Exploiting latent semantic information in statistical language modeling

OF THE IEEE , AUGUST 2000 1 Exploiting Latent Semantic Information in Statistical Language Modeling Jerome

| Statistical language models used in large vocabulary speech recognition must properly encapsulate the various constraints, both local and global, present in the language. While local constraints are readily captured through n-gram modeling, global constraints, such as long-term semantic dependencies, have been more diÆcult to handle within a data-driven formalism. This paper focuses on the us...

On latent semantic language modeling and smoothing

Language modeling plays a critical role for automatic speech recognition. Conventionally, the n-gram language models suffer from lacking good representation of historical words and estimating unseen parameters from insufficient training data. In this work, the latent semantic information is explored for language modeling and parameter smoothing. In language modeling, we present a new representa...

Exploiting Sensitive Information in Background Mode using Latent Semantic Indexing

Access to specific information items often needs to be restricted due to privacy, competitive, legal, or national security issues. However, the growth of problems such as international terrorism creates requirements for increased sharing of information. In general, there is a tradeoff that must be made between protecting information and carrying out needed activities. Numerous information shari...

Statistical language modeling for information retrieval

Latent Maximum Entropy Principle for Statistical Language Modeling

In this paper, w dcscrihe a unified probabilistic framework for statistical lmiguage modeling, latent mmmum mtmpy Wnciple. The salient feature of this approach is that the hidden causal hierarchical dependency structure can be enc d e d into thpsiatwtical model in a principled way hy mixturts of expoocntial families with a rirh expressive power. We first show tlie problem lormulatiun, whition, ...