A recent swarm algorithm, Rat Swarm Optimizer (RSO), has demonstrated its efficiency in solving various mathematical problems. The Modified Binary RSO (MBRSO) is a binary version of RSO designed to address binary problems with enhanced performance. Transfer functions, used in MBRSO, serve as the medium for mapping an algorithm from continuous to discrete binary space. Despite the widespread use of swarm algorithms, studies focusing on transfer functions are still limited in both number and diversity. This study proposes eight new transfer functions, grouped into two families: sinus-based and cosinus-based, for MBRSO. Two standard benchmark datasets for Arabic sentiment analysis were used to evaluate the proposed transfer functions in terms of fitness, accuracy and feature reduction rate. Compared with state-of-the-art transfer functions, the newly introduced functions demonstrate their competitiveness and efficiency with MBRSO algorithm. Additionally, parameter fine-tuning appears to be a crucial factor in improving results.