Labeling Function - Githubissues

achen353 / TransformerSum

BERT-based extractive summarizer for long legal document using a divide-and-conquer approach

GNU General Public License v3.0

3 stars 0 forks source link

Closed achen353 closed 2 years ago

achen353 commented 2 years ago

LABELING FUNCTION: Verify the Labeling Function difference between TransformerSum (https://github.com/achen353/TransformerSum/blob/master/src/convert_to_extractive.py) and BillSum (https://github.com/FiscalNote/BillSum/blob/master/billsum/data_prep/label_sentences.py), adopt BillSum and make it an option in convert_to_extractive.py. Prepare the data with the labeling function

DUE: 11/20 Saturday 11:59pm

achen353 commented 2 years ago

As described in #22

There are 2 labeling functions in TransformerSum (Combination vs Greedy), both are based on established paper, while the labeling function in BillSum is more naive (also based on ROUGE score). As a result, I decided to use TransformerSum's labeling functions instead.
Both Combination Selection and Greedy Selection are available for BillSum dataset now with #22.