/VMWE-Identification

This is the code and data for the system SHOMA participated in the Parseme shared task on identification of verbal MWEs.

Primary LanguagePython

VMWE-Identification

This repository contains the code and documentation for several different neural architectures to identify Verbal Multiword Expressions.

There are three main approaches each in a directory with their code and documentation:

  1. SHOMA: a ConvNet + LSTM (+ CRF) neural network architecture that participated in Parseme 2018 shared task on automatic identification of verbal multiword expressions - edition 1.1.
  2. MTL: a multi-task learning system
  3. TRL: a cross-lingual transfer learning system

Requirements

  • Python 3
  • keras with a tensorflow backend
  • Gensim