jinxuz/swss

Semantically Weighted Sentence Similarity

Semantically Weighted Sentence Similarity (SWSS)

The implementation of Incorporate Semantic Structures into Machine Translation Evaluation via UCCA.

Structure

• model/
  from TUPA pretrained model
• vocab/
  from TUPA pretrained model
• data/
  for data storage
• align.py
  modified scripts from UCCA
• meteor5.py
  a self-implemented Meteor (users may want to use other Meteor instead)
• process.py
  the functions to modify original scores
• uccamt.py
  the main functions (modify this to get model run on specific dataset)

Dependencies

• UCCA 1.2.3
• TUPA 1.4.1
• nltk 3.4.3
• Spacy 2.1.3 (higher version may not work) with datasets Spacy en core web md 2.1.0 & Spacy en core web lg 2.1.0

Download

TUPA BiLSTM Model

Spacy en core web md

Spacy en core web lg