Acquisition and Development of Code-Mixed Corpus for Sentiment Analysis of Resource-Scarce Languages

Code-mixing on social media sites has emerged as the new way of writing in a multilingual country. The researchers are already facing challenges on pre-processing (spell correction, POS tagging, stemming, language identification, etc. of code-mixed multilingual text. However, in this article some code-mixed bilingual phonetic text has been developed for sentiment analysis. The bilingual English-Punj abi phonetic word dictionary is developed using Gurmukhi to Roman transliteration (GRT) system. The training and testing code-mixed dataset has been developed using a bilingual word dictionary and GRT system. The English and bilingual code-mixed sentiment-bearing word dictionary has been developed manually using existing English SentiWordNet 3.0. This article focused on the development of code-mixed corpus and datasets for resource-scarce languages.