README update

[mit456]

I would like to contribute to this project, It would be great if you guide me to set/compile/run this project on my local system, build models and make use of predict.py for prediction for an unknown data.

As far as I know, I need to use Makefile to setup, so I tried running make but got some output which has some errors:

finnpos-train data/twitter.0.config data/twitter.0.train+dev.feats data/twitter.0.dev.feats models/twitter.0.model
make[1]: finnpos-train: Command not found
Makefile:35: recipe for target 'models/twitter.0.model' failed
make[1]: *** [models/twitter.0.model] Error 127

Looking forward to hearing.

[rohitsaluja22]

Hi getting the same error

What we can understand from some of the scripts is:-

1. extract_features.py extract unstructured features (without yt), + yt, from entries like in https://github.com/mpsilfve/ocrpp/blob/master/data/klk.0.dev
2. unlabel.sh removes label: i.e. replace y_t (and o/ps other than x_t related terms in unstructured features) with _ in output of step 1.
3. compile_label_candidate_map.py create confusions map (in a .lc file) with count of each confusion from entries like in https://github.com/mpsilfve/ocrpp/blob/master/data/klk.0.dev
4. compile_fst_model.py compiles the fst model for a file that takes enteries with strings like "STRUCTURED FEATURES" or "UNSTRUCTURED FEATURES"

What is missing is a file which creates the input for compile_fst_model.py (using outputs from steps 1-3)
Any help would be highly appreciated.

Thanks in advance,
Rohit