OrBEAGLE: Integrating Orthography into a Holographic Model of the Lexicon

Kachergis, George; Cox, Gregory E.; Jones, Michael N.

doi:10.1007/978-3-642-21735-7_38

Cited by 8 publications

(5 citation statements)

References 15 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Recent work in this area using the WINDSORS global co-occurrence definition of SND ( Macdonald, 2013 ; Danguecan and Buchanan, 2014, unpublished) found support for the idea that words with many near neighbors are processed more slowly than words with few near neighbors in both lexical decision and semantic categorization tasks. Although the present study uses the WINDSORS model to study semantic neighborhood effects ( Durda and Buchanan, 2008 ), other distributional models such as Hyperspace Analog to Language (HAL; Lund and Burgess, 1996 ), Correlated Occurrence Analog to Lexical Semantics (COALS; Rohde et al, 2004 ), Latent Semantic Analysis (LSA; Landauer and Dumais, 1997 ), Bound Encoding of the AGgregate Language Environment (BEAGLE; Jones and Mewhort, 2007 ), OrBEAGLE ( Kachergis et al, 2011 ), Random Permutation Model ( Sahlgren et al, 2008 ), the Topic model ( Griffiths et al, 2007 ), and HiDEx ( Shaoul and Westbury, 2010 ) have also contributed extensively to our knowledge of semantic phenomena.…”

Section: Introductionmentioning

confidence: 99%

Semantic Neighborhood Effects for Abstract versus Concrete Words

Danguecan

Buchanan

2016

Front. Psychol.

View full text Add to dashboard Cite

Studies show that semantic effects may be task-specific, and thus, that semantic representations are flexible and dynamic. Such findings are critical to the development of a comprehensive theory of semantic processing in visual word recognition, which should arguably account for how semantic effects may vary by task. It has been suggested that semantic effects are more directly examined using tasks that explicitly require meaning processing relative to those for which meaning processing is not necessary (e.g., lexical decision task). The purpose of the present study was to chart the processing of concrete versus abstract words in the context of a global co-occurrence variable, semantic neighborhood density (SND), by comparing word recognition response times (RTs) across four tasks varying in explicit semantic demands: standard lexical decision task (with non-pronounceable non-words), go/no-go lexical decision task (with pronounceable non-words), progressive demasking task, and sentence relatedness task. The same experimental stimulus set was used across experiments and consisted of 44 concrete and 44 abstract words, with half of these being low SND, and half being high SND. In this way, concreteness and SND were manipulated in a factorial design using a number of visual word recognition tasks. A consistent RT pattern emerged across tasks, in which SND effects were found for abstract (but not necessarily concrete) words. Ultimately, these findings highlight the importance of studying interactive effects in word recognition, and suggest that linguistic associative information is particularly important for abstract words.

show abstract

Section: Introductionmentioning

confidence: 99%

Semantic Neighborhood Effects for Abstract versus Concrete Words

Danguecan

Buchanan

2016

Front. Psychol.

View full text Add to dashboard Cite

show abstract

“…For example, while it is obvious to the human reader that the terms “depressive disorder” and “major depressive disorder” are related to one another, the vector representations of the UMLS concepts corresponding to these terms will be similar only to the extent that they occur in similar contexts in SemMedDB. Methods exist to encode such orthographic information into distributed representations [77], including our own [78]. However, the utility this additional information for semantic and predictive modeling remains to be determined.…”

Section: Resultsmentioning

confidence: 99%

Embedding of semantic predications

Cohen

Widdows²

2017

Journal of Biomedical Informatics

View full text Add to dashboard Cite

This paper concerns the generation of distributed vector representations of biomedical concepts from structured knowledge, in the form of subject-relation-object triplets known as semantic predications. Specifically, we evaluate the extent to which a representational approach we have developed for this purpose previously, known as Predication-based Semantic Indexing (PSI), might benefit from insights gleaned from neural-probabilistic language models, which have enjoyed a surge in popularity in recent years as a means to generate distributed vector representations of terms from free text. To do so, we develop a novel neural-probabilistic approach to encoding predications, called Embedding of Semantic Predications (ESP), by adapting aspects of the Skipgram with Negative Sampling (SGNS) algorithm to this purpose. We compare ESP and PSI across a number of tasks including recovery of encoded information, estimation of semantic similarity and relatedness, and identification of potentially therapeutic and harmful relationships using both analogical retrieval and supervised learning. We find advantages for ESP in some, but not all of these tasks, revealing the contexts in which the additional computational work of neural-probabilistic modeling is justified.

show abstract

“…Previous attempts to model orthographic similarity in vector space depended upon using a binding operator to generate near-orthogonal vector representations of sequences of characters within a word, including gapped sequences to allow for flexibility (Cox et al, 2011; Kachergis et al, 2011; Hannagan et al, 2011). However, encoding in this way requires the generation of a large number of vector products.…”

Section: Applications Of Vsas and Psimentioning

confidence: 99%

Reasoning with vectors: A continuous model for fast robust inference

Widdows

Cohen

2014

Logic Journal of IGPL

View full text Add to dashboard Cite

This paper describes the use of continuous vector space models for reasoning with a formal knowledge base. The practical significance of these models is that they support fast, approximate but robust inference and hypothesis generation, which is complementary to the slow, exact, but sometimes brittle behavior of more traditional deduction engines such as theorem provers. The paper explains the way logical connectives can be used in semantic vector models, and summarizes the development of Predication-based Semantic Indexing, which involves the use of Vector Symbolic Architectures to represent the concepts and relationships from a knowledge base of subject-predicate-object triples. Experiments show that the use of continuous models for formal reasoning is not only possible, but already demonstrably effective for some recognized informatics tasks, and showing promise in other traditional problem areas. Examples described in this paper include: predicting new uses for existing drugs in biomedical informatics; removing unwanted meanings from search results in information retrieval and concept navigation; type-inference from attributes; comparing words based on their orthography; and representing tabular data, including modelling numerical values. The algorithms and techniques described in this paper are all publicly released and freely available in the Semantic Vectors open-source software package.1

show abstract

OrBEAGLE: Integrating Orthography into a Holographic Model of the Lexicon

Cited by 8 publications

References 15 publications

Semantic Neighborhood Effects for Abstract versus Concrete Words

Semantic Neighborhood Effects for Abstract versus Concrete Words

Embedding of semantic predications

Reasoning with vectors: A continuous model for fast robust inference

Contact Info

Product

Resources

About