DOCSLIB.ORG

From Statistical Term Extraction to Hybrid Machine Translation

Total Page：16

File Type：pdf, Size：1020Kb

Download full-text PDF Read full-text
From Statistical Term Extraction to Hybrid Machine Translation From Statistical Term Extraction to Hybrid Machine Translation Petra Wolf, Ulrike Bernardi Christian Federmann, Sabine Hunsicker Lucy Software and Services DFKI Neumarkter Str. 81 Stuhlsatzenhausweg 3 81673 Munich, Germany 66123 Saarbrucken,¨ Germany petra.wolf,[email protected] sabine.hunsicker,[email protected] Abstract multi-layered Linguistically augmented Statistical Terminology EXtraction). We decided to impose This study presents a new hybrid approach strict knowledge-enhanced requirements for the for translation equivalent selection within statistical term extraction which consequently was a transfer-based machine translation sys- augmented by several layers of linguistic data re- tem using an intertwined net of traditional finement and automatic feature attribution. LiS- linguistic methods together with statisti- TEX is distinct from other term extractors as it cal techniques. Detailed evaluation reveals has layers that access already during the early term that the translation quality can be improved defining extraction phase the RBMT system com- substantially in this way. ponents for linguistic filtering and later production of knowledge-augmented output. In contrast to 1 Introduction the phrase-table approach, the F-measure is high A promising integration point for statistical which avoids evident deteriorations after integra- techniques into Rule-Based Machine Translation tion into the MT system. (RBMT) systems is the transfer phase. A key prob- The remainder of this paper is organized as fol- lem here is to deal with non-deterministic rules lows: Section 2 gives an overview of related work. and preferences in order to disambiguate and select Details on the terminology extraction are then pro- the most natural expressions in the target language vided in Section 3 followed by a description of the (cf. (Eisele et al., 2008b; Thurmair, 2009)). We results of terminology and translation quality eval- performed studies on a phrase-table-driven trans- uations in section 4. Section 5 summarizes the key fer approach based on an RBMT system (for de- findings and outlines open issues for future work. tails on the RBMT system see (Alonso and Thur- mair, 2003)): A prototype was built accessing sta- 2 Related work tistically generated bilingual phrase tables at run- There have been several studies of hybrid machine time in addition to system lexicons and grammars. translation approaches to overcome the drawbacks This hybrid prototype showed indeed better lexi- of rule-based and statistical MT alone by a com- cal selection, improved coverage and even some- bined approach, starting from RBMT as well times enhanced syntax, but well-known issues in as starting from SMT. Evaluations of SMT vs. morpho-syntax and a very low hit rate of the phrase RBMT systems revealed that one of the weak table module remained as limiting factors. Thus al- points of RBMT systems is the lexical selection in though the data which were accessed and preferred transfer (cf. (Thurmair, 2009; Chen et al., 2009). are obviously the desired ones, the first challenge Since RBMT systems tend to suffer from insuf- was the lack of deeper linguistic knowledge within ficient and too deterministic lexical coverage and the data while the second challenge was the small choice (Eisele et al., 2008b), this study concen- F-measure of the data itself. trates on automatic enlargements of RBMT lexi- This led us to a deeper intertwined hybrid exten- cons and enhanced transfer-generation operations, sion: The LiSTEX approach (Hybrid Transfer by while taking into account the peculiarities of a spe- c 2011 European Association for Machine Translation. cific RBMT system and statistical techniques. Mikel L. Forcada, Heidi Depraetere, Vincent Vandeghinste (eds.) Proceedings of the 15th Conference of the European Association for Machine Translation, p. 225232 Leuven, Belgium, May 2011 The considerable potential of statistical term ex- traction combined with RBMT has been evalu- ated by other researchers, such as (Thurmair, 2003; Dugast et al., 2009). Here we go one step further by already integrating some RBMT techniques into the very early stages of the statistical term extraction process. This helps to avoid the well- known problems found in term guessing and iden- tification (Heid, 1999). Additional linguistic lay- ers assure that the extracted terms and phrases are tailored and augmented for the RBMT system in question. 3 Intelligent Terminology Extraction The underlying term extraction tool extracts term pairs by means of statistical algorithms from ex- isting translation memories or bilingual corpora (Eisele et al., 2008a). As a bilingual corpus, the Europarl corpus (Koehn, 2005) was used for devel- opment. As a test set, we choose the ACL WMT 2008 test set which contains the Q4/2000 portion Figure 1: Term Extraction Workflow of the EuroParl data (2000-10 to 2000-12). The initial design study on peculiarities of a spe- tents of this terminology to be extracted and im- cific statistically based term extraction that satis- ported are determined by the fine-grained data fies the requirements of an RBMT system dealt needed by an RBMT system. The first major modi- with issues like term definition in a strict linguis- fication of the statistical extraction has to fulfill the tic and system-related sense, term identification requirement to extract only well defined linguis- as single words vs. multiword expressions, base tic terms. For this reason, the extraction toolkit form reduction and application of linguistic cate- has been extended in order to access and use the gory patterns for identification or elimination of RBMT analysis, transfer and generation trees to noise at the end of the first multi-layered extrac- find interesting linguistic phrases, i.e. terms. In tion phase. Also the term recognition needed to be addition, it checks whether the translation of a spe- defined in relation with the linguistically based tar- cific term in the transfer tree, be it a single or a get system, i.e. comparison of the extraction result multiword, differs from the reference translation with the target system lexicon in order to identify in the corpus. In this way, the term extraction tool known vs. unknown terms. Thus this painstaking delivers only terms which receive a linguistic iden- specification research provided the basis for the ex- tification and which are translated differently with tended implementation of the LiSTEX term identi- the conventional RBMT system than in the bilin- fication, knowledge-enhanced acquisition and aug- gual memory. mentation modules. The extracted terminology has to contain the For LiSTEX, we concentrated on German- following minimum information: Canonical forms English and Spanish-English. The remaining sub- and categories of the source and target terms, do- sections explain details of the intelligent term ex- main area, frequency and, in case of multiwords, traction techniques and present the term extraction the internal structure of the expression. Finally, workflow as a whole (see Figure 1). the context has to be passed through: One exam- ple source language sentence in which the term 3.1 Term Identification with Initial Meta occurs in the corpus with the corresponding trans- Knowledge Acquisition and Organization lation equivalent sentence, useful for later manual The objective is to have the semantic translation inspection. equivalents decided by the statistical technique. For achieving this first layer of terminology ac- The requirements, however, on the linguistic con- quisition, the source and reference text are tok- 226 enized, tagged with part of speech information and as punctuation marks or integers, are filtered the tokens are lemmatized. Source and reference into separate lists. Since these entries tend to texts are aligned word-by-word, the source text be wrongly aligned and extracted, the sublists is translated by the RBMT Engine and analysis, have to be inspected and only useful terms transfer and generation trees are created. These will be imported. trees are aligned to the words in the source and ref- All entries showing a category change • erence text and all phrases which the RBMT sys- (i.e. source and target categories are not tem translated differently than the reference trans- identical) are excluded from import since lation are selected as potential term candidates. they are likely to be wrong. Sometimes they Since these phrases still include the original sur- are caused by errors during part of speech face forms, the canonical forms are built now, e.g. tagging, sometimes by alignment errors. adjectives are inflected properly to match the head Quality Splitting procedures which allow • noun. Thus the terms receive the proper dictionary further predictions as to the quality of the format. The categories for the entire terms are de- extracted terms. A German multiword term rived from the part of speech sequences. The list for example which results in an English of terms is alphabetically sorted and non-frequent single target word is a likely non-valid term, terms are filtered out. mostly due to alignment errors: Up to this point in the processing chain, the ex- 1. Single words on source and target side traction process is done for one language direction 2. Single words on source, multiword ex- only. Now the other language direction is gener- pressions on target side ated and the intersection of both is created to avoid 3. Multiword expressions on source side, wrong term pairs caused by alignment problems. single words on the target side
Load more

Recommended publications
  • Integrating Optical Character Recognition and Machine
    Integrating Optical Character Recognition and Machine Translation of Historical Documents Haithem Afli and Andy Way ADAPT Centre School of Computing Dublin City University Dublin, Ireland haithem.afli, andy.way @adaptcentre.ie { } Abstract Machine Translation (MT) plays a critical role in expanding capacity in the translation industry. However, many valuable documents, including digital documents, are encoded in non-accessible formats for machine processing (e.g., Historical or Legal documents). Such documents must be passed through a process of Optical Character Recognition (OCR) to render the text suitable for MT. No matter how good the OCR is, this process introduces recognition errors, which often renders MT ineffective. In this paper, we propose a new OCR to MT framework based on adding a new OCR error correction module to enhance the overall quality of translation. Experimenta- tion shows that our new system correction based on the combination of Language Modeling and Translation methods outperforms the baseline system by nearly 30% relative improvement. 1 Introduction While research on improving Optical Character Recognition (OCR) algorithms is ongoing, our assess- ment is that Machine Translation (MT) will continue to produce unacceptable translation errors (or non- translations) based solely on the automatic output of OCR systems. The problem comes from the fact that current OCR and Machine Translation systems are commercially distinct and separate technologies. There are often mistakes in the scanned texts as the OCR system occasionally misrecognizes letters and falsely identifies scanned text, leading to misspellings and linguistic errors in the output text (Niklas, 2010). Works involved in improving translation services purchase off-the-shelf OCR technology but have limited capability to adapt the OCR processing to improve overall machine translation perfor- mance.
    [Show full text]
  • Machine Translation for Academic Purposes Grace Hui-Chin Lin
    Proceedings of the International Conference on TESOL and Translation 2009 December 2009: pp.133-148 Machine Translation for Academic Purposes Grace Hui-chin Lin PhD Texas A&M University College Station Master of Science, University of Southern California Paul Shih Chieh Chien Center for General Education, Taipei Medical University Abstract Due to the globalization trend and knowledge boost in the second millennium, multi-lingual translation has become a noteworthy issue. For the purposes of learning knowledge in academic fields, Machine Translation (MT) should be noticed not only academically but also practically. MT should be informed to the translating learners because it is a valuable approach to apply by professional translators for diverse professional fields. For learning translating skills and finding a way to learn and teach through bi-lingual/multilingual translating functions in software, machine translation is an ideal approach that translation trainers, translation learners, and professional translators should be familiar with. In fact, theories for machine translation and computer assistance had been highly valued by many scholars. (e.g., Hutchines, 2003; Thriveni, 2002) Based on MIT’s Open Courseware into Chinese that Lee, Lin and Bonk (2007) have introduced, this paper demonstrates how MT can be efficiently applied as a superior way of teaching and learning. This article predicts the translated courses utilizing MT for residents of global village should emerge and be provided soon in industrialized nations and it exhibits an overview about what the current developmental status of MT is, why the MT should be fully applied for academic purpose, such as translating a textbook or teaching and learning a course, and what types of software can be successfully applied.
    [Show full text]
  • Machine Translation for Language Preservation
    Machine translation for language preservation Steven BIRD1,2 David CHIANG3 (1) Department of Computing and Information Systems, University of Melbourne (2) Linguistic Data Consortium, University of Pennsylvania (3) Information Sciences Institute, University of Southern California [email protected], [email protected] ABSTRACT Statistical machine translation has been remarkably successful for the world’s well-resourced languages, and much effort is focussed on creating and exploiting rich resources such as treebanks and wordnets. Machine translation can also support the urgent task of document- ing the world’s endangered languages. The primary object of statistical translation models, bilingual aligned text, closely coincides with interlinear text, the primary artefact collected in documentary linguistics. It ought to be possible to exploit this similarity in order to improve the quantity and quality of documentation for a language. Yet there are many technical and logistical problems to be addressed, starting with the problem that – for most of the languages in question – no texts or lexicons exist. In this position paper, we examine these challenges, and report on a data collection effort involving 15 endangered languages spoken in the highlands of Papua New Guinea. KEYWORDS: endangered languages, documentary linguistics, language resources, bilingual texts, comparative lexicons. Proceedings of COLING 2012: Posters, pages 125–134, COLING 2012, Mumbai, December 2012. 125 1 Introduction Most of the world’s 6800 languages are relatively unstudied, even though they are no less im- portant for scientific investigation than major world languages. For example, before Hixkaryana (Carib, Brazil) was discovered to have object-verb-subject word order, it was assumed that this word order was not possible in a human language, and that some principle of universal grammar must exist to account for this systematic gap (Derbyshire, 1977).
    [Show full text]
  • Machine Translation and Computer-Assisted Translation: a New Way of Translating? Author: Olivia Craciunescu E-Mail: Olivia [email protected]
    Machine Translation and Computer-Assisted Translation: a New Way of Translating? Author: Olivia Craciunescu E-mail: [email protected] Author: Constanza Gerding-Salas E-mail: [email protected] Author: Susan Stringer-O’Keeffe E-mail: [email protected] Source: http://www.translationdirectory.com The Need for Translation IT has produced a screen culture that Technology tends to replace the print culture, with printed documents being dispensed Advances in information technology (IT) with and information being accessed have combined with modern communication and relayed directly through computers requirements to foster translation automation. (e-mail, databases and other stored The history of the relationship between information). These computer documents technology and translation goes back to are instantly available and can be opened the beginnings of the Cold War, as in the and processed with far greater fl exibility 1950s competition between the United than printed matter, with the result that the States and the Soviet Union was so intensive status of information itself has changed, at every level that thousands of documents becoming either temporary or permanent were translated from Russian to English and according to need. Over the last two vice versa. However, such high demand decades we have witnessed the enormous revealed the ineffi ciency of the translation growth of information technology with process, above all in specialized areas of the accompanying advantages of speed, knowledge, increasing interest in the idea visual
    [Show full text]
  • Language Service Translation SAVER Technote
    TechNote August 2020 LANGUAGE TRANSLATION APPLICATIONS The U.S. Department of Homeland Security (DHS) AEL Number: 09ME-07-TRAN established the System Assessment and Validation for Language translation equipment used by emergency services has evolved into Emergency Responders commercially available, mobile device applications (apps). The wide adoption of (SAVER) Program to help cell phones has enabled language translation applications to be useful in emergency responders emergency scenarios where first responders may interact with individuals who improve their procurement speak a different language. Most language translation applications can identify decisions. a spoken language and translate in real-time. Applications vary in functionality, Located within the Science such as the ability to translate voice or text offline and are unique in the number and Technology Directorate, of available translatable languages. These applications are useful tools to the National Urban Security remedy communication barriers in any emergency response scenario. Technology Laboratory (NUSTL) manages the SAVER Overview Program and conducts objective operational During an emergency response, language barriers can challenge a first assessments of commercial responder’s ability to quickly assess and respond to a situation. First responders equipment and systems must be able to accurately communicate with persons at an emergency scene in relevant to the emergency order to provide a prompt, appropriate response. The quick translations provided responder community. by mobile apps remove the language barrier and are essential to the safety of The SAVER Program gathers both first responders and civilians with whom they interact. and reports information about equipment that falls within the Before the adoption of smartphones, first responders used language interpreters categories listed in the DHS and later language translation equipment such as hand-held language Authorized Equipment List translation devices (Figure 1).
    [Show full text]
  • Terminology Extraction, Translation Tools and Comparable Corpora Helena Blancafort, Béatrice Daille, Tatiana Gornostay, Ulrich Heid, Claude Méchoulam, Serge Sharoff
    TTC: Terminology Extraction, Translation Tools and Comparable Corpora Helena Blancafort, Béatrice Daille, Tatiana Gornostay, Ulrich Heid, Claude Méchoulam, Serge Sharoff To cite this version: Helena Blancafort, Béatrice Daille, Tatiana Gornostay, Ulrich Heid, Claude Méchoulam, et al.. TTC: Terminology Extraction, Translation Tools and Comparable Corpora. 14th EURALEX International Congress, Jul 2010, Leeuwarden/Ljouwert, Netherlands. pp.263-268. hal-00819365 HAL Id: hal-00819365 https://hal.archives-ouvertes.fr/hal-00819365 Submitted on 2 May 2013 HAL is a multi-disciplinary open access L’archive ouverte pluridisciplinaire HAL, est archive for the deposit and dissemination of sci- destinée au dépôt et à la diffusion de documents entific research documents, whether they are pub- scientifiques de niveau recherche, publiés ou non, lished or not. The documents may come from émanant des établissements d’enseignement et de teaching and research institutions in France or recherche français ou étrangers, des laboratoires abroad, or from public or private research centers. publics ou privés. TTC: Terminology Extraction, Translation Tools and Comparable Corpora Helena Blancafort, Syllabs, Universitat Pompeu Fabra, Barcelona, Spain Béatrice Daille, LINA, Université de Nantes, France Tatiana Gornostay, Tilde, Latva Ulrich Heid, IMS, Universität Stuttgart, Germany Claude Mechoulam, Sogitec, France Serge Sharoff, CTS, University of Leeds, England The need for linguistic resources in any natural language application is undeniable. Lexicons and terminologies
    [Show full text]
  • Acknowledging the Needs of Computer-Assisted Translation Tools Users: the Human Perspective in Human-Machine Translation Annemarie Taravella and Alain O
    The Journal of Specialised Translation Issue 19 – January 2013 Acknowledging the needs of computer-assisted translation tools users: the human perspective in human-machine translation AnneMarie Taravella and Alain O. Villeneuve, Université de Sherbrooke, Quebec ABSTRACT The lack of translation specialists poses a problem for the growing translation markets around the world. One of the solutions proposed for the lack of human resources is automated translation tools. In the last few decades, organisations have had the opportunity to increase their use of technological resources. However, there is no consensus on the way that technological resources should be integrated into translation service providers (TSP). The approach taken by this article is to set aside both 100% human translation and 100% machine translation (without human intervention), to examine a third, more realistic solution: interactive translation where humans and machines co-operate. What is the human role? Based on the conceptual framework of information systems and organisational sciences, we recommend giving users, who are mainly translators for whom interactive translation tools are designed, a fundamental role in the thinking surrounding the implementation of a technological tool. KEYWORDS Interactive translation, information system, implementation, business process reengineering, organisational science. 1. Introduction Globalisation and the acceleration of world trade operations have led to an impressive growth of the global translation services market. According to EUATC (European Union of Associations of Translation Companies), translation industry is set to grow around five percent annually in the foreseeable future (Hager, 2008). With its two official languages, Canada is a good example of an important translation market, where highly skilled professional translators serve not only the domestic market, but international clients as well.
    [Show full text]
  • MULTILINGUAL CHATBOT with HUMAN CONVERSATIONAL ABILITY [1] Aradhana Bisht, [2] Gopan Doshi, [3] Bhavna Arora, [4] Suvarna Pansambal [1][2] Student, Dept
    International Journal of Future Generation Communication and Networking Vol. 13, No. 1s, (2020), pp. 138- 146 MULTILINGUAL CHATBOT WITH HUMAN CONVERSATIONAL ABILITY [1] Aradhana Bisht, [2] Gopan Doshi, [3] Bhavna Arora, [4] Suvarna Pansambal [1][2] Student, Dept. of Computer Engineering,[3][4] Asst. Prof., Dept. of Computer Engineering, Atharva College of Engineering, Mumbai, India Abstract Chatbots - The chatbot technology has become very fascinating to people around the globe because of its ability to communicate with humans. They respond to the user query and are sometimes capable of executing sundry tasks. Its implementation is easier because of wide availability of development platforms and language libraries. Most of the chatbots support English language only and very few have the skill to communicate in multiple languages. In this paper we are proposing an idea to build a chatbot that can communicate in as many languages as google translator supports and also the chatbot will be capable of doing humanly conversation. This can be done by using various technologies such as Natural Language Processing (NLP) techniques, Sequence To Sequence Modeling with encoder decoder architecture[12]. We aim to build a chatbot which will be like virtual assistant and will have the ability to have conversations more like human to human rather than human to bot and will also be able to communicate in multiple languages. Keywords: Chatbot, Multilingual, Communication, Human Conversational, Virtual agent, NLP, GNMT. 1. Introduction A chatbot is a virtual agent for conversation, which is capable of answering user queries in the form of text or speech. In other words, a chatbot is a software application/program that can chat with a user on any topic[5].
    [Show full text]
  • Exploring the Use of Acoustic Embeddings in Neural Machine Translation
    This is a repository copy of Exploring the use of Acoustic Embeddings in Neural Machine Translation. White Rose Research Online URL for this paper: http://eprints.whiterose.ac.uk/121515/ Version: Accepted Version Proceedings Paper: Deena, S., Ng, R.W.M., Madhyashtha, P. et al. (2 more authors) (2018) Exploring the use of Acoustic Embeddings in Neural Machine Translation. In: Proceedings of IEEE Automatic Speech Recognition and Understanding Workshop. 2017 IEEE Automatic Speech Recognition and Understanding Workshop, December 16-20, 2017, Okinawa, Japan. IEEE . ISBN 978-1-5090-4788-8 https://doi.org/10.1109/ASRU.2017.8268971 Reuse Items deposited in White Rose Research Online are protected by copyright, with all rights reserved unless indicated otherwise. They may be downloaded and/or printed for private study, or other acts as permitted by national copyright laws. The publisher or other rights holders may allow further reproduction and re-use of the full text version. This is indicated by the licence information on the White Rose Research Online record for the item. Takedown If you consider content in White Rose Research Online to be in breach of UK law, please notify us by emailing [email protected] including the URL of the record and the reason for the withdrawal request. [email protected] https://eprints.whiterose.ac.uk/ EXPLORING THE USE OF ACOUSTIC EMBEDDINGS IN NEURAL MACHINE TRANSLATION Salil Deena1, Raymond W. M. Ng1, Pranava Madhyastha2, Lucia Specia2 and Thomas Hain1 1Speech and Hearing Research Group, The University of Sheffield, UK 2Natural Language Processing Research Group, The University of Sheffield, UK {s.deena, wm.ng, p.madhyastha, l.specia, t.hain}@sheffield.ac.uk ABSTRACT subsequently used to obtain a topic-informed encoder context Neural Machine Translation (NMT) has recently demon- vector, which is then passed to the decoder.
    [Show full text]
  • Neural Machine Translation System of Indic Languages - an Attention Based Approach
    Neural Machine Translation System of Indic Languages - An Attention based Approach Parth Shah Vishvajit Bakrola Uka Tarsadia University Uka Tarsadia University Bardoli, India Bardoli, India [email protected] [email protected] Abstract—Neural machine translation (NMT) is a recent intervention. This task can be effectively done using machine and effective technique which led to remarkable improvements translation. in comparison of conventional machine translation techniques. Machine Translation (MT) is described as a task of compu- Proposed neural machine translation model developed for the Gujarati language contains encoder-decoder with attention mech- tationally translate human spoken or natural language text or anism. In India, almost all the languages are originated from their speech from one language to another with minimum human ancestral language - Sanskrit. They are having inevitable similari- intervention. Machine translation aims to generate translations ties including lexical and named entity similarity. Translating into which have the same meaning as a source sentence and Indic languages is always be a challenging task. In this paper, we grammatically correct in the target language. Initial work on have presented the neural machine translation system (NMT) that can efficiently translate Indic languages like Hindi and Gujarati MT started in early 1950s [2], and has advanced rapidly that together covers more than 58.49 percentage of total speakers since the 1990s due to the availability of more computational in the country. We have compared the performance of our capacity and training data. Then after, number of approaches NMT model with automatic evaluation matrices such as BLEU, have been proposed to achieve more and more accurate ma- perplexity and TER matrix.
    [Show full text]
  • Terminology Extraction, Translation Tools and Comparable Corpora
    Terminology Extraction, Translation Tools and Comparable Corpora: TTC concept, midterm progress and achieved results Tatiana Gornostay, Anita Gojun, Marion Weller, Ulrich Heid, Emmanuel Morin, Beatrice Daille, Helena Blancafort, Serge Sharoff, Claude Méchoulam To cite this version: Tatiana Gornostay, Anita Gojun, Marion Weller, Ulrich Heid, Emmanuel Morin, et al.. Terminology Extraction, Translation Tools and Comparable Corpora: TTC concept, midterm progress and achieved results. LREC 2012 Workshop on Creating Cross-language Resources for Disconnected Languages and Styles (CREDISLAS), May 2012, Istanbul, Turkey. 4 p. hal-00819909 HAL Id: hal-00819909 https://hal.archives-ouvertes.fr/hal-00819909 Submitted on 9 May 2013 HAL is a multi-disciplinary open access L’archive ouverte pluridisciplinaire HAL, est archive for the deposit and dissemination of sci- destinée au dépôt et à la diffusion de documents entific research documents, whether they are pub- scientifiques de niveau recherche, publiés ou non, lished or not. The documents may come from émanant des établissements d’enseignement et de teaching and research institutions in France or recherche français ou étrangers, des laboratoires abroad, or from public or private research centers. publics ou privés. Terminology Extraction, Translation Tools and Comparable Corpora: TTC concept, midterm progress and achieved results Tatiana Gornostaya, Anita Gojunb, Marion Wellerb, Ulrich Heidb, Emmanuel Morinc, Beatrice Daillec, Helena Blancafortd, Serge Sharoffe, Claude Méchoulamf Tildea, Institute
    [Show full text]
  • N-Gram-Based Machine Translation
    N-gram-based Machine Translation ∗ JoseB.Mari´ no˜ ∗ Rafael E. Banchs ∗ Josep M. Crego ∗ Adria` de Gispert ∗ Patrik Lambert ∗ Jose´ A. R. Fonollosa ∗ Marta R. Costa-jussa` Universitat Politecnica` de Catalunya This article describes in detail an n-gram approach to statistical machine translation. This ap- proach consists of a log-linear combination of a translation model based on n-grams of bilingual units, which are referred to as tuples, along with four specific feature functions. Translation performance, which happens to be in the state of the art, is demonstrated with Spanish-to-English and English-to-Spanish translations of the European Parliament Plenary Sessions (EPPS). 1. Introduction The beginnings of statistical machine translation (SMT) can be traced back to the early fifties, closely related to the ideas from which information theory arose (Shannon and Weaver 1949) and inspired by works on cryptography (Shannon 1949, 1951) during World War II. According to this view, machine translation was conceived as the problem of finding a sentence by decoding a given “encrypted” version of it (Weaver 1955). Although the idea seemed very feasible, enthusiasm faded shortly afterward because of the computational limitations of the time (Hutchins 1986). Finally, during the nineties, two factors made it possible for SMT to become an actual and practical technology: first, significant increment in both the computational power and storage capacity of computers, and second, the availability of large volumes of bilingual data. The first SMT systems were developed in the early nineties (Brown et al. 1990, 1993). These systems were based on the so-called noisy channel approach, which models the probability of a target language sentence T given a source language sentence S as the product of a translation-model probability p(S|T), which accounts for adequacy of trans- lation contents, times a target language probability p(T), which accounts for fluency of target constructions.
    [Show full text]