Empirical Analysis of Sentence Embedding Techniques for Answer Retrieval in Marathi Question Answering

Kāraka-Based Answer Retrieval for Question Answering in Indic Languages

“…Several other approaches were proposed, including a natural language interface to relational databases using Paninian grammar and kārakas (Gupta et al, 2012), the use of Universal Networking Language (UNL) for representing the meaning of text in the source language without translation (Shukla et al, 2004), and rule-based systems for Hindi QA (Sahu et al, 2012). Additionally, there were developments in web-based QA systems (Stalin et al, 2012), pattern matching algorithms for QA (Gupta and Gupta, 2014), question classification models (Banerjee and Bandyopadhyay, 2012), answer sentence selection models for QA (Verma et al, 2021;Joshi et al, 2022) and deep learning-based frameworks for cross-lingual (Gupta et al, 2018) and multi-lingual QA (Gupta et al, 2019). Recent experiments explored the use of transformer models pre-trained on multiple languages, with a focus on Hindi and Tamil QA, achieving improved performance in extractive QA tasks (Thirumala and Ferracane, 2022;Namasivayam and Rajan, 2023).…”