Báo cáo khoa học: "Adaptive Transformation-based Learning for Improving Dictionary Tagging" pdf

... Adaptive Transformation-based Learning for Improving Dictionary TaggingBurcu Karagol-Ayan, David Doermann, and Amy WeinbergInstitute for Advanced Computer Studies (UMIACS)University ... by 2.77% for tokens and 2.27% for phrases for the Cebuano dictionary. As for the Iraqi Ara-bic dictionary, using TBL(font) and GT(font) re-sulted in an improvement of 0.57% and 0.85% for tokens ... allowing for a relaxed matching of OCRed output to tags. For each dictionary, a human operator must spec-ify the lexicographic information used in that par-ticular dictionary, along with the clues for...

Báo cáo khoa học: "Heterogeneous Transfer Learning for Image Clustering via the Social Web" docx

... hetero-geneous transfer learning relates to other types of learning, Figure 1 presents an intuitive illustrationof four learning strategies, including traditionalmachine learning, transfer learning across ... in Transfer Learning In the past, several other works made use of trans-fer learning for cross-feature-space learning. Wuand Oard (2008) proposed to handle the cross-language learning problem ... distributions, multi-view learning and hetero-geneous transfer learning. As we can see, animportant distinguishing feature of heterogeneoustransfer learning, as compared to other types of learning, is...

Báo cáo khoa học: "Semi-supervised Learning for Natural Language Processing" pptx

...

Báo cáo khoa học: "Advanced Online Learning for Natural Language Processing" pot

...

Báo cáo khoa học: "Structural Correspondence Learning for Parse Disambiguation" pot

... Greece, 2 April 2009.c2009 Association for Computational LinguisticsStructural Correspondence Learning for Parse DisambiguationBarbara PlankAlfa-informaticaUniversity of Groningen, The Netherlandsb.plank@rug.nlAbstractThe ... ofStructural Correspondence Learning (SCL) (Blitzer et al., 2006) for domainadaptation of a stochastic attribute-valuegrammar (SAVG). So far, SCL hasbeen applied successfully in NLP for Part-of-Speech ... need for domain adaptation arises in manyNLP tasks: Part-of-Speech tagging, SentimentAnalysis, Semantic Role Labeling or StatisticalParsing, to name but a few. For example, the per-formance...

Báo cáo khoa học: "Hierarchical Sequential Learning for Extracting Opinions and their Attributes" ppt

... construc-tion. For instance, one can add λδgI(δ, x, i)and λ′δ,ˆδg′I(δ,ˆδ, x, i) to Equation (1) for δ ∈{0, 1, , 9}, in order to allow more individualized learning for each label.Notice ... g′Oare feature vectors deﬁned for Opinion extraction, gPand g′Pare feature vectorsdeﬁned for Polarity extraction, and gSand g′Sarefeature vectors deﬁned for Strength extraction, andα, ... pages 269–274,Uppsala, Sweden, 11-16 July 2010.c2010 Association for Computational LinguisticsHierarchical Sequential Learning for Extracting Opinions and theirAttributesYejin Choi and Claire...

Báo cáo khoa học: "Semi-supervised Learning for Automatic Prosodic Event Detection Using Co-training Algorithm" doc

... data.Finally we compared our co-training perfor-mance with supervised learning. For supervised learning, all labeled utterances except for the testset are used for training. We used most conﬁ-dent ... co-training is how toselect informative samples. Active learning ap-proaches, such as Muslea et al. (2000), can gener-ally select more informative samples, for example,samples for which two classiﬁers ... anaccuracy of 84% for accent, 71% for IPB, and84% for break index detection at the syllablelevel. Chen et al. (2004) used a Gaussian mix-ture model for acoustic-prosodic information andneural...

Báo cáo khoa học: "Weakly Supervised Learning for Hedge Classiﬁcation in Scientiﬁc Literature" pot

... such ideas can be adapted for this task isone of many avenues for future research.10.6 Error AnalysisSome errors are due to the variety of hedge forms. For example, the learning models were unsuccess-ful ... avenues for future research.1 IntroductionThe automatic processing of scientiﬁc papers usingNLP and machine learning (ML) techniques is anincreasingly important aspect of technical informat-ics. ... probabilistic weakly supervised learning model and use it to motivate our ap-proach.4. We analyze our learning model experimentallyand report promising results for the task on anew publicly-available...

Báo cáo khoa học: "Unsupervised Multilingual Learning for Morphological Segmentation" potx

... pages 737–745,Columbus, Ohio, USA, June 2008.c2008 Association for Computational LinguisticsUnsupervised Multilingual Learning for Morphological SegmentationBenjamin Snyder and Regina BarzilayComputer ... of information for automatic lin-guistic analysis as well. In this paper we investi-gate two questions: (i) Can we exploit cross-lingualcorrespondences to improve unsupervised language learning? ... (the basicunits of meaning). For example, the English wordmisunderstanding would be segmented into mis -understand - ing. This task is an informative testbed for our exploration, as strong...

Báo cáo khoa học: "Adaptive Language Modeling for Word Prediction" potx

... interpolation,we must also choose whether smoothing (a prereq-uisite for backoff) is performed before or after theinterpolation. If we smooth before the interpolation,then the frequencies will be overly ... approach for LSA-adapted modeling, how-ever, we found this approach to be inferior to di-rect estimation of the posterior probability for wordprediction (Trnka et al., 2006). Part of the reason for ... re-search by forcing trigram predictions to occur overbigrams and so on (rather than backoff) and usingthe topic-adapted model for re-ranking within eachset of predictions, but found that the forced...

Xem thêm