]Latent semantic indexing (LSI) is a promising information
]retrieval technique for searching and organizing large data
]collections. LSI finds patterns in unstructured data (documents
]without descriptors such as keywords or special semantic
]tagging), and can return relevant results for a query even when
]there is no keyword match.

]Data collections don't have to be in English, or even in any
]human language at all. We have had good success in searching
]protein databases with the technique, as well as chemical mass
]spectra.

I'm going to have to look this over in some detail later..

Projects: Latent Semantic Indexing