Virtual Assistant Using Lstm Networks In Indonesian

Researches into the development of virtual assistants that are human-like today were an active research that has been developed over the past few decades. Currently the generative model virtual assistant is a few steps towards artificial general intelligence, the smart computer with complex feelings and characteristics. One method for building the generative model virtual assistant is by using the LSTM Networks. This method has been popular for building English virtual assistant. Our aim is to test whether this method could be applied in Indonesian or not. We used the dataset from many movie subtitles in Indonesian languages. The LSTM variant that we used was the sequence-to-sequence model embedded with word2vec. The results show that the model could answer appropriately to the simple questions such as greetings. Although the models were failed to answer some complex questions, the results still give potential works on the future.