Unified framework for sentiment analysis in multiple languages

Multilingual sentiment analysis plays a critical role in comprehending customer sentiment, feedback, and emotional responses. This study introduces a comprehensive framework designed to augment the efficacy of sentiment analysis across multiple languages. The research utilizes renowned machine translation services, namely Google Translate and Yandex Translate, to carry out sentiment analysis in several languages including English, Turkish, Arabic, and French. The outcomes underline the advantage of deploying a single, comprehensive framework for multilingual sentiment analysis. Furthermore, they underscore the crucial role machine translation services play in simplifying sentiment analysis across various languages. The insights gained from the results are beneficial to both researchers and practitioners in the sentiment analysis sphere. The proposed framework underwent testing on multiple datasets, exhibiting encouraging results with an improvement in accuracy between 1% and 22% depending on the language. Our method outperforms language-specific models and substantiates the efficiency of the proposed translation- based multilingual framework. Additionally, the study revealed that the efficacy of sentiment analysis fluctuates between different languages. Google Translate demonstrated superior performance in Turkish and Arabic sentiment analysis translations, whereas Yandex Translate excelled in English and French sentiment analysis translations. Keyword: Sentiment analysis, multilingual sentiment analysis, deep learning, translation-based sentiment analysis, LSTM.