Article Structure
Abstract
We address the problem of translating from morphologically poor to morphologically rich languages by adding per-word linguistic information to the source language.
Introduction
Traditional statistical machine translation methods are based on mapping on the lexical level, which takes place in a local window of a few words.
Methods for enriching input
We selected to focus on noun cases agreement and verb person conjugation, since they were the most frequent grammatical errors of our baseline SMT system (see full error analysis in Figure 2).
Factored Model
The factored statistical machine translation model uses a log-linear approach, in order to combine the several components, including the language model, the reordering model, the translation models and the generation models.
Experiments
This preprocessing led to annotated source data, which were given as an input to a factored SMT system.
Conclusion
In this paper we have shown how SMT performance can be improved, when translating from English into morphologically richer languages, by adding linguistic information on the source.
Topics
translation model
- Our method is based on factored phrase-based statistical machine translation models .Page 1, “Introduction”
- Traditional statistical machine translation models deal with this problems in two ways:Page 2, “Introduction”
- Then, contrary to the methods that added only output features or altered the generation procedure, we used this information in order to augment only the source side of a factored translation model , assuming that we do not have resources allowing factors or specialized generation in the target language (a common problem, when translating from English into under-resourced languages).Page 3, “Introduction”
- Considering such annotation, a factored translation model is trained to map the word-case pair to the correct inflection of the target noun.Page 3, “Methods for enriching input”
- The factored statistical machine translation model uses a log-linear approach, in order to combine the several components, including the language model, the reordering model, the translation models and the generation models.Page 4, “Factored Model”
- Opposed to other factored translation model approaches that require target language factors, that are not easily obtainable for many languages, our approach only requires English syntax trees, which are acquired with widely available automatic parsers.Page 7, “Conclusion”
See all papers in Proc. ACL 2008 that mention translation model.
See all papers in Proc. ACL that mention translation model.
Back to top.
language model
- 0 The basic SMT approach uses the target language model as a feature in the argument maximisation function.Page 2, “Introduction”
- This language model is trained on grammatically correct text, and would therefore give a good probability for word sequences that are likely to occur in a sentence, while it would penalise ungrammatical or badly ordered formations.Page 2, “Introduction”
- Thus, with respect to these methods, there is a problem when agreement needs to be applied on part of a sentence whose length exceeds the order of the of the target n-gram language model and the size of the chunks that are translated (see Figure 1 for an exam-Page 2, “Introduction”
- The factored statistical machine translation model uses a log-linear approach, in order to combine the several components, including the language model , the reordering model, the translation models and the generation models.Page 4, “Factored Model”
- For testing the factored translation systems, we used Moses (Koehn et al., 2007), along with a 5-gram SRILM language model (Stolcke, 2002).Page 5, “Experiments”
See all papers in Proc. ACL 2008 that mention language model.
See all papers in Proc. ACL that mention language model.
Back to top.
phrase-based
- Our method is based on factored phrase-based statistical machine translation models.Page 1, “Introduction”
- 1.1 Morphology in Phrase-based SMTPage 1, “Introduction”
- 0 Meanwhile, in phrase-based SMT models, words are mapped in chunks.Page 2, “Introduction”
- The approach was clearly motivated by the problems occurring by a single-word-based SMT and have been solved by adopting a phrase-based model.Page 2, “Introduction”
- The Czech noun cases which appear only in prepositional phrases were ignored, since they are covered by the phrase-based model.Page 6, “Experiments”
See all papers in Proc. ACL 2008 that mention phrase-based.
See all papers in Proc. ACL that mention phrase-based.
Back to top.
machine translation
- Traditional statistical machine translation methods are based on mapping on the lexical level, which takes place in a local window of a few words.Page 1, “Introduction”
- Our method is based on factored phrase-based statistical machine translation models.Page 1, “Introduction”
- Traditional statistical machine translation models deal with this problems in two ways:Page 2, “Introduction”
- The factored statistical machine translation model uses a log-linear approach, in order to combine the several components, including the language model, the reordering model, the translation models and the generation models.Page 4, “Factored Model”
See all papers in Proc. ACL 2008 that mention machine translation.
See all papers in Proc. ACL that mention machine translation.
Back to top.
NIST
- Results were evaluated with both BLEU (Papineni et al., 2001) and NIST metrics ( NIST , 2002).Page 5, “Experiments”
- BLEU NIST set devtest test07 devtest test07 baseline 18.13 18.05 5.218 5.279 person 18.16 18.17 5.224 5.316Page 5, “Experiments”
- The NIST metric clearly shows a significant improvement, because it mostly measures difficult n-gram matches (e. g. due to the long-distance rules we have been dealing with).Page 5, “Experiments”
- BLUE NIST set devtest test devtest test baseline 12.08 12.34 4.634 4.865 person+case altpachPOS 11.98 11.99 4.584 4.801 person 12.23 12.11 4.647 4.846 altpathzword casePage 6, “Experiments”
See all papers in Proc. ACL 2008 that mention NIST.
See all papers in Proc. ACL that mention NIST.
Back to top.
POS tag
- In one of the first efforts to enrich the source in word-based SMT, Ueffing and Ney (2003) used part-of-speech (POS) tags, in order to deal with the verb conjugation of Spanish and Catalan; so, POS tags were used to identify the pronoun+verb sequence and splice these two words into one term.Page 2, “Introduction”
- In their presentation of the factored SMT models, Koehn and Hoang (2007) describe experiments for translating from English to German, Spanish and Czech, using morphology tags added on the morphologically rich side, along with POS tags .Page 2, “Introduction”
- The POS tag of this noun is then used to identify if it is plural or singular.Page 4, “Methods for enriching input”
- The word “aspects” is found, which has a POS tag that shows it is a plural noun.Page 4, “Methods for enriching input”
See all papers in Proc. ACL 2008 that mention POS tag.
See all papers in Proc. ACL that mention POS tag.
Back to top.
statistical machine translation
- Traditional statistical machine translation methods are based on mapping on the lexical level, which takes place in a local window of a few words.Page 1, “Introduction”
- Our method is based on factored phrase-based statistical machine translation models.Page 1, “Introduction”
- Traditional statistical machine translation models deal with this problems in two ways:Page 2, “Introduction”
- The factored statistical machine translation model uses a log-linear approach, in order to combine the several components, including the language model, the reordering model, the translation models and the generation models.Page 4, “Factored Model”
See all papers in Proc. ACL 2008 that mention statistical machine translation.
See all papers in Proc. ACL that mention statistical machine translation.
Back to top.
n-gram
- Thus, with respect to these methods, there is a problem when agreement needs to be applied on part of a sentence whose length exceeds the order of the of the target n-gram language model and the size of the chunks that are translated (see Figure 1 for an exam-Page 2, “Introduction”
- The NIST metric clearly shows a significant improvement, because it mostly measures difficult n-gram matches (e. g. due to the long-distance rules we have been dealing with).Page 5, “Experiments”
- In n-gram based metrics, the scores for all words are equally weighted, so mistakes on crucial sentence constituents may be penalized the same as errors on redundant or meaningless words (Callison-Burch et al., 2006).Page 6, “Experiments”
See all papers in Proc. ACL 2008 that mention n-gram.
See all papers in Proc. ACL that mention n-gram.
Back to top.
noun phrases
- Take the example of translating noun phrases from English to Greek (or German, Czech, etc.).Page 1, “Introduction”
- However, Greek words in noun phrases are inflected based on their role in the sentence.Page 1, “Introduction”
- A purely lexical mapping of English noun phrases to Greek noun phrases suffers from the lack of information about its role in the sentence, making it hard to choose the right inflected forms.Page 1, “Introduction”
See all papers in Proc. ACL 2008 that mention noun phrases.
See all papers in Proc. ACL that mention noun phrases.
Back to top.
