EMNLP Assignment 1

code structure

• Pay attention that the directory of data(named "data") should be placed in the same directory as .sh and .py files, and it should contain four files, namely sst_train.csv, sst_test.csv, yelp_train.csv, yelp_test.csv.

• The structure of working directory should be:

-main.py

-model.py

-config.py

-cleandata.py

-utils.py

-run.sh

-data

|----sst_train.csv

|----sst_test.csv

|----yelp_train.csv

|----yelp_test.csv

main.py contains the main routine of the procedure, which includes loading data, pre-processing data, training model and evaluation.

model.py contains the DIY model, and many sub-functions defined in it.

config.py contains the operation of getting options by using argparse.ArgumentParser. --dataset and --alpha could be defined by users in shell.

cleandata.py contains functions of doing data pre-processing.

utils.py defines metrics and other basic functions.

Dependency

Python 3.8.9(64-bit)

NLTK 3.5

numpy 1.20.2

Run it now!

Please run the shell to check the program by typing as follows:

• If you are MAC user, then:

bash run_mac.sh {sst, yelp}

• Otherwise:

bash run.sh {sst, yelp}

The argument(chosen dataset) for .sh file will be passed to the program. By default, it'll run on sst-5.