Stance Detection in Hinglish Data using the BART-large-MNLI Integration Model

Author:

Giddaluru Somasekhar,Maturi Sreerama Murty,Ooruchintala Obulesu,Munirathnam Mahendra

Abstract

Real-time stance detection can be used in a wide range of applications like debate analysis, sentiment analysis, system feedback, etc. This study focuses on stance detection in political speeches, discerning whether the speaker is in favor, against, neutral, or lacking any stance on a given topic. The problem with this type of speeches dwells in the modification of the existing methods of stance detection to allow for the fine distinctions of Hinglish, a language mixture blending Hindi and English, as conveyed in human-edited texts. The proposed method utilizes the Bidirectional Auto-Regressive Transformers Multi-Genre Natural Language Inference (BART-large-MNLI) model with zero-Shot, few-Shot and N-Shot learning approaches. The proposed model is compared with the existing models of stance detection on Hinglish texts. For the pre-trained BART-based models, a limited number of labeled examples are utilized to determine the labels of test instances. For the other models, the train-test split method is adopted to get accurate results. The results indicate that the model surpasses the previous models. 

Publisher

Engineering, Technology & Applied Science Research

同舟云学术

1.学者识别学者识别

2.学术分析学术分析

3.人才评估人才评估

"同舟云学术"是以全球学者为主线,采集、加工和组织学术论文而形成的新型学术文献查询和分析系统,可以对全球学者进行文献检索和人才价值评估。用户可以通过关注某些学科领域的顶尖人物而持续追踪该领域的学科进展和研究前沿。经过近期的数据扩容,当前同舟云学术共收录了国内外主流学术期刊6万余种,收集的期刊论文及会议论文总量共计约1.5亿篇,并以每天添加12000余篇中外论文的速度递增。我们也可以为用户提供个性化、定制化的学者数据。欢迎来电咨询!咨询电话:010-8811{复制后删除}0370

www.globalauthorid.com

TOP

Copyright © 2019-2024 北京同舟云网络信息技术有限公司
京公网安备11010802033243号  京ICP备18003416号-3