Feature-based Decipherment for Large Vocabulary Machine Translation

TL;DR

This paper introduces a log-linear decipherment model that leverages orthographic similarities for large vocabulary machine translation, improving performance over existing models by using approximate inference techniques.

Contribution

It presents a novel log-linear model with orthographic features and an efficient inference method for large vocabulary decipherment tasks.

Findings

Outperforms existing generative decipherment models

Scales effectively to large vocabularies

Utilizes orthographic features for better translation accuracy

Abstract

Orthographic similarities across languages provide a strong signal for probabilistic decipherment, especially for closely related language pairs. The existing decipherment models, however, are not well-suited for exploiting these orthographic similarities. We propose a log-linear model with latent variables that incorporates orthographic similarity features. Maximum likelihood training is computationally expensive for the proposed log-linear model. To address this challenge, we perform approximate inference via MCMC sampling and contrastive divergence. Our results show that the proposed log-linear model with contrastive divergence scales to large vocabularies and outperforms the existing generative decipherment models by exploiting the orthographic features.