Variable-order CRFs

Install

We assume that you are running the Anaconda Python distribution.

Type the following in your terminal and you should be good to go!

$ make
$ make install
$ make test

To grab some POS tagging data run, e.g., English UD data,

$ make data/UD/English

Here is an example invocation for training a model

$ python vocrf/pos/tagfail.py --tag-type upos --lang English \
    --C .01 --budget 1500 \
    --inner-iterations 1 --outer-iterations 2 \
    --initial-order 0 --max-order 0 --context-count 5 \
    --dump /tmp/foo

(To quickly test the model on some real data, pass in the --quick flag, which will run on a small subset of data.)

Publications

Tim Vieira*, Ryan Cotterell*, and Jason Eisner. Speed-Accuracy Tradeoffs in Tagging with Variable-Order CRFs and Structured Sparsity. EMNLP 2016.
Tim Vieira*, Ryan Cotterell*, and Jason Eisner. Forward-Backward with Failure Arcs: Faster Inference for Variable-Order Conditional Random Fields. In preparation.

timvieira/vocrf

Variable-order CRFs