BibSonomyburst/user/seandalai/2009BibSonomy RSS feed for /user/seandalai/20092012-02-16T10:34:22+01:00http://www.bibsonomy.org/bibtex/2776b23d582963bc6a08a0950337c29c5/seandalaiseandalai2010-02-09T16:38:34+01:002009 compounds mt norwegian <span class="authorEditorList"><a href="/author/Bungum">Lars Bungum</a>, and <a href="/author/Oepen">Stephan Oepen</a> </span><em>Proceedings of the 13th Annual Meeting of the European Association for Machine Translation EAMT-09, </em><em>Barcelona, Spain, </em>(<em>2009</em>)Tue Feb 09 16:38:34 CET 2010Barcelona, SpainProceedings of the 13th Annual Meeting of the European Association for Machine Translation (EAMT-09)Automatic Translation of Norwegian Noun Compounds20092009 compounds mt norwegian http://www.bibsonomy.org/bibtex/2768d3af222fc607fbc4d70cc2c0d3e93/seandalaiseandalai2009-08-30T13:34:15+02:002009 compounds prosody <span class="authorEditorList"><a href="/author/Giegerich">Heinz J. Giegerich</a> </span><em>Word Structure</em> <em>2(1):1--17</em> (<em>2009</em>)Sun Aug 30 13:34:15 CEST 2009Word Structure11--17The English compound stress myth220092009 compounds prosody This study investigates the distribution of end-stress and fore-stress among English NN and NNN compounds. It finds that end-stress in NNs is not 'exceptional', as many researchers have claimed, but confined to a reasonably well defined class of attribute-head NNs within which it is (at least optionally) grammatical and often predictable. In NNNs–NNs with embedded NNs–both fore-stress and end-stress can occur in both the embedding and the embedded NN, giving rise to eight possible stress patterns, all of which are attested. Moreover, the distribution of fore-stress and end-stress in embedding and embedded NNs follows the regularities identified in free-standing NNs. There is therefore no reason to accept the generalization whereby in NNNs, the second element is always stressed under right-branching and the first element under left-branching. While such patterns are perhaps particularly frequent, all others are also grammatical: the Compound Stress Rule known in the literature for some fifty years, deriving stress patterns from structural geometry, is wrong.http://www.bibsonomy.org/bibtex/241a91f180bfa7d64b32fbb66859fb0cd/seandalaiseandalai2009-08-18T14:42:49+02:002009 analogy compound <span class="authorEditorList"><a href="/author/Krott">Andrea Krott</a> </span><em>Analogy in Grammar: Form and Acquisition, </em><em>Oxford University Press, </em><em>Oxford, </em>(<em>2009</em>)Tue Aug 18 14:42:49 CEST 2009OxfordAnalogy in Grammar: Form and AcquisitionThe Role of Analogy for Compound Words20092009 analogy compound http://www.bibsonomy.org/bibtex/235871b4140b33ba56c1a164de8228b07/seandalaiseandalai2009-07-20T14:04:32+02:002009 compounds typology <span class="authorEditorList"><a href="/author/Wälchli">Bernhard Wälchli</a> </span><em>Oxford University Press, </em><em>Oxford, </em>(<em>2009</em>)Mon Jul 20 14:04:32 CEST 2009OxfordCo-Compounds and Natural Coordination 20092009 compounds typology This book presents a typological survey and analysis of the co-compound construction. This understudied phenomenon is essentially a compound whose meaning is the result of coordinating the meanings of its components, as when in some varieties of English 'father-mother' denotes 'parents'. During the course of the book Dr Wälchi examines and discusses topics of great theoretical and linguistic interest. These include the notion of word, markedness, the syntax and semantics of coordination, grammaticalization, lexical semantics, the distinction between compounding and phrase formation, and the constructional meanings languages can deploy. The book makes many observations and points about typology and areal features and includes a wealth of unfamiliar data. It will be invaluable for typologists and of considerable interest to a variety of specialists including lexicologists, morphologists, construction grammarians, cognitive linguists, semanticists, field linguists, and syntacticians.http://www.bibsonomy.org/bibtex/2c86e99c1c207979ddd8584d43dd2b7ba/seandalaiseandalai2009-07-08T15:48:11+02:002009 compounds <span class="authorEditorList"><a href="/author/Kirchner">Robert Kirchner</a>, and <a href="/author/Nicoladis">Elena Nicoladis</a> </span><em>Canadian Journal of Linguistics</em> <em>54(1):91--116</em> (<em>2009</em>)Wed Jul 08 15:48:11 CEST 2009Canadian Journal of Linguistics191--116A Level Playing-Field: Perceptibility and Inflection in English Compounds5420092009 compounds http://www.bibsonomy.org/bibtex/2ed850fc149d9880c71f63dc334dc1c99/seandalaiseandalai2009-06-11T15:47:32+02:002009 graphs icml <span class="authorEditorList"><a href="/author/Shaw">Blake Shaw</a>, and <a href="/author/Jebara">Tony Jebara</a> </span><em>Proceedings of the 26th International Conference on Machine Learning ICML-09, </em><em>Montreal, Canada, </em>(<em>2009</em>)Thu Jun 11 15:47:32 CEST 2009Montreal, CanadaProceedings of the 26th International Conference on Machine Learning (ICML-09)Structure Preserving Embedding20092009 graphs icml Structure Preserving Embedding (SPE) is an algorithm for embedding graphs in Euclidean space such that the embedding is low-dimensional and preserves the global topological properties of the input graph. Topology is preserved if a connectivity algorithm, such as k-nearest neighbors, can easily recover the edges of the input graph from only the coordinates of the nodes after embedding. SPE is formulated as a semidefinite program that learns a low-rank kernel matrix constrained by a set of linear inequalities which captures the connectivity structure of the input graph. Traditional graph embedding algorithms do not preserve structure according to our definition, and thus the resulting visualizations can be misleading or less informative. SPE provides significant improvements in terms of visualization and lossless compression of graphs, outperforming popular methods such as spectral embedding and Laplacian eigenmaps. We find that many classical graphs and networks can be properly embedded using only a few dimensions. Furthermore, introducing structure preserving constraints into dimensionality reduction algorithms produces more accurate representations of high-dimensional data.http://www.bibsonomy.org/bibtex/27814db4799ea4cb31bae36f121187372/seandalaiseandalai2009-06-11T15:45:14+02:002009 icml kernels tree <span class="authorEditorList"><a href="/author/Aiolli">Fabio Aiolli</a>, <a href="/author/Martino">Giovanni Da San Martino</a>, and <a href="/author/Sperduti">Alessandro Sperduti</a> </span><em>Proceedings of the 26th International Conference on Machine Learning ICML-09, </em><em>Montreal, Canada, </em>(<em>2009</em>)Thu Jun 11 15:45:14 CEST 2009Montreal, CanadaProceedings of the 26th International Conference on Machine Learning (ICML-09)Route Kernels for Trees20092009 icml kernels tree Almost all tree kernels proposed in the literature match substructures without taking into account their relative positioning with respect to one another. In this paper, we propose a novel family of kernels which explicitly focus on this type of information. Specifically, after defining a family of tree kernels based on routes between nodes, we present an efficient implementation for a member of this family. Experimental results on four different datasets show that our method is able to reach state of the art performances, obtaining in some cases performances better than computationally more demanding tree kernels.http://www.bibsonomy.org/bibtex/231d9d2bba9b8df065b6efdf5fdf25c41/seandalaiseandalai2009-06-11T15:38:44+02:002009 icml markovlogic multitask <span class="authorEditorList"><a href="/author/Davis">Jesse Davis</a>, and <a href="/author/Domingos">Pedro Domingos</a> </span><em>Proceedings of the 26th International Conference on Machine Learning ICML-09, </em><em>Montreal, Canada, </em>(<em>2009</em>)Thu Jun 11 15:38:44 CEST 2009Montreal, CanadaProceedings of the 26th International Conference on Machine Learning (ICML-09)Deep Transfer via Second-Order Markov Logic20092009 icml markovlogic multitask Standard inductive learning requires that training and test instances come from the same distribution. Transfer learning seeks to remove this restriction. In shallow transfer, test instances are from the same domain, but have a different distribution. In deep transfer, test instances are from a different domain entirely (i.e., described by different predicates). Humans routinely perform deep transfer, but few learning systems, if any, are capable of it. In this paper we propose an approach based on a form of second-order Markov logic. Our algorithm discovers structural regularities in the source domain in the form of Markov logic formulas with predicate variables, and instantiates these formulas with predicates from the target domain. Using this approach, we have successfully transferred learned knowledge between molecular biology, social network and Web domains. The discovered patterns include broadly useful properties of predicates, like symmetry and transitivity, and relations among predicates, like various forms of homophily.http://www.bibsonomy.org/bibtex/247a5d4e2a1e7fbfd37fe98374b800f51/seandalaiseandalai2009-06-11T15:34:36+02:002009 icml lm <span class="authorEditorList"><a href="/author/Wood">Frank Wood</a>, <a href="/author/Archambeau">Cedric Archambeau</a>, <a href="/author/Gasthaus">Jan Gasthaus</a>, <a href="/author/James">Lancelot James</a>, and <a href="/author/Teh">Yee Whye Teh</a> </span><em>Proceedings of the 26th International Conference on Machine Learning ICML-09, </em><em>Montreal, Canada, </em>(<em>2009</em>)Thu Jun 11 15:34:36 CEST 2009Montreal, CanadaProceedings of the 26th International Conference on Machine Learning (ICML-09)A Stochastic Memoizer for Sequence Data20092009 icml lm We propose an unbounded-depth, hierarchical, Bayesian nonparametric model for discrete sequence data. This model can be estimated from a single training sequence, yet shares statistical strength between subsequent symbol predictive distributions in such a way that predictive performance generalizes well. The model builds on a specific parameterization of an unbounded-depth hierarchical Pitman-Yor process. We introduce analytic marginalization steps (using coagulation operators) to reduce this model to one that can be represented in time and space linear in the length of the training sequence. We show how to perform inference in such a model without truncation approximation and introduce fragmentation operators necessary to do predictive inference. We demonstrate the sequence memoizer by using it as a language model, achieving state-of-the-art results.http://www.bibsonomy.org/bibtex/20f6fc60015312c8bf27a3d0d100fd0f7/seandalaiseandalai2009-05-09T21:33:55+02:002009 compounds psycholinguistics <span class="authorEditorList"><a href="/author/Gagné">Christina L. Gagné</a>, <a href="/author/Spalding">Thomas L. Spalding</a>, <a href="/author/Figueredo">Lauren Figueredo</a>, and <a href="/author/Mullaly">Allison C. Mullaly</a> </span><em>The Mental Lexicon</em> <em>4(1):41--76</em> (<em>2009</em>)Sat May 09 21:33:55 CEST 2009The Mental Lexicon141--76Does snow man prime plastic snow? The effect of constituent position in using relational information during the interpretation of modifier-noun phrases420092009 compounds psycholinguistics Three experiments were conducted to determine the extent to which relational and morphosyntactic information influence the processing of modifier-noun phrases. Processing of the target was faster when the shared constituent was in the same position in both the prime and the target, regardless of whether the relation was the same or different. In contrast, relation priming was contingent on the morphosyntactic role of the shared constituent; repeating the relation with the constituent in a different morphosyntactic role did not speed processing of the target (Experiments 1–3) whereas repeating the relation with the constituent in the same role did speed processing (Experiments 3). These results suggest that conceptual information is accessed in light of the constituent’s particular morphosyntactic role.http://www.bibsonomy.org/bibtex/2c65bf6e9491c3a949ed1b7cf8ba18165/seandalaiseandalai2009-05-04T19:58:04+02:002009 compounds psycholinguistics <span class="authorEditorList"><a href="/author/Rawson">Katherine A. Rawson</a>, and <a href="/author/Middleton">Erica L. Middleton</a> </span><em>Journal of Experimental Psychology: Learning, Memory, and Cognition</em> <em>35(2):353-370</em> (<em>2009</em>)Mon May 04 19:58:04 CEST 2009Journal of Experimental Psychology: Learning, Memory, and Cognition2353-370Memory-Based Processing as a Mechanism of Automaticity in Text Comprehension3520092009 compounds psycholinguistics A widespread theoretical assumption is that many processes involved in text comprehension are automatic, with automaticity typically defined in terms of properties (e.g., speed, effort). In contrast, the authors advocate for conceptualization of automaticity in terms of underlying cognitive mechanisms and evaluate one prominent account, the memory-based processing account, which states that one mechanism underlying automatization involves a shift from algorithm-based interpretation of stimuli to retrieval of prior interpretations of those stimuli. During practice, participants repeatedly read short stories containing novel conceptual combinations that were disambiguated with either their dominant or subordinate meaning. During transfer, the combinations were embedded in new sentences that either preserved or changed the disambiguated meaning. The primary dependent variable was reading time in the disambiguating region of target sentences. Supporting the memory-based processing account, speed-ups with practice were larger for repeated versus unrepeated items of the same type, reading times for subordinate versus dominant meanings of the combinations converged on later trials, and practiced meanings were retrieved when items appeared in a transfer context.http://www.bibsonomy.org/bibtex/2020c247edec9ace2f904a1d95b1d3d00/seandalaiseandalai2009-04-22T15:18:28+02:002009 compounds productivity <span class="authorEditorList"><a href="/author/Fernández-Domínguez">Jesús Fernández-Domínguez</a> </span><em>Peter Lang, </em><em>Bern, Switzerland, </em>(<em>2009</em>)Wed Apr 22 15:18:28 CEST 2009Bern, SwitzerlandProductivity in English Word-formation: An approach to N+N compounding20092009 compounds productivity This book is a contribution to the study of morphological productivity, that is, the property of word-formation processes whereby new words are created to satisfy a naming need. It presents an up-to-date picture of this phenomenon, characterising its major attributes and addressing neighbouring theoretical concepts like availability, profitability or lexicalisation. Links are also established between those notions and N+N compounding, a word-formation process regarded as very productive but traditionally overlooked in studies of this type. Unlike other productivity surveys, mostly directed at affixation, a corpus of N+N compounds is here compiled to which the mainstream models of productivity are applied. This allows to detect the pros and cons of those proposals and to propose a model of productivity. Two measures, Indicator of Profitability (p) and Trend of Profitability (P), are introduced which can be applied across word-formation processes and are able to compute their productivity based on semantic categories.http://www.bibsonomy.org/bibtex/26fd263fc0cae9b503c67f08abaaf57be/seandalaiseandalai2009-04-16T15:31:15+02:002009 association compounds eacl <span class="authorEditorList"><a href="/author/Peirsman">Yves Peirsman</a>, and <a href="/author/Geeraerts">Dirk Geeraerts</a> </span><em>Proceedings of the 12th Conference of the European Chapter of the Association for Computational Linguistics EACL-09, </em><em>Athens, Greece, </em>(<em>2009</em>)Thu Apr 16 15:31:15 CEST 2009Athens, GreeceProceedings of the 12th Conference of the European Chapter of the Association for Computational Linguistics (EACL-09)Predicting Strong Associations on the Basis of Corpus Data20092009 association compounds eacl http://www.bibsonomy.org/bibtex/22aaa6224432d5b2d5e9776783def32c3/seandalaiseandalai2009-02-11T17:51:07+01:002009 compounds eacl kernels <span class="authorEditorList"><a href="/author/Séaghdha">Diarmuid Ó Séaghdha</a>, and <a href="/author/Copestake">Ann Copestake</a> </span><em>Proceedings of the 12th Conference of the European Chapter of the Association for Computational Linguistics EACL-09, </em><em>Athens,Greece, </em>(<em>2009</em>)Wed Feb 11 17:51:07 CET 2009Athens,GreeceProceedings of the 12th Conference of the European Chapter of the Association for Computational Linguistics (EACL-09)Using lexical and relational similarity to classify semantic relations20092009 compounds eacl kernels Many methods are available for computing semantic similarity between individual words, but certain NLP tasks require the comparison of word pairs. This paper presents a kernel-based framework for application to relational reasoning tasks of this kind. The model presented here combines information about two distinct types of word pair similarity: lexical similarity and relational similarity. We present an efficient and flexible technique for implementing relational similarity and show the effectiveness of combining lexical and relational models by demonstrating state-of-the-art results on a compound noun interpretation task.http://www.bibsonomy.org/bibtex/2c97bcff8e33618f87120894e864e57fc/seandalaiseandalai2009-02-09T15:53:41+01:002009 compounds theory <span class="authorEditorList"><a href="/author/Mukai">Makiko Mukai</a> </span><em>Word Structure</em> <em>1(2):178--198</em> (<em>2009</em>)Mon Feb 09 15:53:41 CET 2009Word Structure2178--198Recursive compounds120092009 compounds theory I propose a structure for recursive compounds in Japanese, English and Mainland Scandinavian within the frameworks of the Minimalist Program (Chomsky 1995, 2000, 2001) and Distributed Morphology (Halle and Marantz 1993). Within the proposed theory, Collins’ (2002) definition of head is used: a head is a category which has one or more unsaturated features. Secondly, it is assumed that discharging a theta-feature from a simple nominal is also possible, since nouns are predicates (Higginbotham 1985) and an undischarged theta-feature counts as an unsaturated feature. In addition, the non-head of a compound is argued to be a root without word class features.http://www.bibsonomy.org/bibtex/27ad41f0c4e1e66f73d0f4d2bb7444871/seandalaiseandalai2009-02-07T14:08:23+01:002009 context kernels <span class="authorEditorList"><a href="/author/Pahikkala">Tapio Pahikkala</a>, <a href="/author/Pyysalo">Sampo Pyysalo</a>, <a href="/author/Boberg">Jorma Boberg</a>, <a href="/author/Järvinen">Jouni Järvinen</a>, and <a href="/author/Salakoski">Tapio Salakoski</a> </span><em>Machine Learning</em> <em>74(2):133--158</em> (<em>2009</em>)Sat Feb 07 14:08:23 CET 2009Machine Learning2133--158Matrix representations, linear transformations, and kernels for disambiguation in natural language7420092009 context kernels In the application of machine learning methods with natural language inputs, the words and their positions in the input text are some of the most important features. In this article, we introduce a framework based on a word-position matrix representationof text, linear feature transformations of the word-position matrices, and kernel functions constructed from the transformations.We consider two categories of transformations, one based on word similarities and the second on their positions, which canbe applied simultaneously in the framework in an elegant way. We show how word and positional similarities obtained by applyingpreviously proposed techniques, such as latent semantic analysis, can be incorporated as transformations in the framework.We also introduce novel ways to determine word and positional similarities. We further present efficient algorithms for computingkernel functions incorporating the transformations on the word-position matrices, and, more importantly, introduce a highlyefficient method for prediction. The framework is particularly suitable to natural language disambiguation tasks where theaim is to select for a single word a particular property from a set of candidates based on the context of the word. We demonstratethe applicability of the framework to this type of tasks using context-sensitive spelling error correction on the ReutersNews corpus as a model problem.http://www.bibsonomy.org/bibtex/2b4d7c48052aff2265db89d77d43dfabc/seandalaiseandalai2009-01-05T11:37:10+01:002009 evaluation <span class="authorEditorList"><a href="/author/García">Salvador García</a>, and <a href="/author/Herrera">Francisco Herrera</a> </span><em>Journal of Machine Learning Research</em> (<em>2009</em>)Mon Jan 05 11:37:10 CET 2009Journal of Machine Learning Research2677--2694An Extension on "Statistical Comparisons of Classifiers over Multiple Data Sets" for all Pairwise Comparisons 920092009 evaluation In a recently published paper in JMLR, Demšar (2006) recommends a set of non-parametric statistical tests and procedures which can be safely used for comparing the performance of classifiers over multiple data sets. After studying the paper, we realize that the paper correctly introduces the basic procedures and some of the most advanced ones when comparing a control method. However, it does not deal with some advanced topics in depth. Regarding these topics, we focus on more powerful proposals of statistical procedures for comparing n × n classifiers. Moreover, we illustrate an easy way of obtaining adjusted and comparable p-values in multiple comparison procedures.