DMCA
Unsupervised Learning by Probabilistic Latent Semantic Analysis (2001)
Cached
Download Links
| Venue: | Machine Learning |
| Citations: | 616 - 4 self |
BibTeX
@INPROCEEDINGS{Hofmann01unsupervisedlearning,
author = {Thomas Hofmann},
title = {Unsupervised Learning by Probabilistic Latent Semantic Analysis},
booktitle = {Machine Learning},
year = {2001},
pages = {2001}
}
Share
 |
 |
 |
 |
||
OpenURL
Abstract
Abstract. This paper presents a novel statistical method for factor analysis of binary and count data which is closely related to a technique known as Latent Semantic Analysis. In contrast to the latter method which stems from linear algebra and performs a Singular Value Decomposition of co-occurrence tables, the proposed technique uses a generative latent class model to perform a probabilistic mixture decomposition. This results in a more principled approach with a solid foundation in statistical inference. More precisely, we propose to make use of a temperature controlled version of the Expectation Maximization algorithm for model fitting, which has shown excellent performance in practice. Probabilistic Latent Semantic Analysis has many applications, most prominently in information retrieval, natural language processing, machine learning from text, and in related areas. The paper presents perplexity results for different types of text and linguistic data collections and discusses an application in automated document indexing. The experiments indicate substantial and consistent improvements of the probabilistic method over standard Latent Semantic Analysis.
Keyphrases
probabilistic latent semantic analysis singular value decomposition related area many application information retrieval probabilistic mixture decomposition perplexity result standard latent semantic analysis count data novel statistical method co-occurrence table natural language processing probabilistic method factor analysis expectation maximization algorithm excellent performance consistent improvement model fitting linguistic data collection different type linear algebra machine learning latent semantic analysis generative latent class model principled approach automated document indexing statistical inference solid foundation latter method
