Analyzing Knowledge Distillation in Neural Machine Translation
Knowledge distillation has recently been successfully applied to neural machine translation. It basically allows for building shrunk networks while the resulting systems retain most of the quality of the original model. Despite that many authors report on the benefits of knowledge distillation, few works discuss the actual reasons why it works, especially in the context of neural MT. In this paper, we conduct several experiments aiming at understanding why and how distillation impacts accuracy on an English-German translation task. We show that translation complexity is actually reduced when building a distilled/synthesized bi-text when compared to the reference bi-text. We further remove noisy data from synthesized translations and merge filtered synthesized data together with original reference, thus achieving additional accuracy gains.
15th International Workshop on Spoken Language Translation, October 29-30 2018, Bruges, Belgium
Corpus-based approaches to machine translation rely on the availability of clean parallel corpora. Such resources are scarce, and because of the automatic processes involved in their preparation, they are often noisy. % may contain sentence pairs that are not as parallel as one would expect. This paper describes an unsupervised method for detecting translation divergences in parallel sentences. We rely on a neural network that computes cross-lingual sentence similarity scores, which are then used to effectively filter out divergent translations. Furthermore, similarity scores predicted by the network are used to identify and fix some partial divergences, yielding additional parallel segments. We evaluate these methods for English-French and English-German machine translation tasks, and show that using filtered/corrected corpora actually improves MT performance.
2018 Conference on Empirical Methods in Natural Language Processing, October 31 – November 4 2018, Brussels, Belgium
This paper describes the participation of SYSTRAN to the shared task on parallel corpus filtering at the Third Conference on Machine Translation (WMT 2018). We participate for the first time using a neural sentence similarity classifier which aims at predicting the relatedness of sentence pairs in a multilingual context. The paper describes the main characteristics of our approach and discusses the results obtained on the data sets published for the shared task.
Third Conference on Machine Translation (WMT18), October 31 - November 1 2018, Brussels, Belgium
Comments: Published in "Proceedings of the Fifth Workshop on NLP for Similar Languages, Varieties and Dialects" (pages 128–136) - Publisher: Association for Computational Linguistics
SYSTRAN competes this year for the first time to the DSL shared task, in the Arabic Dialect Identification subtask. We participate by training several Neural Network models showing that we can obtain competitive results despite the limited amount of training data available for learning. We report our experiments and detail the network architecture and parameters of our 3 runs: our best performing system consists in a Multi-Input CNN that learns separate embeddings for lexical, phonetic and acoustic input features (F1: 0.5289); we also built a CNN-biLSTM network aimed at capturing both spatial and sequential features directly from speech spectrograms (F1: 0.3894 at submission time, F1: 0.4235 with later found parameters); and finally a system relying on binary CNN-biLSTMs (F1: 0.4339).
2018, New Mexico, USA, August 20
Comments: Published in "Proceedings of the 2nd Workshop on Neural Machine Translation and Generation" (pages 122–128) - Publisher: Association for Computational Linguistics
We present a system description of the OpenNMT Neural Machine Translation entry for the WNMT 2018 evaluation. In this work, we developed a heavily optimized NMT inference model targeting a high-performance CPU system. The final system uses a combination of four techniques, all of them leading to significant speed-ups in combination: (a) sequence distillation, (b) architecture modifications, (c) pre-computation, particularly of vocabulary, and (d) CPU targeted quantization. This work achieves the fastest performance of the shared task, and led to the development of new features that have been integrated to OpenNMT and made available to the community.
2018, Melbourne, Australia, July 20
Comments: 24e Conférence sur le Traitement Automatique des Langues Naturelles (TALN) - pages 19-20
Cet article présente un système d'alertes fondé sur la masse de données issues de Tweeter. L'objectif de l'outil est de surveiller l'actualité, autour de différents domaines témoin incluant les événements sportifs ou les catastrophes naturelles. Cette surveillance est transmise à l'utilisateur sous forme d'une interface web contenant la liste d'événements localisés sur une carte.
Actes de TALN 2017, volume 3 : démonstrations
Comments: Published in "Proceedings of the Second Conference on Machine Translation" (pages 265-270) - Publisher: Association for Computational Linguistics
This paper describes SYSTRAN's systems submitted to the WMT 2017 shared news translation task for English-German, in both translation directions. Our systems are built using OpenNMT1, an opensource neural machine translation system, implementing sequence-to-sequence models with LSTM encoder/decoders and attention. We experimented using monolingual data automatically back-translated. Our resulting models are further hyperspecialised with an adaptation technique that finely tunes models according to the evaluation test sentences.
2017, Copenhagen, Denmark
Comments: Published in "Proceedings of ACL 2017, System Demonstrations" (pages 67-72) - Publisher: Association for Computational Linguistics
We describe an open-source toolkit for neural machine translation (NMT). The toolkit prioritizes efficiency, modularity, and extensibility with the goal of supporting NMT research into model architectures, feature representations, and source modalities, while maintaining competitive performance and reasonable training requirements. The toolkit consists of modeling and translation support, as well as detailed pedagogical documentation about the underlying techniques.
2017, Vancouver, Canada
Comments: Published in "Proceedings of the Eighth International Joint Conference on Natural Language Processing" (Volume 2: Short Papers) - Publisher: Asian Federation of Natural Language Processing
Training efficiency is one of the main problems for Neural Machine Translation (NMT). Deep networks need for very large data as well as many training iterations to achieve state-of-the-art performance. This results in very high computation cost, slowing down research and industrialisation. In this paper, we propose to alleviate this problem with several training methods based on data boosting and bootstrap with no modifications to the neural network. It imitates the learning process of humans, which typically spend more time when learning "difficult" concepts than easier ones. We experiment on an English-French translation task showing accuracy improvements of up to 1.63 BLEU while saving 20% of training time.
2017, Taipei, Taiwan
Comments: Published in "Proceedings of the International Conference Recent Advances in Natural Language Processing, RANLP 2017", Varna, Bulgaria, Sep 4–6 2017 - Publisher: INCOMA Ltd.
Machine translation systems are very sensitive to the domains they were trained on. Several domain adaptation techniques have been deeply studied. We propose a new technique for neural machine translation (NMT) that we call domain control which is performed at runtime using a unique neural network covering multiple domains. The presented approach shows quality improvements when compared to dedicated domains translating on any of the covered domains and even on out-of-domain data. In addition, model parameters do not need to be re-estimated for each domain, making this effective to real use cases. Evaluation is carried out on English-to-French translation for two different testing scenarios. We first consider the case where an end-user performs translations on a known domain. Secondly, we consider the scenario where the domain is not known and predicted at the sentence level before translating. Results show consistent accuracy improvements for both conditions.
[v2] Tue, 12 Sep 2017 12:01:40 GMT
Comments: 24e Conférence sur le Traitement Automatique des Langues Naturelles (TALN) - Actes de TALN 2017, volume 2 : articles courts - pages 218-225
L'adaptation au domaine est un verrou scientifique en traduction automatique. Il englobe généralement l'adaptation de la terminologie et du style, en particulier pour la post-édition humaine dans le cadre d'une traduction assistée par ordinateur. Avec la traduction automatique neuronale, nous étudions une nouvelle approche d'adaptation au domaine que nous appelons “spécialisation” et qui présente des résultats prometteurs tant dans la vitesse d'apprentissage que dans les scores de traduction. Dans cet article, nous proposons d'explorer cette approche.
Orléans, France – 26-30 juin 2017
Neural Machine Translation: Torniamo alle origini
Each of us have experienced or heard of deep learning in day-to-day business applications. What are the fundamentals of this new technology and what new opportunities does it offer?
Jan 31, 2017
Subjects: Computation and Language (cs.CL)
Abstract: Text simplification aims at reducing the lexical, grammatical and structural complexity of a text while keeping the same meaning. In the context of machine translation, we introduce the idea of simplified translations in order to boost the learning ability of deep neural translation models. We conduct preliminary experiments showing that translation complexity is actually reduced in a translation of a source bi-text compared to the target reference of the bi-text while using a neural machine translation (NMT) system learned on the exact same bi-text. Based on knowledge distillation idea, we then train an NMT system using the simplified bi-text, and show that it outperforms the initial system that was built over the reference data set. Performance is further boosted when both reference and automatic translations are used to learn the network. We perform an elementary analysis of the translated corpus and report accuracy results of the proposed approach on English-to-French and English-to-German translation tasks.
[v1] Mon, 19 Dec 2016 11:50:58 GMT
Subjects: Computation and Language (cs.CL)
Domain adaptation is a key feature in Machine Translation. It generally encompasses terminology, domain and style adaptation, especially for human post-editing workflows in Computer Assisted Translation (CAT). With Neural Machine Translation (NMT), we introduce a new notion of domain adaptation that we call "specialization" and which is showing promising results both in the learning speed and in adaptation accuracy. In this paper, we propose to explore this approach under several perspectives.
[v1] Mon, 19 Dec 2016 11:52:08 GMT
Subjects: Computation and Language (cs.CL)
Abstract: Since the first online demonstration of Neural Machine Translation (NMT) by LISA, NMT development has recently moved from laboratory to production systems as demonstrated by several entities announcing roll-out of NMT engines to replace their existing technologies. NMT systems have a large number of training configurations and the training process of such systems is usually very long, often a few weeks, so role of experimentation is critical and important to share. In this work, we present our approach to production-ready systems simultaneously with release of online demonstrators covering a large variety of languages (12 languages, for 32 language pairs). We explore different practical choices: an efficient and evolutive open-source framework; data preparation; network architecture; additional implemented features; tuning for production; etc. We discuss about evaluation methodology, present our first findings and we finally outline further work.
Our ultimate goal is to share our expertise to build competitive production systems for "generic" translation. We aim at contributing to set up a collaborative framework to speed-up adoption of the technology, foster further research efforts and enable the delivery and adoption to/by industry of use-case specific engines integrated in real production workflows. Mastering of the technology would allow us to build translation engines suited for particular needs, outperforming current simplest/uniform systems.
[v1] Tue, 18 Oct 2016 11:32:42 GMT
Abstract: This paper describes the joint submission by RWTH Aachen University and SYSTRAN in the Chinese-English Patent Machine Translation Task at the 10th NTCIR Workshop. We specify the statistical systems developed by RWTH Aachen University and the hybrid machine translation systems developed by SYSTRAN. We apply RWTH Aachen’s combination techniques to create consensus hypotheses from very different systems: phrase-based and hierarchical SMT, rule-based MT (RBMT) and MT with statistical post-editing (SPE). The system combination was ranked second in BLEU and second in the human adequacy evaluation in this competition.
June 18-21, 2013, Tokyo, Japan
Abstract: This report describes SYSTRAN’s Chinese-English and English-Chinese machine translation systems that participated in the CWMT 2011 machine translation evaluation tasks. The base systems are SYSTRAN rule-based machine translation systems, augmented with various statistical techniques. Based on the translations of the rule-based systems, we performed statistical post-editing with the provided bilingual and monolingual training corpora. In this report, we describe the technology behind the systems, the training data, and finally the evaluation results in the CWMT 2011 evaluation. Our primary Chinese-English system was ranked first in BLEU in the translation tasks.
Proceedings of the 7th China Workshop on Machine Translation (CWMT), September 2011.
Abstract: We present two methods that merge ideas from statistical machine translation (SMT) and translation memories (TM). We use a TM to retrieve matches for source segments, and replace the mismatched parts with instructions to an SMT system to fill in the gap. We show that for fuzzy matches of over 70%, one method outperforms both SMT and TM base- lines.
JEC, November 2010.
Abstract: We present a novel exact solution to the approximate string matching problem in the context of translation memories, where a text segment has to be matched against a large corpus, while allowing for errors. We use suffix arrays to detect exact n-gram matches, A* search heuristics to discard matches and A* parsing to validate candidate segments. The method outperforms the canonical baseline by a factor of 100, with average lookup times of 4.3–247ms for a segment in a realistic scenario.
AMTA, October 2010.
Abstract: This report describes both of SYSTRAN's Chinese-English and English-Chinese machine translation systems that participated in the CWMT2009 machine translation evaluation tasks. The base systems are SYSTRAN rule-based machine translation systems, augmented with various statistical techniques. Based on the translations of the rule-based systems, we perform statistical post-editing with the provided bilingual and monolingual training corpora. In this report, we describe the technology behind the systems, the training data, and finally the evaluation results in the CWMT2009 evaluation. Our primary systems were top-ranked in the evaluation tasks.
November 2009, CWMT
Selective addition of corpus-extracted phrasal lexical rules to a rule-based machine translation system
Abstract: In this work, we show how an existing rule-based, general-purpose machine translation system may be improved and adapted automatically to a given domain, whenever parallel corpora are available. We perform this adaptation by extracting dictionary entries from the parallel data. From this initial set, the application of these rules is tested against the baseline performance. Rules are then pruned depending on sentence-level improvements and deteriorations, as evaluated by an automatic string-based metric. Experiments using the Europarl dataset show a 3% absolute improvement in BLEU over the original rule-based system.
MT Summit, August 2009.
Abstract: We describe here the two Systran/University of Edinburgh submissions for WMT2009. They involve a statistical post-editing model with a particular handling of named entities (English to French and German to English) and the extraction of phrasal rules (English to French).
Abstract: This paper describes the development of several machine translation systems for the 2009 WMT shared task evaluation. We only consider the translation between French and English. We describe a statistical system based on the Moses decoder and a statistical post-editing system using SYSTRAN’s rule-based system. We also investigated techniques to automatically extract additional bilingual texts from comparable corpora.
Abstract: This paper describes an initial version of a general purpose French/English statistical machine translation system. The main features of this system are the open-source Moses decoder, the integration of a bilingual dictionary and a continuous space target language model. We analyze the performance of this system on the test data of the WMT'08 evaluation.
Abstract: This paper describes SYSTRAN submissions for the shared task of the third Workshop on Statistical Machine Translation at ACL. Our main contribution consists in a French-English statistical model trained without the use of any human-translated parallel corpus. In substitution, we translated a monolingual corpus with SYSTRAN rule-based translation engine to produce the parallel corpus. The results are provided herein, along with a measure of error analysis.
Abstract: XSL Transformation stylesheets are usually used to transform a document described in an XML formalism into another XML formalism, to modify an XML document, or to publish content stored into an XML document to a publishing format (XSL-FO, (X)HTML...). SYSTRAN Translation Stylesheets (STS) use XSLT to drive and control the machine translation of XML documents (native XML document formats or XML representations — such as XLIFF — of other kinds of document formats).
Abstract: SYSTRAN started the design and the development of Arabic, Farsi and Urdu to English machine translation systems in July 2002. This paper describes the methodology and implementation adopted for dictionary building and morphological analysis. SYSTRAN's IntuitiveCoding® technology (ICT) facilitates the creation, update, and maintenance of Arabic, Farsi and Urdu lexical entries, is more modular and less costly. ICT for Arabic, Farsi, and Urdu requires the implementation of stem-based lexical entries, the authentic scripts for each language, a statistical Arabic stem-guesser, and separate declarative modules for internal and external morphology.
MT Summit IX; September 22-26, 2003.
Abstract: Customization of Machine Translation (MT) is a prerequisite for corporations to adopt the technology. It is therefore important but nonetheless challenging. Ongoing implementation proves that XML is an excellent exchange device between MT modules that efficiently enables interaction between the user and the processes to reach highly granulated structure-based customization. Accomplished through an innovative approach called the SYSTRAN Translation Stylesheet, this method is coherent with the current evolution of the “authoring process”. As a natural progression, the next stage in the customization process is the integration of MT in a multilingual tool kit designed for the "authoring process".
MT Summit IX, September 22-26, 2003.
Abstract: The SYSTRAN Review Manager (SRM) is one of the components that comprise the SYSTRAN Linguistics Platform (SLP), a comprehensive enterprise solution for managing MT customization and localization projects. The SRM is a productivity tool used for the review, quality assessment and maintenance of linguistic resources combined with a SYSTRAN solution. The SRM is used in-house by SYSTRAN’s development team and is also licensed to corporate customers as it addresses leading linguistic challenges, such as terminology and homographs, which makes it a key component of the QA process. Extremely flexible, the SRM adapts to localization and MT customization projects from small to large-scale. Its Web-based interface and multi-user architecture enable a centralized and efficient work environment for local and geographically disbursed individual users and teams. Users segment a given corpus to fluidly review and evaluate translations, as well as identify the typology of errors. Corpus metrics, terminology extraction and detailed reporting capabilities facilitate prioritizing tasks, resulting in immediate focus on those issues that significantly impact MT quality. Data and statistics are tracked throughout the customization process and are always available for regression tests and overall project management. This environment is highly conducive to increased productivity and efficient QA in the MT customization effort.
MT Summit IX; September 22-26, 2003.
Abstract: Customizing a general-purpose MT system is an effective way to improve machine translation quality for specific usages. Building a user-specific dictionary is the first and most important step in the customization process. An intuitive dictionary-coding tool was developed and is now utilized to allow the user to build user dictionaries easily and intelligently. SYSTRAN's innovative and proprietary IntuitiveCoding® technology is the engine powering this tool. It is comprised of various components: massive linguistic resources, a morphological analyzer, a statistical guesser, finite-state automaton, and a context-free grammar. Methodologically, IntuitiveCoding® is also a cross-application approach for high quality dictionary building in terminology import and exchange. This paper describes the various components and the issues involved in its implementation. An evaluation frame and utilization of the technology are also presented. Future plans for further advancing this technology forward are projected.
MT Summit IX; September 22-26, 2003.
Abstract: SYSTRAN's SLP (SYSTRAN Linguistics Platform) is a comprehensive enterprise solution for managing a full range of translation and localization project tasks. The SLP consists of the SYSTRAN machine translation (MT) technology, linguistic resources and tools for project management, corpus analysis and quality evaluation. The underlying platform that supports the SLP is the SYSTRAN WebServer, a client/server application that can be accessed transparently through most common software applications. It supports document formats including HTML, RTF, XML, and SGML. The SYSTRAN WebServer is hosted at the customer’s site and can be integrated with internal translation workflow systems. The SYSTRAN WebServer is a robust and high-volume platform that can support an unlimited number of users, and millions of translation jobs per day.
Abstract: In this article we present the concept of "implicit transfer" rules. We will show that they represent a valid compromise between huge direct transfer terminology lists and large sets of transfer rules, which are very complex to maintain. We present a concrete, real-life application of this concept in a customization project (TOLEDO project) concerning the automatic translation of Autodesk (ADSK) support pages. In this application, the alignment is moreover combined with a graph representation substituting linear dictionaries. We show how the concept could be extended to increase coverage of traditional translation dictionaries as well as to extract terminology from large existing multilingual corpora. We also introduce the concept of "alignment dictionary" which seems promising in its ability to extend the pragmatic limits of multilingual dictionary management.
MT Summit 8, September 18-22, 2001.
Abstract: In this paper, we present the design of the new generation Systran translation systems, currently utilized in the development of English-Hungarian, English-Polish, English-Arabic, French-Arabic, Hungarian-French and Polish-French language pairs. The new design, based on the traditional Systran machine translation expertise and the existing linguistic resources, addresses the following aspects: efficiency, modularity, declarativity, reusability, and maintainability. Technically, the new systems rely on intensive use of state-of-the-art finite automaton and formal grammar implementation. The finite automata provide the essential lookup facilities and the natural capacity of factorizing intuitive linguistic sets. Linguistically, we have introduced a full monolingual description of linguistic information and the concept of implicit transfer. Finally, we present some by-products that are directly derived from the new architecture: intuitive coding tools, spell checker and syntactic tagger.
MT Summit 8, September 18-22, 2001.