Probabilistic Latent Semantic Analysis. Hofmann, T.
Probabilistic Latent Semantic Analysis [pdf]Paper  Probabilistic Latent Semantic Analysis [pdf]Website  abstract   bibtex   
Probabilistic Latent Semantic Analysis is a novel statistical technique for the analysis of twoomode and co-occurrence data, which has applications in information retrieval and [ltering, natural language processing, ma-chine learning from text, and in related ar-eas. Compared to standard Latent Semantic Analysis which stems from linear algebra and performs a Singular Value Decomposition of co-occurrence tables, the proposed method is based on a mixture decomposition derived from a latent class model. This results in a more principled approach which has a solid foundation in statistics. In order to avoid overrtting, we propose a widely applicable generalization of maximum likelihood model by tempered EM. Our approach yields substantial and consistent improvements over Latent Semantic Analysis in a number of ex-periments.
@article{
 title = {Probabilistic Latent Semantic Analysis},
 type = {article},
 websites = {http://www.iro.umontreal.ca/~nie/IFT6255/Hofmann-UAI99.pdf},
 id = {f26ef594-349c-3ec3-888e-dafbf9e3cad1},
 created = {2018-02-05T16:56:49.934Z},
 accessed = {2018-02-05},
 file_attached = {true},
 profile_id = {371589bb-c770-37ff-8193-93c6f25ffeb1},
 group_id = {f982cd63-7ceb-3aa2-ac7e-a953963d6716},
 last_modified = {2018-02-05T16:56:51.867Z},
 read = {false},
 starred = {false},
 authored = {false},
 confirmed = {false},
 hidden = {false},
 private_publication = {false},
 abstract = {Probabilistic Latent Semantic Analysis is a novel statistical technique for the analysis of twoomode and co-occurrence data, which has applications in information retrieval and [ltering, natural language processing, ma-chine learning from text, and in related ar-eas. Compared to standard Latent Semantic Analysis which stems from linear algebra and performs a Singular Value Decomposition of co-occurrence tables, the proposed method is based on a mixture decomposition derived from a latent class model. This results in a more principled approach which has a solid foundation in statistics. In order to avoid overrtting, we propose a widely applicable generalization of maximum likelihood model by tempered EM. Our approach yields substantial and consistent improvements over Latent Semantic Analysis in a number of ex-periments.},
 bibtype = {article},
 author = {Hofmann, Thomas}
}
Downloads: 0