purpose of Penn TreeBank and PCFG Model in Stanford Parser

I was confused by the purpose of english开发者_StackOverflowPCFG Model and Penn treebank annotation, the package of Standford Parser only included all kind of Models, it always questions me how does this model works if we already have annotation from Peen treebank. Simply, what is Peen Treebank Annaotation works for Parser and how does Model come out? if a raw text come for parser, does it need to query Treebank to predict trees agian?

I am reading some materials, but still dont know when did Model being generate at below steps. 1, Choose an available treebank. 2, Choose a parser engine suitable for the treebank annotation. 3, Select training and test data. 4, Train the parser on the training set. 5, Evaluate the parser's accuracy on the test set. 6, Write a report on the project with experimental results.

anyone can help?

It is saved state after step 4, which you can use to evaluate the parser or to parse text at any later time, without needing to retrain.