NITK-UoH: Tamil-Telugu Machine Translation Systems for the WMT21 Similar Language Translation Task

Proceedings of the Sixth Conference on Machine Translation (WMT), pages 299–303 November 10–11, 2021. ©2021 Association for Computational Linguistics 299 NITK-UoH: Tamil-Telugu Machine Translation Systems for the WMT21 Similar Language Translation Task Richard Saldanha, Ananthanarayana V. S and Anand Kumar M Department of Information Technology, National Institute of Technology Karnataka NH 66, Srinivasnagar, Surathkal, Mangalore Karnataka 575025, India richardsaldanha.207it005@nitk.edu.in anvs@nitk.edu.in m_anandkumar@nitk.edu.in Parameswari Krishnamurthy Centre for Applied Linguistics and Translation Studies, University of Hyderabad Prof. CR Rao Road Gachibowli, Hyderabad Telangana 500046, India pksh@uohyd.ac.in Abstract In this work, two Neural Machine Transla- tion (NMT) systems have been developed and evaluated as part of the bidirectional Tamil- Telugu similar languages translation subtask in WMT21. The OpenNMT-py toolkit has been used to create quick prototypes of the systems, following which models have been trained on the training datasets containing the parallel corpus and finally the models have been evalu- ated on the dev datasets provided as part of the task. Both the systems have been trained on a DGX station with 4 - V100 GPUs. The first NMT system in this work is a Trans- former based 6 layer encoder-decoder model, trained for 100000 training steps, whose con- figuration is similar to the one provided by OpenNMT-py and this is used to create a model for bidirectional translation. The sec- ond NMT system contains two unidirectional translation models with the same configuration as the first system, with the addition of utiliz- ing Byte Pair Encoding (BPE) for subword tok- enization through the pre-trained MultiBPEmb model. Based on the dev dataset evaluation metrics for both the systems, the first system i.e. the vanilla Transformer model has been submitted as the Primary system. Since there were no improvements in the metrics during training of the second system with BPE, it has been submitted as a contrastive system. 1 Introduction Tamil is a language, predominantly spoken in Tamil Nadu, a state in Southern India, along with coun- tries with a large Tamil speaking diaspora such as Sri Lanka, Malaysia and Singapore, to name a few. Telugu on the other hand is the official language of two Southern states in India, namely Andhra Pradesh and Telangana. It is also spoken among the Telugu speaking immigrant population in the USA, Canada and the UK. Both languages belong to the Dravidian family of languages which com- prise of Tamil, Telugu, Kannada and Malayalam as the major languages spoken in South India. Despite belonging to the same family of languages, there are many differences between Tamil and Telugu, such as the script used for writing and linguistic dif- ferences in terms of phonology, morphology, syn- tax among others. Tamil belongs to the Southern branch of Dravidian languages, which has a rich literary tradition spanning more than 2000 years. Telugu, on the other hand, belongs to the South Central branch of Dravidian languages and has a considerable amount of different linguistic charac- teristics when compared to Tamil as described by Krishnamurthy (2019). As part of the similar language translation’s sub- task …

Content truncated to 500 words