Towards deeper understanding of the latent semantic analysis performance
Preslav Nakov, Elena Valchanova, Galia Angelova · Amsterdam studies in the theory and history of linguistic science. Series 4, Current issues in linguistic theory · 2004
The paper studies the factors influencing the performance of the Latent Semantic Analysis. Unlike previous related research that concentrates on parameters such as matrix elements weighting, space dimensionality, similarity measure etc., we address the impact of another fundamental factor: the definition of “word”. For the purpose, series of experiments were performed on two corpora in order to compare (with respect to the task of text categorisation) six word variants with different linguistic quality. The results show that while the linguistic processing influences the performance, the traditional factors are more important.