IJLLL 2017 Vol.3(4): 226-229 ISSN: 2382-6282
DOI: 10.18178/IJLLL.2017.3.4.137

A Karaka Based Approach to Cross Lingual Sentiment Analysis

Vartika Rai, Sakshee Vijay, and Dipti Misra Sharma
Abstract—Sentence level translation and projection of polarity is way more challenging for english-hindi pair as hindi is relatively free order language and translation cost adds up to the error rate. Instead of that, if we assume each sentiment bearing sentences as combination of certain karaka relations and labels, and narrow down our observation to phrase level for every sentence, there can be a dependency between english and hindi phrases through phrase level translation and finding sentiments for those particular english phrases can lead us to predicting sentiments of their respective hindi sentences as well, with very less translation errors and no dependency on hindi labelled corpus.

Index Terms—Cross lingual, dependency parser, karaka roles, sentiment analysis.

The authors are with International Institute of Information Technology, Hyderabad, India (e-mail: vartika.rai@research.iiit.ac.in, sakshee.vijay@research.iiit.ac.in, dipti@iiit.ac.in).

[PDF]

Cite:Vartika Rai, Sakshee Vijay, and Dipti Misra Sharma, "A Karaka Based Approach to Cross Lingual Sentiment Analysis," International Journal of Languages, Literature and Linguistics vol. 3, no. 4, pp. 226-229, 2017.

Copyright©2008-2017. International Journal of Languages, Literature and Linguistics. All rights reserved.
E-mail: ijlll@ejournal.net