@author Dipanjan Das

Read a simplified version of the CoNLL format. Two columns Word \t POSTag With a blank line separating sentences. Returns 'null' for the input re

Extracts features and numberizes them Also numberizes other things if necessary (e.g. label numberizations for MEMM training)

String normalizations and other shared utilities for feature computation.

We should test if these are obsolete yet

This contains (1) Feature and label vocabularies (therefore knowledge of numberization) (2) Model coefficients (and knowledge how to flattenize them

One sequence structure -- typically, for one sentence This is the model's view of a sentence -- only deals with non-textual numberized versions of eve

Class implementing the Orthant-Wise Limited-memory Quasi-Newton algorithm (OWL-QN). OWN-QN is a numerical optimization procedure for finding the optim

Commandline interface to run the Twitter POS tagger with a variety of possible input and output formats. Also does basic evaluation if given labeled i

Holds textual and linguistic information for a sentence. Theoretically could add additional textual,syntactic,etc. annotations as inputs

One token and its tag.

Tagger object -- wraps up the entire tagger for easy usage from Java. To use: (1) call loadModel(). (2) call tokenizeAndTag() for every tweet. See

Twokenize -- a tokenizer designed for Twitter text in English and some other European languages. This is the Java version. If you want the old Python

Could scrap this and use ark-regression's version -- that one has CheapStrings

Brown word clusters: features are path prefixes down the tree.