Machine Transliteration

Language transliteration is one of the important areas in NLP. Transliteration is very useful for converting the named entities (NEs) written in one script to another script in NLP applications like Cross Lingual Information Retrieval (CLIR), Multilingual Voice Chat Applications and Real Time Machine Translation (MT). The most important requirement of Transliteration system is to preserve the phonetic properties of source language after the transliteration in target language. In this paper, we have proposed the named entity transliteration for Hindi to English and Marathi to English language pairs using Support Vector Machine (SVM). In the proposed approach, the source named entity is segmented into transliteration units; hence transliteration problem can be viewed as sequence labeling problem. The classification of phonetic units is done by using the polynomial kernel function of Support Vector Machine (SVM). Proposed approach uses phonetic of the source language and n-gram as two features for transliteration.

У монографії подано, обґрунтовано та розвинено комплексний пiдxiд до вживання і творення українських тeрмiнiв, заснований на використанні методів природничих наук і мовознавчих аспектів дослідження. Такий підхід створює передумови для того, щоб розглядати термінологію не лише як розділ лексики, сукупність термінів і вчення про них, а насамперед як науку про утворення, розвиток і функціонування термінів. За допомогою послідовного застосування статистичного та аналітичного методів досліджено теоретичні й практичні термінографічні проблеми, розкрито правописні аспекти відтворення запозичених та іншомовних слів. На основі методу акустичних інваріантів досліджено акустичні характеристики звуків української (а також англійської) мови та знайдено їх інваріантні акустичні параметри. У монографії відображено результати автора, одержані в процесі його наукової та викладацької роботи в Державній академії керівних кадрів освіти, Міжрегіональному інституті вдосконалення вчителів, Київському національному лінгвістичному університеті, Київському національному університеті імені Тараса Шевченка, Техаському університеті в Арлінгтоні (США), Єнському університеті імені Фрідріха Шіллера (Німеччина), Гіссенському університеті імені Юстуса Лібіга (Німеччина), Лейпцізькому університеті (Німеччина), Дрезденському технічному університеті (Німеччина), Університеті Країни басків (Іспанія), Українському мовно-інформаційному фонді Національної академії наук України.

Для науковців, викладачів, аспірантів і студентів філологічних факультетів вищих навчальних закладів, а також для лексикографів, фонетистів, перекладачів і фахівців у галузі прикладної лінгвістики, української термінології і термінографії.

CONTENTS
LIST OF SYMBOLS AND ABBREVIATIONS 8
INTRODUCTION 9
PART 1
THEORETICAL AND METHODOLOGICAL PRINCIPLES OF TERMINOLOGY 18
1.1. Term and terminology: underlying statements 18
1.2. Successful term criteria 30
1.3. General tendencies of development of terminological systems 30
1.4. Main lexical semantic relations in term vocabulary (polysemy, homonymy, paronymy, synonymy) 40
1.5. National and international trends of term fund updating 62
PART 2
SEMANTIC, GRAMMATICAL, AND PHONETIC PROBLEMS OF
UKRAINIAN TERMS CODIFICATION 75
Section 2.1. Semantic and morphological qualities of Ukrainian
term vocabulary 75
2.1.1. Alien borrowings as a linguistic and terminological problem 75
2.1.2. Semantic and stylistic features of use of paronyms and pseudosynonyms in the professional language 83
2.1.3. Scientific approach to  terminological use of verbal nouns 112
2.1.4. Principles of representation of active participial forms in the Ukrainian scientific language 125
2.1.5. Possibilities and prospects of national word coinage
in terminology 127
Section  2.2. Acoustic characteristics of Ukrainian speech sounds as a ground
for scientific approach to render the borrowings 135
2.2.1. Phonetics, acoustic analysis, speech recognition 135
2.2.2. Phonetical and phonological language elements 138
2.2.2.1. Basic concepts of phonetics and phonology 138
2.2.2.2. Nonlinearity and non-invariance of speech elements 147
2.2.3. Speech acoustics 152
2.2.3.1. Acoustic theory of speech production 152
2.2.3.2. Speech sounds formants 157
2.2.3.3. Resonators in a human speech tract 159
2.2.3.4. Invariant acoustic speech analysis 162
2.2.3.5. Acoustic characteristics and invariants of
speech sounds 165
2.2.3.6. Interaction between sounds 179
PART 3
PECULIARITIES OF ORTHOGRAPHIC NORM APPLICATION IN SCIENTIFIC LEXICOGRAPHY AND WAYS TO PERFECT
MODERN SPELLING 189
3.1. Modern spelling strategy problems 189
3.2. Problems of modern spelling of terminological
and common vocabulary 191
3.2.1. Genitive case of masculine nouns of second reversal 181
3.2.2. Use of genitive and accusative cases 195
3.2.3. Spelling of compiled adjectives 196
3.2.4. The features of use of Ukrainian prepositions “в” (“у”) and
“до” for denotation of motion aim and direction 198
3.2.5. Stress problem in three-syllable nouns with the prefix “за ” 202
3.3. Denotation of time moments in Ukrainian 206
3.4. Problem of adaptation of alien terms to Ukrainian spelling 211
3.4.1. Study of phonemes /ґ/, /г/, and /х/ 211
3.4.2. “The nine” (“dev’jatka”) rule in the context of Slavic and non Slavic parallels of the Ukrainian language 214
3.4.3. Doubling simplification problems 220
3.4.4. Some “trifles” of spelling perfection 221
3.4.5. Discussion suggests to spelling 222
3.5. Scholar principles to render borrowed and foreign words:
invariant transcription and transliteration 229
3.5.1. Pronunciation and transcription 229
3.5.2. Writing and transliteration 234
CONCLUDING STATEMENTS 255
LIST OF LITERATURE 260
APPENDIX A: Visual representation of speech sounds
(spectrograms, oscillograms) 397
APPENDIX B: Inculcation of the research results
(main documentation) 337
APPENDIX V: Use of the term Χυμεία in Greek 361

Let me present to you my study “Word Order Errors in English and Bulgarian Machine-Translated Texts”.
My work contains Introduction, Three chapters, Conclusion, Appendix and Bibliography.
The object, the main goal and tasks of the research are defined into the Introduction. The object is a specific kind of errors in English and Bulgarian machine-translated texts, defined as word order errors.
The main goal is to give theoretical and practical description of how and why the word order errors appear in the machine-translated texts.
The main tasks are to examine and analyze the word order errors in machine-translated texts, in order to facilitate their editing and to improve the use of a machine translator.
The results and the analysis could be helpful to the users of Google translate, especially students who often take advantage of its services. Nowadays more and more examples of ridiculous machine-translated texts can be found. So, if one knows better the types of errors which are made by the translation systems, one can easily recognize and edit them.
To achieve these tasks two scientific texts have been selected – English and a Bulgarian one, and have been translated via Google Translate, as it is one of the most popular machine translators.

This paper proposes a novel approach for multilingual query processing, wherein we propose a phonetic distance based measure, for searching proper name data in Indian language scripts. The systems allows query in a language of user's choice. A cross-lingual search is conducted with the query being in one language and the documents being searched for, in another. Grapheme-to-phoneme converters are used to convert the user's query into an intermediate language-independent common ground (CG) representation. A dynamic time warping algorithm, wherein the substitution cost is based on a weighted phonetic distance measure, is used to match and rank the query results. In turn, a phoneme-to-grapheme converter is used to convert the search results in CG representation to the user's query language. We also discuss in detail the various issues particular to cross-lingual search on proper name data and address the same using the proposed approach.

AWEJ Volume.5 Number.2, 2014   
                                                                                                                                                                    Pp. 142-152



The Impact of Translation on the Moroccan Political Discourse


Mohamed MAROUANE
                                              Department of English
                          Chouaib Doukkali University, Morocco



Abstract
The present article investigates the influence of concept transfer and translation on the political discourse used in Morocco. It has been attested that the Moroccan political register suffers from an acute lexical gap to express a number of aspects of the modern political system. Translation from European languages has, thus, represented an important source of the political register used in the political texts and speeches. This is explained by the adoption of a modern political system which has developed in different historical and social contexts totally different from the Arab traditional communities. Yet, this transferred register is very heterogeneous as it consists of items which have been translated following different translation strategies ranging from providing native equivalents to mere transliteration.
Keywords: Neologism, Loan translation, Political discourse , Political register, Transliteration,

Globalization and growth of Internet users truly demands for almost all internet based applications to support local languages. Support of local languages can be given in all internet based applications by means of Machine Transliteration and Machine Translation. This paper provides the thorough survey on machine transliteration models and machine learning approaches used for machine transliteration over the period of more than two decades for internationally used languages as well as Indian languages. Survey
shows that linguistic approach provides better results for the closely related languages and probability based statistical approaches are good when one of the languages is phonetic and other is nonphonetic.Better accuracy can be achieved only by using Hybrid and Combined models.

In Sanskrit study and research in Thailand, transliteration from Sanskrit documents and proofreading is a difficult and time-consuming task since Sanskrit documents contain a great deal of content. To find solutions, this study presented the Romanized Sanskrit – Thai Machine Transliteration, the rules development and algorithm of Romanized Sanskrit transliteration based on the IAST standard to Thai alphabets. There are two patterns. Transliteration patterns are composed of stable pattern and common pattern. The experimental results showed that the proposed former could perform accurately up to 98.16 % from 21,399 samples whereas the latter could perform accurately up to 95.82 % from 30,049 samples. The expert in Sanskrit assessed it and commented that this system could reasonably transliterate Sanskrit and saved a great deal of time in transliteration.

This paper describes our submission for FIRE 2014 Shared Task on Transliterated Search. The shared task features two subtasks: Query word labeling and Mixed-script Ad hoc retrieval for Hindi Song Lyrics.
Query Word Labeling is on token level language identification of query words in code-mixed queries and  back-transliteration of identified Indian language words into their native scripts. We have developed letter based language models for the token level language identification of query words and a decision tree classifier for transliteration.
The second subtask for Mixed-script Ad hoc retrieval for Hindi Song Lyrics is to retrieve a ranked list of songs from a corpus of Hindi song lyrics given an input query in Devanagari or transliterated Roman script. We have used edit distance based query expansion and language modeling followed by relevance based reranking for the retrieval of relevant Hindi Song lyrics for a given query.

This paper presents a design for rule-based machine translation system for English to Marathi language pair. The machine translation system will take input script as English sentence and parse with the help of Stanford parser. The Stanford parser will be used for main purposes on the source side processing, in the machine translation system. English to Marathi Bilingual dictionary is going to be created. The system will
take the parsed output and separate the source text word by word and searches for their corresponding target words in the bilingual dictionary. The hand coded rules are written for Marathi inflections and also reordering rules are there. After applying the reordering rules, English sentence will be syntactically reordered to suit Marathi language

This article discusses basic approaches to transcribe foreign and borrowed words in Ukrainian, and Russian, Belarusian, and Ukrainian words in Latin script. It is argued that the adopted and foreign words should be rendered on different bases, namely by invariant transcription and transliteration. Also, current problems of implementation of the Ukrainian Latinics as an international graphical presentation of Ukrainian, are analyzed. The scholarly grounded simple-correspondent transliteration system for Belarusian, Russian, and Ukrainian, is given.

e-Governance and Web based online commercial multilingual applications has given utmost importance to
the task of translation and transliteration. The Named Entities and Technical Terms occur in the source
language of translation are called out of vocabulary words as they are not available in the multilingual
corpus or dictionary used to support translation process. These Named Entities and Technical Terms need
to be transliterated from source language to target language without losing their phonetic properties. The
fundamental problem in India is that there is no set of rules available to write the spellings in English for
Indian languages according to the linguistics. People are writing different spellings for the same name at
different places. This fact certainly affects the Top-1 accuracy of the transliteration and in turn the
translation process. Major issue noticed by us is the transliteration of named entities consisting three
syllables or three phonetic units in Hindi and Marathi languages where people use mixed approach to
write the spelling either by orthographical approach or by phonological approach. In this paper authors
have provided their opinion through experimentation about appropriateness of either approach

Transliteration may be defined as the alteration of text from one language to another. In this paper, we have discussed how transliteration is useful in handling unknown words in Named Entity Recognition (NER). We have shown some of our results on unknown words handling in NER using transliteration.

Hindi is the lingua-franca of India. Although all non-native speakers can communicate well in Hindi, there are only a few who can read and write in it. In this work, we aim to bridge this gap by building transliteration systems that could transliterate Hindi into at-least 7 other Indian languages. The transliteration systems are developed as a reading aid for non-Hindi readers. The systems are trained on the transliteration pairs extracted automatically from a parallel corpora. All the transliteration systems perform satisfactorily for a non-Hindi reader to understand a Hindi text.

We have developed a finite state transducer based transliteration engine called Konkanverter that performs statistical machine transliteration between three different scripts used to write the Konkani language. The statistical machine transliteration system consists of cascading finite state transducers combining both rule-based and statistical approaches. Based on the limited availability of parallel corpora, this cascading approach is found to perform significantly better than a pure rule-based approach or pure statistical approach.

The Web provides the largest, exploitable collection of language use. If we can mine the Web to build abstract models of language use, these models may have many applications. Here we present one example of using the implicit intelligence of language use to solve an important problem for machine translation programs and cross-lingual applications. This problem involves the translation of words written in katakana characters in Japanese. In this paper, we describe techniques of discovering katakana transliteration of English names and of finding English translations of multiword katakana sequences using implicit language models of English and Japanese found on the Web. These techniques were evaluated against human-constructed English-katakana glosses.

Translatability in natural languages today supports, and can only be understood in the context of, economic globalization, and the universalization of market logic. ‘Untranslatability’, as it is most often construed, does not provide a critical alternative to this logic: it bolsters it. A different account of untranslatability (and, by extension, of what it means for expressions in natural languages to be ‘translatable’) is required: this essay seeks to provide such. It finds in passages in Marx and in Derrida's Monolinguisme de l'autre, and in different translations of those texts, an untranslatability which is not one, irreducible to mathematical identity.

Transliteration may be defined as the process of mapping sounds in a text written in one language to another language. Current paper discusses about transliteration and its use in Named Entity Recognition. We have designed a code that executes Transliteration and assist in the process of Named Entity Recognition. We have presented some of the results of Named Entity Recognition (NER) using Transliteration.

Much of the user generated content on the internet is written in their transliterated form instead of in their indigenous script. Due to this search engines receive a large number of transliterated search queries.

This paper presents our approach to handle labelling of queries and ad hoc retrieval of documents based on these queries, as part of the FIRE2014 shared task on transliterated search. The content of each document is written in either the native Devanagari script or its transliterated form in Roman script or a combination of both. The queries to retrieve these documents can also be in mixed script. The task is challenging primarily due to the spelling variations that occur in the transliterated form of search queries. This particular problem is addressed by using back transliteration to reduce spelling variations, and a set of hand-tailored rules for consonant mapping. Sub-word indexing is done to take care of breaking and joining of transliterated words. Implementation of query labelling of the mixed script content was done using a supervised learning approach where an SVM classifier was trained using character $n$-grams as features for language identification. A Naïve Bayes classifier was used for classifying transliterated words that can belong to both Hindi and English when looked at individually.

The 2 runs submitted by our team BITS-Lipyantaran performs best across all metrics for Subtask 2 among all the teams that participated, with a MRR score of 0.8171 and MAP score of 0.6421.

Only the working notes have been added here. The full paper can be viewed at this link: http://dl.acm.org/citation.cfm?doid=2824864.2824873

Machine Translation for Indian languages is an emerging research area. Transliteration is one such module that we design while designing a translation system. Transliteration means mapping of source language text into the target language. Simple mapping decreases the efficiency of overall translation system. We propose the use of stemming and part-of-speech tagging for transliteration. The effectiveness of translation can be improved if we use part-of-speech tagging and stemming assisted transliteration. We have shown that much of the content in Gujarati gets transliterated while being processed for translation to Hindi language.

Translatability in natural languages today supports, and can only be understood in the context of, economic globalization, and the universalization of market logic. ‘Untranslatability’, as it is most often construed, does not provide a critical alternative to this logic: it bolsters it. A different account of untranslatability (and, by extension, of what it means for expressions in natural languages to be ‘translatable’) is required: this essay seeks to provide such.
It finds in passages in Marx and in Derrida’s Monolinguisme de l’autre, and in different translations of those texts, an untranslatability which is not one, irreducible to mathematical identity.

We describe ScriptTranscriber, an open source toolkit for extracting transliterations in comparable corpora from lan- guages written in different scripts. The system includes various methods for extracting potential terms of interest from raw text, for providing guesses on the pronunciations of terms, and for comparing two strings as possible transliterations using both phonetic and temporal measures. The system works with any script in the Unicode Basic Multilingual Plane and is easily extended to include new modules. Given comparable corpora, such as newswire text, in a pair of languages that use different scripts, ScriptTranscriber provides an easy way to mine transliterations from the comparable texts. This is particularly useful for underresourced languages, where training data for transliteration may be lacking, and where it is thus hard to train good transliterators. ScriptTranscriber provides an open source package that allows for ready incorporation of more sophisticated modules — e.g. a trained transliteration model for a particular language pair. ScriptTranscriber is available as part of the nltk contrib source tree at http://code.google.com/p/nltk/

The present article investigates the influence of concept transfer and translation on the political discourse used in Morocco. It has been attested that the Moroccan political register suffers from an acute lexical gap to express a number of aspects of the modern political system. Translation from European languages has, thus, represented an important source of the political register used in the political texts and speeches. This is explained by the adoption of a modern political system which has developed in different historical and social contexts totally different from the Arab traditional communities. Yet, this transferred register is very heterogeneous as it consists of items which have been translated following different translation strategies ranging from providing native equivalents to mere transliteration.

Transliteration may be defined as the alteration of text from one language to another. In this paper, we have discussed how transliteration is useful in handling unknown words in Named Entity Recognition (NER). We have shown some of our results on unknown words handling in NER using transliteration.

Computational LFG grammar in development in Konstanz Aim: large-scale LFG grammar for parsing Urdu/Hindi Grammar is part of the ParGram project Collaborative, world-wide research project Devoted to developing parallel LFG grammars for a variety of languages Features and analyses are kept parallel for easy transfer between languages

In this paper, we propose efficient and less resource-intensive strategies for parsing of code-mixed data. These strategies are not constrained by in-domain annotations , rather they leverage pre-existing monolingual annotated resources for training. We show that these methods can produce significantly better results as compared to an informed baseline. Besides, we also present a data set of 450 Hindi and English code-mixed tweets of Hindi multilingual speakers for evaluation. The data set is manually annotated with Universal Dependencies.

Computational LFG grammar in development in Konstanz Aim: large-scale LFG grammar for parsing Urdu/Hindi Grammar is part of the ParGram project Collaborative, world-wide research project Devoted to developing parallel LFG grammars for a variety of languages Features and analyses are kept parallel for easy transfer between languages

Machine Transliteration has come out to be an emerging and a very important research area in the field of machine translation. Transliteration basically aims to preserve the phonological structure of words. Proper transliteration of name entities plays a very significant role in improving the quality of machine translation. In this paper we are doing machine transliteration for English-Punjabi language pair using rule based approach. We have constructed some rules for syllabification.  Syllabification is the process to extract or separate the syllable from the words. In this we are calculating the probabilities for name entities (Proper names and location). For those words which do not come under the category of name entities, separate probabilities are being calculated by using relative frequency through a statistical machine translation toolkit known as MOSES. Using these probabilities we are transliterating our input text from English to Punjabi.

e-Governance and Web based online commercial multilingual applications has given utmost importance to the task of translation and transliteration. The Named Entities and Technical Terms occur in the  source language of translation  are called out of vocabulary words as they are not available in the  multilingual corpus or dictionary used to support translation process. These Named Entities and Technical Terms need to be transliterated from source language to target language without losing their phonetic properties. The fundamental problem in India is that there is no set of rules available to write the spellings in English for Indian languages according to the linguistics. People are writing different spellings for the same name at different places. This fact certainly affects the Top-1 accuracy of the transliteration and in turn the translation process. Major issue noticed by us is the transliteration of named entities consisting three syllables or three phonetic units in Hindi and Marathi languages where people use mixed approach to write the spelling either by orthographical approach or by phonological approach. In this paper authors have provided their opinion through experimentation about appropriateness of either approach.

Our work is based on the system “Transscriba” [11]. This is a rule-based system, that means that transliteration model is a set of rules, constructed manually by expert. Such approach provides very high accuracy of transliteration, but it takes from two weeks to six months of an expert's work to construct a system of rules for one pair of languages. Moreover, “Transscriba” has one more drawback. Its method of strings transformation is ineffective as speed of parsing depends on amount of rules in the system. In Section 4 we demonstrate effective method of strings processing with generated rules. Section 5 introduces the method of automatic learning of cross-language transcription rules from small training set.

An important prerequisite for data-driven Machine Translation (MT) systems is the availability of high-quality training data. Corpus-based MT systems requiring domain specificity additionally require the selection of a large training dataset and application of proper domain adaptation techniques. This paper demonstrates the portability of MT systems to multiple domains under low-resource conditions for the target language, where target-language domain data are not available. We refer to this case as a bilingually low-resource language pair. We compare the performance of an attention-based Neural MT (NMT) system to that of a phrase-based Statistical MT (SMT) system under this condition, training each on parallel corpora consisting of different domains. Experimental results on Spanish-Farsi as a bilingually low-resource language pair show that the SMT paradigm still outperforms that of NMT.

Phonetic typing using the English alphabet has become widely popular nowadays for social media and chat services. As a result, a text containing various English and Bangla words and phrases has become increasingly common. Existing transliteration tools display poor performance for such texts. This paper proposes a robust Three-stage Hybrid Transliteration (THT) framework that can transliterate both English words and phonetic typed Bangla words satisfactorily. This is achieved by adopting a hybrid approach of dictionary-based and rule-based techniques. Experimental results confirm superiority of THT as it significantly outperforms the benchmark transliteration tool.