Seismic activity at Soufri¿re Hills volcano is characterized by a variety of transients, such as tectonic earthquakes, long-period events, hybrid events, and rockfalls. The huge quantity of seismic data daily recorded on the volcano makes the application of automatic processing highly recommendable. We propose a method of supervised classification of the transients based on Artificial Neural Networks (ANN), which may be useful for processing the large data sets piled up in the past. Particularly, data sets recorded before the climactic eruptions from 1995 to 2002 may allow us to reconstruct the distribution of the different classes of seismic transients in time. We believe that this analysis may give useful insights into impending eruptive scenarios. The good performance of the ANN with 70% of transients correctly classified in a test set of 156 data, along with the opportunity to revise the misfits, make ANN a powerful tool for data processing.