Tanmai Khanna have got AIR 61 in UPSC CSE EXAM 2022.
Tanmai Khanna have completed his education from IIT Hyderabad. He is a researcher also. He likes to play guitar. He is living with his family. He have cleared UPSC CSE Exam through self study.
In this post, I have shared some research related work of Tanmai Khanna.
The Computational Linguistics (CL) MS Dual Degree was awarded to Tanmai Khanna. Professor Dipti M. Sharma oversaw his research. An overview of Tanmai Khanna’s dissertation, Rule-based pre-processing of idioms and non-compositional constructs to improve black-box machine translation, is provided below.
Machine Translation is a sub-field of Computational Linguistics that deals with systems that automatically translate text or an utterance from one language into another. The field has made huge strides, from rule-based translators to the now state-of-the-art neural machine translators. However, translators in the general domain are far from achieving human parity. At this stage, even for state-of-the-art translators, a human needs to check the output and post-edit it. The aim therefore, is to continuously improve the machine translator so that the post-editing effort reduces. Guided by this aim and a will to make translators more complete, I identify one area where even state-of-the-art machine translators perform poorly – translating idiomatic and non-compositional constructions.
I demonstrate that these structures are uncommon in the data that translators are taught and assessed on, which may be the main cause of the translations’ inadequateness. I suggest a rule-based pre-processor that recognises non-compositional constructs in the input sentence and turns them into more compositional constructions, which are far more likely to translate adequately, in order to increase the translation adequacy of these constructions.
I begin by creating a list of constructions, ranging from entirely lexical and inflexible constructs to constructions with slots restricted by parts of speech to fully syntactic constructions.
I assess five English-Hindi NMT systems using samples of sentences using these constructs and find that they perform woefully poorly when translating non-compositional constructions. I analyse these constructions in order to understand the capabilities that a pre-processor would require to address this problem, and I identify the qualities that a rule-based pre-processor needs to identify and streamline the constructions. The actual pre-processor I developed for this project, together with its rule formalism, is described after this theoretical examination of rules. After that, the pre-processor is thoroughly assessed for English-Hindi translation.
I report both a significant improvement in the quality of Hindi translation following the preprocessing of the non-compositional constructions in the English input text into more compositional constructions and a high accuracy of construction recognition in English based on manually established rules. In my final section, I go over the evaluation’s findings, the pre-processor’s mistakes, some of this solution’s limitations, and potential future work to expand and enhance this system.
IAS Topper Tanmai Khanna’s mock interview is presented to you by iasbio.com. Tanmai Khanna received an AIR of -61 in the UPSC CSE 2022 Examination. We sincerely congratulate Tanmai Khanna on her outstanding accomplishment. We are very grateful to have been a part of his inspiring journey, and we once again wish him nothing but the best for the future. Visit our YouTube channel, please.
