This is the anonymous code sample for our submission titled ``Investigating Non-local Features for Neural Constituency Parsing''.
The code sample contains several key scripts in our experiments: 
    1) main.py
        -- main script to train/dev/test the parser
    2) parse_chart.py
        -- contains the char-based parser that is augmented with the pattern loss (pattern_loss in the script) and the consistency loss (compatible_loss in the script)
    3) get_pattern_constituent_pair.py
        -- to obtain non-local pattern features by processing constituency parse trees in the train set
We plan to release the full version of our code on GitHub after the anonymous period. 

Note that our code is developed on the basis of the publicly relseased code of Berkeley Neural Parser (https://github.com/nikitakit/self-attentive-parser). 
All required packages can be found in their github repo. 
