The new Semantic Search feature of SQL Server Denali enhances full-text search by letting you do a search on the meaning of documents. To use this feature, its required that you install the SemanticsDB. The SemanticsDB ships with SQL Server or can be downloaded as part of the SQL Server Denali CTP3 Feature Pack. One you have attached the SemanticsDB to your SQL Server instance and provided appropriate access to it, you can use the STATISTICAL_SEMANTICS keyword of CREATE FULLTEXT INDEX or ALTER FULLTEXT INDEX to build semantic indices, and use them with the three table-valued functions I mentioned in an earlier blog post. So what does the SemanticsDB provide?

The SemanticsDB contains Statistical Language Models. A starter description of this can be found in Wikipedia, but these models are used to build the semantic indexes that are used with the semantic search table-valued functions. There is actually one language model internal table for each supported language. The SemanticDB is static and you can't customize it or provide your own. It's not updated during semantic index processing, in fact, you can make the SemanticDB read-only with no ill effects to the feature. It's used to make the index-building scalable, as the speed of semantic index building is not dependant on the size of the corpus of documents.

Hope this gets you intrigued about this interesting new feature.

@bobbeauch