Báo cáo khoa học: "Pattern Learning for Relation Extraction with a Hierarchical Topic Model" pptx

... notgeneralizable across all pairs with the same relation. Finally φ A contains the patterns that are observedacross most pairs related with the same relation. φ A is the topic model of interest for ... that are expressingthe relation and those that are ambiguous and can beapplied across relations. In this way, high-precision extraction patterns can be learned without the needof any manual ... for Relation Extraction with a Hierarchical Topic ModelEnrique Alfonseca Katja Filippova Jean-Yves DelortGoogle ResearchBrandschenkestrasse 1108002 Zurich, Switzerland{ealfonseca,katjaf,jydelort}@google.comGuillermo...

Tài liệu Báo cáo khoa học: "Composite Kernels For Relation Extraction" pdf

... innatural text can be expressed in many ways, e.g.“Obama was educated at Harvard”, “Obama is a graduate of Harvard Law School”, or, “Obamawent to Harvard College”. Relation extraction aims at ... for Relation Extraction Relation extraction aims at learning a relation from a number of positive and negative instancesin natural language sentences. As a classiﬁer weuse Support Vector Machines ... GermanyHannes KorteFraunhofer IAISSt. Augustin, Germany{frank.reichartz,hannes.korte,gerhard.paass}@iais.fraunhofer.deGerhard PaassFraunhofer IAISSt. Augustin, GermanyAbstractThe automatic...

Báo cáo khoa học: "Distant supervision for relation extraction without labeled data" docx

... Freebase relation dataduring training, and comparing newly discovered relation instances against this held-out data, andmanually, having humans who look at each posi-tively labeled entity pair ... extracting relational facts fromtext (for example, learning that a person is em-ployed by a particular organization, or that a ge-ographic entity is located in a particular region).In supervised approaches, ... evaluation to a small number of relation instances and corpora of less than a mil-lion words.Many early algorithms for relation extraction used little or no syntactic information. For ex-ample,...

Báo cáo khoa học: "Automated Whole Sentence Grammar Correction Using a Noisy Channel Model" pptx

... erroneous01 a: a/0.62an:an/0.43cat:cat/14ear:ear/15ε:ε/1ε:ε/10 a: a/0.95[0.95,0] a: an/0.05[0,0.05]an:an/1cat:cat/1ear:ear/101 a: a/0.57[0.57,0] a: an/0.03[0,0.03]2an:an/0.43cat:cat/14ear:ear/15ε:ε/1ε:ε/1Figure ... thetwelfth national conference on Artificial intelli-gence (vol. 1), AAAI ’94, pages 779–784, MenloPark, CA, USA. American Association for Artifi-cial Intelligence.Lavie, A. and Agarwal, A. (2007). ... extent, but many approaches are notsufficient for real-world automated grammar correc-tion for various of reasons. First, as we have alreadymentioned, classification systems and generation-based systems...

Tài liệu Báo cáo khoa học: "Guided Learning for Bidirectional Sequence Classiﬁcation" ppt

... spans in S as con-text, and replace it with a new candidate span takingp(and another accepted span) as context. Wealwaysmaintain B different states for each span.Back to the previous example, ... the learning phase.Guided learning is not as hard as reinforcement learning. At each local step in learning, we alwaysknow the undesirable labeling actions according tothe gold standard, although ... Iright) with a left and a rightinterface is called a state. We partition the hypothe-ses associated with span p into sets compatible with the same state. In practice, for span p, we use a ma-trix...

Tài liệu Báo cáo khoa học: "Machine Learning for Coreference Resolution: From Local Classiﬁcation to Global Ranking" ppt

... creating training instances.In McCarthy and Lehnert’s method, a positiveinstance is created for each anaphoric NP paired with each of its antecedents, and a negative instanceis created by pairing ... the ACE research program.Our approach. Recall that our approach uses la-beled data to train both the coreference classiﬁersand the ranking model. To ensure a fair comparisonof our approach with ... feature sets, and the training instance creation methods discussed in Section 3.1.3.2 Learning to Rank Candidate PartitionsWe train an SVM-based ranker for ranking candidatepartitions by means...

Báo cáo khoa học: "Accurate Learning for Chinese Function Tags from Minimal Features" pdf

... Merlo andMusillo, 2005; Gildea and Palmer, 2002) havealready been proposed. (Blaheta and Charniak,2000; Blaheta, 2004) described a statistical sys-tem trained on the data of Penn Treebank to au-tomatically ... predicate-argument structure,while adverbials are actually semantically orientedlabels (though not true for all cases, see (Merloand Palmer, 2006)) that carry semantic role infor-mation.As for ... that are decorated with the same function label.55tags. Punctuation marks like comma, semi-colonand period that separate sentences are also denotedas “O”. But the punctuation that appear within...

Báo cáo khoa học: "Transductive learning for statistical machine translation" potx

... statistical machine translationNicola UefﬁngNational Research Council CanadaGatineau, QC, Canadanicola.ueffing@nrc.gc.caGholamreza Haffari and Anoop SarkarSimon Fraser UniversityBurnaby, BC, Canada{ghaffar1,anoop}@cs.sfu.caAbstractStatistical ... BC, Canada{ghaffar1,anoop}@cs.sfu.caAbstractStatistical machine translation systems areusually trained on large amounts of bilin-gual text and monolingual text in the tar-get language. In ... co-training is appliedto MT. This approach requires several source lan-guages which are sentence-aligned with each otherand all translate into the same target language. Onelanguage pair creates...

Báo cáo khoa học: "Spectral Learning for Non-Deterministic Dependency Parsing" ppt

... Córdobaand CONICETCórdoba X5000HUA, Argentinafrancolq@famaf.unc.edu.arAriadna Quattoni and Borja Balle and Xavier CarrerasUniversitat Politècnica de CatalunyaBarcelona E-08034{aquattoni,bballe,carreras}@lsi.upc.eduAbstractIn ... ReferencesRaphael Bailly. 2011. Quadratic weighted automata:Spectral algorithm and likelihood maximization.JMLR Workshop and Conference Proceedings –ACML.James K. Baker. 1979. Trainable grammars for ... E-08034{aquattoni,bballe,carreras}@lsi.upc.eduAbstractIn this paper we study spectral learning methods for non-deterministic split head-automata grammars, a powerful hidden-state formalism for dependency parsing.We...

Báo cáo khoa học: "Reinforcement Learning for Mapping Instructions to Actions" pdf

... patterns. In contrast, our emphasisis on learning language by proactively interacting with an external environment.Reinforcement Learning for Language Pro-cessing Reinforcement learning has ... ac-tions that emit natural language utterances. Thereinforcement learning state space encodes infor-mation about the goals of the user and what theysay at each time step. The learning problem ... immediate reward,which is available after each action, and delayedreward, which does not provide feedback until thelast action. For example, task completion is a de-layed reward that produces a...

Xem thêm