Gui at MixMT 2022 : English-Hinglish : An MT Approach for Translation of Code Mixed Data

Proceedings of the Seventh Conference on Machine Translation (WMT), pages 1126 - 1130 December 7-8, 2022 ©2022 Association for Computational Linguistics Gui at MixMT 2022 : English-Hinglish : An MT approach for translation of code mixed data Akshat Gahoi Jayant Duneja Anshul Padhi Shivam Mangale Saransh Rajput Tanvi Kamble Dipti Misra Sharma Vasudeva Varma International Institute of Information Technology, Hyderabad {akshat.gahoi,anshul.padhi,saransh.rajput,tanvi.kamble}@research.iiit.ac.in {dunejajayant,shivammangale}@gmail.com Abstract Code-mixed machine translation has become an important task in multilingual communities and extending the task of machine translation to code mixed data has become a common task for these languages. In the shared tasks of WMT 2022, we try to tackle the same for both English + Hindi to Hinglish and Hinglish to English. The first task dealt with both Roman and Devanagari script as we had monolingual data in both English and Hindi whereas the second task only had data in Roman script. To our knowledge, we achieved one of the top ROUGE-L and WER scores for the first task of Monolingual to Code-Mixed machine translation. In this paper, we discuss the use of mBART with some special pre-processing and post-processing (transliteration from Devanagari to Roman) for the first task in detail and the experiments that we performed for the second task of translating code-mixed Hinglish to monolingual English. 1 Introduction Code Mixing occurs when a multi-lingual individual uses two or more languages while communicating with others. It is the most natural form of conversation for multilinguals. It is often confused with code-switching but there is a slight difference between the two. Both these phenomena include communicating in multiple languages but code switching usually takes place within multiple sentences while code mixing usually refers to words of different languages used in the same sentence. In code mixing, phrases, words and morphemes of one language may be embedded within an utterance of another language. Code mixing is extensively observed on social media sites like Facebook and twitter. With the rapid growth of social media and consequently, increase in the use of code-mixed data, it becomes important to develop systems to process such text. Machine Translation, also known as automated translation, is the process where a software trans- lates text from one language to another without any human involvement. There are multiple forms of machine translation, however, over the past few years, neural machine translation has become extremely popular. The WMT shared task had two subtasks. The first subtask consisted of the translation of Hindi-English parallel sentence pairs to Hindi-English code mixed sentences through machine translation. The second subtask consisted of the translation of Hindi-English code mixed sentences to English. 2 Background While there is a growing interest in code-mixed text analysis as a research problem, there is one bottleneck that has hindered the growth of such works, and that is the lack of data. Due to this, there aren’t many robust models for code-mixed text. To build standardized datasets of code-mixed text, we need to come up with ways of text genera- tion of these code-mixed texts. …

Content truncated to 500 words