Application of Word2vec in Phoneme Recognition

TL;DR

This paper introduces a hybrid phoneme recognition system combining Word2vec embeddings with an attention-based model, improving performance on TIMIT by innovative training and data augmentation techniques.

Contribution

It proposes a novel integration of Word2vec embeddings into an end-to-end speech recognition model and a new training method to address overfitting in phoneme recognition.

Findings

Achieved 16.5% PER on TIMIT dataset.

Enhanced phoneme vector separation with Word2vec initialization.

Implemented a phoneme mapping method for data augmentation.

Abstract

In this paper, we present how to hybridize a Word2vec model and an attention-based end-to-end speech recognition model. We build a phoneme recognition system based on Listen, Attend and Spell model. And the phoneme recognition model uses a word2vec model to initialize the embedding matrix for the improvement of the performance, which can increase the distance among the phoneme vectors. At the same time, in order to solve the problem of overfitting in the 61 phoneme recognition model on TIMIT dataset, we propose a new training method. A 61-39 phoneme mapping comparison table is used to inverse map the phonemes of the dataset to generate more 61 phoneme training data. At the end of training, replace the dataset with a standard dataset for corrective training. Our model can achieve the best result under the TIMIT dataset which is 16.5% PER (Phoneme Error Rate).