--- title: "Using a Dependency Parser to Improve SMT for" description: "Phrase-based SMT struggles when the source and target languages order verbs and… 5% BLEU. 6 references, cited by 62 papers here." authors: ["Peng Xu", "Jaeho Kang", "Michael Ringgaard", "Franz Och"] year: 2009 venue: "Proceedings of Human Language Technologies: The 2009 Annual Conference of the North American Chapter of the Association for Computational Linguistics" source: "https://aclanthology.org/N09-1028/" licence: "CC BY-NC-SA 3.0 (ACL Anthology, before 2016)" licence_url: "https://creativecommons.org/licenses/by-nc-sa/3.0/" summary_licence: "Elamur Cosmic License" canonical: "https://research.elamur.ai/free/en/papers/xu-2009-dependency-parser-improve-smt-subject/" --- [Elamur Research](https://md.research.elamur.ai/free/en/) · [this page in HTML](https://research.elamur.ai/free/en/papers/xu-2009-dependency-parser-improve-smt-subject/) [home](https://md.research.elamur.ai/free/en/) / [papers](https://md.research.elamur.ai/free/en/papers/) / [2009](https://md.research.elamur.ai/free/en/papers/2009/) / Using a Dependency Parser to Improve SMT for Subject-Object-Verb Languages # Using a Dependency Parser to Improve SMT for Subject-Object-Verb Languages **Authors:** [Peng Xu](https://md.research.elamur.ai/free/en/authors/peng-xu/), Jaeho Kang, Michael Ringgaard and [Franz Och](https://md.research.elamur.ai/free/en/authors/franz-josef-och/) · **Year:** 2009 · **Venue:** [Proceedings of Human Language Technologies: The 2009 Annual Conference of the North American Chapter of the Association for Computational Linguistics](https://md.research.elamur.ai/free/en/venues/naacl/) ([NAACL 2009](https://md.research.elamur.ai/free/en/venues/naacl/2009/)) · **Publisher:** [Association for Computational Linguistics](https://md.research.elamur.ai/free/en/publishers/association-for-computational-linguistics/) · **Licence:** [CC BY-NC-SA 3.0 (ACL Anthology, before 2016)](https://creativecommons.org/licenses/by-nc-sa/3.0/) **Links:** [Publisher's page](https://aclanthology.org/N09-1028/) · [PDF](https://aclanthology.org/N09-1028.pdf) · [OpenAlex](https://openalex.org/W2108460050) **Citations:** 118 citations (OpenAlex, 2026-10-14) · [62](https://md.research.elamur.ai/free/en/papers/xu-2009-dependency-parser-improve-smt-subject/cited-by/) in our library, of the 141,840 papers whose bibliographies we read ## Summary Phrase-based SMT struggles when the source and target languages order verbs and arguments very differently. Translating from English (SVO) into SOV languages forces the decoder to make long jumps, and decoders usually cap those jumps for speed. Xu et al. avoid this by rearranging English sentences before translation. Each sentence is parsed with a dependency parser, and hand-written "precedence rules" then reorder each head's children so the sentence roughly follows SOV order. The rules were written from English–Korean examples by a bilingual speaker. The same reordering is applied to both training and input data, so decoding cost does not increase, and the standard reordering models can still handle local word shifts. The authors tested the method on English→Korean, Japanese, Hindi, Urdu and Turkish, with training data ranging from about 17M to over 350M words. It gave statistically significant BLEU gains over a strong phrase-based baseline for all five languages. Combined with a hierarchical (SAMT) system, it also improved four of the five languages. This makes cheap, linguistically informed preprocessing an attractive choice for fast, real-time MT. ## Key points - **Rule format:** each rule associates a head POS tag T with a collection of (L, W, O) triples. - L is a child's dependency label, W is a real-valued weight, and O is NORMAL or REVERSE. - Children are sorted by descending weight; O only decides how ties are ordered. - A special `self` label lets the head itself be positioned. - Children with unlisted labels default to weight 0 and NORMAL order. - **Verb rule (VB\*):** - `advcl` moves to the front. - `nsubj`, `prep` and unlisted children stay in their original order. - `dobj` comes next. - The verb group (`prt`, `aux`, `auxpass`, `neg`, `self`) goes last, with its internal order reversed. - Separate rules cover adjectives (JJ\*), nouns (NN/NNS, e.g. `prep`, `rcmod` before the head) and prepositions (IN/TO, `pobj` before the head). - **Reordering distance (English→Korean):** without the rules, BLEU keeps rising as the allowed reordering distance grows. With the rules, the system gains almost 5% BLEU at the same decoding speed. - **Hierarchical setup (SAMT):** rules had at most 6 source-side items, initial phrases up to 12 words, and 4-gram LMs. Blindtest BLEU, PR+Hier vs. Hier: | Language | Hier | PR+Hier | Significant? | |---|---|---|---| | Korean | 27.9 | 28.8 | yes | | Japanese | 30.5 | 31.1 | yes | | Urdu | 10.0 | 10.7 | yes | | Turkish | 10.5 | 10.9 | yes | | Hindi | 19.3 | 19.3 | no gain | - **Main weakness:** the authors acknowledge that parser errors hurt some sentences. ## Interesting ideas - **Weight-sorted ordering of children.** Ordering a head's children by numeric weight, with a tie-breaking flag, is far more compact than listing explicit permutations. Default weights make the rules degrade gracefully on labels nobody anticipated. The same design could suit any tree-based linearization or word-order transformation task. - **Reorder training data too, not just inputs.** Making the source look like the target's word order also simplified EM training and, by the authors' reasoning, improved word alignment. This trick can help alignment-dependent pipelines for any language pair with divergent word order. - **Preprocessing stacks with existing models.** Because the rules run before training, they can be combined with phrase-based reordering models or a hierarchical system without changing the decoder. This is useful when a change to the core engine is costly. - **Evaluate quality against the speed knob.** Plotting BLEU against the maximum allowed reordering distance shows the quality–speed trade-off directly, instead of a single score. It is a good evaluation pattern for any system with a search-budget parameter. - **Rules from one language pair, applied to a typological class.** Rules built from English–Korean examples carried over to four other SOV languages. This suggests that typology-level rules can bootstrap low-resource language pairs before any automatic rule extraction is attempted. *Our summary of the paper, in our own words — not its text. The work belongs to its authors; read it at the links above.* ## Institutions - [Google](https://md.research.elamur.ai/free/en/companies/google/) — Company, Mountain View, United States · 353 works · 364,806 citations 📚 Shelves (4): [🖥️ Computer Science](https://md.research.elamur.ai/free/en/shelves/computer-science/) · 9,086 works · [🤖 Artificial Intelligence](https://md.research.elamur.ai/free/en/shelves/computer-science/artificial-intelligence/) · 8,889 works · [🎭 Arts and Humanities](https://md.research.elamur.ai/free/en/shelves/arts-and-humanities/) · 9,296 works · [🗣️ Language and Linguistics](https://md.research.elamur.ai/free/en/shelves/arts-and-humanities/language-and-linguistics/) · 9,206 works 🧭 Topics (2): [🔁 Machine translation](https://md.research.elamur.ai/free/en/topics/machine-translation/) · 958 works · [🌳 Dependency parsing](https://md.research.elamur.ai/free/en/topics/dependency-parsing/) · 120 works ## Cite Cite as: Peng Xu, Jaeho Kang, Michael Ringgaard and Franz Och. 2009. Using a Dependency Parser to Improve SMT for Subject-Object-Verb Languages. Proceedings of Human Language Technologies: The 2009 Annual Conference of the North American Chapter of the Association for Computational Linguistics. The paper itself, not this page. BibTeX from the library's record of the work. ```bibtex @inproceedings{xu2009using, title = {{Using a Dependency Parser to Improve SMT for Subject-Object-Verb Languages}}, author = {Peng Xu, Jaeho Kang, Michael Ringgaard and Franz Och}, booktitle = {Proceedings of Human Language Technologies: The 2009 Annual Conference of the North American Chapter of the Association for Computational Linguistics}, year = {2009}, doi = {10.3115/1620754.1620790}, url = {https://doi.org/10.3115/1620754.1620790} } ``` - [← previous A Unified Model of Phrasal and Sentential Evidence for Information Extraction](https://md.research.elamur.ai/free/en/papers/patwardhan-2009-unified-model-phrasal-sentential-evidence/) - [next → Shared Logistic Normal Distributions for Soft Parameter Tying in Unsupervised Grammar Induction](https://md.research.elamur.ai/free/en/papers/cohen-2009-shared-logistic-normal-distributions-soft/) - [↑ back to papers of 2009](https://md.research.elamur.ai/free/en/papers/2009/) ## References 6 references, 2 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet. 1. Yaser Al-Onaizan and Kishore Papineni 2006. Distortion Models for Statistical Machine Translation In Proceedings of ACL Pi-Chuan Chang and Kristina Toutanova 2007. A Discriminative Syntactic Word Order Model for Machine Translation In Proceedings of ACL David Chiang 2005. A Hierarchical Phrase-based… [find](https://scholar.google.com/scholar?q=Yaser+Al%2DOnaizan+and+Kishore+Papineni+2006%2E+Distortion+Models+for+Statistical+Machine+Translation+In+Proceedings+of+ACL+Pi%2DChuan+Chang+and+Kristina+Toutanova+2007%2E+A+Discriminative+Syntactic+Word+Orde) 2. Joakim Nivre and Mario Scholz 2004. Deterministic Dependency Parsing for English Text. In Proceedings of COLING Franz J. Och 2002. Statistical Machine Translation: From Single Word Models to Alignment Template Ph.D. [find](https://scholar.google.com/scholar?q=Joakim+Nivre+and+Mario+Scholz+2004%2E+Deterministic+Dependency+Parsing+for+English+Text%2E+In+Proceedings+of+COLING+Franz+J%2E+Och+2002%2E+Statistical+Machine+Translation%3A+From+Single+Word+Models+to+Alignment) 3. Thesis, RWTH Aachen, Germany Franz J. Och. 2003. Minimum Error Rate Training in Statistical Machine Translation. In Proceedings of ACL Franz J. Och and Hermann Ney 2004. The Alignment Template Approach to Statistical Machine Translation. Computational Linguistics, 30:417-449 [find](https://scholar.google.com/scholar?q=Thesis%2C+RWTH+Aachen%2C+Germany+Franz+J%2E+Och%2E+2003%2E+Minimum+Error+Rate+Training+in+Statistical+Machine+Translation%2E+In+Proceedings+of+ACL+Franz+J%2E+Och+and+Hermann+Ney+2004%2E+The+Alignment+Template+Approac) 4. Kishore Papineni, Roukos, Salim et al. 2002. [BLEU: A Method for Automatic Evaluation of Machine Translation](https://md.research.elamur.ai/free/en/papers/papineni-2002-bleu-method-automatic-evaluation-machine/). In Proceedings of ACL Chris Quirk, Arul Menezes and Colin Cherry 2005. Dependency Tree Translation: Syntactically Informed Phrasal SMT In Proceedings of ACL Christoph Tillmann 2004. A Block O… 5. Fei Xia and Michael McCord 2004. Improving a Statistical MT System with Automatically Learned Rewrite Patterns In Proceedings of COLING Deyi Xiong, Qun Liu and Shouxun Lin 2006. Maximum Entropy Based Phrase Reordering Model for Statistical Machine Translation In Proceedings of COLING-ACL Kenji Yamad… [find](https://scholar.google.com/scholar?q=Fei+Xia+and+Michael+McCord+2004%2E+Improving+a+Statistical+MT+System+with+Automatically+Learned+Rewrite+Patterns+In+Proceedings+of+COLING+Deyi+Xiong%2C+Qun+Liu+and+Shouxun+Lin+2006%2E+Maximum+Entropy+Based+) 6. Andreas Zollmann and Ashish Venugopal 2006. [Syntax Augmented Machine Translation via Chart Parsing](https://md.research.elamur.ai/free/en/papers/zollmann-2006-syntax-augmented-machine-translation-chart/) In Proceedings of NAACL 2006 - Workshop on Statistical Machine Translation Andreas Zollmann, Ashish Venugopal, Franz Och and Jay Ponte 2008. A Systematic Comparison of Phrase-Based, Hierarchical and Sy… *Updated 24 October 2026*