| Resumen |
In this paper it is presented a model of text representation based on graphs. The model is applied in the particular case study of au-thorship attribution. The experiments were performed by using a corpus made up of 500 documents written by 10 di erent authors (50 documents per author). The obtained results highlight the bene t of using text fea-tures at di erent levels of language description in tasks associated to automatic processing of information. In particular, we have obtained a performance of 57% of accuracy for the authorship attribution task. |
