There are many native languages in Latin America, over the decades the number of speakers was reduced by the strong influence of the Spanish language. There is a continuous concern for the preservation of these languages, such as: Aymara, Quechua, Guaraní. To create Neural Machine Translator (NMT) models, there is no data set of translations from the native language Aymara - Spanish. Therefore, this document presents a data set of conversations in native Aymara language and the respective translations into Spanish. The first translation tests with the seq2seq model are also carried out. The first initial results are promising, considering that it is the first application of Natural Language Processing (NLP) and translation machine for the native Aymara language.