More Publications (104)

1Scaled Log Likelihood Ratios for the Detection of Abbreviations in Text Corpora2002  Tibor Kiss,Jan StrunkScaled log likelihood ratios for the detection of abbreviations in text corpora Tibor Kiss Sprachwissenschaftliches...

2Identifying Multiword Expressions by Leveraging Morphological and Syntactic Idiosyncrasy2010  Hassan AlHaj,Shuly Wintner2001): Pointwise mutual information (PMI); TScore; loglikelihood; and the raw frequency of N1 N2 in the corpus.3 3A detailed...

3Fast Computation of Lexical Affinity Models2004  Egidio Terra,Charles L. A. ClarkeDistance watermelon,seeds watermelon,fruits Figure 3: Loglikelihood – WATERMELON 1 0 1 2 3 1 2 3 4 5 6 ...

4SMRCmp: SquareMeanRoot Approach to Comparison of Monolingual Contrastive Corporastatistical measures, including chisquare (χ2) and loglikelihood ratio (LLR). KEYWORDS : SquareMeanRoot evenness...

5LinguisticallyBased SubSentential Alignment for Terminology Extraction from a Bilingual Automotive Corpusalgorithms: LogLikelihood for single word entries and Mutual Expectation Measure for multiword entries. 4.2.1 LogLikelihood Measure...

6Collocation Extraction Based on Modifiability Statistics2004  Joachim Wermter,Udo Hahnentropy) statistical measures (e.g., chisquare, ttest, loglikelihood, Dice’s coefficient) The corresponding metrics have...

7Comparative Quality Estimation: Automatic SentenceLevel Ranking of Multiple Machine Translation Outputs2012  Eleftherios Avramidisincluded the basic parsing statistics of the overall parse loglikelihood, the confidence for the best parse tree and the average...

8Bilingual lexicon extraction from comparable corpora using indomain terms2010  Azniah Ismail,Suresh Manandhar(WS) of a source word WS , as in (Rapp, 1999), we use loglikelihood ratio (LL) Dunning (1993). We choose all words with LL...

9A comparison of unsupervised methods for PartofSpeech Tagging in Chinesealternate between calculating the expectation of the log likelihood of the model given the parameters: ...

10A MWE Acquisition and Lexicon Builder Web Service(Pointwise) Mutual Information, Dice, Pearson’s chisquared, LogLikelihood Ratio, Odds Ratio, Fisher’s Exact tests, and various...

11The Computation of Word Associations: Comparing Syntagmatic and Paradigmatic Approaches2002  Reinhard Rappdata is an important issue, it is better to use the loglikelihood ratio. It would then be assumed that the strongest ...

12Exploring variation across biomedical subdomainsby the feature’s loglikelihood between the subdomain’s data and the rest of the corpus. Loglikelihood has been shown to...

13"Got You!": Automatic Vandalism Detection in Wikipedia with Webbased Shallow SyntacticSemantic ModelingNtag Log Likelihood and Perplexity Semantic Normalized Topicspecific Syntactic Ngram Log Likelihood and Perplexity...

14Unsupervised FeatureRich Clustering2012  Vladimir Eidelmanobjective function, L (θ), which is commonly maximizing the loglikelihood of generating the data D under the current parameters...

15Latent Community Discovery with Network Regularization for Core Actors Clusteringcorrespond to actors in an association document. So, the log likelihood of a network n to be generated with PLSA model is given...

16A Classbased Probabilistic approach to Structural Disambiguation2000  Stephen Clark,David J. Weiris the same. The loglikelihood X 2 statistic corresponding to TM)le 2 is 4.8. The loglikelihood X 2 statistic is used...

17Looking for Candidate Translational Equivalents in Specialized, Comparable Corporaadditional, alternative weights are computed: tf:idf and log likelihood. 1We shall see below that actually, only a subset of...

18Evaluating performance of grammatical error detection to maximize learning effect2010  Ryo Nagata,Kazuhide Nakatanithe left of the NP; (iii) three words to its right. The loglikelihood ratio (Yarowsky, 1995) decides in which order rules in...

19Revisiting Contextbased Projection Methods for TermTranslation Spotting in Comparable Corpora2010  Audrey Laroche,Philippe Langlaissyntaxbased methods. Morin et al. (2007) consider both the loglikelihood and the mutual information association scores as well...

20Chinese Segmentation and New Word Detection using Conditional Random Fieldseach given their corresponding input sequences. The loglikelihood of training set {(xi, yi) : i = 1, ...M} is written ...

21Using Hidden Markov Random Fields to Combine Distributional and PatternBased Word Clustering2008  Nobuhiro Kaji,Masaru Kitsuregawaprior distribution on the hidden variables. Computing the loglikelihood of the complete data (n, z), we found log p(n, z) =...

22Structural Feature Selection For EnglishKorean Statistical Machine Translationentropy is equivalent to the model that maximizes the log likelihood of the training samples like (2) if we can assume it...

23Multilingual and crosslingual news topic trackingSchultz’ advice. Unlike Schultz, however, we use the loglikelihood test for term weighting as this measure seems to be...

24Classifying Chart Cells for Quadratic Complexity ContextFree Inference2008  Brian Roark,Kristy Hollingsheadstring. At each word position wi for 1log likelihood ratio of tag S1 as follows: LLR(wi ∈ S1) = log P(wi...

25Automatic Text Categorization by Unsupervised Learning2000  Youngjoong Ko,Jungyun Seothe five highest frequencies in total sentences. 2.Loglikelihood faclor: In general, the words that are indicative of...

26A Comparative Evaluation of Datadriven Models in Translation Selection of Machine Translation(wz), and P (dz) are estimated by maximization of the loglikelihood function L = ∑ d∈D ∑ w∈W n(d,w) logP (d,w), (4) and...

27Linguistic correlates of style: authorship classification with deep linguistic analysis features2004  Michael Gamonfeature vector size. We have begun experimenting with log likelihood ratio (Dunning 1993) as a thresholding technique. ...

28Sentiment classification on customer feedback data: noisy data, large feature vectors, and the role of linguistic analysis2004  Michael Gamonexperiments. The measure of “predictiveness” we employed is log likelihood ratio with respect to the target variable (Dunning ...

29Neural Network Approach To Word Category Prediction For English TextsTo avoid the multit)lication of probabilities% tbe log likelihood, STi, is defined as : STi = IolIP(Ci/Ci2 Cil) ÷...

30Towards Automatic Extraction of Monolingual and Bilingual Terminology(LOG) (the use of a likelihood ratio test leads to a log likelihood statistics, which explains tile LOG abbrevi:ttion) shows...

31Subcorpora Sampling with an Application to Bilingual Lexicon Extraction2012  Ivan Vulić,MarieFrancine Moens2003) or obtained by associative methods such as the loglikelihood score or the Dice coefficient. They are then used in...

32A Discriminative Alignment Model for Abbreviation Recognitionx(1),y(1)), ..., (a(N),x(N),y(N)) ) , we maximize the loglikelihood of the conditional probability distribution by using...

33Effective Constituent Projection across Languagesshows the loglikelihood on the projected treebank after each EM iteration. It is obvious that the loglikelihood increases...

34A Linguistically Grounded Graph Model for Bilingual Lexicon Extractioncalculate association scores for all relationships using the loglikelihood measure (Dunning, 1993). For noun pairs, we discard all...

35Bridging Topic Modeling and Personalized Searchparameters to be estimated are {θj} and {pid,j}. The loglikelihood of document d is: log p(d) = ∑ w∈V c(w, d) log[λBp(wθB)...

36Jointly Identifying Entities and Extracting Relations in Encyclopedia Text via A Graphical Model Approach2010  Xiaofeng Yu,Wai Lamassumption, we ignore the summation operator ∑N i=1 in the loglikelihood during the following derivations. To reduce overfitting...

37Underspecified Query Refinement via Natural Language Question Generationstatistics such as pointwise mutual information (PMI) and loglikelihood. In this paper, we use PMI. Given an ngram n ∈ S, we...

38A Method of Measuring Term Representativeness  Baseline Method Using Cooccurrence Distributioninstance, mutual information (Church ct al. 1990) and the loglikelihood (Dunning 1993) methods for extracting word bigrams have...

39Learning Verb Argument Structure from Minimally Annotated Corpora2002  Anoop Sarkar,Woottiporn Tripasaigiven that v is not present (written as !v). We use the log likelihood test statistic (Bickel and Doksum, 1977, 209) as a measure...

40A Measure of Term Representativeness Based on the Number of Cooccurring Salient Words2002  Toru Hisamitsu,Yoshiki Niwainstance, the mutual information (Church et al. 1990) and loglikelihood ratio (Dunning 1993; Cohen 1995) have been widely used...

41Integrating CrossLingually Relevant News Articles and Monolingual Web Documents in Bilingual Lexicon Acquisitioninformation, the φ2 statistic, the dice coeﬃcient, and the loglikelihood ratio. tJ ¬tJ tE df(tE , tJ ) = a df(tE ,¬tJ ) = b ¬tE...

42Translating Queries into Snippets for Improved Query Expansionvarious signals such as cooccurrence in similar sessions or loglikelihood ratio of original and expansion phrases. Other approaches...

43Simultaneous Ranking and Clustering of Sentences: A Reinforcement Approach to MultiDocument SummarizationTaking into account the hidden variable zC , the complete loglikelihood can be written as ( ) ( ) ( )∑∑ ∏∏ ∏∏ = = = =...

44Sentiment Translation through MultiEdge Graphsn with human ratings the weight of each link to the loglikelihood ratio of the two words it connects according to the corpus...

45Tag Dispatch Model with Social Network Regularization for Microblog User Tag Suggestionall t /∈ au. In other words,∑ t∈au Pr(tu,au) = 1. The log likelihood of generating a collection DU in TDM is formalized as...

46Finitestate phrase parsing by rule sequences1996  Marc Vilain,David D. MCDONALDP&R Arithmetic (ys) 88.8 8I.z 8+8 87.2 79.0 82. 9 Log likelihood 81.9 85.7 78.4 8t.o 73.4 77.0 F measure, ~:o.8 86. 3...

47Improved Iterative Scaling Can Yield Multiple Globally Optimal Models with Radically Differing Performance Levels2002  Iain Bancarz,Miles OsborneB(Æj ), which provides a lower bound on the change in loglikelihood of the model. Each adjustment i i + Æ i is...

48HomotopyBased SemiSupervised Hidden Markov Models for Sequence Labeling2008  Gholamreza Haffari,Anoop Sarkarthe labeled and unlabeled data is to parameterize the log likelihood function as Lλ(Θ) defined by 1− λ L ∑ (x,y)∈L log...

49Robust Measurement and Comparison of Context Similarity for Finding Translation Pairsapplication, to measure association robustly, often the LogLikelihood Ratio (LLR) measurement is suggested (Rapp, 1999; Morin...

50Webbased and combined language models: a case study on noun compound identificationon contingency table, as discussed in section 2. The loglikelihood association measure (LL, alternatively called expected...

51Corpusdependent Association Thesauri for Information Retrievaldetermine whether two terms are related to each other by a loglikelihood ratio test, and we filter out pairs of terms that do...

52Automatic Extraction of Subcategorization Frames for Czech2000  Anoop Sarkar,Daniel Zemanthese probabilities to be binomially distributed, the log likelihood statistic (Dunning, 1993) is given by:  2 log A =...

53Subcategorization Acquisition and Evaluation for Chinese Verbsfor hypothesis filtering mainly include the BHT, the Log Likelihood Ratio (LLR), the Ttest and the MLE, and the most popular...

54A Complete and Modestly Funny System for Generating and Performing Japanese StandUp Comedy2008  Jonas Sjöbergh,Kenji Arakihint and the dirty word, also using web frequencies. The loglikelihood ratios are then compared, and if the hint is more closely...

55Collocation Extraction using Parallel Corpus1993) which assumes the data to be normally distributed. Log likelihood ratio (LLR) is another measure that is used in the...

56Decoderbased Discriminative Training of Phrase Segmentation for Statistical Machine Translation2012  HyoungGyu Lee,HaeChang Rimof two adjacent words as the feature set. We use the log likelihood ratio, which is widely used to measure the association...

57An Empirical Etudy of NonLexical Extensions to Delexicalized Transfer2012  Anders Søgaard,Julie Wulffpredictive inference under covariate shift by weighting the loglikelihood function. Journal of Statistical Planning and Inference...

58Efficient Feedbackbased Feature Learning for Blog Distillation as a Terabyte ChallengeSeveral information theoretic methods, such as Chisquare, loglikelihood ratio and mutual information, are applicable for this...

59Harnessing the CRF Complexity with DomainSpecific Constraints. The Case of Morphosyntactic Tagging of a Highly Inflected Language2012  Jakub Waszczukthat training is performed with respect to the standard loglikelihood function. 2790 We show that by integrating morph...

60Unsupervised Discriminative Induction of Synchronous Grammar for Machine TranslationIn particular: • We approximate the exact conditional loglikelihood objective inspired by contrastive estimation (Smith and...

61Automatic Thesaurus Generation through Multiple Filteringdoes not seem to be radically different. We adopted loglikelihood ratio (Danning 1993), which gave the best pertbrmance...

62Measuring the Similarity between Compound Nouns in Different Languages Using NonParallel Corpora2002  Takaaki Tanakathe feature value of r, µw(t, r) is defined by the log likelihood ratio (Dunning, 1993) 1 as follows. µw(t, r) = { L(t...

63Syntactic Features for High Precision Word Sense Disambiguation(Kilgarriff & Palmer, 2000). Features are weighted with a loglikelihood measure, and arranged in an ordered list according to...

64Japanese Unknown Word Identification by Characterbased Chunking2004  Masayuki Asahara,Yuji Matsumotodetermined by the Viterbi algorithm. In practice, we use log likelihood as cost. Maximizing probabilities means minimizing costs...

65Semantic Similarity Applied to Spoken Dialogue Summarization2004  Iryna Gurevych,Michael Strubesubsume both c1 and c2 and − log p(c) is the negative log likelihood (information content). The probability p is computed...

66Significance tests for the evaluation of ranking methods2004  Stefan Evertthe scores assigned by four association measures: the loglikelihood ratio G2 (Dunning, 1993), Pearson’s chisquared statistic...

67Acquiring Sense Tagged Examples using Relevance Feedbackbigrams: Salient bigrams within the abstract with high loglikelihood scores, as described by Pedersen (2001). Unigrams: Lemmas...

68Extractive Summarization Using Supervised and SemiSupervised Learningwhere A and B are estimated by minimizing the negative loglikelihood function using training data and their decision values...

69Weakly Supervised Morphology Learning for Agglutinating Languages Using Small Training Sets.2010  Ksenia Shalonova,Bruno Goleniadivergence is used in order to analyse the increase of loglikelihood among all possible models. The JensenShannon divergence...

70Improving RelativeEntropy Pruning using Statistical Significance(Papineni et al., 2002). After computing the negative log likelihood of both scores, we also rescale both score’s values by...

71Effects of Adjective Orientation and Gradability on Sentence Subjectivitysentences in the corpus on the basis (, . The proba of the loglikelihood ratio test statistic ,2 bility of a sentence being...

72Unsupervised Word Sense Disambiguation Using Bilingual Comparable Corpora2002  Hiroyuki Kaji,Yasutsugu Morimotorelation is judged to be statistically significant through a loglikelihood ratio test. 3.3 Alignment of pairs of related words...

73The Importance of Supertagging for WideCoverage CCG Parsing2004  Stephen Clark,James R. Currannormalform derivations, d1, . . . , dm. L(Λ) is the loglikelihood of model Λ, and G(Λ) is a Gaussian prior term used to...

74Towards Terascale Semantic Acquisitionrelatively high occurrence and high precision. We apply the log likelihood principle (Dunning 1993) to compute this score. The...

75Understanding and Summarizing Answers in CommunityBased Question Answering Servicesimproved when answer quality measure was integrated in a log likelihood retrieval model. As mentioned in Section 1, cQA services...

76Semantic Role Assignment for Event Nominalisations by Leveraging Verbal Datadimensions. As similarity measure, we use cosine distance on loglikelihood transformed counts. Lexical level model. The lexical...

77Modeling LatentDynamic in Shallow Parsing: A Latent Conditional Model with Imrpoved Inference(3) The first term of this equation is the conditional loglikelihood of the training data. The second term is the regularizer...

78Training Conditional Random Fields Using Incomplete Annotationsestimator for this model can be obtained by maximizing the log likelihood function: LL(θ) = N∑ n=1 lnP„(YL(n) x(n)) (3) =...

79An Integrated Probabilistic and Logic Approach to Encyclopedia Relation Extraction with Multiple Features2008  Xiaofeng Yu,Wai Lamdirectly exploited. To avoid overfitting, we penalized the loglikelihood by the commonly used zeromean Gaussian prior over the...

80Towards an optimal weighting of context words based on distancecan be found by using Newton’s method, maximizing the log likelihood via leaveoneout estimation: L−1(μW, C) =∑ i ∑ x∈V...

81A Discriminative Latent VariableBased "DE" Classifier for ChineseEnglish SMT(4) The first term of this equation is the conditional loglikelihood of the training data. The second term is a regularizer...

82StructureAware Review Mining and Summarizationis to determine the parameters based on maximizing the loglikelihood !" = # $%& ( ( ) ( )) . In Linear CRFs model...

83Probabilistic TreeEdit Models with Structured Latent Variables for Textual Entailment and Question Answeringe: e.a⊆Ss+S∗z P(e  τt,τh) (2) The L2norm penalized loglikelihood over n training examples (L) is our training objective...

84Resolving Surface Forms to Wikipedia Topicsour setting, the loss function is a negative binomial loglikelihood, xi is the feature vector for a surfaceform and Wi...

85Learning to Model DomainSpecific Utterance Sequences for Extractive Summarization of Contact Center Dialogues2010  Ryuichiro Higashinaka,Yasuhiro Minami,Hitoshi Nishikawa,Kohji Dohsaka,Toyomi Meguro,Satoshi Takahashi,Genichiro Kikuiprobability of wi in all domains except for DMk. This log likelihood ratio estimates how much a word is characteristic of...

86Incremental Chinese Lexicon Extraction with Minimal Resources on a DomainSpecific Corpus2010  Gaël Patin(PMI), PoissonStriling (PS) (Quasthoff and Wolff, 2002), Loglikelihood (LL), Pointwise Mutual Information Cube (PMI3) (Daille...

87Extraction of Multiword Expressions from Small Parallel Corpora2010  Yulia Tsvetkov,Shuly Wintnersuggest that some collocation measures (especially PMI and Loglikelihood) are superior to others for identifying MWEs. Soon, however...

88Accelerated Training of Maximum Margin Markov Models for Sequence Labeling: A Case Study of NP Chunking2010  Xiaofeng Yu,Wai Lamparameters for CRFs. To avoid overfitting, we penalized the loglikelihood by the commonly used zeromean Gaussian prior over the...

89Sentence Ordering with EventEnriched Semantics and TwoLayered Clustering for MultiDocument News Summarizationcovariance matrix Ci’. We iterate the two steps until the loglikelihood converges within a threshold = 106. The performance...

90Active Deep Networks for SemiSupervised Sentiment Classification sigm 1 1 12e The derivative of the loglikelihood with respect to the model parameter w k can be obtained...

91Improvements to Training an RNN parserthe CKY algorithm to find the parse with the highest log likelihood score. For longer sentences, the scoring model may give...

92Extraction of Russian Sentiment Lexicon for Product MetaDomainto appear in each sentiment class we define a scaled loglikelihood: )( )( log)( wP cwP wLhc Scalability is required...

93Mining Words in the Minds of Second Language Learners: LearnerSpecific Word Difficultyw0 2 + ηa2 ∑u∈U a2u . (13) We define the negative log likelihood function of the proposed model as nll y,u, v def...

94Structured Term Recognition in Medical Text2012  Michael Glass,Alfio Gliozzo3: Logical description of the PCFG model i to j with log likelihood x . The proposal predicates are isRule, inCluster. The...

95Finding Thoughtful Comments from Social Media2012  Swapna Gottipati,Xing Jiangreference corpus. The lexical feature is defined as the log likelihood of the comment based on θr , calculated as:∑ w n(w...

96Exploiting CategorySpecific Information for MultiDocument Summarizationcategory. To quantify this difference, we applied the loglikelihood ratio test (LLR) (Dunning, 1993). The LLR of a word w...

97Semisupervised Representation Learning for Domain Adaptation using Dynamic Dependency Networksg., POS tags, for X i . Given the training data, its loglikelihood is L(θ) = N∑ i=1 log P(X i , Z i θ) where θ denotes...

98Quantifying Semantics using Complex Network Analysiswould be distributed independently. Here, we use the loglikelihood test (Dunning, 1993) to prune the network: We only draw...

99Adjective Deletion for Linguistic Steganography and Secret Sharing2012  ChingYun Chang,Stephen Clarknoun types and 792,914 adjective types. We also use the log likelihood ratio (LLR), an alternative to PMI, which is reported...

100Extraction of DomainSpecific Bilingual Lexicon from Comparable Corpora: Compositional Translation and RankingThe number of cooccurrences is normalized with the loglikelihood ratio (Dunning, 1993). Then, the vector of the source...