Structural classifiers of text types: Towards a novel model of text representation

Alexander Mehler, Peter Geibel, Olga Pustylnikov

2007 · 18 citations · 27 references

Concepts

Abstract

Texts can be distinguished in terms of their content, function, structure\nor layout (Brinker, 1992; Bateman et al., 2001; Joachims, 2002; Power\net al., 2003). These reference points do not open necessarily orthogonal\nperspectives on text classification. As part of explorative data analysis,\ntext classification aims at automatically dividing sets of textual objects into\nclasses of maximum internal homogeneity and external heterogeneity. This\npaper deals with classifying texts into text types whose instances serve more\nor less homogeneous functions. Other than mainstream approaches, which\nrely on the vector space model (Sebastiani, 2002) or some of its descendants\n(Baeza-Yates and Ribeiro-Neto, 1999) and, thus, on content-related lexical\nfeatures, we solely refer to structural differentiae. That is, we explore\npatterns of text structure as determinants of class membership. Our starting\npoint are tree-like text representations which induce feature vectors and\ntree kernels. These kernels are utilized in supervised learning based on\ncross-validation as a method of model selection (Hastie et al., 2001) by\nexample of a corpus of press communication. For a subset of categories we\nshow that classification can be performed very well by structural differentia\nonly.

References

27