Skip to content

Latest commit

 

History

History
16 lines (14 loc) · 1022 Bytes

README.md

File metadata and controls

16 lines (14 loc) · 1022 Bytes
Raw

Dependency Parser [Hindi]

done as a course project of Intro to NLP | Spring 2021
In this project, a dependency parser (Hindi) for which the training dataset was given. A dependency parser identifies the syntactic dependency between words in a sentence. There are few well-developed tools for Indian languages.

data

http://ltrc.iiit.ac.in/treebank_H2014/

Preprocessing

  • coding + manual editing for processing before getting to the final data
  • on initial data, use extract_data.py, extract_head.py, extract_dependency.py in order
  • then get the indexes from arceager.py
  • put the indexes in chean.py and run extract_k_u_dependency.py

for running the model

  • Datasets to be used are in src folder, directly loaded into notebooks.
  • All experiments + model fitting + predictions + classification reports are being done in the jupyter notebooks inside. To re run, simply restart and run all. To change the parameters such as list_file (contains tokens) and input files - train and test, refer to cell 3.