OpenNMT/OpenNMT-tf

Coverage Mechanism and Coverage Loss

Open

#180 opened on Jul 25, 2018

View on GitHub
 (8 comments) (5 reactions) (0 assignees)Python (1,428 stars) (394 forks)batch import
enhancementhelp wanted

Description

May I ask if there is any plan adding the coverage attention mechanism (https://arxiv.org/pdf/1601.04811.pdf) and coverage loss (https://arxiv.org/pdf/1704.04368.pdf) to the decoder, as these could potentially help alleviating the repetition problem in generation?

Or, any hints on a quick implementation? Thanks!

Contributor guide

Tech stack
pythontensorflow
Domain
machine learning
Issue type
feature
DifficultyEstimated implementation difficulty for a new contributor, from 1 for very small changes to 5 for expert-level work.
4
Estimated timeA rough time range for an experienced contributor to investigate, implement, test, and prepare a pull request.
over 1 week
Activity statusHow available the issue appears right now: fresh, active, stale, blocked, or waiting on maintainer input.
stale
ClarityHow clearly the issue explains the expected change, acceptance criteria, and next step.
mostly clear
Prerequisites
Understanding of attention mechanismsTensorFlow basicsseq2seq models
Newbie friendlinessA 1-100 score estimating how approachable this issue is for first-time contributors.
25
Research direction
Review the decoder implementation in OpenNMT tf (likely in the 'opennmt/decoders' directory) to understand the current attention mechanism, then implement coverage attention as described in See et al. (2016) and coverage loss as in Tu et al. (2017). Check the issue comments for any prior discussion on implementation details or code references.