A Computational Model of Unsupervised Speech Segmentation for Correspondence Learning

Abstract

In this paper, we develop a new conceptual framework for an important problem in language acquisition, the correspondence problem: the fact that a given utterance has different manifestations in the speech and articulation of different speakers and that the correspondence of these manifestations is difficult to learn. We put forward the Correspondence-by-Segmentation Hypothesis, which states that correspondence is primarily learned by first segmenting speech in an unsupervised manner and then mapping the acoustics of different speakers onto each other. We show that a rudimentary segmentation of speech can be learned in an unsupervised fashion. We then demonstrate that, using the previously learned segmentation, different instances of a word can be mapped onto each other with high accuracy when trained on utterance-label pairs for a small set of words.

Elements of Methodology of Precision Phonetic Analysis of Oral Phonograms

“…Ця аксіома правильна навіть для окремих фонем, не кажучи про масивніші мовні одиниці. За таких умов адекватну математичну модель дискретного джерела мовних повідомлень слід створювати на основі визначення фонем за виразом (5), чітко кластеризованих у параметричному просторі:…”

Unsupervised optimal phoneme segmentation: theory and experimental evaluation

“…Roughly speaking, this is because a Gaussian distribution is the one with the largest entropy when mean and variance are given, and Gaussian variables are the most difficult to code.]. We use (8) to define the following RD objective function of X under segmentation S…”

“…ISA can be related to the scenario of infants’ speech acquisition. Psychological researches indicate that infants do not have acoustic models of the phonemes of their native languages, but they have the ability to discriminate sounds [8]. This discriminant ability resembles the metric we used for segmentation, which may enable infants to preliminarily segment speech signals.…”

“…This paper studies the unsupervised segmentation problem without knowledge on linguistic contents and acoustic models. The unsupervised segmentation is similar to the phenomenon that infants perceive speech [8][9][10]. Infants have no language and acoustic models at the beginning, but they have to segment speech into words or syllables through language acquisition.…”

“…Infants have no language and acoustic models at the beginning, but they have to segment speech into words or syllables through language acquisition. Recent studies indicate that infants use computational strategies to detect the statistical and prosodic patterns in speech [8].…”