Data.CRF.Chain2.Pair

Data types

data Word a b

A word with a representing the observation type and b representing the compound label type.

Instances

(Eq a, Eq b) => Eq (Word a b)

(Eq (Word a b), Ord a, Ord b) => Ord (Word a b)

(Show a, Show b) => Show (Word a b)

mkWord :: Set a -> Set b -> Word a b

A word constructor which checks non-emptiness of the potential set of labels.

type Sent a b = [Word a b]

data Dist a

A probability distribution defined over elements of type a. All elements not included in the map have probability equal to 0.

mkDist :: Ord a => [(a, Double)] -> Dist a

Construct the probability distribution.

type WordL a b = (Word a b, Dist b)

A WordL is a labeled word, i.e. a word with probability distribution defined over labels. We assume that every label from the distribution domain is a member of the set of potential labels corresponding to the word. TODO: Ensure the assumption using the smart constructor.

type SentL a b = [WordL a b]

A sentence of labeled words.

CRF

data CRF a b c

Constructors

CRF

Fields

codec :: Codec a b c

model :: Model Ob Lb Feat

Instances

(Ord a, Ord b, Ord c, Binary a, Binary b, Binary c) => Binary (CRF a b c)

Training

train

Arguments

:: (Ord a, Ord b, Ord c)

=> SgdArgs

Args for SGD

-> IO [SentL a (b, c)]

Training data IO action

-> Maybe (IO [SentL a (b, c)])

Maybe evalation data

-> IO (CRF a b c)

Resulting codec and model

Train the CRF using the stochastic gradient descent method. When the evaluation data IO action is Just, the iterative training process will notify the user about the current accuracy on the evaluation part every full iteration over the training part. TODO: Add custom feature extraction function.

Tagging

tag :: (Ord a, Ord b, Ord c) => CRF a b c -> Sent a (b, c) -> [(b, c)]

Find the most probable label sequence.