131x Filetype PDF File size 0.20 MB Source: www.ijettcs.org
International Journal of Emerging Trends & Technology in Computer Science (IJETTCS) Web Site: www.ijettcs.org Email: editor@ijettcs.org, editorijettcs@gmail.com Volume 9, Issue 5, September - October 2020 ISSN 2278-6856 Systematic Review on Techniques of Machine Translation for Indian Languages 1 2 3 Simarn N. Maniyar , Sonali B. Kulkarni , Pratibha R. Bhise 1 Departemet of Computer Sciece and Information Techonology Dr.Babasaheb Ambedker Marathwada University,Auragabad. 2 Departemet of Computer Sciece and Information Techonology Dr.Babasaheb Ambedker Marathwada University,Auragabad. 3 Departemet of Computer Sciece and Information Techonology Dr.Babasaheb Ambedker Marathwada University,Auragabad. Abstract: Machine Translation is the branch of Natural linguistics (CL) and Natural Language Processing (NLP) Language Processing, which deals the use of Machine Translation that explores the use of computer/mobile application to software to convert from one language to other natural languages translate text or speech from one natural language known as with the help of machine Translation Approaches. The objective is the source language to another known as target language. to fill the language gap between two different languages speaking Like translation done by human, MT does not simply people, communities or countries. In India we have multiple and substituting words but the application of complex linguistic greatly distinct language scripts, hence scope of need of language knowledge; morphology, grammar, meaning all this things translation. In this purpose we present Literature survey on is taken into consideration. Generally, MT is classified into machine translation on the current scenario of research machine various categories: Direct based, rule-based, corpus based, translation in India. There is various Machine Translations statistical-based, hybrid-based, example-based, knowledge- system. Machine translation is considered an important task that can be used to attain information from documents written in based, principle-based, and online interactive based different languages. In this paper examine the rule based machine methods. At present, most of the MT related research is translation is useful for translate for English to Indian languages. based on Rule based approaches because rule based is always extensible and maintainable. Morphological Keywords: NLP, Machine Translation, Techniques, analysis, part of speech tagging, chunking, parsing and RBMT word sense disambiguation this is the major goals of 1. INTRODUCTION machine translation. In this paper, we are present a systematic literature review of the techniques used for Natural language processing can be classified as a subset of machine translation. We have also mentioned recent work the broader field of speech and language processing. in the table from. Because of this, NLP shares similarities with parallel 2. Literature Review disciplines such as computational linguistics, which is concerned with modeling language using rule-based Table 1: Review for Machine Translation Techniques models. The goal of natural language processing is to build computational models of natural language for its analysis SR. PAPER NAME AUTHER YE LAGUA TECHNI and generation. First, there is technological motivation of NO AR GE PAIR QUES building intelligent computer system such as machine 1 TRANSLATION OF DR. 2014 TELU RULE- translation system natural language interfaces to databases, TELUGU- SIDDHARTHA GU- BASED MARATHI AND GHOSH, MAR AND man-machine interfaces to computer in general, speech VICE-VERSA SUJATA ATHI STATIST understanding system, text analysis and understanding USING RULE THAMKE ICAL- system, computer aided instruction systems, systems that BASED MACHINE AND KALYANI BASED read and understand printed or handwritten text. Second, TRANSLATION U.R.S 2 RULE BASED NAILA ATA, - ENGLI RULE there is a cognitive and linguistic motivation to gain better ENGLISH TO BUSHRA SH TO BASED insights into how humans communicate using natural URDU MACHINE JAWAID, AMIR URDU MACHIN language. The tools of work in NLP are grammar TRANSLATION KAMRAN E formalisms, algorithms for representing world knowledge, TRANSL ATION reasoning mechanisms, etc. Natural language interfaces to 3 TRANSLATION OF PROF. 2014 ENGLI RULE databases, natural language interfaces to computer, SIMPLE ENGLISH GORAKSH SH TO BASED question answering system, story understanding and INTERROGATIVE V.GARJE, MARA MACHIN machine translation system these all the application of SENTENCES TO MANISHA THI E MARATHI MARATHE, TRANSL natural language processing. We are focusing in the SENTENCES URMILA ATION, machine translation application and these approaches. ADSULE TRANSF Machine Translation (MT) is the field of computational ER MT 4 HYBRID 2018 MARA HYBRID Volume 9, Issue 5, September - October 2020 Page 44 International Journal of Emerging Trends & Technology in Computer Science (IJETTCS) Web Site: www.ijettcs.org Email: editor@ijettcs.org, editorijettcs@gmail.com Volume 9, Issue 5, September - October 2020 ISSN 2278-6856 MACHINE PROF. THI MACHIN TECHNIQUE TRANSLATION ABHAPATHAK TO E 15 WEB BASED VISHAL GOYAL 2010 HINDI DIRECT FROM MARATHI , ANCHAL ENGLI TRANSL HINDI TO AND GURPREET TO TRANSL TO ENGLISH: A KUMARI SH ATION PUNJABI SINGH LEHAL PUNJ ATION RESEARCH , AKANKSHA MACHINE ABI EVOLUTION IN PRASAD TRANSLATION MACHINE , ASHWINI SYSTEM TRANSLATION TOPRE 16 A HEURISTIC PRIYANKA 2017 ENGLI GRAPH , APPROACH FOR MALVIYA, SH TO BASED RUTUJALONDH GRAPH BASED GAURI RAO, HINDI APPROA E MACHINE ROHINI B. CH TRANSLATION JADHAV, 5 TRANSMUTER: G V GARJE, G 2014 ENGLI RULE MAYURI H. AN APPROACH K KHARATE, SH TO BASED MOLAWADE TO RULE-BASED HARSHAD MARA MACHIN 17 DEVELOPMENT K. 2015 TELU TRANSF ENGLISH TO KULKARNI THI E OF TELUGU- PARAMESWARI GU- ER- MARATHI TRANSL TAMIL TAMI BASED MACHINE ATION TRANSFER- L MACHIN TRANSLATION BASED MACHINE E 6 MARATHI TO G.V. GARJE, 2016 MARA RULE TRANSLATION TRANSL ENGLISH PHD, AKSHAY THI BASED SYSTEM: WITH ATION SENTENCE BANSODE, TO TRANSL SPECIAL SYSTEM TRANSLATOR SUYOG ENGLI ATION REFERENCE TO FOR SIMPLE GANDHI, ADITA SH APPROA DIVERGENCE ASSERTIVE AND KULKARNI CH INDEX INTERROGATIVE 18 AN ENGLISH TO R.M.K. SINHA, 2003 ENGLI RULE- SENTENCES HINDI MACHINE- A. JAIN SH TO BASED 7 ENGLISH TO G.V. GARJE, 2015 ENGLI RULE AIDED HINDI AND MARATHI RULE- G.K. KHARATE SH TO BASED TRANSLATION EXAMPL BASED MACHINE AND M.L. MARA MACHIN SYSTEM E- TRANSLATION OF DHORE THI E BASED SIMPLE TRANSL 19 ENGLISH–HINDI ONDŘEJ BOJAR, 2008 ENGLI RULE- ASSERTIVE ATION TRANSLATION IN PAVEL SH TO BASED SENTENCES 21 DAYS STRAŇÁK, HINDI AND 8 NEURAL SANDEEP SAINI, 2018 ENGLI NEURAL DANIEL ZEMAN EXAMPL MACHINE VINEET SH TO MACHIN E- TRANSLATION SAHULA HINDI E BASED FOR ENGLISH TO TRANSL HINDI ATION SIMPLE ANANTHAKRIS 9 NEURAL KARTHIK 2017 INDIA NEURAL 20 SYNTACTIC AND HNAN 2008 ENGLI STATIST MACHINE REVANURU, N MACHIN MORPHOLOGICA RAMANATHAN, SH- ICAL TRANSLATION OF KAUSHIK LANG E L PROCESSING PUSHPAK HINDI MACHIN INDIAN TURLAPATY, UAGE TRANSL CAN HELP BHATTACHARY E LANGUAGES SHRISHA RAO S ATION ENGLISH-HINDI YA, JAYPRASAD TRANSL 10 NEURAL HIMANSHU - ENGLI NEURAL STATISTICAL HEGDE, RITESH ATION MACHINE CHOUDHARY, SH- MACHIN MACHINE M. TRANSLATION ADITYA TAMI E TRANSLATION SHAH,SASIKUM FOR ENGLISH- KUMAR L TRANSL AR M TAMIL PATHAK ATION 21 EXAMPLE BASED F.H.A.M. 2015 ENGLI EXAMPL 11 STATISTICAL SUBALALITH, 2018 ENGLI STATIST MACHINE SILVA , SH- E MACHINE AARTHI SH TO ICAL TRANSLATION A.R.WEERASIN SINHA BASED TRANSLATION VENKATARAMA HINDI MACHIN FOR GHE AND LA MACHIN FROM ENGLISH N,BASIMSHAHI E ENGLISH- H.L.PREMARAT E TO HINDI DBAQUI TRANSL SINHALA ENE TRANSL ATION TRANSLATIONS ATION 12 G.SURYAKALA 2018 ENGLI STATIST 22 A NOVEL PROF.KRUSHNA 2014 ENGLI EXAMPL MACHINE ESWARI, SH TO ICAL . APPROCH FOR DEO.T.BELERA SH TO E- TRANSLATION N.V.S.SOWJAN HINDI MACHIN INTERLINGUAL O, MARA BASED FROM ENGLISH YA,P.SURYAPR E EXAMPLE-BASED PROF.VINOD. S. THI MACHIN TO HINDI ABHAKAR RAO TRANSL TRANSLATION OF WADNE E ATION ENGLISH TO ,PROF. S. V. TRANSL 13 ETRANS- PROMILA 2012 ENGLI RULE MARATHI PHULARI, ATION ENGLISH TO BAHADUR, SH TO BASED PROF. B.S. SANSKRIT A.K.JAIN, SANS MACHIN KANKATE MACHINE D.S.CHAUHAN KRIT E 23 DESIGN AND LATHA R NAIR, 2012 MALA A TRANSLATION TRANSL DEVELOPMENT DAVID PETER YALA TRANSF ATION OF A S, RENJITH P M TO ER 14 TRANSFORMATIO MR.UDAY C. 2012 ENGLI RULE MALAYALAM TO RAVINDRAN ENGLI BASED N OF MULTIPLE PATKAR, SH TO BASED ENGLISH SH APPROA ENGLISH TEXT PROF.PRAKASH SANS MACHIN TRANSLATOR- A CH SENTENCES TO R. DEVALE, KRIT E TRANSFER VOCAL SANSKRIT PROF.DR.SUHA TRANSL BASED USING RULE S.H.PATIL ATION APPROACH BASED 24 MALAYALAM TO ANISREE P G1, 2016 MALA HYBRID Volume 9, Issue 5, September - October 2020 Page 45 International Journal of Emerging Trends & Technology in Computer Science (IJETTCS) Web Site: www.ijettcs.org Email: editor@ijettcs.org, editorijettcs@gmail.com Volume 9, Issue 5, September - October 2020 ISSN 2278-6856 ENGLISH RADHIKA K T2 YALA APPROA 3.1 Rule Based Translation MACHINE M TO CH A Rule-Based Machine Translation (RBMT) system TRANSLATION: A ENGLI consists of collection of various rules, called grammar HYBRID SH rules, a bilingual lexicon or dictionary, and software APPROACH 25 A HYBRID NITHYA B, 2013 ENGLI HYBRID programs to process the rules. Rule-Based Machine APPROACH TO SHIBILY JOSEPH SH TO APPROA Translation (RBMT), also known as Humaistic methods of ENGLISH TO MALA CH MT, is a general term that indicates machine translation MALAYALAM YALA systems based on linguistic information about source and MACHINE M TRANSLATION target languages basically recover from (bilingual) dictionaries and grammars covering the main semantic, 26 A RULE BASED T.K. BIJIMOL1* 2018 MALA A RULE morphological, and syntactic consistencies of each APPROACH FOR , JOHN T. YALA BASED language respectively. Having source language sentences , TRANSLATION OF ABRAHAM2 M TO APPROA CAUSATIVE ENGLI CH an RBMT system generates them to target language CONSTRUCTION SH sentences on the support of morphological, syntactic, and OF ENGLISH AND AND advantages of RBMT system is that the interlingua develop MALAYALAM ENGLI into more valuable as the amount of target languages it can FOR THE SH TO DEVELOPMENT MALA be turned into development. Interlingual machine OF YALA translation system has been built operational at the PROTOTYPE FOR M economical level is the KANT system (Nyberg and MALAYALAM TO Mitamura, 1992), which is develop to translate Caterpillar ENGLISH AND ENGLISH TO Technical English (CTE) into other languages. The MALAYALAM interlingual technique is clearly attractive for multilingual BILINGUAL systems. All other analysis modules and of all generation MACHINE modules both of analysis module can be independent. TRANSLATION SYSTEM 3.2 Direct Translation: 27 RULE BASED R. REMYA, S. 2009 ENGLI RULE One of the simplest machine translation techniques is MACHINE REMYA, R. SH TO BASED Direct Machine Translation in which technique with the TRANSLATION REMYA, AND K. MALA MACHIN help of bilingual dictionary direct word to word translation FROM ENGLISH P. YALA E is done. Starting with the shallowest level at the bottom of TO MALAYALAM SOMAN M TRANSL ATION the pyramid is the Direct Machine Translation Technique. DMT technique is the oldest technique and also less popular technique. Direct translation is made at the word 3. Overview of machine translation level. Machine translation systems that use this approach Approaches are capable of translating a language, source language (SL) Researchers proposed many approaches for the Machine directly to target language (TL). Words of the source Translation. Overview of main approaches is presented language are translated without passing through an here. There are two broad categories of Machine additional/intermediary representation. The analysis of Translation Systems, namely Rule-Based and Empirical source language texts is oriented to only one target Based Machine Translation Systems. Hybrid Machine language. Direct translation systems are basically bilingual Translation system takes the benefits from both Rule-Based and uni-directional. Direct translation technique needs only Machine Translation System and Empirical Based Machine a little syntactic and semantic analysis. SL analysis is Translation System. Rule-Based Machine Translation oriented specifically to the production of representations System is further classified into Direct, Transfer and applicable for one particular Target language . DMT is a Interlingua, while Empirical Based Translation System is word-by-word translation technique with some simple classified into Statistical and Example based machine grammatical adjustments. translation system Figure 2: Direct Machine translation Approach Figure1. Technique of Machine Translation Volume 9, Issue 5, September - October 2020 Page 46 International Journal of Emerging Trends & Technology in Computer Science (IJETTCS) Web Site: www.ijettcs.org Email: editor@ijettcs.org, editorijettcs@gmail.com Volume 9, Issue 5, September - October 2020 ISSN 2278-6856 3.3. Statistical-based Approach: example-based machine translation is the idea of translation SMT can be described as the process of finding and by analogy. The ethic of translation by analogy is encoded matching identical pairs from SL and TL in parallel to example-based machine translation through the example corpora. The goal of SMT is to make an optimal decision in translations that are used to train such a system. language translation by using statistical decision theory, 3.5. Knowledge-Based MT: based on probability distribution function. The important Knowledge-Based Machine Translation (KBMT) requires feature of SMT is the presence of statistical table, which complete understanding of the source text prior to the can be built by using supervised or unsupervised statistical translation into the target text. KBMT is implemented on machine learning algorithms. Statistical table generally the Interlingua 56 International Journal on Natural contains statistical information pertaining to sentences or Language Computing (IJNLC) Vol. 4, No.2,April 2015 57 languages. SMT relies on a statistical calculation of the architecture. KBMT must be supported by world probabilities of a match [17] by using two probabilistic knowledge and by linguistic semantic knowledge around models: Language model and Translation model, rather meanings of words and their combinations. than relying on linguistic translation algorithms. The idea of SMT is that document can be translated on the basis of 3.6. Hybrid-based Translation: probability distribution function P(t/s), where P(t/s) is the Hybrid-based technique is develop using the advantages of probability of translating a sentence, say ’s’ in SL to a SMT and EBMT methodologies. The hybrid technique sentence ’t’ in TL. And this function is generated easily by used in a number of different ways. Translations are using Bayes theorem. In Bayes theorem probability performed in the first level using a rule-based technique distribution p(t/s) is obtained from the product of P(s/t) and which is followed by adjusting or correcting the output p(t), where P(s/t) is the probability that the source sentence using statistical information. other way in which rules are is a translation of the target sentence, and P(t) is the used to pre-process the input data and for post-process the probability of the TL statistical output of a statistical-based translation system. By taking the advantage of both statistical and rule-based translation methodologies, a new approach was developed, called hybrid-based approach, which has confirm to have better efficiency in the area of MT systems. Now days, several governmental and private based MT fields use this hybrid-based technique to develop covert from source to target language, which is based on both rules and statistics. The hybrid technique can be used in different ways. In some cases, translations are performed in the first level using a rule-based technique followed by adjusting or correcting the output using statistical information. Other way, rules are used to pre-process the input data as well as post-process the statistical output of a statistical Machine translation system. Hybrid based technique is better than the previous and has more power, flexibility, and control in translation. Hybrid technique integrating more than one MT paradigm are receiving increasing attention. The METIS-II MT system is an example of hybridization about the EBMT Figure 3: Statistical Machine Translation Approach framework; it avoids the current need for parallel corpora by using a bilingual dictionary (similar to that found in 3.4. Example-based translation: most RBMT systems) and a monolingual corpus in the TL Basic idea of this MT is to reuse the examples of already (Dirix et al., 2005). An example of hybridization about the existing translations. An example-based translation is uses rule-based paradigm is given by Open. It integrates a bilingual corpus as its main knowledge base and it is statistical methods within an RBMT system to choose the essentially translation by analogy. Example-based machine best translation from a set of competing hypotheses translation (EBMT) is define by its use of bilingual corpus (translations) generated using rule-based methods. with parallel texts as its main knowledge, in which translation by analogy is the main idea. An EBMT system 3.7. Neural Machine Translation: is given a set of sentences in the source language (from Neural Machine Translation is an approach to MT that uses which one is translating) and corresponding translations of a neural network which directly models the conditional each sentence in the target language with point to point probability of translating a given source sentence to a target mapping. These examples are used to covert similar types sentence. of sentences of source language to the target language. Example acquisition, example base and management, 4. Conclusion: example application and synthesis this is four tasks in In this paper explain the various standardized approaches in Example based machine translation system. At the base of the field of Machine translation word wide and especially Volume 9, Issue 5, September - October 2020 Page 47
no reviews yet
Please Login to review.