The Vietnamese deaf community communicates via a special language called Vietnamese Sign Language (VSL). Three-dimensional space and hand gesture are primarily used to convey meanings that allow deaf people to communicate among themselves and with non-deaf people around them. It maintains syntax, grammar, and vocabulary which is completely different from regular spoken and/or written Vietnamese. The normal procedure of transformation from spoken and/or written language (SWL) to VSL consists of consecutive steps: (i) Vietnamese word tokenization, (ii) machine translation into written sign language sentences, and (iii) conversion of these written sign language sentences into visual sign gesture. In this procedure, the second step gets the most attention due to the completion of the conveyed message. The basic challenge is that sign language, in general, has limited vocabulary compared to spoken/written language. If the machine translation is poorly performed, the complete message might not be successfully communicated, or in some cases, the conveyed message has a different meaning from the original. Consequently, the high accuracy translation should be maintained in any circumstances. This research is the efforts of evaluating an effective classification algorithm that the authors recommend to be integrated into the workflow of VSL translation. We believe that this is the first showing a quantitative comparison of several classification algorithms used in the workflow. The experimental results show that the translation accuracy rate is 96.7% which strongly support our recommendation.