Survey of machine translation systems in india

International Journal on Natural Language Computing (IJNLC) Vol. 2, No.4, October2013

SURVEY OF MACHINE TRANSLATION SYSTEMS IN
INDIA
G V Garje1 and G K Kharate2
1

Department of Computer Engineering and Information Technology PVG’s College of
Engineering and Technology, Pune, India
2

Principal, Matoshri College of Engineering and Research Centre, Nashik, India

ABSTRACT
The work in the area of machine translation has been going on for last few decades but the promising
translation work began in the early 1990s due to advanced research in Artificial Intelligence and
Computational Linguistics. India is a multilingual and multicultural country with over 1.25 billion
population and 22 constitutionally recognized languages which are written in 12 different scripts. This
necessitates the automated machine translation system for English to Indian languages and among Indian
languages so as to exchange the information amongst people in their local language. Many usable machine
translation systems have been developed and are under development in India and around the world. The
paper focuses on different approaches used in the development of Machine Translation Systems and also
briefly described some of the Machine Translation Systems along with their features, domains and
limitations.

KEYWORDS
Machine Translation, Example-based MT, Transfer-based MT, Interlingua-based MT

1. INTRODUCTION
India is a multilingual country where the spoken language changes after every 50 miles. There are
22 official languages; and approximately 2000 dialects are spoken by different communities in
India. English and Hindi are used for official work in most states of India. The state governments
in India predominantly carry out their official work in their respective regional language whereas
the official work of Union government is carried out in English and/or Hindi. All the official
documents and reports of Union government are published in English or Hindi or in both English
and Hindi. Many newspapers are also published in regional languages. Translating these
documents manually is very time consuming and costly. Hence there is need to develop good
machine translation (henceforth referred as MT) systems to address all these issues, in order to
establish a better communication between states and Union governments and exchange of
information amongst the people of different states with different regional languages. Indian
languages are divided into five language families; viz. Indo-Aryan (76.87% speakers), Dravidian
(20.82% speakers), Austro-Asiatic (1.11% speakers), Tibeto-Burman (1% speakers) and
Andmanese (less than 0.001% speakers). Many Indian languages being low resource languages
become a major hurdle in the development of MT systems for Indian languages. [1][37].
Many researchers, institutions and research organizations in India have started working on MT
systems for English to Indian languages and among Indian languages have succeeded in obtaining
DOI : 10.5121/ijnlc.2013.2504

47

International Journal on Natural Language Computing (IJNLC) Vol. 2, No.4, OCtober 2013

very satisfactory results. The Government of India has decided to give more thrust to Language
Technology for Indian languages during VIIIth plan and to initiate a programme that would
emphasize on quality, national relevance and participation of traditional knowledge and R & D
efforts in the area of information processing in Indian languages. The Department of Electronics
of Government of India launched a National level programme during the year 1990-91 on
Technology Development for Indian Languages (TDIL) [16]. Other institutions like IIT Kanpur,
IIT Bombay, IIIT Hyderabad, University of Hyderabad, NCST Mumbai, CDAC Pune, CDAC
Noida, Department of Computer Science and Engineering Jadavpur University, Kolkata, JNU
New Delhi etc are playing a major role in developing the MT systems in India. Many MT systems
have been developed and are being developed. The MT systems have been developed using
different machine translation approaches. This paper provides brief information about
development year, source & target language, translation approach, domain, salient features, and
translation accuracy of major Machine Translation systems in India [2]. There is an immense
need to translate these documents in respective state’s local language for proper communication
with common people of the state. More than 95% of the Indian population is deprived of the
benefits of Information Technology due to language barrier [16].
This paper is organized into 6 sections. Section 2 gives a brief history of MT system; section 3
gives an idea of the different approaches to build a MT system. Section 4 discusses major MT
systems in India based on translation approaches along with their features, translation quality,
domain etc. Section 5 describes the summary of literature review in brief for major MT systems.
Section 6 gives the comparison of major translation approaches.

2. BRIEF HISTORY OF MT
Table 1 shows the time line chart of Machine translation at International level
[3][23][24][25][27].
Table 1. Time Line Chart of MT
Period

Year
1949

1948
to
1960

1952
The
beginning

Activity
Warren Weaver proposed the first idea on the use of
computers in translation by adopting the term computer
translation.
The first symposium of MT was held at MIT under
leadership of Yehoshua Bar-Hillel.
The first basic automatic Russian-English translator was
developed by a group of researchers from Georgetown
University in collaboration with IBM that translated more
than sixty Russian sentences.

1954
Victor Yngve published the first journal on MT, entitled
Mechanical translation devoted to the translation of
languages by the aid of machines.
The computational linguistics was born due to weekly
lectures organized by David G. Hays at the Rand Corporation
in Los Angeles.
1960
to
1966

Parsing
and
disillusionment

1961
First International Conference on MT of Languages and
Applied Language Analysis of Teddington was held with the
participation of linguists and computer scientists.
The scientists involved in the translation work were Paul
Garvin, Sydney M. Lamb, Kenneth E. Harper, Charles
48


1964
1960
to
1966

Parsing
and
disillusionment

1966

Hockett, Martin Kay and Bernard Vauquois.
Creation of committee ALPAC (Automatic Language
Processing Advisory Committee) with American government
to study the perspectives and the chances of machine
translation.
ALPAC published its famous rapport in which it concluded
that its work on machine translation was just waste of time
and money. Conclusion of this rapport made a negative
impact on the MT research for number of years.
Start of the project REVERSO by a group of Russian
researchers.

1970

1966
to
1980

New birth
and
hope

1976

1978

1981

1980
to
1990

1983
Japanese invaders

Development of System SYSTRAN1 (Russian-English) by
Peter Toma, who was member of a group search for
Georgetown at that time.
Developed a MT system WEATHER in the project TAUM
(machine translation in the university of Montreal) under the
direction of Alai Colmerauer for the machine translation
weather-forecasts for the general public. This system was
created by group of researchers.
Developed a MT system ATLAS2 by the Japanese firm
FUJITSU. This translator was based on rules. It was able to
translate from Korean to Japanese and vice versa.
The Japanese firm SHARP developed Automatic translator
DUET (English - Japanese), which was based on rules and
transfer approach.
NEC developed a system based on algorithm called PIVOT
named as Honyaku Adaptor II, used for Interlingua approach.
OKI3 Developed a Japanese-English system PENSEE, which
was rule based translator.

1986

1990
to
2000

Web & new
vague of
translators

1993
1998

2000
2000
to
2010

2005

2007
2008
2000
to
2010

2009
2010

Hitachi developed a translation Japanese-English system
HICATS (Hitachi Computer Aided Translation System).
The project C-STAR (Consortium for Speech Translation
Advanced Research) was initiated. It was trilingual (English,
German & Japanese) MT system for the parole in the field of
tourism (dialogue client travel agent)
Marketing of machine translator REVERSO was done by
Softissimo.
Japanese Laboratory ATR developed a (Japanese-English &
Chinese - English) system ALPH. This system used Example
based approach of MT.
The first web site for automatic translation by Google was
launched
A hybrid MT METIS-II was developed that used the SMT,
EBMT, and RBMT machine translation approaches.
23% of Internet users used the MT and 40 % considering
doing so.
30% the professionals have used the MT and 18% perform a
proofreading.
28% of Internet users used the MT and 50% planned to do
so.

49


3. MACHINE TRANSLATION APPROACHES
Many MT systems across the globe have already been developed for the most commonly used
natural languages such as English, Russian, Japanese, Chinese, Spanish, Hindi and other Indian
languages etc. Figure 1 depicts the existing machine translation systems and various approaches
used in developing these systems.

Figure 1. Machine Translation Systems

4. INDIAN MACHINE TRANSLATION SYSTEMS
4.1. Direct Machine Translation Systems
4.1.1. Anusaaraka systems among Indian Languages (1995)
Anusaaraka project started at IIT Kanpur by Rajeev Sangal is now being continued at IIIT
Hyderabad. The purpose of the project was the MT of one Indian language to another Indian
language. The project is being funded by Technology Development for Indian Languages (TDIL),
Ministry of Information Technology, Government of India and Satyam Computers Private
Limited. The source languages are {Telugu, Kannada, Bengali, Punjabi and Marathi} and the
target language is Hindi. It is not domain specific but the system has been tested mainly for
translating children’s’ stories. The system was mainly developed for the purpose of perfect
“information preservation”. The output of the system followed the grammar of the source
language only.
For Example, a Bengali to Hindi translation can take a Bengali text and produce output in Hindi
that can be understood by the user but may not be grammatically perfect. For 80% of the Kannada
words in the Anusaaraka dictionary [26], 1997) of 30,000 root words, there is a single equivalent
Hindi word which covers the sense of the original Kannada word.
The focus of Anusaaraka was not mainly on MT, but it was on language access between Indian
languages. It is currently attempting an English-Hindi machine translation. It uses a Paninian
Grammar (PG) and exploits the close similarity of Indian languages [26][27].

4.1.2. Punjabi to Hindi MT System (2007, 2008)
G S Josan and G S Lehal have developed a system which is based on direct word-to-word MT
approach. This system comprised of modules such as pre-processing, word-to-word translation
using Punjabi-Hindi lexicon, morphological analysis, word sense disambiguation, transliteration
50


and post processing. Accuracy of the translation produced by this system is 90.67%. Word Error
Rate is 2.34% and SER is 24.26% [19][38].

4.1.3. Web based Hindi-to-Punjabi MT System (2010)
Goyal V and Lehal G S developed the extended version of Hindi-to-Punjabi MT System to Web.
The system has several facilities like website translation, email translation, etc. [6].

4.1.4 Hindi-to-Punjabi MT System (2009, 2011)
Goyal V and Lehal G S developed a system that uses direct word to word translation approach at
Punjabi University, Patiala. The translation accuracy of the system is 95.40% on the basis of
intelligibility test and 87.60% on the basis of accuracy test. In the quantitative tests the Word
Error Rate is 4.58% whereas Sentence Error Rate is 28.82% and BLUE score found is
0.7801[1][10][21][32].

4.2. Transfer-Based MT Systems
4.2.1. Mantra MT (1997)
Mantra is English to Hindi MT system developed by Bharati for information preservation. The
text available in one Indian language is made accessible in another Indian language with the help
of this system. It uses XTAG based super tagger and light dependency analyzer developed at
University of Pennsylvania for performing the analysis of the input English text. It distributes the
load on man and machine in novel ways. The system produces several outputs corresponding to a
given input. Output based on the most detailed analysis of the English input text, uses a full parser
and a bilingual dictionary. The parsing system is based on XTAG (Bandyopadhyay 2002,
consisting of super tagger and parser) with minor modification for the task at hand. A user may
read the output produced after the full analysis, but when she finds that the system has
“obviously” gone wrong or failed to produce the output, she can always switch to a simpler
output [8].

4.2.2. MANTRA MT(1999)
Hemant Darbari and Mahendra Kumar Pandey developed a MAchiNe assisted TRAnslation tool
(MANTRA). It translates English text into Hindi in a specific domain of personal administration
that includes gazette notifications, office orders, office memorandums and circulars. The system
was tested for the translation of administrative documents such as appointment letters,
notification and circular issued in central government from English to Hindi in initial stage. It
uses the Tree Adjoining Grammar (TAG) formalism developed by University of Pennsylvania
and Lexicalized Tree Adjoining Grammar (LTAG) (Bandyopadhyay, 2004) to represent the
English and Hindi grammar. It is based on synchronous Tree Adjoining Grammar and uses tree
transfer for translating from English to Hindi. It has become a part of “The 1999 Innovation
Collection” on information technology at Smithsonian institution’s National museum of
American history, Washington DC, USA. The project was funded by the Rajya Sabha Secretariat.
The grammar of the system was specifically designed to accept, analyze and generate sentential
constructions in “Official” domain and the lexicon was restricted to deal with meanings of
English words that are used in its subject domain.
The system is developed for the Rajya Sabha Secretariat, the Upper House of Parliament of India
and used to translate the proceedings of parliament such as study to be laid on the Table, Bulletin
Part-I and Part-II. This system can also be used for other language pairs such as English- Bengali,
51


English-Telugu, English-Gujarati and Hindi-English and also among Indian languages such as
Hindi-Bengali and Hindi-Marathi. The Mantra approach is general, but the dictionary (lexicon)
and grammar rules developed were limited to the sub-language of the domain [8][28].

4.2.3. An English–Hindi Translation System (2002)
Gore L and Patil N developed a system based on transfer based translation approach, which uses
different grammatical rules of source and target languages and a bilingual dictionary for
translation. The translation module consists of pre-processing, English tree generator, postprocessing of English tree, generation of Hindi tree, Post-processing of Hindi tree and generating
output. The domain of the system was weather narration [14].

4.2.4. MAT (2002)
Murthy K developed a machine assisted translation system for translating English texts into
Kannada, which used morphological analyzer and generator for Kannada. The English input
sentence is parsed by Universal Clause Structure Grammar (UCSG) parser and outputs the
number, type and inter-relationships amongst various clauses in the sentence and the word
groups. For each word, suitable target language equivalence is obtained from the bilingual
dictionary. Finally, the target language sentence is generated by placing the clauses and the word
groups in appropriate linear order, according to the target language grammar. Post editing tool is
provided for editing the translated text. MAT System 1.0 had shown about 40-60% of fully
automatic accuracy. The domain of the translation system was government circulars [13].

4.2.5. Shakti (2003)
Bharati, R Moona, P Reddy, B Sankar, D M Sharma and R Sangal have developed a system
which translates English to any Indian languages with simple system architecture. It combines
linguistic rule-based approach with statistical approach. The system consists of large number of
modules (69 modules). Nine modules are used for analyzing the source language (English), 24
modules are used for performing bilingual tasks, and the remaining modules are used for
generating target Indian language [16].

4.2.6. English-Telugu MT System (2004)
Bandyopadhyay S developed a system that uses English - Telugu dictionary containing 42,000
words. A word form synthesizer for Telugu is developed and incorporated in the system. The
system handles various complex English sentences [30].

4.2.7. Telugu-Tamil MT System (2004)
Bandyopadhyay S developed a system that uses the Telugu Morphological analyzer and Tamil
generator for translation. The system makes use of Telugu-Tamil dictionary developed as a part
of MAT Lexica. It also uses verb sense disambiguation based on verbs argument structure to take
care of ambiguity in the meaning of the verb [30].

4.2.8. OMTrans(2004)
Mohanty S, Balabantaray R C developed a system that translates text from English to Oriya based
on grammar and semantics of the source and target language. Word Sense Disambiguation
52


(WSD) is also handled in this system. OMTrans is designed and developed using principles of
object-oriented approach [17][39].

4.2.9. The MaTra System (2004, 2006)
Ananthakrishnan R, Kavitha M, Hegde J J, Chandra Shekhar, Ritesh Shah, Sawani Bade, and
Sasikumar M have developed a system that uses transfer-based approach using a frame-like
structured representation. It also uses heuristics to resolve ambiguities. The domain of the system
is news, annual reports and technical phrases. It has a text categorization component which
determines the type of news story (political, terrorism, economic, etc.) before operating on the
given story. It has different dictionaries for different domains like political, terrorism etc. and
chooses an appropriate dictionary depending on the type of news. It requires considerable human
assistance in analyzing the input. Another novel component is sentence splitter which breaks a
complex English sentence into simpler sentences. These simple sentences are further analyzed
and used to generate Hindi sentences [18][32][33].

4.2.10. English-Kannada machine-aided translation system (2009)
K Narayana Murthy has developed a system English-Kannada Machine Assisted Translation at
Resource Centre for Indian Language Technology Solutions (RC-ILTS), University of
Hyderabad. The system uses a transfer-based approach and it has been applied to the domain of
government circulars. English-Kannada machine translation system uses Universal Clause
Structure Grammar (UCSG) formalism. The system is funded by the Karnataka government
[4][5].

4.2.11. Tamil-Hindi Machine-Aided Translation system (2009)
Sobha L, Pralayankar P and Kavitha V developed a system which is based on Anusaaraka (started
in 1984). MT system architecture is developed by Prof. C N Krishnan. It uses a lexical-level
translation and has 80-85% coverage. Both stand-alone and web-based on-line versions have been
developed. Tamil morphological analyzer and Tamil-Hindi bilingual dictionary (36K) are the biproducts of this system. It performs exhaustive syntactical analysis. They have also developed a
prototype of English-Tamil MAT system. Currently, it has limited vocabulary (100-150
sentences) and small set of Transfer rules [1][4][31].

4.2.12. Sampark System: Automated Translation among Indian Languages (2009)
A consortium of 11 institutions in India have developed a multipart machine translation system
for Indian Language to India Language Machine Translation (ILMT) funded by TDIL program
of Department of Electronics and Information Technology (DeitY), Govt. of India. It uses
Computational Paninian Grammar (CPG) for analyzing language and combines it with machine
learning. It is developed using both traditional rules-based and dictionary-based algorithms with
statistical machine learning. This consortium has developed language technology for 9 Indian
languages resulting in Machine Translation for 18 Indian language pairs [25].

4.3. Interlingua Machine Translation Systems
4.3.1. ANGLABHARTI (2001)
R M K Sinha, Jain R, Jain A developed a machine aided translation system designed for
translating English to Indian languages. It is developed using pseudo-interlingua approach. The
53


interlingua approach made it possible to use the same system for translating English to more than
one Indian language and has eliminated the need of developing separate translation system for
English to each Indian language. The analysis of English as a source language is done only once
and it creates intermediate structure – PLIL (Pseudo Lingua for Indian Languages). The PLIL is
then converted to each Indian language through a process of text-generation. The effort for PLIL
generation is 70% and text generation is 30%. Only with an additional 30% effort, new English to
Indian language translation system can be built. The attempt has been made whereby has to do
90% translation task and remaining 10% is left for the human post-editing. The domain of this
machine translation system has been public health [4].

4.3.2. UNL-based English-Hindi MT System (2001)
Dave S, Parikh J and Bhattacharyya P developed a translation system using Universal Networking
Language (UNL) as the Interlingua structure. The UNL is an international project aimed to create
an Interlingua for all major human languages. IIT Mumbai is the Indian participant in UNL
project. English-Hindi, Hindi-UNL, UNL-Hindi, English-Marathi and English-Bengali were also
developed using UNL formalism [2][7][40][41].

4.3.3. AnglaHindi (2003)
AnglaHindi is a derivative of AnglaBharti MT System developed by R M K Sinha and Jain A for
English to Indian languages, which is a pseudo interlingual rule-based English to Hindi MachineAided Translation System. It uses all the modules of AnglaBharti and also uses abstracted
example-base for translating frequently encountered noun phrases and verb phrases. The accuracy
of the translation is 90% [12].

4.4. Hybrid Machine Translation Systems
4.4.1. Anubharti Technology (1995, 2004)
Anubharti (Sinha, 2004) is developed using a hybridized example-based machine translation
approach i.e. a combination of example-based, corpus-based approaches and some elementary
grammatical analysis. The example-based approaches follow human-learning process for storing
knowledge from past experiences and to be used it in future. In Anubharti, the traditional EBMT
(Gupta and Chatterjee, 2003) approach has been modified to reduce the requirement of a large
example-base. The modification in traditional EBMT is achieved by generalizing the constituents
and replacing them with abstracted form from the raw examples. The abstraction is achieved by
identifying the syntactic groups. Matching of the input sentence with abstracted examples is done
based on the syntactic category and semantic tags of the source language structure. The
architectures of both AnglaBharti and AnuBharti, have undergone a considerable change from
their initial conceptualization. In 2004 these systems were named as AnglaBharti-II and
AnuBharti-II respectively. AnglaBharti-II uses a generalized example-base for hybridization
besides a raw example-base and the AnuBharti-II makes use of Hindi as source language for
translation to any other language. The generalization of the example-base is dependent upon the
target language [2][37].

4.4.2. ANUBHARTI-II (2004),
R M K Sinha developed a MT system using Generalized Example-Base (GEB) along with Raw
Example-Base (REB) MT approach for hybridization. The combination of example-based
approach and traditional rule-based approach is used in this system. The example based approach
emulates human-learning process for storing knowledge from past experiences and to be used in
54


future. The source language is Hindi. The inputted Hindi sentence is converted into a standard
form to handle the word-order variations. The Hindi sentences converted into standard form are
matched with a top level standard form of example-base. If no match is found then a shallow
chunker is used to fragment the input sentence into small units and then they are matched with a
hierarchical example-base. The small chunks obtained by shallow chunker are translated and
positioned by matching with sentence level example base [2][37].

4.4.3. Bengali to Hindi MT System (2009)
Chatterji S, Roy D, Sarkar S and Basu A developed a hybrid Machine Translation system. It uses
an integration of SMT with a lexical transfer based system (RBMT) i.e. multi-engine Machine
Translation approach. The experimentation shows that BLEU scores of SMT and lexical transfer
based system when evaluated separately are 0.1745 and .0424 respectively. The performance of
hybrid system is better and its BLEU score is 0.2275 [20].

4.4.4. Lattice Based Lexical Transfer in Bengali Hindi MT Framework (2011),
Sanjay Chatterji, Praveen Sonare, Sudeshna Sarkar, and Anupam Basu described a method for
making proper lexical translation in Bengali to Hindi Machine Translation Framework and used a
transfer based MT approach. In the baseline system, Bengali word is replaced by the most
frequent word in Hindi. However, the most frequent translation may not be correct if we consider
the context of the word(s). The proposed method finds a better lexical choice amongst the
dictionary options with the help of the contextual information of a monolingual corpus of Hindi.
The system takes Bengali sentence and converts it to Hindi sentence with the help of lattice-based
data structure. The baseline system used for comparison and proposed translation systems are
evaluated using the BLEU automatic evaluation tool and human evaluation process. It is observed
that the proposed system performs better. Training corpus size is 500K Hindi corpus and is tested
on Hindi to Bengali [34].

4.5. Example Based Machine Translation (EBMT) Systems
4.5.1. ANUBAAD (2000, 2004)
Bandyopadhyay S developed a MT system which translates news headlines from English to
Bengali using example based Machine Translation approach. An English news headline given to
the system as an input is initially searched in the direct example-base for an exact match. If a
match is found, the Bengali headline from the example-base is produced as output. If match is not
found, the headline is tagged and the tagged headline is searched in the Generalized Tagged
example-base. If a match is found in Generalized Tagged Example-Base, the Bengali headline is
to be generated after appropriate synthesis. If a match is not found, the Phrasal example-base will
be used to generate the target translation. If the headline still cannot be translated, the heuristic
translation strategy is applied where translation of the individual words or terms in their order of
appearance in the input headline will be generated. Appropriate dictionaries have been consulted
for translation of the news headlines [11].

4.5.2. VAASAANUBAADA (2002)
Vijayanand K, Choudhury S I and Ratna P developed an Automatic Machine Translation system
for Bengali-Assamese News Texts using Example Based Machine Translation (EBMT) approach.
It involves Bengali-Assamese sentence level Machine Translation for Bengali text. It includes
preprocessing and post-processing tasks. The bilingual corpus has been constructed and aligned
55


manually by feeding the real examples using pseudo code. Longer sentences are fragmented at
punctuations to obtain better quality translation. When the exact match is not found at
sentence/fragment level in Example-Base, the backtracking is used and further fragmentation of
the sentence is done[15].

4.5.3. Shiva and Shakti MT System (2003)
MT system ‘Shiva’ is designed using an Example-based and the system Shakti is designed using
combination of rule based and statistical approaches. The Shakti system is working for three
target languages like Hindi, Marathi and Telgu and can produce machine translation systems for
new languages rapidly. Shiva and Shakti are the two Machine Translation systems from English
to Hindi developed jointly by Carneige Mellon University of USA, International Institute of
Information Technology, Hyderabad and Indian Institute of Science, Bangalore, India. The
system is used for translating English sentences into an appropriate target Indian language. The
rules used for target language generation are mostly linguistic in nature and the statistical
approach tries to infer or use linguistic information. Semantic information is also used by some
modules in the system. Currently the system is working for three languages (Hindi, Marathi and
Telugu) [2][4].

4.5.4. ANGLABHARTI-II (2004)
R M K Sinha suggested a generalized example-base (GEB) approach for hybridization besides a
Raw Example-Base (REB). It is found that the modification in the rule-base system is difficult
during development phase and may result in unpredictable results; the example-base approach is
grown interactively by augmenting rule-base base. The system first attempts a match in REB and
GEB before invoking the rule-base at the time of actual use. It also provides provisions for
automated pre-editing and paraphrasing, generalized and conditional multi-word expressions as
well as recognition of named-entities. It also contains the modules for an error-analysis and
statistical language-model for automated post-editing. The automatic pre-editing module is used
to transform/paraphrase the input sentence to a form which can be easily translated. Automatic
pre-editing may even fragment an input sentence if the fragments are easily translatable and
positioned in the final translation. The system also contains a ‘failure analysis’ module. The
failure analysis module consists of heuristics on speculating the reasons for wrong translation.
The system includes various sub-modules [2][4].

4.5.5. Hinglish MT System (2004)
Sinha and Thakur developed Hinglish - a machine translation system for pure Hindi to pure
English forms. It incorporates additional level to the existing English to Hindi translation
(AnglaBharti-II) and Hindi to English translation (AnuBharti-II) systems developed by Sinha.
The system has produced satisfactory acceptable results in more than 90% of the cases. The
system is not capable of resolving the meaning of polysemous verbs due to a very shallow
grammatical analysis used in the process [29].

4.5.6. English to {Hindi, Kannada, Tamil} and Kannada to Tamil Language-Pair
Example Based MT (2006)
Balajapally P, P Pydimarri, M Ganapathiraju, N Balakrishnan and R Reedy developed a MT
system based on a bilingual dictionary comprising of sentence dictionary, phrases dictionary,
words dictionary and phonetic dictionary. Each of the dictionaries contains parallel corpus of
sentences, phrases, words and phonetic mappings of words in their respective files. Example-Base
has a set of 75000 most commonly spoken sentences that are originally available in English. All
56


the sentences in Example-Base have been manually translated into three target Indian languages,
namely Hindi, Kannada and Tamil [2][4].

4.5.7. The MATREX (MT using Example) System (2008)
Ankit Kumar Srivastava, Rejwanul Haque, Sudip Kumar Naskar and Andy Way developed a
DCU Machine Translation System for ICON 2008. The MATREX system makes use of markerbased chunking, which is based on the Marker Hypothesis (Green, 1979), a psycholinguistic
constraint which posits that all languages are marked for surface syntax by a specific closed set of
lexemes or morphemes which signify context. Using a set of closed-class (or “marker”) words,
such as determiners, conjunctions, prepositions, possessive and personal pronouns, aligned
source-target sentences are segmented into chunks (Gough and Way, 2004) during a preprocessing step. A chunk is created at each new occurrence of a marker word in such a way that
each chunk must contain at least one content (or non-marker) word. In order to align the chunks
obtained by the chunking procedures, the system makes use of an “edit-distance style” dynamic
programming alignment algorithm [22][36].

4.6. Statistical Machine Translation Systems
4.6.1. Shakti (2003)
Bharati, R Moona, P Reddy, B Sankar, D M Sharma and R Sangal developed a MT system
which translates English text to any Indian language with simple system architecture. It combines
linguistic rule based approach with statistical approach. The system contains 69 different
modules. Nine modules are used for analyzing the source language (English), 24 modules are
used for performing bilingual tasks, and the remaining modules are used for generating target
Indian language [2][4].

4.6.2. English to Indian Languages MT System (E-ILMT) (2006)
The EILMT is a MT System for English to Indian Languages in Tourism and Healthcare
Domains. It is developed by a Consortium of Nine institutions namely C-DAC Mumbai, IISc
Bangalore, IIIT Hyderabad, C-DAC Pune, IIT Mumbai, Jadavpur University Kolkata, IIIT
Allahabad, Utkal University Bangalore, Amrita University Coimbatore and Banasthali
Vidyapeeth, Banasthali. The project is funded by Department of Information Technology, MCIT
Government of India. The role of C-DAC Mumbai is to develop statistical models and resources
for a statistical MT (SMT) system from English to Hindi/Marathi/Bengali. The engine was
initially developed as a baseline system using the state-of-art statistical techniques and the
contemporary tools that include the POS tagger (fnTBL), parser (Bikel), decoder (Pharaoh) etc.
The primary objective is to initially build an English-Hindi translation system capable of
translation of free flow text as found on the web and gradually adapt it to other Indian language
pairs as well.
The training corpus (translation model) consisted of 5000 sentences and 800 sentences were split
for testing and tuning. The baseline techniques used in this system were inadequate in producing a
good quality translation. Therefore, pre-processing stage was included in the system which takes
care of syntactic re-ordering on the source language to reduce long distance movements through
SMT. It has helped to obtain a better phrase alignment table which resulted in a good
improvement in the translation quality using Moses decoder with Giza++ alignment tool. The
corpus (translation model) training size for achieving this effort was 12299 sentences with
additional 1570 sentences split for testing and tuning.
57


Some degradation in the output even after the syntactic processing was observed due to
unavailability of sufficient corpus. The syntactically processed corpus was morphologically
processed and used for training to counteract the problem of degradation in translation quality. A
rule based suffix separation approach was used to separate the root word and the affixes due to
the unavailability of sophisticated morphological analyzers. The system is extended and tested for
English-Marathi and English-Bengali pairs with the statistics shown in Table 2 [35].
Table 2. Performance of E-ILMT
Language Pair

Training size

Testing + Tuning size

English-Marathi

13598

1500 (750+750)

English-Bengali

13015

1550

5. SUMMARY OF LITERATURE REVIEW FOR MAJOR INDIAN MT SYSTEMS
5.1. Direct Machine Translation Approach
The following table describes the summary of MT systems developed using Direct MT
approach and their salient features and/or limitations
Table 3. Summary of Direct MT Approach
SN
01

02

03

04

MT System name

Languages

Domain/main
application
Domain free
but the system
has been
applied mainly
for translating
children’s
stories

Anusaaraka systems
among Indian
languages (1995)

{Telugu,
Kannada,
Bengali,
Punjabi and
Marathi}
to
Hindi

Punjabi to Hindi
Machine Translation
System (2007, 2008)

Punjabi
to
Hindi

General

Web based Hindi-toPunjabi Machine
Translation
System(2010)
Hindi-to-Punjabi
Machine Translation
System (2009, 2011)

Hindi
to
Punjabi

Web
email

Hindi
to
Punjabi

General

Salient Features/Limitations
-Translation quality is very coarse
-The focus is not mainly on MT but
on language access between Indian
languages.
-It is currently attempting an
English-Hindi MT

-Requires post processing
-Accuracy 90.7%

pages,

-Translation quality is very coarse ,
extension of Punjabi to Hindi MT
-Web-based
-Accuracy 95.4%
-BLEU score 0.7804
-Word Error Rate is 4.58% and
Sentence Error Rate is 28.82%

58


5.2. Transfer-Based Machine Translation Approach
The following table describes the summary of MT systems developed using TransferBased MT approach and their salient features and/or limitations
Table 4. Summary of Transfer-Based MT Approach
SN

MT System name

Languages

01

MANTRA
System(1997)

English
and
Hindi

MANTRARajayasabha (1999)

English
and
Hindi

Domain/main
application
Office
administration
documents

- Uses a TAG and LTAG
- Distributes a load on man and
machine in a novel way

Proceeding of - Uses a TAG, Synchronous TAG
and Tree Transfer
Rajyasabaha
- The system produces several
outputs corresponding to a given
input
- Output based on the most detailed
analysis of the English input text
uses a full parser and bilingual
dictionary

02

03

04

05

An English–Hindi
Translation
System(2002)

English
And
Hindi

Weather
narration

MAT System
(2002)

English
and
Kannada

General

Shakti (2003)

English
and
IL

General

English-Telugu
Machine
Translation System
(2004)

English
and
Telugu

General

-The Mantra approach is general but
the lexicon/grammar has been
limited to sub-language of the
domain
-Rule-based
-The translation module consists of
Pre-processing, English tree
generator, post-processing of
English tree, generation of Hindi
tree, Post-processing of Hindi tree
and generating output
-UCSG(Universal Clause Structure
Grammar), morphological analyzer
& post-editing
-MAT System 1.0 has shown about
40-60% fully automatic accuracy
-The system is tested for
government circulars
-Linguistic rule-base with
Statistical processing
-The system consists of 69 different
modules, of which 9 modules are
used for analyzing the source
language, 24 modules are used for
performing bilingual tasks, and the
remaining modules are used for
generating target Indian language.
-Rule-based
-Lexicon consists of 42,000 words
-Handles various complex English
sentences
59


5.3. Interlingua-Based Machine Translation Approach
The following table describes the summary of MT systems developed using InterlinguaBased MT approach and their salient features and/or limitations
Table 5. Summary of Interlingua-based MT Approach
SN

MT System name L

01

ANGLABHARTI
(2001)

English to IL

UNL-based
English-Hindi MT
System (2001)

English to
Hindi, UNL to
Hindi, Hindi
to UNL

General

AnglaHindi (2003)

English
to
Hindi

General

02

03

Languages

Domain/main
application
Public health

-Uses pseudo-interlingua
-Possible to use the same system for
translating English to more than
one Indian language due to
interlingua approach
-No need to develop separate
translation system for English to
each Indian language
English (source) language creates
intermediate structure – PLIL
(Pseudo Lingua for Indian
Languages)
-The effort for PLIL generation is
70% and text generation is 30%
-Only with an additional 30% effort,
new English to Indian language
translation system can be built
90% translation task is done by
machine and 10% left to the human
post-editing.
-Universal Networking Language
(UNL) as Interlingua
-Developed for English-Hindi,
Hindi-UNL, UNL-Hindi, EnglishMarathi and English-Bengali
-Easy to add new language for
translation
-UNL is an international project
with an aim to create inerlingua for
all major human languages
-Pseudo interlingual rule-based
-Uses all the modules of
AnglaBharti
-Makes use of an abstracted
example-base for translating
frequently encountered noun
phrases and verb phrases
-The translation accuracy is 90%.

5.4. Statistical Machine Translation Approach
The following table describes the summary of MT systems developed using Statistical
MT approach and their salient features and/or limitations

60

Table 6. Summary of Statistical MT Approach
SN
01

MT System name

Languages

English to Indian
Languages
Machine
Translation System
(E-ILMT 2006)

English
to
Indian
Languages

Domain/main
application
Tourism
and
Healthcare

-Rule-based and Statistical
-The engine was developed using
the statistical techniques and tools
which includes the POS tagger
(fnTBL), parser (Bikel), decoder
(Pharaoh) etc
-The objective was to build an
English-Hindi translation system
capable of translation of free flow
text as found on the web and
gradually adapt it to other Indian
language pairs
-The training corpus consisted of
5000 sentences and 800 sentences
were split for testing and tuning
-Pre-processing phase was included
to take care of syntactic re-ordering
on the source language to reduce
long distance movements through
SMT
-The syntactically processed corpus
was morphologically processed and
used for training to tackle the
problem of degradation in
translation quality.
-A rule based suffix separation
approach was used to separate the
root word and the affixes

5.5. Example-Based Machine Translation Approach
The following table describes the summary of MT systems developed using ExampleBased MT approach and their salient features and/or limitations
Table 7. Summary of Example-based MT Approach
SN
01

MT System name

Languages

ANUBAAD
(2000, 2004)

English
to
Bengali

Domain/main
application
News Headlines

-Example-base, Generalized Tagged
examplebase
and
Phrasal
example-base
are
separately
maintained
- Bengali headline is generated after
appropriate synthesis if the headline
is found in Generalized Tagged
example-base
61


02

03

04

05

06

VAASAANUBAA
DA (2002)

Bengali
to
Assamese

News Text

Shiva and Shakti
machine
translation system
(2003)

English
to
{Hindi,
Telugu,
Marathi}

General

ANGLABHARTIII (2004)

English
to
Indian
languages

General

Hinglish machine
translation system
(2004)

Hindi
to
English

General

English to {Hindi,
Kannada, Tamil}
and Kannada to
Tamil LanguagePair Example
Based Machine
Translation (2006)

English
to
{Hindi,
Kannada,
Tamil} and
Kannada
to
Tamil

General

If the headline cannot be translated
using Example-base, Generalized
Tagged example-base or Phrasal
example-base then the heuristic
translation strategy is used
-It includes pre-processing and
post-processing tasks
-The
bilingual
corpus
is
constructed and aligned manually
-Longer sentences are fragmented
at punctuations to obtain better
quality translation
-Shiva is designed using an
Example-based and Shakti is
designed using combination of rule
based and statistical approaches
-Easy to extend this system for new
target language
-The rules used for target language
generation are mostly linguistics in
nature
-Semantic information is also used
by some modules in the system.
-Currently system is working for
three languages
-Uses a generalized example-base
(GEB) approach along with a Raw
Example-Base (REB)
-Provides provisions for automated
pre-editing
&
paraphrasing,
generalized and conditional multiword expressions as well as
recognition of named-entities
-Contains the modules for an erroranalysis, statistical language-model
for automated post-editing and
failure analysis module
-Based on AnubBarti-II and
AnglaBharti-II
-Produced satisfactory acceptable
results in more than 90% of the
cases
-Performs
very
shallow
grammatical analysis
-The system is not capable of
resolving
the
meaning
of
polysemous verbs
-Maintains a bilingual sentence
dictionary, phrases dictionary,
words dictionary and phonetic
dictionary
-Example-Base of 75000 most
commonly spoken sentences that
are originally available in English
-Sentences in the Example-Base
have been manually translated into
three Indian languages namely
62


07

The MATREX
(Machine
Translation using
Example) System
(2008) The DCU
Machine
Translation System
for ICON 2008

English
to
Hindi

Hindi, Kannada and Tamil.
-Makes use of marker-based
chunking, which is based on the
Marker
Hypothesis,
a
psycholinguistic constraint which
signifies context
-A set of “marker” words, such as
determiners,
conjunctions,
prepositions,
possessive
and
personal pronouns is used to split
sentences into chunks during preprocessing
-In order to align the chunks the
system makes use of an “editdistance
style”
dynamic
programming alignment algorithm

Conference
papers

5.6. Hybrid Machine Translation Approach
The following table describes the summary of MT systems developed using Hybrid MT
approach and their salient features and/or limitations
Table 8. Summary of Hybrid MT Approach
SN

MT System name

Languages

01

Anubharti
Technology (1995)

Hindi to IL

Domain/main
application
General

- Hybrid Example-based
-

Combination of example-based,
corpus-based approaches and some
elementary grammatical analysis

- Reduced the requirement of a large
example-base

02

03

ANUBHARTI-II
(2004)

Hindi to IL

General

Bengali to Hindi
Machine
Translation System
(2009)

Bengali to
Hindi

General

-The generalization of the examplebase is dependent on the target
language.
-Generalized
Example-Base(GEB)
along with Raw Example-Base(REB)
-Emulates human-learning process for
storing
knowledge
from
past
experiences to use it in future
-Shallow chunker is used to fragment
the input sentence into small units and
then they are matched with a
hierarchical example-base
-Multi-engine Machine Translation
approach
-Uses an integration of SMT with a
lexical transfer based system (RBMT)
-The BLEU scores of SMT and lexical
transfer based system when evaluated
separately are 0.1745 and .0424
respectively
-The BLEU score of hybrid system is
63


04

Lattice Based
Lexical Transfer in
Bengali Hindi
Machine
Translation
Framework

Bengali
to
Hindi

better and it is 0.2275
-Lattice based integrated with transfer
based
-The lattice based lexical translation
system has been integrated with
transfer based
-Uses a lattice-based data structure i.e.
a weighted directed acyclic graph with
one start node and one end node.
-BLEU score of proposed system is
better than baseline system
-It is tested for 500k Hindi sentences

General

6. COMPARISON OF MT APPROACHES
The following table shows the comparison of major MT approaches
Table 9. Comparison of MT Approaches
MT approach
Rule-based

Advantages

1.
2.
3.
4.
5.

Knowledge-based

1.
2.
3.

Easy to build an initial system
Based on linguistic theories
Effective for core phenomena
Better choice for domain specific
translation
The quality of translation is good
for domain specific systems
Based on taxonomy of knowledge
Contains an inference engine
Interlingual representation

Disadvantages

1. Rules are formulated by experts
2. Difficult to maintain and extend
3. Ineffective for managerial
phenomena

4. The number of rules will grow
drastically in case of general
translation systems

1. Hard to build knowledge hierarchy
2. Hard to define granularity of
3.

Example-based

1.
2.
3.

Statistical

1.
2.
3.
4.

Extracts knowledge from corpus
Based on translation patterns in
corpus
Reduces the human cost
Does not consider language
grammar for translation
Extracts knowledge from corpus
Reduces the human errors
Model is mathematically
grounded

1.
2.
3.
1.
2.
3.
4.
5.

knowledge
Hard to represent knowledge
Similarity measure is sensitive to
system
Search cost is more
Knowledge acquisition problem
still persists
No linguistic background
Search cost is expensive
Hard to capture long distance
phenomena
Require huge amount of parallel
corpora
The translation quality will be
very coarse due to lack of
sufficient corpora

7. CONCLUSIONS
This paper described MT techniques in a longitudinal and latitudinal way with an emphasis on the
MT development for Indian languages as well as non-Indian languages. From the study, we found
that almost all existing Indian language MT systems are based on rule-based, hybrid and
statistical approaches. We identified the following reasons to justify as to why most of the
64


developed MT systems for Indian languages have followed the rule-based, hybrid and statistical
approach.
•

•

•
•
•

•

Most of the Indian and non-Indian MT systems developed so far are developed for specific
domains such as tourism, health care, children stories, medical, news headlines, technical
documents, government circulars and notifications etc. They have used rule based approach
as it provides better performance and accuracy if the set of rules is under control.
The Indian languages are morphologically rich in features and agglutinative in nature, hence
rule-based approaches may fail in situations where full-fledged general purpose MT systems
are to be developed because the number of rules would be very high.
Support of linguistic experts is essential for developing rule-based MT systems hence many
researchers are now working on statistical and hybrid approaches.
Statistical and Example-based MT systems require huge bilingual parallel corpora but even
monolingual corpus in not available for many Indian languages including Marathi.
Not much of the work is done for developing English to Marathi translation system. The only
MT system available for English to Marathi is the β-version of Anuvadak developed and
made available by TDIL (Technology Development for Indian Languages), New Delhi.
Most of the Indian MT systems requires post editing for producing better results

REFERENCES
[1]

[2]

[3]

[4]
[5]
[6]

[7]
[8]
[9]
[10]

[11]
[12]
[13]
[14]

Sitender & Seema Bawa, (2012) “Survey of Indian Machine Translation Systems”, International
Journal Computer Science and Technolgy, Vol. 3, Issue 1, pp. 286-290, ISSN : 0976-8491
(Online) | ISSN : 2229-4333 (Print)
Sanjay Kumar Dwivedi & Pramod Premdas Sukhadeve, (2010) “Machine Translation System in
Indian Perspectives”, Journal of Computer Science 6 (10): 1082-1087, ISSN 1549-3636, © 2010
Science
John Hutchins, (2005) “Current commercial machine translation systems and computer-based
translation tools: system types and their uses”, International Journal of Translation vol.17, no.1-2,
pp.5-38.
Vishal Goyal & Gurpreet Singh Lehal, (2009) “Advances in Machine Translation Systems”,
National Open Access Journal, Volume 9, ISSN 1930-2940 http://www.languageinindia.
Latha R. Nair & David Peter S., (2012) “Machine Translation Systems for Indian Languages”,
International Journal of Computer Applications (0975 – 8887) Volume 39– No.1
Vishal Goyal & Gurpreet Singh Lehal, (2010) “Web Based Hindi to Punjabi Machine Translation
System”, International Journal of Emerging Technologies in Web Intelligence, Vol. 2, no. 2, pp.
148-151, ACADEMY PUBLISHER
Shachi Dave, Jignashu Parikh & Pushpak Bhattacharyya, (2002) “Interlingua-based English-Hindi
Machine Translation and Language Divergence”, Journal of Machine Translation, pp. 251-304.
Sudip Naskar & Shivaji Bandyopadhyay, (2005) “Use of Machine Translation in India: Current
status” AAMT Journal, pp. 25-31.
Sneha Tripathi & Juran Krishna Sarkhel, (2010) “Approaches to Machine Translation”,
International journal of Annals of Library and Information Studies, Vol. 57, pp. 388-393
Gurpreet Singh Josan & Jagroop Kaur, (2011) “Punjabi To Hindi Statistical Machine
Transliteration”, International Journal of Information Technology and Knowledge Management ,
Volume 4, No. 2, pp. 459-463.
S. Bandyopadhyay, (2004) "ANUBAAD - The Translator from English to Indian Languages", in
proceedings of the VIIth State Science and Technology Congress. Calcutta. India. pp. 43-51
R.M.K. Sinha & A. Jain, (2002) “AnglaHindi: An English to Hindi Machine-Aided Translation
System”, International Conference AMTA(Association of Machine Translation in the Americas)
Murthy. K, (2002) “MAT: A Machine Assisted Translation System”, In Proceedings of
Symposium on Translation Support System( STRANS-2002), IIT Kanpur. pp. 134-139.
Lata Gore & Nishigandha Patil, (2002) “English to Hindi - Translation System”, In proceedings of
Symposium on Translation Support Systems. IIT Kanpur. pp. 178-184.
65

[15]

[16]
[17]
[18]

[19]

[20]

[21]

[22]

[23]
[24]

[25]
[26]

[27]
[28]
[29]

[30]

[31]

[32]

[33]
[34]

Kommaluri Vijayanand, Sirajul Islam Choudhury & Pranab Ratna “VAASAANUBAADA Automatic Machine Translation of Bilingual Bengali-Assamese News Texts”, in proceedings of
Language Engineering Conference-2002, Hyderabad, India © IEEE Computer Society.
Bharati, R. Moona, P. Reddy, B. Sankar, D.M. Sharma & R. Sangal, (2003) “Machine
Translation: The Shakti Approach”, Pre-Conference Tutorial, ICON-2003.
S. Mohanty & R. C. Balabantaray, (2004) “English to Oriya Translation System (OMTrans)”
cs.pitt.edu/chang/cpol/c087.pdf
Ananthakrishnan R, Kavitha M, Jayprasad J Hegde, Chandra Shekhar, Ritesh Shah, Sawani Bade
& Sasikumar M., (2006) “MaTra: A Practical Approach to Fully- Automatic Indicative EnglishHindi Machine Translation”, In the proceedings of MSPIL-06.
G. S. Josan & G. S. Lehal, (2008) “A Punjabi to Hindi Machine Translation System”, in
proceedings of COLING-2008: Companion volume: Posters and Demonstrations, Manchester,
UK, pp. 157-160.
Sanjay Chatterji, Devshri Roy, Sudeshna Sarkar & Anupam Basu, (2009) “A Hybrid Approach
for Bengali to Hindi Machine Translation”, In proceedings of ICON-2009, 7th International
Conference on Natural Language Processing, pp. 83-91.
Vishal Goyal & Gurpreet Singh Lehal, (2011) “Hindi to Punjabi Machine Translation System”, in
proceedings of the ACL-HLT 2011 System Demonstrations, pages 1–6, Portland, Oregon, USA, 21
June 2011.
Ankit Kumar Srivastava, Rejwanul Haque, Sudip Kumar Naskar & Andy Way, (2008) “The
MATREX (Machine Translation using Example): The DCU Machine Translation System for
ICON 2008”, in Proceedings of ICON-2008: 6th International Conference on Natural Language
Processing, Macmillan Publishers, India, http://ltrc.iiit.ac.in/proceedings/ICON-2008.
hutchinsweb.me.uk/Nutshell-2005.pdf
John Hutchins “Historical survey of machine translation in Eastern and Central Europe”, Based
on an unpublished presentation at the conference on Crosslingual Language Technology in
service of an integrated multilingual Europe, 4-5 May 2012, Hamburg, Germany.
(www.hutchinsweb.me.uk/Hamburg-2012.pdf)
Sampark: Machine Translation System among Indian languages (2009) http://tdildc.in/index.php?option=com_vertical&parentid=74, http://sampark.iiit.ac.in/
Akshar Bharti, Chaitanya Vineet, Amba P. Kulkarni & Rajiv Sangal, (1997) ”ANUSAARAKA:
Machine Translation in stages’, Vivek, a quarterly in Artificial Intelligence, Vol. 10, No. 3, NCST
Mumbai, pp. 22-25
Akshar Bharti, Chaitanya Vineet, Amba P. Kulkarni & Rajiv Sangal, (2001) ”ANUSAARAKA:
overcoming the language barrier in India”, published in Anuvad: approaches to Translation
Hemant Darabari, (1999) “Computer Assisted Translation System- An Indian Perspective”, in
proceedings of MT Summit VII, Thialand
R. Mahesh K. Sinha & Anil Thakur, (2005) “Machine Translation of Bi-lingual Hindi-English
(Hinglish) Text”, in proceedings of 10th Machine Translation Summit organized by Asia-Pacific
Association for Machine Translation (AAMT), Phuket, Thailand
Parameswari K, Sreenivasulu N.V., Uma Maheshwar Rao G & Christopher M, (2012)
“Development of Telugu-Tamil Bidirectional Machine Translation System: A special focus on
case divergence”, in proceedings of 11th International Tamil Internet conference, pp 180-191
Salil Badodekar, (2004) “Translation Resources, Services and Tools for Indian Languages”, a
report of Centre for Indian Language Technology, IITB, http://www.cfilt.iitb.ac.in/Translationsurvey/survey.pdf
Ananthakrishnan R, Kavitha M, Jayprasad J Hegde, Chandra Shekhar, Ritesh Shah, Sawani Bade
& Sasikumar M, (2006) “MaTra: A Practical Approach to Fully-Automatic Indicative EnglishHindi Machine Translation”, in proceedings of the first national symposium on Modelling and
shallow parsing of Indian languages (MSPIL-06) organized by IIT Bambay,
202.141.152.9/clir/papers/matra_mspil06.pdf
CDAC Mumbai, (2008) “MaTra: an English to Hindi Machine Translation System”, a report by
CDAC Mumbai formerly NCST.
Sanjay Chatterji, Praveen Sonare, Sudeshna Sarkar & Anupam Basu, (2011) “Lattice Based
Lexical Transfer in Bengali Hindi Machine Translation Framework”, in Proceedings of ICON2011: 9th International Conference on Natural Language Processing, Macmillan Publishers,
India. Also accessible from ltrc.iiit.ac.in/proceedings/ICON-2011.
66

[35]

[36]

[37]
[38]
[39]

[40]
[41]
[42]
[43]
[44]
[45]
[46]
[47]
[48]
[49]
[50]
[51]

R. Ananthakrishnan, Jayprasad Hegde, Pushpak Bhattacharyya, Ritesh Shah & M. Sasikumar,
(2008) “Simple Syntactic and Morphological Processing Can Help English-Hindi Statistical
Machine Translation”, in proceedings of International Joint Conference on NLP (IJCNLP08),
Hyderabad, India.
Yanjun Ma, John Tinsley, Hany Hassan, Jinhua Du & Andy Way, (2008) “Exploiting Alignment
Techniques in MATREX: the DCU Machine Translation System for IWSLT 2008’, in
proceedings of IWSLT 2008, Hawaii, USA
projects.uptuwatch.com/cs-it/anubharti-an-hybrid-example-based-approach-for-machine-aidedtrapnslation/
Sugata Sanyal & Rajdeep Borgohain, (2013) “Machine Translation Systems in India”, Cornel
University Library, arxiv.org/ftp/arxiv/papers/1304/1304.7728.pdf
Antony P. J., (2013) “Machine Translation Approaches and Survey for Indian Languages”,
International journal of Computational Linguistics and Chinese Language Processing Vol. 18,
No. 1, pp. 47-78.
Manoj Jain & Om P. Damani, (2009) “English to UNL (Interlingua) Enconversion”, in
proceedings of 4th Language and Translation Conference (LTC-09).
Smriti Singh, Mrugank Dalal, Vishal Vachhani, Pushpak Bhattacharyya & Om P. Damani, (2007)
“Hindi Generation from Interlingua (UNL)”, in proceedings of MT Summit, 2007
language.worldofcomputing.net
sampark.iiit.ac.in
www.cdacmumbai.in/xlit
www.cdacmumbai.in/rupantar
translationjournal.net/journal/29computers.htm
www.cfilt.iitb.ac.in/resources/surveys/MT-Literature%20Survey-2012-Somya.pdf
www.cdacmumbai.in/e-ilmt
www.iiit.net/ltrc/Anusaaraka/anu_home.html
cdac.in/html/aai/mantra.asp
translate.google.com/about/intl/en_ALL/

Authors
G V Garje (gvg_comp@pvgcoet.ac.in) has completed ME in Computer Science and
Engineering from NITTR, Chandigarh, India in 1998. Currently he is working as Associate
Professor and Head of Computer Engineering and Information Technology Department at
Pune Vidyarthi Griha’s College of Engineering and Technology, Pune. Presently, he is a
Chairman, Board of Studies in Information Technology, University of Pune and formerly,
chairman, Board of Studies in Computer Engineering, University of Pune. Presently he is pursuing his
Ph.D. from University of Pune, Maharashtra, India, in the area of Machine Translation. His area of research
are NLP, Machine Translation specifically English-Marathi Language Pair. He has developed a tool for
translating simple English interrogative sentences to Marathi sentences funded by University of Pune. His
areas of interest are Data Structures, Operating Systems and Software Architecture.
G K Kharate (gkkharate@rediffmail.com) has completed his Ph.D. in Electronics and
Telecommunication Engineering from University of Pune and ME Electronics and
Communication from Walchand College of Engineering, Sangali, Maharashtra. Currently he
is a Principal at Matoshri College of Engineering and Research Centre, Nashik, Maharshtra.
He is a Dean, Faculty of Engineering and Member of Man agement Council, University of
Pune. He is former Chairman, Board of Studies in Electronics Engineering, University of Pune. His areas of
research are Image Processing, Pattern Recognition, and Artificial Intelligence. His areas of interest are
Digital Electronics, Computer Networks, Image Processing and Natural Language Processing.

67

Survey of machine translation systems in india

More Related Content

Similar to Survey of machine translation systems in india (20)

Recently uploaded (20)

Survey of machine translation systems in india