study the acquisition of auditory and articulatory patterns during infancy and the ways in which the learning of the first words emerges from the linkage of auditory, articulatory, and conceptual systems. Neural network models have also been used to study the learning of inflectional markings and basic syntactic patterns. Using both neural network modeling and concepts from the study of dynamic systems, it is possible to analyze language learning as the integration of emergent dynamic systems.
While working on this project I have received institutional support of several kinds, for which I am most grateful. I thank the Institute for Advanced Study at Stanford University, and the Spencer Foundation. for a stimulating environment in which the basic idea of this book was developed. The Max Planck Institute for Psycho linguistics at Nijmegen enabled me to spend several months working on the the manuscript. A National Science Foundation grant to develop Discourse Representation Theory, and a grant from The University Research Institute of the University of Texas, also gave me time to pursue this project. I thank Helen Aristar-Dry for reading early drafts of the manuscript, Osten Dahl for penetrating remarks on a preliminary version, and my collaborator Gilbert Rappaport for relentless comments and questions throughout. People with whom I have worked on particular languages are mentioned in the relevant chapters. lowe a special debt of gratitude to the members of my graduate seminar on aspect in the spring of 1990: they raised many questions of importance which made a real differ ence to the final form of the theory. I have benefitted from presenting parts of this material publicly, including colloquia at the University of California at Berkeley, the University of California at San Diego, the University of Pennsylvania, Rice University, the University of Texas, and University of Tel Aviv. I thank Adrienne Diehr and Marjorie Troutner for their efficient and good-humored help throughout the work on the first edition.
This study reports three experiments on how children learning Mandarin Chinese comprehend and use aspect markers. These experiments examine the role of lexical aspect in children's acquisition of grammatical aspect. Results provide converging evidence for children's early sensitivity to (1) the association between atelic verbs and the imperfective aspect markers zai, -zhe, and -ne, and (2) the association between telic verbs and the perfective aspect marker -le. Children did not show a sensitivity in their use or understanding of aspect markers to the difference between stative and activity verbs or between semelfactive and activity verbs. These results are consistent with Slobin's (1985) basic child grammar hypothesis that the contrast between process and result is important in children's early acquisition of temporal morphology. In contrast, they are inconsistent with Bickerton's (1981, 1984) language bioprogram hypothesis that the distinctions between state and process and between punctual and nonpunctual are preprogrammed into language learners. We suggest new ways of looking at the results in the light of recent probabilistic hypotheses that emphasize the role of input, prototypes and connectionist representations.
This study reports three experiments on how children learning Mandarin Chinese comprehend and use aspect markers. These experiments examine the role of lexical aspect in children’s acquisition of grammatical aspect. Results provide converging evidence for children’s early sensitivity to (1) the association between atelic verbs and the imperfective aspect markers zai, -zhe, and -ne, and (2) the association between telic verbs and the perfective aspect marker -leChildren did not show a sensitivity in their use or understanding of aspect markers to the difference between stative and activity verbs or between semelfactive and activity verbs. These results are consistent with Slobin’s (1985) basic child grammar hypothesis that the contrast between process and result is important in children’s early acquisition of temporal morphology. In contrast, they are inconsistent with Bickerton’s (1981, 1984) language bioprogram hypothesis that the distinctions between state and process and between punctual and nonpunctual are preprogrammed into language learners. We suggest new ways of looking at the results in the light of recent probabilistic hypotheses that emphasize the role of input, prototypes and
In a recent issue of this journal, Pinker and Prince (1988) and Lachter and Bever (1988) presented detailed critiques of Rumelhart and McClelland's (1986) connectionist model of the child's learning of the phonological form of the English past tense. In order to address these criticisms, a new connectionist model was constructed using the back-propagation algorithm, a larger input corpus, a fuller paradigm, and a new phonological representation. This new implementation successfully addressed the criticisms of the phonological representation used by Rumelhart and McClelland. It did a much better job of learning the past tense using a fuller input set with realistic frequencies of occurrence. Ancillary simulations using the same network were able to deal with the homonymy problem and the generation of forms like “ated” from “ate”. The one feature not provided by the new model was a way of modeling early correct production of irregular forms. The success of the new model can be used to help clarify the extent to which the published critiques apply to a particular connectionist implementation as opposed to fundamental principles underlying the broader connectionist conceptualization.
This study examined the role of covert semantic classes or 'cryptotypes' in determining children's overgeneralizations of reversive prefixes such as un - in *unsqueeze or *unpress . A training corpus of 160 English verbs was presented incrementally to a backpropagation network. In three simulations, we showed that the network developed structured representations for the semantic cryptotype associated with the use of the reversive prefix un- . Overgeneralizations produced by the network, such as *unbury or *unpress , match up well with actual overgeneralizations observed in human children, showing that structured cryptotypic semantic representations underlie this overgeneralization behaviour. Simulation 2 points towards a role of lexical competition in morphological acquisition and overgeneralizations. Simulation 3 provides insight into the relationship between plasticity in network learning and the ability to recover from overgeneralizations. Together, these analyses paint a dynamic picture in which competing morphological devices work together to provide the best possible match to underlying covert semantic structures.
Este libro ofrece un nuevo enfoque a la auto-organización, adapta -ción, aprendizaje y memoria, que da lugar a cursos para postgraduados en ciencias de la información, ciencias computacionales, psicología, biología teórica, y física.
DISLEX is an artificial neural network model of the mental lexicon. It was built to test computationally whether the lexicon could consist of separate feature maps for the different lexical modalities and the lexical semantics, connected with ordered pathways. In the model, the orthographic, phonological, and semantic feature maps and the associations between them are formed in an unsupervised process, based on cooccurrence of the lexical symbol and its meaning. After the model is organized, various damage to the lexical system can be simulated, resulting in dyslexic and category-specific aphasic impairments similar to those observed in human patients.
Generalization, representation, and recovery in a self-organizing feature-map model of language acquisition
- P Li
Li, P. (1999). Generalization, representation, and recovery in
a self-organizing feature-map model of language acquisition. In M. Hahn & S.C. Stoness (eds.), Proceedings of
the 21st Annual Conference of the Cognitive Science Society (pp.308-313). Mahwah, NJ: Lawrence Erlbaum.