FRINK Linked Data Fragments server

Matches in SemOpenAlex for { <https://semopenalex.org/work/W73850480> ?p ?o ?g. }

• W73850480 abstract "The Active Role of Partial Knowledge in Cross-Situational Word Learning Daniel Yurovsky, Damian Fricker, Chen Yu, and Linda B. Smith {dyurovsk, dfricker, chenyu, smith4} @indiana.edu Department of Psychological and Brain Science, and Cognitive Science Program 1101 East 10 th Street Bloomington, IN 47405 USA Abstract A number of modern word learning theories posit statistical processes in which knowledge is accumulated across many exposures to a word and its potential referents. Accordingly, words do not go directly from unknown to known, but rather pass through intermediate stages of partial knowledge. This work presents empirical evidence for the existence of such partial knowledge, and further demonstrates its active driving role in cross-situational word learning. Subsequently, an incremental model which leverages its partial knowledge of word-object mappings from trial to trial is shown to account well for the data. In contrast, models which do not do so cannot explain the data. These results confirm crucial assumptions made by statistical word learning models and shed light on the representations underlying the acquisition of word meanings. Keywords: word learning; language computational modeling; statistical learning acquisition; Introduction We have a tendency to characterize word learning as an all-or-none process: either a child knows a given word, or she has not yet learned it. This is apparent in our methodology (e.g. forced-choice tests, preferential looking), and assessment of vocabulary size via MCDI (Fenson, Dale, Reznick, Bates, Thal, & Pethick, 1994), as well as some theoretical claims. But this implicit all-or-none characterization may stymie our thinking about potential word-meaning representations. For almost a century we have known that human learning and memory are not binary phenomena (Ebbinghaus, 1913). In learning lists of paired associates, for instance, a failure to recall the correct pair for a prompt does not imply no knowledge of the mapping. Evidence of this knowledge can be recovered using a different test paradigm (e.g. recognition or savings). The knowledge is not absent, but rather partial or sub-threshold. The central idea motivating this work is that such sub-threshold knowledge may play a profound in the course of language acquisition. Several recent theoretical and computational approaches to word learning have made explicit use of partial knowledge. For instance, McMurray (2007) modeled the learning of a word’s meaning as the acquisition of partial meaning tokens. Yu and Smith (2007) argued that early word learning can be thought of as the accumulation of co- occurrence statistics between words and objects across multiple situations. These theories suggest that a word can be learned in bits rather than in a single perfect moment. Other models make an even stronger claim: not only can one build lexical knowledge by accumulating parts; this partial knowledge is an active driver of the learning system (Blythe, Smith, & Smith, in press, McMurray, Horst, Toscano, & Samuelson, in press, Fazly, Alishahi, & Stevenson, in press, Yu, 2008). These models have been tested predominantly on large corpora, reproducing qualitative patterns found in children’s word learning. If they are correct about the presence and role of partial knowledge, however, then we should be able to find empirical evidence for the role of partial knowledge in human word learners. Yurovsky and Yu (2008) presented indirect evidence of the active role of partial knowledge in cross-situational learning. They exposed participants to a series of individually ambiguous learning trials consisting of multiple words and multiple objects. At the end of each trial, participants were asked to indicate how sure they were (1- 10) that they knew the correct label for each object. Yurovsky and Yu showed that a given object’s rating could be predicted from the ratings given to the other objects on the same trial, even after the object’s rating on its previous exposure was taken into account. Thus, participants seemed to be using partial knowledge of word-object mappings to reduce the set of candidates for other labels. This analysis, while promising, was performed on participants’ subjective knowledge ratings. In the present work, we propose to offer stronger and more direct evidence that partial knowledge plays an active role in word learning. To this end, we expose participants to two consecutive blocks of cross-situational learning. Crucially, half of the words and objects in the second block are those which participants failed to learn in the first block. Comparing the results of block 2 to those of several control conditions, we can determine the role of partial knowledge in cross- situational learning. First, we can ask whether partial knowledge exists in the system, whether learners are really accumulating bits of sub-threshold knowledge. At a deeper level, we pursue a more interesting question: does partial knowledge of individual word-referent pairs – interacting in a system with partial knowledge of other word-referent pairs – facilitate the acquisition of new words. To answer this question, a set of computational models are fit to the data to understand the underlying learning mechanisms which give rise to the empirical results. We compare a simple associative model, a biased associative model which increments associations in proportion to their current strength, and a competitive associative model which adds within-trial competition. In the simple associative model, partial knowledge is not used in learning. In the biased associative model, partial knowledge of a word-" @default.
• W73850480 created "2016-06-24" @default.
• W73850480 creator A5047568087 @default.
• W73850480 creator A5070692772 @default.
• W73850480 creator A5075533695 @default.
• W73850480 creator A5085240636 @default.
• W73850480 date "2010-01-01" @default.
• W73850480 modified "2023-09-27" @default.
• W73850480 title "The active role of partial knowledge in cross-situational word learning" @default.
• W73850480 cites W1983578042 @default.
• W73850480 cites W2006969979 @default.
• W73850480 cites W2016429292 @default.
• W73850480 cites W2025156626 @default.
• W73850480 cites W2030007850 @default.
• W73850480 cites W2050985708 @default.
• W73850480 cites W2098428317 @default.
• W73850480 cites W2141038596 @default.
• W73850480 cites W2147579759 @default.
• W73850480 cites W2155889433 @default.
• W73850480 cites W2165878470 @default.
• W73850480 cites W2166571083 @default.
• W73850480 cites W2319178748 @default.
• W73850480 cites W3092316137 @default.
• W73850480 cites W3169044938 @default.
• W73850480 cites W40243891 @default.
• W73850480 cites W2028520710 @default.
• W73850480 cites W2493479876 @default.
• W73850480 cites W2593723150 @default.
• W73850480 hasPublicationYear "2010" @default.
• W73850480 type Work @default.
• W73850480 sameAs 73850480 @default.
• W73850480 citedByCount "1" @default.
• W73850480 countsByYear W738504802012 @default.
• W73850480 crossrefType "journal-article" @default.
• W73850480 hasAuthorship W73850480A5047568087 @default.
• W73850480 hasAuthorship W73850480A5070692772 @default.
• W73850480 hasAuthorship W73850480A5075533695 @default.
• W73850480 hasAuthorship W73850480A5085240636 @default.
• W73850480 hasConcept C138885662 @default.
• W73850480 hasConcept C145420912 @default.
• W73850480 hasConcept C154945302 @default.
• W73850480 hasConcept C15744967 @default.
• W73850480 hasConcept C180747234 @default.
• W73850480 hasConcept C188147891 @default.
• W73850480 hasConcept C204321447 @default.
• W73850480 hasConcept C2777601683 @default.
• W73850480 hasConcept C2986548447 @default.
• W73850480 hasConcept C41008148 @default.
• W73850480 hasConcept C41895202 @default.
• W73850480 hasConcept C74672266 @default.
• W73850480 hasConcept C77805123 @default.
• W73850480 hasConcept C90805587 @default.
• W73850480 hasConcept C9114305 @default.
• W73850480 hasConceptScore W73850480C138885662 @default.
• W73850480 hasConceptScore W73850480C145420912 @default.
• W73850480 hasConceptScore W73850480C154945302 @default.
• W73850480 hasConceptScore W73850480C15744967 @default.
• W73850480 hasConceptScore W73850480C180747234 @default.
• W73850480 hasConceptScore W73850480C188147891 @default.
• W73850480 hasConceptScore W73850480C204321447 @default.
• W73850480 hasConceptScore W73850480C2777601683 @default.
• W73850480 hasConceptScore W73850480C2986548447 @default.
• W73850480 hasConceptScore W73850480C41008148 @default.
• W73850480 hasConceptScore W73850480C41895202 @default.
• W73850480 hasConceptScore W73850480C74672266 @default.
• W73850480 hasConceptScore W73850480C77805123 @default.
• W73850480 hasConceptScore W73850480C90805587 @default.
• W73850480 hasConceptScore W73850480C9114305 @default.
• W73850480 hasIssue "32" @default.
• W73850480 hasLocation W738504801 @default.
• W73850480 hasOpenAccess W73850480 @default.
• W73850480 hasPrimaryLocation W738504801 @default.
• W73850480 hasRelatedWork W166519878 @default.
• W73850480 hasRelatedWork W1948131710 @default.
• W73850480 hasRelatedWork W2068251364 @default.
• W73850480 hasRelatedWork W2295667662 @default.
• W73850480 hasRelatedWork W2298217069 @default.
• W73850480 hasRelatedWork W2395554419 @default.
• W73850480 hasRelatedWork W2789935384 @default.
• W73850480 hasRelatedWork W2896062597 @default.
• W73850480 hasRelatedWork W2922560562 @default.
• W73850480 hasRelatedWork W2923749282 @default.
• W73850480 hasRelatedWork W3005300799 @default.
• W73850480 hasRelatedWork W3022013674 @default.
• W73850480 hasRelatedWork W3047183891 @default.
• W73850480 hasRelatedWork W3192452456 @default.
• W73850480 hasRelatedWork W3192596574 @default.
• W73850480 hasRelatedWork W3199139428 @default.
• W73850480 hasRelatedWork W3201481672 @default.
• W73850480 hasRelatedWork W3211073792 @default.
• W73850480 hasRelatedWork W53236192 @default.
• W73850480 hasRelatedWork W2608704154 @default.
• W73850480 hasVolume "32" @default.
• W73850480 isParatext "false" @default.
• W73850480 isRetracted "false" @default.
• W73850480 magId "73850480" @default.
• W73850480 workType "article" @default.