Implementation of the Expectation Maximation algorithm to calculate probabilities of rules in Context-Free Grammar (CFG) in order to create Probabilistic Context-Free Grammar (PCFG). It also generates new sentences in that grammar using a PCFG (.gen file).

• .cfg file

A file with a context-free grammar in Chomsky normal form, but with "#" instead of an arrow (no spaces around).

Rule example:

S#NPa ViADV

• .train file

A text file with .train extension with sentences you want to train your grammar on. Sentences should be in separate lines.

In the command line, type:

python train.py [YOUR_GRAMMAR_FILE].cfg [YOUR_TRAIN_FILE].train

You can use some examples in the test folder, or try your own.

• Inside–outside algorithm

• The Inside-Outside Algorithm

• Note on the Inside-Outside Algorithm