Non-linear correspondence analysis in text retrieval: a kernel view
Davide Picca, François Bavaud · 2006
Classical factorial treatments applied on words-documents counts matrices (such as Correspondence Analysis (FCA), Latent Semantic Indexing (LSI), as well as non-linear generalizations of FCA (NLCA)) can be described in the framework of kernels associated to Support Vector Machines (SVM). This paper exposes the relationships between those formalisms, and demonstrates how textual pre-processing by a “power kernel” can improve (with respect to the classical FCA kernel) the documents classification in the Reuters-21578 corpus.