Examine the Impact of Normalizing and Using Amharic Informal Opinionated Features in Sentiment Analysis

Abstract Social media users currently express their ideas, opinions, and feelings using informal vocabulary. The majority of social media users commonly speak informally, using slang, misspellings, grammar mistakes, and abbreviations. Nowadays, people express their opinions most of the time informally. We tackle the issue of Amharic sentiment analysis by using informal opinionated words in Amharic as a feature and preprocessing it using normalization. We also study the impact of using reduced minimum word frequency parameter in an automated feature extractor that incorporates word and character n-gram embedding. Compared to earlier work approaches, the study’s highest recall result was 91.67 %; an average recall improvement of 2.8 was gained.