Giter VIP home page Giter VIP logo

multitask-ft-fsmt's Introduction

Multi-Task Neural Models for FT and FSMT

Multi-Task Neural Models for Translating Between Styles Within and Across Languages:

  • Formality Transfer (FT)
  • Formality-Sensitive Machine Translation (FSMT)
  • FSMT with Synthetic Supervision

This repository contains implementations for

Usage Instructions

Set-Up

Read and follow the instructions in setup.sh to obtain the necessary software and data.

Machine Translation Models


Model
 
WMT
Informal
BLEU
WMT
Formal
BLEU
WMT

LeD
WMT

PoD
MSLT
Informal
BLEU
MSLT
Formal
BLEU
MSLT

LeD
MSLT

PoD
NMT 28.63 28.63 0 0 47.83 47.83 0 0
NMT DS-Tag 28.24 28.95 9.27 6.44 47.60 47.24 8.18 1.10
Multi-Task 27.75 28.39 10.89 7.76 47.55 45.08 11.97 1.41
Multi-Task DS-Tag 27.65 29.12 11.51 8.35 47.46 46.66 10.29 1.54
Online Target Inference 27.70 28.53 10.97 7.25 46.64 43.23 12.40 1.63
Online Style Inference 26.67 28.65 14.53 12.58 45.46 44.16 14.52 2.19
NMT
> bash scripts/main.sh -m 1 -t translation -y none
NMT DS-Tag
> bash scripts/main.sh -m 1 -t translation-unk-ds
Multi-Task
> bash scripts/main.sh -m 1 -t translation-unk-transfer
Multi-Task DS-Tag
> bash scripts/main.sh -m 1 -t translation-unk-transfer-ds
Online Target Inference
> bash scripts/main.sh -m 1 -t translation-unk-transfer -i oti
Online Style Inference
> bash scripts/main.sh -m 1 -t translation-unk-transfer -i osi

Formality Transfer Benchmark Models

Model TYPE (cmd) Informal
->Formal
Formal
->Informal
Informal
->Informal
Formal
->Formal
None none 70.63 ± 0.23 37.00 ± 0.18 54.54 ± 0.44 58.98 ± 0.93
Tag-Src tag-src 72.16 ± 0.34 37.67 ± 0.11 66.87 ± 0.58 78.78 ± 0.37
Tag-Src-Block block-tag-src 72.00 ± 0.05 37.38 ± 0.12 65.46 ± 0.29 76.72 ± 0.39
Tag-Src-Tgt tag 72.29 ± 0.23 37.62 ± 0.37 67.81 ± 0.41 79.34 ± 0.55
Factor-Concat factor-concat 72.47 ± 0.11 37.62 ± 0.26 67.03 ± 0.36 79.80 ± 0.38
Factor-Sum factor-sum 72.43 ± 0.29 37.78 ± 0.26 67.24 ± 0.56 80.34 ± 0.46
Pred-Concat decoder-concat 72.35 ± 0.16 37.62 ± 0.13 66.69 ± 0.21 77.85 ± 0.31
Pred-Sum decoder-sum 72.02 ± 0.30 37.41 ± 0.17 66.15 ± 0.41 77.62 ± 0.28
BOS decoder-bos 72.08 ± 0.22 37.56 ± 0.13 66.40 ± 0.23 77.43 ± 0.34
Bias decoder-bias 71.58 ± 0.31 37.52 ± 0.15 63.66 ± 0.51 73.24 ± 0.55
> bash scripts/main.sh -m -5 -t transfer -y TYPE

Lexical and Positional Differences (LePoD)

See the LePoD directory.

multitask-ft-fsmt's People

Contributors

xingniu avatar

Stargazers

 avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar  avatar

Watchers

 avatar  avatar  avatar

Recommend Projects

  • React photo React

    A declarative, efficient, and flexible JavaScript library for building user interfaces.

  • Vue.js photo Vue.js

    🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.

  • Typescript photo Typescript

    TypeScript is a superset of JavaScript that compiles to clean JavaScript output.

  • TensorFlow photo TensorFlow

    An Open Source Machine Learning Framework for Everyone

  • Django photo Django

    The Web framework for perfectionists with deadlines.

  • D3 photo D3

    Bring data to life with SVG, Canvas and HTML. 📊📈🎉

Recommend Topics

  • javascript

    JavaScript (JS) is a lightweight interpreted programming language with first-class functions.

  • web

    Some thing interesting about web. New door for the world.

  • server

    A server is a program made to process requests and deliver data to clients.

  • Machine learning

    Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.

  • Game

    Some thing interesting about game, make everyone happy.

Recommend Org

  • Facebook photo Facebook

    We are working to build community through open source technology. NB: members must have two-factor auth.

  • Microsoft photo Microsoft

    Open source projects and samples from Microsoft.

  • Google photo Google

    Google ❤️ Open Source for everyone.

  • D3 photo D3

    Data-Driven Documents codes.