Sign language (SL) is based on the movement of hands, body gestures and motions to express information instead of speech or written forms. SL has been used since 384-322 BC and developed in many countries, including Vietnam. In some countries, there are several applications of using technology to develop SL and communicate within the deaf community. However, there is lack of studies in these areas in Vietnam. This study is approaching of using 3D avatar of the HamNoSys and building a dictionary for Vietnamese SL animation. The study builds a data set and applies the machine learning decision tree (ID3) to convert the Vietnamese normal sentences into the short sentences of the deaf. The conversion process generates SiGML language to express the SL by a virtual signer leading to present the television news. The assessment results by experts in SL institutions indicate high precision of proposed solution and show the needs of the study.