Easily Labelling Hierarchical Document Clusters

Maria Fernanda MouraRicardo Marcondes MarcaciniSolange Oliveira Rezende

One of the problems of automatic models that generate topic taxonomies is the process of creating the most significant term list that discriminates each document group. In this paper, a new method to label document hierarchical clusters is proposed, which is completely independent from the clustering method. This method automatically decides the number of the words in each label list, avoids word repetitions in a tree branch and provides a kind of cutting for the cluster tree. The obtained results were tested as search queries in a retrieval process and showed a very good performance. Additionally, the use of the method was experimented by some specialists in the text collection domain, trying to evaluate their understanding and expectations over the results.

Caso o link acima esteja inválido, faça uma busca pelo texto completo na Web: Buscar na Web

Biblioteca Digital Brasileira de Computação - Contato:
     Mantida por: