Phrase-based Statistical Machine Translation System Research Articles

A Critique of Statistical Machine Translation

Phrase-Based Statistical Machine Translation (PB-SMT) is clearly the leading paradigm in the field today. Nevertheless—and this may come as some surprise to the PB-SMT community—most translators and, somewhat more surprisingly perhaps, many experienced MT protagonists find the basic model extremely difficult to understand. The main aim of this paper, therefore, is to discuss why this might be the case. Our basic thesis is that proponents of PB-SMT do not seek to address any community other than their own, for they do not feel any need to do so. We demonstrate that this was not always the case; on the contrary, when statistical models of trans-lation were first presented, the language used to describe how such a model might work was very conciliatory, and inclusive. Over the next five years, things changed considerably; once SMT achieved dominance particularly over the rule-based paradigm, it had established a position where it did not need to bring along the rest of the MT community with it, and in our view, this has largely pertained to this day. Having discussed these issues, we discuss three additional issues: the role of automatic MT evaluation metrics when describing PB-SMT systems; the recent syntactic embellishments of PB-SMT, noting especially that most of these contributions have come from researchers who have prior experience in fields other than statistical models of translation; and the relationship between PB-SMT and other models of translation, suggesting that there are many gains to be had if the SMT community were to open up more to the other MT paradigms.

Evaluation of English–Slovak Neural and Statistical Machine Translation

This study is focused on the comparison of phrase-based statistical machine translation (SMT) systems and neural machine translation (NMT) systems using automatic metrics for translation quality evaluation for the language pair of English and Slovak. As the statistical approach is the predecessor of neural machine translation, it was assumed that the neural network approach would generate results with a better quality. An experiment was performed using residuals to compare the scores of automatic metrics of the accuracy (BLEU_n) of the statistical machine translation with those of the neural machine translation. The results showed that the assumption of better neural machine translation quality regardless of the system used was confirmed. There were statistically significant differences between the SMT and NMT in favor of the NMT based on all BLEU_n scores. The neural machine translation achieved a better quality of translation of journalistic texts from English into Slovak, regardless of if it was a system trained on general texts, such as Google Translate, or specific ones, such as the European Commission’s (EC’s) tool, which was trained on a specific-domain.

Semantically Smooth Bilingual Phrase Embeddings Based on Recursive Autoencoders

In this paper, we propose Semantically Smooth Bilingual Recursive Autoencoders to learn bilingual phrase embeddings. The intuition behind our work is to exploit the intrinsic geometric structure of the embedding space and enforce the learned phrase embeddings to be semantically smooth. Specifically, we extend the conventional bilingual recursive autoencoders by preserving the translation and paraphrase probability distributions via regularization terms to simultaneously exploit richer explicit and implicit similarity constraints for bilingual phrase embeddings. To examine the effectiveness of our model, we incorporate two phrase-level similarity features based on the proposed model into a state-of-the-art phrase-based statistical machine translation system. Experiments on NIST Chinese–English test sets show that our model achieves substantial improvements over the baseline.

Deep learning-based techniques to enhance the precision of phrase-based statistical machine translation system for Indian languages

The paper focuses on improving the existing phrase-based statistical machine translation (PB-SMT) system by integrating deep learning knowledge to it. In this paper, a deep learning-based PB-SMT system for Indian languages is developed, so as to improve the conditional probability of the phrase-table and replaced the neural probabilistic language model with the existing back off algorithm of n-gram language model to improve the performance of language model. It is shown that the deep feature-based PB-SMT is better than the standard PB-SMT system. It is shown the significance of integrating manually created dictionaries that has been trained as separate translational model can enhance the result of statistical machine translation system when decoding. For automatic evaluation, it is shown that RIBES being a better evaluation metric for Indian languages compared to BLEU, a standard one.

Deep learning-based techniques to enhance the precision of phrase-based statistical machine translation system for Indian languages

Deep learning-based techniques to enhance the precision of phrase-based statistical machine translation system for Indian languages

Neural Machine Translation for Low-resource English-Bangla

Neural machine translation has recently been able to gain state-of-the-art translation quality for many language pairs. However, neural machine translation has been less tested for English-Bangla language pair, two linguistically distant and widely spoken languages. In this paper, we apply neural machine translation to the task of English-Bangla translation in both directions and compare it against a standard phrase-based statistical machine translation system. We obtain up to +0.30 and +4.95 BLEU improvement over phrase-based statistical machine translation for English-to-Bangla and Bangla-to-English respectively. Due to low-resource and morphological richness of Bangla, English-Bangla translation task produces a large number of rare words. We apply subword segmentation with byte pair encoding to handle this rare words issue. We obtain up to +0.69 and +0.30 BLEU improvement over baseline neural machine translation for English-to-Bangla and Bangla-to-English respectively. We further investigate our system output for several challenging linguistic properties like subject-verb agreement, noun inflection, long distance reordering and rare words translation. We observe that neural machine translation with and without subword segmentation significantly outperform the phrase-based statistical machine translation system, thus establishing itself as the state-of-the-art technology for low-resource English-Bangla machine translation.

Google Translate Gets Voltaire: Literary Translation and the Age of Artificial Intelligence

In the fall of 2016, users of Google Translate working from French to English and English to French saw a marked improvement in the quality of the translated texts. Google Translate was beginning to develop from a purely phrase-based statistical machine translation system into a system that uses deep learning and artificial neural networks that can adapt and learn in reaction to all they encounter. The Google Translate development team has been working on expanding and strengthening the system’s capability of pragmatic analysis. This capability entails extracting useful information from a text so that its meaning can be accurately inferred in an attempt to overcome the vagueness and ambiguity inherent in language, which have remained the greatest hurdles to effective machine translation. This article explores the potential of Google Translate to provide increasingly competent translations of literary texts. Using passages from the works of Voltaire translated by Google Translate into English, this article explores both the successes and the ongoing problems faced by new generations of Google Translate, and of machine translation systems in general, as their developers seek to overcome the hurdles of identifying and rendering style, nuance, and ambiguity in language.

Shu-torjoma: An English↔Bangla Statistical Machine Translation System

An efficient and publicly open machine translation system is in dire need to get the maximum benefits of Information and Communication Technology through removing the language barrier in this era of globalization. In this study, we present a Phrase-Based Statistical Machine Translation (PBMT) system between English and Bangla languages in both directions. To the best of our knowledge, the system is trained on the largest dataset of more than three million tokens each side in English↔Bangla translation task. In the system, we perform data preprocessing and use optimized parameters to produce efficient system output. We analyze our system output from several viewpoints: overall results, comparisons with the available systems, sentence type and length effect, and behaviour of two challenging linguistic properties– prepositional phrase and noun inflection. Our analysis provides useful insights that translating into morphologically richer language is harder than translating from them and this is mainly due to the difficulties of translating noun inflections. Comparisons with the available systems show that our system outperforms the other systems significantly and gain 10.84 BLEU, 2.18 NIST and 19.02 TER points over the next best system. The analysis of the sentence type and length effect shows that simple sentences are easier to translate and the sentences longer than 15 words are harder to translate for English↔Bangla translation task. To foster the English↔Bangla machine translation research, we have developed development and test datasets, which are representative in sentence length and balanced in genre to be used as a benchmark and are made publicly available.

Developing Statistical Machine Translation System for English and Nigerian Languages


 
 
 
 The global demand for translation and translation tools currently surpasses the capacity of available solutions. Besides, there is no one-solution-fits-all, off-the-shelf solution for all languages. Thus, the need and urgency to increase the scale of research for the development of translation tools and devices continue to grow, especially for languages suffering under the pressure of globalisation. This paper discusses our experiments on translation systems between English and two Nigerian languages: Igbo and Yorùbá. The study is setup to build parallel corpora, train and experiment English-to-Igbo, (), English-to-Yorùbá, () and Igbo-to-Yorùbá, () phrase-based statistical machine translation systems. The systems were trained on parallel corpora that were created for each language pair using text from the religious domain in the course of this research. A BLEU score of 30.04, 29.01 and 18.72 respectively was recorded for the English-to-Igbo, English-to-Yorùbá and Igbo-to-Yorùbá MT systems. An error analysis of the systems’ outputs was conducted using a linguistically motivated MT error analysis approach and it showed that errors occurred mostly at the lexical, grammatical and semantic levels. While the study reveals the potentials of our corpora, it also shows that the size of the corpora is yet an issue that requires further attention. Thus an important target in the immediate future is to increase the quantity and quality of the data.
 
 
 
 

Statistical machine translation of Indian languages: a survey

In this study, performance analysis of a state-of-art phrase-based statistical machine translation (SMT) system is presented on eight Indian languages. State of the art in SMT on different Indian languages to English language has also been discussed briefly. The motivation of this study was to promote the development of SMT and linguistic resources for these Indian language pairs, as the current systems are in infancy stage due to sparse data resources. EMILLE and crowdsourcing parallel corpora have been used in this study for experimental purposes. The study is concluded by presenting the performance of baseline SMT system for Indian languages (Bengali, Gujarati, Hindi, Malayalam, Punjabi, Tamil, Telugu and Urdu) into English with average 10–20 % accurate results for all the language pairs. As a result of this study, both of these annotated parallel corpora resources and SMT system will serve as benchmarks for future approaches to SMT in Hindi → English, Urdu → English, Punjabi → English, Telugu → English, Tamil → English, Gujarati → English, Bengali → English and Malayalam → English.

Dependency-based Pre-ordering For English-Vietnamese Statistical Machine Translation

Reordering is a major challenge in machine translation (MT) between two languages with significant differences in word order. In this paper, we present an approach as pre-processing step based on a dependency parser in phrase-based statistical machine translation (SMT) to learn automatic and manual reordering rules from English to Vietnamese. The dependency parse tree and transformation rules are used to reorder the source sentences and applied for systems translating English to Vietnamese. We evaluated our approach and compared on English-Vietnamese machine translation tasks, and showed that it outperforms the baseline phrase-based SMT system.

Toward Building a Comprehensive Phrase-based English-Arabic Statistical Machine Translation System

This paper explores a phrase-based statistical machine translation (PBSMT) pipeline for English-Arabic (En-Ar)language pair. The work surveys the most recent experiments conducted to enhance Arabic machine translation in the En-Ar direction. It also focuses on free datasets and linguistically motivated ideas that enhance phrase-based En-Ar statistical machine translation (SMT) as it is as aims to use those only in order to build a large scale En-Ar SMT system. In addition, the paper highlights Arabic linguistic challenges in Machine Translation (MT) in general. This paper can be considered a guide for building an En-Ar PBSMT system. Furthermore, the presented pipeline can be generalized to any language pairs.

Detecting and Integrating Multiword Expression into English-Arabic Statistical Machine Translation

In this paper we introduce a new method for detecting a type of English Multiword Expressions (MWEs), which is phrasal verbs, into an English-Arabic phrase-based statistical machine translation (PBSMT) system. The detection starts with parsing the English side of the parallel corpus, detecting various linguistic patterns for phrasal verbs and finally integrate them into the En-Ar PBSMT system.In addition, the paper explores the effect of cliticizing specific words in English that have no Arabic equivalent. The results, which reported with the BLEU scores, showed that some patterns achieved significant improvements compared to other patterns and still the baseline achieves the highest score.This paper shows that, by detecting more linguistic patterns and integrating them into En-Ar SMT system, translation quality could be improved with other integration methods. Yet, the results show which path is worth to follow and clarifies the perspective that linguistic features are not handled properly in the statistically learned models.

Template-Based Model for Mongolian-Chinese Machine Translation

Mongolian and Chinese statistical machine translation (SMT) system has its limitation because of the complex Mongolian morphology, scarce resource of parallel corpus and the significant syntax differences. To address these problems, we propose a template-based machine translation (TBMT) system and combine it with the SMT system to achieve a better translation performance. The TBMT model we proposed includes a template extraction model and a template translation model. In the template extraction model, we present a novel method of aligning and abstracting static words from bilingual parallel corpus to extract templates automatically. In the template translation model, our specially designed method of filtering out the low quality matches can enhance the translation performance. Moreover, we apply lemmatization and Latinization to address data sparsity and do the fuzzy match. Experimentally, the coverage of TBMT system is over 50%. The combined SMT system translates all the other uncovered source sentences. The TBMT system outperforms the baselines of phrase-based and hierarchical phrase-based SMT systems for +3.08 and +1.40 BLEU points. The combined system of TBMT and SMT systems also performs better than the baselines of +2.49 and +0.81 BLEU points.

Efficient Word Alignment with Markov Chain Monte Carlo

Abstract We present EFMARAL, a new system for efficient and accurate word alignment using a Bayesian model with Markov Chain Monte Carlo (MCMC) inference. Through careful selection of data structures and model architecture we are able to surpass the fast_align system, commonly used for performance-critical word alignment, both in computational efficiency and alignment accuracy. Our evaluation shows that a phrase-based statistical machine translation (SMT) system produces translations of higher quality when using word alignments from EFMARAL than from fast_align, and that translation quality is on par with what is obtained using GIZA++, a tool requiring orders of magnitude more processing time. More generally we hope to convince the reader that Monte Carlo sampling, rather than being viewed as a slow method of last resort, should actually be the method of choice for the SMT practitioner and others interested in word alignment.

End-to-end statistical machine translation with zero or small parallel texts

AbstractWe use bilingual lexicon induction techniques, which learn translations from monolingual texts in two languages, to build an end-to-end statistical machine translation (SMT) system without the use of any bilingual sentence-aligned parallel corpora. We present detailed analysis of the accuracy of bilingual lexicon induction, and show how a discriminative model can be used to combine various signals of translation equivalence (like contextual similarity, temporal similarity, orthographic similarity and topic similarity). Our discriminative model produces higher accuracy translations than previous bilingual lexicon induction techniques. We reuse these signals of translation equivalence as features on a phrase-based SMT system. These monolingually estimated features enhance low resource SMT systems in addition to allowing end-to-end machine translation without parallel corpora.

A deep source-context feature for lexical selection in statistical machine translation

This paper presents a methodology to address lexical disambiguation in a standard phrase-based statistical machine translation system. Similarity among source contexts is used to select appropriate translation units. The information is introduced as a novel feature of the phrase-based model and it is used to select the translation units extracted from the training sentence more similar to the sentence to translate. The similarity is computed through a deep autoencoder representation, which allows to obtain effective low-dimensional embedding of data and statistically significant BLEU score improvements on two different tasks (English-to-Spanish and English-to-Hindi).

Integrating Rules and Dictionaries from Shallow-Transfer Machine Translation into Phrase-Based Statistical Machine Translation

We describe a hybridisation strategy whose objective is to integrate linguistic resources from shallow-transfer rule-based machine translation (RBMT) into phrase-based statistical machine translation (PBSMT). It basically consists of enriching the phrase table of a PBSMT system with bilingual phrase pairs matching transfer rules and dictionary entries from a shallow-transfer RBMT system. This new strategy takes advantage of how the linguistic resources are used by the RBMT system to segment the source-language sentences to be translated, and overcomes the limitations of existing hybrid approaches that treat the RBMT systems as a black box. Experimental results confirm that our approach delivers translations of higher quality than existing ones, and that it is specially useful when the parallel corpus available for training the SMT system is small or when translating out-of-domain texts that are well covered by the RBMT dictionaries. A combination of this approach with a recently proposed unsupervised shallow-transfer rule inference algorithm results in a significantly greater translation quality than that of a baseline PBSMT; in this case, the only hand-crafted resource used are the dictionaries commonly used in RBMT. Moreover, the translation quality achieved by the hybrid system built with automatically inferred rules is similar to that obtained by those built with hand-crafted rules.

Exploring Diverse Features for Statistical Machine Translation Model Pruning

In phrase-based and hierarchical phrase-based statistical machine translation systems, translation performance depends heavily on the size and quality of the translation table. To meet the requirements of making a real-time response, some research has been performed to filter the translation table. However, most existing methods are always based on one or two constraints that act as hard rules, such as not allowing phrase-pairs with low translation probabilities. These approaches sometimes make constraints rigid because they consider only a single factor instead of composite factors. Based on the considerations above, in this paper, we propose a machine learning-based framework that integrates multiple features for translation model pruning. Experimental results show that our framework is effective by pruning 80% of the phrase-pairs and 70% of the hierarchical rules, while retaining the quality of the translation models when using the BLEU evaluation metric. Our study further shows that our method can select the most useful phrase-pairs and rules, including those that are low in frequency but still very useful.

Statistical Post Editing System (SPES) Applied to Hindi-Punjabi PB-SMT System

Post Editing System is an important module for improving the translation accuracy of any Machine Translation System. It is necessary to post-edit the output generated from machine translation system to correct the errors (word errors, syntactic errors, agreement errors, omission, addition and deletion errors, miscellaneous errors). Post Editing can be done extrinsically and intrinsically. Extrinsic Post Editing is done manually with the help of language experts, and is very costly, time consuming and laborious activity, whereas Intrinsic Post Editing refers to automatic post-editing which can be either rule-based or statistical approach-based. The rule-based post editing is language dependent, and statistical post-editing is language independent. The Statistical Post Editing System described in this paper is tested on phrase-based Hindi-Punjabi Statistical Machine Translation System and has been found to improve the translation accuracy from 1.5% to 12% on BLEU score.