TAILIEUCHUNG - Báo cáo khoa học: "Word Alignment via Submodular Maximization over Matroids"

We cast the word alignment problem as maximizing a submodular function under matroid constraints. Our framework is able to express complex interactions between alignment components while remaining computationally efficient, thanks to the power and generality of submodular functions. We show that submodularity naturally arises when modeling word fertility. Experiments on the English-French Hansards alignment task show that our approach achieves lower alignment error rates compared to conventional matching based approaches. . | Word Alignment via Submodular Maximization over Matroids Hui Lin Dept. of Electrical Engineering University of Washington Seattle WA 98195 USA hlin@ Jeff Bilmes Dept. of Electrical Engineering University of Washington Seattle Wa 98195 USA bilmes@ Abstract We cast the word alignment problem as maximizing a submodular function under matroid constraints. Our framework is able to express complex interactions between alignment components while remaining computationally efficient thanks to the power and generality of submodular functions. We show that submodularity naturally arises when modeling word fertility. Experiments on the English-French Hansards alignment task show that our approach achieves lower alignment error rates compared to conventional matching based approaches. 1 Introduction Word alignment is a key component in most statistical machine translation systems. While classical approaches for word alignment are based on generative models . IBM models Brown et al. 1993 and HMM Vogel et al. 1996 word alignment can also be viewed as a matching problem where each word pair is associated with a score reflecting the desirability of aligning that pair and the alignment is then the highest scored matching under some constraints. Several matching-based approaches have been proposed in the past. Melamed 2000 introduces the competitive linking algorithm which greedily constructs matchings under the one-to-one mapping assumption. In Matusov et al. 2004 matchings are found using an algorithm for constructing a maximum weighted bipartite graph matching Schrijver 2003 where word pair scores come from alignment posteriors of generative models. Similarly Taskar et al. 2005 cast word alignment as a maximum weighted matching problem and propose a 170 framework for learning word pair scores as a function of arbitrary features of that pair. These approaches however have two potentially substantial limitations words have fertility of at most .

• Báo cáo khoa học
• Word Alignment
• Submodular Maximization over Matroids
• Hui Lin
• báo cáo khoa học
• báo cáo ngôn ngữ
• ngôn ngữ tự nhiên
• Smaller Alignment Models for Better Translations
• Unsupervised Word Alignment with the 0
• Ashish Vaswani Liang Huang David Chiang
• Alignment Model Adaptation
• Domain Specific Word Alignment
• WU Hua
• mô hình ngôn ngữ
• xử lý ngôn ngữ tự nhiên
• Word Alignment Combination
• Multiple Word Segmentation
• Ning Xi
• Guangchao Tang
• Bootstrapping Word Alignment
• Word Packing
• Yanjun Ma
• Nicolas Stroppa
• Sentence alignment
• Parallel corpora
• word clustering
• Natural language processing
• Alignment algorithms
• IBM model
• Hierarchical Search
• Long Papers
• Diversify and Combine
• Improving Word Alignment
• Machine Translation
• Low Resource Languages
• Synonym Regularization
• Hiroyuki Shindo
• Supervised Word Alignment
• Jinxi Xu and Jinying Chen
• Active Learning Based Elicitation
• Semi Supervised Word Alignment
• Vamshi Ambati
• Dealing with Spurious Ambiguity
• Learning ITG based Word Alignment
• Stephan Vogel
• Confidence Measure
• Fei Huang
• Unsupervised Word Alignment
• Arbitrary Features
• Chris Dyer Jonathan
• An Algorithm
• Unsupervised Transliteration Mining
• Application to Word Alignment
• Data Cleaning
• Tsuyoshi Okita
• Mining Parenthetical Translations
• Web by Word Alignment
• Dekang Lin
• Discriminative Word Alignment
• Conditional Random Fields
• Phil Blunsom and Trevor Cohn
• Improved Discriminative
• Bilingual Word Alignment
• Robert C
• Moore Wen tau Yih
• Word Alignment in English Hindi Parallel Corpus
• Recency Vector Approach
• Some Studies
• Soft Syntactic Constraints
• Discriminative Training
• Semi Supervised Training
• Statistical Word Alignment
• Alexander Fraser
• Yet Another Word Alignment Tool
• Ulrich Germann
• Optimizing Word Alignment Combination
• Phrase Table Training
• Yonggang Deng and Bowen Zhou
• Finding Synonyms
• Automatic Word Alignment
• Measures of Distributional Similarity