Báo cáo khoa học: "Deep Learning for NLP" pptx

...

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

... our learning algorithm is the strategyused to select samples for training.In general, this novel learning framework lies be-tween supervised learning and reinforcement learn-ing. Guided learning ... classiﬁcation are dynamically incorporated in the learning phase.Guided learning is not as hard as reinforcement learning. At each local step in learning, we alwaysknow the undesirable labeling ... performs at most n iterations, and its run-ning time is therefore O(B2n), that is, linear in thelength of the input sequence.2.3 Learning AlgorithmIn this section, we propose guided learning, ...

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

... systemsadopting the standard machine learning approach,outperforming them by as much as 4–7% on thethree data sets for one of the performance metrics.2 Related WorkAs mentioned before, our approach differs ... and Cardie (2002b) for details. Learning algorithms. We consider three learning algorithms, namely, the C4.5 decision tree inductionsystem (Quinlan, 1993), the RIPPER rule learning algorithm ... we do not relyon additional labeled data for learning the ranker;instead, we use half of the training texts for trainingclassiﬁers and the other half for ranking purposes.Results using our...

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

... applying it toour speciﬁc task.3.1 Learning ModelThe framework from which HM-SVM are derivedis a maximum margin formulation for joint fea-ture functions in kernel learning setting. Given nlabeled ... satisfactory for practical use.The limitation is even more pertinent for the lan-guages that do not have sophisticated parsing re-sources, or languages that have inherent linguisticchallenges for ... Chinese). It is there-fore worthwhile to investigate alternatives to func-tion labeling for languages under the parsing bot-tleneck, both in terms of features used and effec-tive learning algorithms.In...

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

... Linguistics, pages 25–32,Prague, Czech Republic, June 2007.c2007 Association for Computational LinguisticsTransductive learning for statistical machine translationNicola UeffingNational Research Council ... Algorithm 1 we run it for a fixed numberof iterations and instead focus on finding useful def-initions for Estimate, Score and Select that can beexperimentally shown to improve MT performance.3.2 The ... translations contained in the N-best list. For details, see (Ueffing and Ney, 2007).Again, the single values are multiplied to obtain ascore for the whole sentence. For NIST, the lan-guage model score...

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

... paper we study spectral learning methods for non-deterministic split head-automata grammars, a powerful hidden-state formalism for dependency parsing.We present a learning algorithm that, ... parsing and learning algorithms for grammatical bigram models. Techni-cal Report UCB/CSD-01-1148, University of Cali-fornia, Berkeley.Slav Petrov and Dan Klein. 2007. Improved infer-ence for unlexicalized ... spectral algorithm for learning hidden markovmodels. In COLT 2009 - The 22nd Conference on Learning Theory.Gabriel Infante-Lopez and Maarten de Rijke. 2004.Alternative approaches for generating...

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

... contrast, our emphasisis on learning language by proactively interactingwith an external environment.Reinforcement Learning for Language Pro-cessing Reinforcement learning has been previ-ously ... divided into 70 for training, 18 for development, and 40 for test. Inthe puzzle game domain, we use 50 tutorials,divided into 40 for training and 10 for test.9Statistics for the datasets ... user.Reinforcement learning is applied very differ-ently in dialogue systems compared to our setup.In some respects, our task is more easily amenableto reinforcement learning. For instance,...

Báo cáo khoa học: "Active Learning for Statistical Natural Language Parsing" potx

... clusteringresults in a better learning curve.4.4 Summary ResultFigure 8 shows the best active learning result comparedwith that of random selection. The learning curve for ac-tive learning is obtained ... avoid the unwanted bias toward smallclusters. For cluster, the weight for sample is proportional to .Weight by Performance: The idea of weight byperformance is to focus the model on its weaknesswhen ... toachieve a satisfactory performance using active learning. Active learning has been studied in the context of manynatural language processing (NLP) applications such asinformation extraction(Thompson...

Báo cáo khoa học: "Active Learning for Multilingual Statistical Machine Translation∗" ppt

... 2: The performance of different sentence selection strategies as the iteration of AL loop goes on for three translationtasks. Plots show the performance of sentence selection methods for single ... for system combination, thesource languages for each of our MT systems aredifferent, which rules out a set of popular methods for obtaining consensus translations which assumetranslation for ... importance of the different transla-tion tasks anddαd= 1. AL-SMT formulation for single language pair is a special case of thisformulation where only one of the αd’s in the ob-jective function...

Báo cáo khoa học: "Discriminative Learning for Joint Template Filling" pptx

... the Association for Computational Linguistics, pages 845–853,Jeju, Republic of Korea, 8-14 July 2012.c2012 Association for Computational LinguisticsDiscriminative Learning for Joint Template ... for each tar-get relation, an integrated model to weight the dif-ferent extraction options, including for example thelikely lengths for events, or the fact that start timesshould come before ... features. In addition, weenforce label consistency and semantic coherence atrecord level.Other related works perform structured relationdiscovery for different settings of information ex-traction....

Xem thêm