Segmental Audio Word2Vec: Representing Utterances as Sequences of Vectors with Applications in Spoken Term Detection

TL;DR

This paper introduces segmental audio Word2Vec, a method that jointly learns to segment and represent utterances as sequences of phonetic vectors, improving spoken term detection across multiple languages.

Contribution

It extends audio Word2Vec from word-level to utterance-level by jointly learning segmentation and vector representation using a segmental sequence-to-sequence autoencoder with reinforcement learning.

Findings

Significantly improved unsupervised spoken word segmentation.

Enhanced spoken term detection performance over frame-based DTW.

Effective across multiple languages including English, Czech, French, and German.

Abstract

While Word2Vec represents words (in text) as vectors carrying semantic information, audio Word2Vec was shown to be able to represent signal segments of spoken words as vectors carrying phonetic structure information. Audio Word2Vec can be trained in an unsupervised way from an unlabeled corpus, except the word boundaries are needed. In this paper, we extend audio Word2Vec from word-level to utterance-level by proposing a new segmental audio Word2Vec, in which unsupervised spoken word boundary segmentation and audio Word2Vec are jointly learned and mutually enhanced, so an utterance can be directly represented as a sequence of vectors carrying phonetic structure information. This is achieved by a segmental sequence-to-sequence autoencoder (SSAE), in which a segmentation gate trained with reinforcement learning is inserted in the encoder. Experiments on English, Czech, French and German show very good performance in both unsupervised spoken word segmentation and spoken term detection applications (significantly better than frame-based DTW).