... analogous similarity me-tric for γ. They range from 0 to 1. These two metrics both evaluate the similarityfor two vec-tors in the same language, so using cosine dis-tance to compute the similarity ... performance over the baseline. The Alg2 cosine similarity function got 0.7 BLEU-score (p<0.01) improvement over the baseline for NIST 2006 test set, and a 0.5 BLEU-score (p<0.05) for ... of the Association for Computational Linguistics, pages 834–843,Uppsala, Sweden, 11-16 July 2010.c2010 Association for Computational LinguisticsBilingual Sense Similarityfor Statistical Machine...
... any information about repair is strippedfrom the input, including partial words, repair sym-bols3, and interruption point information. While anintegrated system for processing and parsing ... results from Haleet al. RCT results are on the right-corner transformedgrammar (transformed back to flat treebank-style trees for scoring purposes). CYK and TAG lines show relevantresults from ... syntactic information tofind repairs, and thus may have access to some ofthis information about where interruptions occur,this experiment is intended to evaluate the use of theright corner transform...
... any other domains, domainj seems to be the domain of unit~. The system se- lects the domain which is the largest of all sim- ilarities in N of domains as the domain of the unit (formula (6)) ... vectors. 5.3 Domain identification experiment The system selects suitable domain of each unit for keyword extraction. Table I shows the results of domain identification. We con- ducted domain identification ... kinds of domains, i.e. 141 domains and 9 large domains. We also compared the results and the result us- ing previous method (Suzuki et al., 1997). For comparison, we selected 5 domains which...
... Statistical Pattern Recognition. Doctoral dissertation. Stanford University, Stanford, Cali- fornia. 283 Statistical Decision-Tree Models for Parsing* David M. Magerman Bolt Beranek and Newman ... sentence length for Wall Street Journal experiments. 5 Conclusion Regardless of what techniques are used forparsing disambiguation, one thing is clear: if a particular piece of information is ... and 7 illustrate the performance of SPATTER as a function of sentence length. SPAT- TER's performance degrades slowly for sentences up to around 28 words, and performs more poorly and more...
... corpus-based measures of similarity perform comparably when used for the task ofshort answer grading. However, since the corpus-based measures can be improved by account-ing fordomain and corpus ... unsupervisedtechniques for the task of automatic shortanswer grading. We compare a number ofknowledge-based and corpus-based mea-sures of text similarity, evaluate the effectof domain and size on ... improve the performance of thesystem by integrating automatic feedbackfrom the student answers. Overall, oursystem significantly and consistently out-performs other unsupervised methods for short...
... gatheredfrom parallel texts and the evaluation data for thetwo SENSEVAL tasks. This gave a set of 6 nouns for SENSEVAL-2 and 9 nouns for SENSEVAL-3. For each noun, we gathered a maximum of 500parallel ... we use for our experimentsbefore presenting our experimental results. Next,we propose using the well calibrated probabilitiesof logistic regression to estimate the sense priors,and perform ... improves performance. For example,row 1 of Table 4 shows that adjusting the pre-dictions of multiclass naive Bayes classifiers bysense priors estimated by logistic regression (NB-EM ) performs significantly...
... target domain labeled instances. We chosethis number since we believe it to be a reasonableamount for a single engineer to label with minimaleffort. For reasons of space, for each target domain dom ... different domains, and annotatingcorpora for every possible domain of interestis impractical. We investigate domain adap-tation for sentiment classifiers, focusing ononline reviews for different ... domainsof discourse makes it an ideal candidate for domain adaptation. This work addressed two importantquestions of domain adaptation. First, we showedthat for a given source and target domain, ...
... reductionPakistanScotland-subj-of-losePakistan-subj-of-win similarity semantic classhead similarity role similarity Pakistanhad won the World Cup lost in the semi-finalScotlandFigure 1: Context reduction and similarity levelsdraw ... 1997;Stern, 1931)). In a place -for- people pattern,a place stands for any persons/organisations associ-ated with it, e.g., for sports teams in (2), (3), and (4),and for the government in (7).4(7) ... a place -for- event pattern, a locationname refers to an event that occurred there (e.g., us-ing the word Vietnam for the Vietnam war). In aplace -for- product pattern a place stands for a product...
... designed:5¢-TGAGATGTGCCAGCTGAGGTTCA-3¢ for I282Q(forward), 5¢-CAACGCCCAGCATACCCAGCAGT-3¢ for Q404H (forward), 5¢-CAACGCCCAGGCAACCCAGCAGT-3¢ for Q404A (forward), 5¢-TGAACCTCAGCTGGCACATCTA-3¢ for I282Q (reverse), ... were obtained by Transformer Site-directedmutagenesis Kit (Clontech). The following primerswere used: 5¢-TCGAGCTGTGTATACTGAGATTCA-3¢ for Q285I, 5¢-TCAATGCTCAGCAGACCCAGCGGC-3¢ for H407Q, 5¢-TCAATGCTCAGGCCACCCAGCGGC-3¢ ... luciferase activity. All experiments wereperformed at least three times in duplicates and luciferaseactivity was normalized for alkaline phosphatase activity. For curve fitting and EC50 calculations,XLFITversion...
... in a specific domain (medicine) and as we are not domain experts, we are in lack of domain knowl-edge. This missing domain knowledge shall be acquired from external resources, for example UMLS. ... (c) is it normal or is it ab-normal? Therefore, when a radiologist looks for information, his search queries most likely con-tain terms from various information sources that provide this kind ... noun)Using a transformation rule of the form,82Ontologies (OBO)5 framework. The OBO con-sortium establishes a set of principles to which the biomedical ontologies shall conform to for purposes...
... Association for Computational LinguisticsExploiting Multiple Treebanks forParsing with Quasi-synchronousGrammarsZhenghua Li, Ting Liu∗, Wanxiang CheResearch Center for Social Computing and Information ... punctuation. Weadopt Dan Bikel’s randomized parsing evaluationcomparator for significance test (Noreen, 1989).7 For all models used in current work (POS taggingand parsing) , we adopt averaged perceptron ... algorithm for pro-jective dependency parsing. In Proceedings of the8th International Workshop on Parsing Technologies(IWPT), pages 149–160.Eric W. Noreen. 1989. Computer-intensive methods for testing...
... gains in performance when moving fromParliament domain to News domain. 3 DataOur source domain is European Parliamentproceedings (http://www.statmt.org/europarl/). We use three target domains: ... methods for mining dic-tionaries from comparable corpora to the domain adaptation setting, by “bootstrapping” them basedon known translations from the source domain. (3)Develop methods for integrating ... establishbaseline performance for the domains. In these ex-periments, we built a translation model based onlyon the Parliament proceedings. We then tune it us-ing the small amount of target -domain tuning...
... that the use ofprobability information from the parser for treeconversion helps target grammar parsing. 4.3 Using Unlabeled Data for Parsing Recent studies on parsing indicate that the use ... treebanks with same grammar for- malism fordomain adaptation of parsers. Roarkand Bachiani (2003) presented count merging andmodel interpolation techniques fordomain adap-tation of parsers. ... Parsing Through grammar formalism conversion, we havesuccessfully turned the problem of using hetero-geneous treebanks forparsing into the problem of parsing on homogeneous treebanks. Before usingconverted...
... (Fig-ure 1): one for the live “instant translation” userinterface, one for demonstrating the different com-ponents of the system and algorithmic visualiza-tions, and one designated for technical ... and David Chiang. 2005. Better k-best parsing. In Proceedings of the International Work-shop on Parsing Technologies.27We will rely on 3 workstations: one for the instant translation demo, ... Opensource toolkit for statistical machine translation. InProceedings of the ACL-2007 Demo and Poster Ses-sions.Zhifei Li and Sanjeev Khudanpur. 2008. A scalabledecoder for parsing- based machine...
... Dong, 2006) and WordNet pro-vide domain information for Chinese and English,but there has been no domain resource for Japanesethat are publicly available.8 Domain dictionary construction methods ... Preparing key-words for each domain (§3.1). 2 Associating JFWswith domains (§3.2). 3 Reassociating JFWs withNODOMAIN (§3.3). 4 Manual correction (§3.5).3.1 Preparing Keywords for each Domain About ... NODOMAIN was prepared for thosewords that do not belong to any particular domain. As for the latter issue, you might use keyword ex-traction techniques; identifying words that representa domain...