A Hybrid Morpheme-Word Representation for Machine Translation of Morphologically Rich Languages

By Minh-Thang Luong, Preslav Nakov, Min-Yen Kan. 2010

49 citations recorded by the source corpus.

We propose a language-independent approach for improving statistical machine translation for morphologically rich languages using a hybrid morpheme-word representation where the basic unit of translation is the morpheme, but word boundaries are respected at all stages of the translation process. Our model extends the classic phrase-based model by means of (1) word boundary-aware morpheme-level phrase extraction, (2) minimum error-rate training for a morpheme-level translation model using word-level BLEU, and (3) joint scoring with morpheme- and word-level language models. Further improvements are achieved by combining our model with the classic one. The evaluation on English to Finnish using Europarl (714K sentence pairs; 15.5M English words) shows statistically significant improvements over the classic model based on BLEU and human judgments.

Subfields: Artificial Intelligence, Machine Translation

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Machine Translation

Citations in the corpus

What do Neural Machine Translation Models Learn about Morphology? (2017, methodology): “Characters and other sub-word units have become increasingly popular in neural MT, although they had also been used in phrase-based MT for handling morphologically-rich (Luong et al., 2010) or closely related language pairs (Durrani et al., 2010; Nakov and Tiedemann, 2012).”
Sheffield Systems for the Finnish-English WMT Translation Task (2015, background): “Luong et al. (2010) also investigate morpheme-level extraction, but integrate this into the decoding process itself, instead of the pre-processing step we have.”
On internal language representations in deep learning: an analysis of machine translation and speech recognition (2018, methodology): “2An earlier approach combining word and sub-word units for statistical machine translation is found in [220].”
Character-Based Pivot Translation for Under-Resourced Languages and Domains (2012, result): “Morpheme-based translation models have been explored in several studies with similar motivations as in our approach, a better generalization from sparse training data (Fishel and Kirik, 2010; Luong et al., 2010).”
Unsupervised Morphology Rivals Supervised Morphology for Arabic MT (2012, methodology): “Some of these approaches rely on morphological analysis for pre-and post-processing, while others modify the core of a translation system to incorporate morphological information (Habash, 2008; Luong et al., 2010; Nakov and Ng, 2011).”
Morfessor 2.0: Python Implementation and Extensions for Morfessor Baseline (2013, background): “, 2009; Turunen and Kurimo, 2011), and statistical machine translation (e.g. Virpioja et al., 2007; Luong et al., 2010; Mermer and Akın, 2010; Clifton and Sarkar, 2011; Popović, 2011).”
Hybrid Morphological Segmentation for Phrase-Based Machine Translation (2016, methodology): “Stymne et al. (2013) investigate several methods for splitting and merging compounds when translating into Germanic languages, and provide an extensive reading list on the topic.”
Understanding and Improving Morphological Learning in the Neural Machine Translation Decoder (2017, methodology): “…popular in Neural MT, for addressing the rare word problem – and they have been used previously also to beneﬁt MT for morphologically rich (Luong et al., 2010; Belinkov and Glass, 2016; Costa-juss ` a and Fonollosa, 2016) and closely related languages (Durrani et al., 2010; Sajjad et…”