Sequence-to-Sequence Neural Net Models for Grapheme-to-Phoneme Conversion

TL;DR

This paper investigates sequence-to-sequence neural models for grapheme-to-phoneme conversion, demonstrating that bi-directional LSTM models can outperform traditional methods in this small-vocabulary task.

Contribution

It introduces the application of bi-directional LSTM neural networks with alignment information to improve grapheme-to-phoneme conversion accuracy.

Findings

Bi-directional LSTM models outperform traditional methods.

Neural models rival state-of-the-art performance.

Alignment information enhances model accuracy.

Abstract

Sequence-to-sequence translation methods based on generation with a side-conditioned language model have recently shown promising results in several tasks. In machine translation, models conditioned on source side words have been used to produce target-language text, and in image captioning, models conditioned images have been used to generate caption text. Past work with this approach has focused on large vocabulary tasks, and measured quality in terms of BLEU. In this paper, we explore the applicability of such models to the qualitatively different grapheme-to-phoneme task. Here, the input and output side vocabularies are small, plain n-gram models do well, and credit is only given when the output is exactly correct. We find that the simple side-conditioned generation approach is able to rival the state-of-the-art, and we are able to significantly advance the stat-of-the-art with bi-directional long short-term memory (LSTM) neural networks that use the same alignment information that is used in conventional approaches.