Beruflich Dokumente
Kultur Dokumente
Abstract—Natural Language Processing(NLP) involves the devel- language i.e Source text into other natural language i.e Target
opment of computational models that aid in the development of text. Various approaches can be followed to automate this
automated tools for processing and generating natural language. process of translation such as linguistic rich approach i.e
Human developing these computational models require deep
insight of linguistic knowledge and are a time consuming process. Rule-based [3] involving Transfer-based mechanism [4] and
Hence, to automate this process and accelerate the computational Interlingua mechanism [5] to corpus-based approach that is
science we use a data-driven approach i.e Statistical learning entirely based on corpus i.e Statistical Phrase-based [6] and
and Deep Learning. For devolving and sharing of information Neural-based [7]. Even there are some approaches that is less
in natural language and making it accessible in other natu- used nowadays i.e Example-based approach [8], Knowledge-
ral languages, Machine Translation(MT) is entailed. It is an
application of NLP. Sanskrit being ’father of informatics’ [1] based [9].
was considered as ”lingua franca” of world intellectuals [2]. According to the census of India 2018, India States and Union
It is also an important language in the Indo-European family territories Sanskrit is mother tongue of 24,821 and Hindi of
and considered as truly ”donor” language of India. It has vast 52,83,47,193 i.e 43% of total languages in India. Sanskrit is
knowledge reserves in different discipline of studies such as considered as the mother or ’donor’ of all Indian languages
Ayurveda, astronomy, literature etc. MT makes this rich language
available to others with help of the computer. We have proposed with holding a rich grammar confined by Panini near 2500
and presented the prominent Deep Neural-based MT system for years ago formulating 3,949 rules, which were extended later
translation of Sanskrit to Hindi. We also present a comparison on.Sanskrit has strongest and simple non-ambiguous grammar.
of Neural MT outperforming Statistical baseline system for this Reasons for choosing Sanskrit for translation purpose is the
language pair. richness of scientific literature with extensiveness and com-
Index Terms—Neural Machine Translation, Statistical Machine prehensive analysis, structured approach and the traditional
Translation, Keras,LSTM,MT-Hub, Sanskrit,Hindi grammar. Many people have attempted to write a grammar
for Sanskrit language using the Paninian framework and used
I. I NTRODUCTION it to develop translation system [10]. The Panini focused on
decoding the information contained in the language string of
Natural Language Processing(NLP) involves the development particularly given input language by Karaka(syntax-semantic)
of computational models that aid in the development of relations and not thematic roles [2]. It also highlights the
automated tools for processing and generating natural lan- importance of case markers, postpositions and word-order. The
guage. Human developing these computational models require central element of semantic model in Paninian framework is
deep insight of linguistic knowledge and is time-consuming that every verbal root (dhaatu) denotes an action consisting of
process. Hence, to automate this process and accelerate the activity (or vyaapara), and a result (or phala). The result is a
computational science we use a data-driven approach i.e Statis- state which is reached after completion of an action. An activ-
tical learning and Deep Learning. NLP involves technological ity consists of steps that are performed by different participants
motivation for developing intelligent computer systems such or Karakas involved in action. The concept of Karaka relation
as machine translation, understanding printed and handwritten is central theme to Paninian grammar. These Karaka relations
text, text analysis, speech understanding system, text under- are referred as syntactic-semantic relations, and on surface
standing system and natural language interface to database etc. level, it highlights syntactic information and also captures
There is linguistic and cognitive motivation for development semantic information at a deeper level. The Sanskrit grammar
of this approach. It has various applications such as Machine is termed as ’Father of Informatics’ as it builds a relationship
Translation(MT), Text summarization, Information extraction between speech and utterance of speaker and meaning derived
etc. Machine Translation is one such applications of Natural by listener. Hence, primary objective of Paninian Grammar is
Language Processing(NLP). It aims to translate one natural to form a theory of human natural language communication.
60%
[12] V. Mishra and R. Mishra, “Divergence patterns between english and
sanskrit machine translation,” INFOCOMP, vol. 8, no. 3, pp. 62–71,
50%
2009.
[13] ——, “Ann and rule based model for english to sanskrit machine
40%
translation,” INFOCOMP, vol. 9, no. 1, pp. 80–89, 2009.
30% [14] D. Mane, P. Devale, and S. SURYAWANS, “A design towards english
to sanskrit machine translatio and sy thesizer syste 1 si grule base
20% approach,” 2010.
[4] G. Noone, “Machine translation a transfer approach,” Computer Science, [26] R. Pandey, A. K. Ojha, and G. N. Jha, “Demo of sanskrit-hindi statistical
Linguistics and a Language (CSLL) Department, University of Dublin, machine translation system,” arXiv preprint arXiv:1804.06716, 2018.
Trinity College, Final Rep, 2003. [27] S. News. (2018) Sanskrit News,”Department of Public Health Rela-
[5] S. Dave, J. Parikh, and P. Bhattacharyya, “Interlingua-based english– tions”, url = http://mpinfo.org/News/SanskritNews.aspx, urldate = 13-
hindi machine translation and language divergence,” Machine Transla- 1-2019.
tion, vol. 16, no. 4, pp. 251–304, 2001. [28] D. Kenny and S. Doherty, “Statistical machine translation in the trans-
[6] P. F. Brown, V. J. D. Pietra, S. A. D. Pietra, and R. L. Mercer, “The lation curriculum: overcoming obstacles and empowering translators,”
mathematics of statistical machine translation: Parameter estimation,” The Interpreter and translator trainer, vol. 8, no. 2, pp. 276–294, 2014.
Computational linguistics, vol. 19, no. 2, pp. 263–311, 1993. [29] P. Koehn, H. Hoang, A. Birch, C. Callison-Burch, M. Federico,
[7] D. Bahdanau, K. Cho, and Y. Bengio, “Neural machine translation by N. Bertoldi, B. Cowan, W. Shen, C. Moran, R. Zens et al., “Moses: Open
jointly learning to align and translate,” arXiv preprint arXiv:1409.0473, source toolkit for statistical machine translation,” in Proceedings of the
2014. 45th annual meeting of the ACL on interactive poster and demonstration
sessions. Association for Computational Linguistics, 2007, pp. 177–
[8] H. Somers, “Example-based machine translation,” Machine translation, 180.
vol. 14, no. 2, pp. 113–157, 1999.
[30] F. Chollet et al., “Keras: Deep learning library for theano and tensor-
[9] S. Nirenburg, “al.(1989) kbmt-89 project report,” Center for Machine flow.(2015),” There is no corresponding record for this reference, 2015.
Translation, Carnegie Mellon University, Pittsburg, p. 286, 1989.
[31] M. Abadi, P. Barham, J. Chen, Z. Chen, A. Davis, J. Dean, M. Devin,
[10] A. Bharati, V. Chaitanya, and R. Sangal, “Paninian framework and its S. Ghemawat, G. Irving, M. Isard et al., “Tensorflow: a system for large-
application toanusaraka,” Sadhana, vol. 19, no. 1, pp. 113–127, 1994. scale machine learning.” in OSDI, vol. 16, 2016, pp. 265–283.
[32] P. Koehn, F. J. Och, and D. Marcu, “Statistical phrase-based translation,”
in Proceedings of the 2003 Conference of the North American Chapter
of the Association for Computational Linguistics on Human Language
Technology-Volume 1. Association for Computational Linguistics, 2003,
pp. 48–54.
[33] F. J. Och, C. Tillmann, and H. Ney, “Improved alignment models for
statistical machine translation,” in 1999 Joint SIGDAT Conference on
Empirical Methods in Natural Language Processing and Very Large
Corpora, 1999.
[34] K. Yamada and K. Knight, “A syntax-based statistical translation model,”
in Proceedings of the 39th annual meeting on association for compu-
tational linguistics. Association for Computational Linguistics, 2001,
pp. 523–530.
[35] L. Tian, F. Wong, and S. Chao, “Word alignment using giza++ on
windows,” Machine Translation, 2011.
[36] M. Federico, N. Bertoldi, and M. Cettolo, “Irstlm: an open source toolkit
for handling large scale language models,” in Ninth Annual Conference
of the International Speech Communication Association, 2008.
[37] Microsoft. (2018) Microsoft Translator Hub, url =
https://hub.microsofttranslator.com/, urldate = 13-1-2019.
[38] K. Papineni, S. Roukos, T. Ward, and W.-J. Zhu, “Bleu: a method
for automatic evaluation of machine translation,” in Proceedings of
the 40th annual meeting on association for computational linguistics.
Association for Computational Linguistics, 2002, pp. 311–318.
[39] S. Hochreiter and J. Schmidhuber, “Long short-term memory,” Neural
computation, vol. 9, no. 8, pp. 1735–1780, 1997.
[40] I. J. Goodfellow, D. Warde-Farley, M. Mirza, A. Courville, and Y. Ben-
gio, “Maxout networks,” arXiv preprint arXiv:1302.4389, 2013.
[41] R. Pascanu, C. Gulcehre, K. Cho, and Y. Bengio, “How to construct
deep recurrent neural networks,” arXiv preprint arXiv:1312.6026, 2013.
[42] K. Cho, B. Van Merriënboer, C. Gulcehre, D. Bahdanau, F. Bougares,
H. Schwenk, and Y. Bengio, “Learning phrase representations using
rnn encoder-decoder for statistical machine translation,” arXiv preprint
arXiv:1406.1078, 2014.
[43] D. P. Kingma and J. Ba, “Adam: A method for stochastic optimization,”
arXiv preprint arXiv:1412.6980, 2014.
[44] Y. Bengio, N. Boulanger-Lewandowski, and R. Pascanu, “Advances in
optimizing recurrent networks,” in 2013 IEEE International Conference
on Acoustics, Speech and Signal Processing. IEEE, 2013, pp. 8624–
8628.