The human-computer interfaces exploits the annotated corpora to identify the meaning of the statements during dialogue management. The construction of an annotated corpus is an arduous task, usually manual involving high costs and limitations on the number of texts to annotate. This paper presents the automatic construction and evaluation of an annotated corpora for strengthening human-computer interfaces. The proposal is inexpensive and independent of domain information; therefore, it applies to other contexts with a low investment of resources.