Training a parts-of-speech tagger
We will use a Haskell library, sequor, to train our own parts of speech tagger. Then we can use this newly trained model on our own input.
Getting ready
Please refer to the Getting ready section of the previous recipe.
How to do it…
In a new file, which we name Main.hs, enter the following source code:
Use the
sequorexecutable to train the parts of speech tagger:The first argument to
sequorwill betrain, to indicate that we are about to train a taggerThe next argument is the template-file,
data/all.featuresThen we provide the train-file,
data/train.conllThe last file path we need to provide is the location of where to save the trained model
We can specify a learning rate using the
-rateflagThe beam size can be modified using the
-beamflagChange the number of iterations using the
-iterflagUse hashing instead of a feature dictionary using the
-hashflagProvide a path to the held out data using the
-heldoutflagAn example of the sequor command in use is as...