Coder Social home page Coder Social logo

masterthesis's Introduction

Cross-lingual Word Embeddings beyond Zero-shot Machine Translation

This is the thesis work for the Language Technology Programm in Uppsala University.

Link to Diva (Where the thesis was officially archived)

There is also a folder called extended_abstract, which contains this SLTC 2020 paper of the same title.

Abstract

Zero-shot translation is a transfer learning setup that refers to the ability of neural machine translation to generalize translation information into unseen language pairs. It provides an appealing solution to the lack of available materials for low-resource languages by transferring knowledge from high-resource languages.

So far, zero-shot translation mainly focuses on unseen language pairs whose individual component is still known to the system. There are fewer reports on transfer learning in machine translation being carried out on completely unknown test languages. This thesis pushes the boundary of zero-shot translation and explores the possibility of transferring learning from training languages to unknown test languages in a multilingual Neural Machine Translation (NMT) system.

Based on the fact that zero-shot translation systems primarily learn language invariant features, we use cross-lingual word embeddings as the only knowledge source since they are good at capturing the semantic similarity of words from different languages in the same vector space. By conducting experiments on an encoder-decoder multilingual NMT model with an attention module, we have examined the relationship of language similarity and the transferability of unseen languages.

We hypothesize that our multilingual NMT model with cross-lingual word embeddings should transfer reasonably even to completely unknown languages. However, we observe little transferability from the training languages to unseen test languages due to the transformed output vector space. Such minor transferability only happens between highly-related languages with a large number of shared vocabularies.

masterthesis's People

Contributors

chenshifei avatar

Watchers

 avatar  avatar

Recommend Projects

  • React photo React

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

  • Vue.js photo Vue.js

    ๐Ÿ–– 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. ๐Ÿ“Š๐Ÿ“ˆ๐ŸŽ‰

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 โค๏ธ Open Source for everyone.

  • D3 photo D3

    Data-Driven Documents codes.