The Impact of Translating Resource-Rich Datasets to Low-Resource Languages Through Multi-Lingual Text Processing

Urdu is still considered a low-resource language despite being ranked as world’s $10^{th}$ most spoken with nearly 230 million speakers. The scarcity of benchmark datasets in languages has led researchers to utilize more ingenious techniques curb the issue. One such option widely adopted use translation services replicate existing from resource-rich English languages, Urdu. For natural processing tasks, including polarity assessment, words translated via Google translator one another often change meaning. It results shift causing system’s performance degradation, particularly for sentiment classification and emotion detection tasks. This study evaluates effect on task language. identifies enlists into five distinct categories. further finds correlation between similar roots. Our shows 2-3 percentage points degradation due result languages.