Exploring how deep neural networks form phonemic categories

By Tasha Nagamine, M. Seltzer, Nima Mesgarani. 2015

86 citations recorded by the source corpus.

Subfields: Artificial Intelligence, Representation & Self-Supervised Learning, Speech & Audio, Signal Processing

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Representation & Self-Supervised Learning

In Speech & Audio

In Signal Processing

Citations in the corpus

Analysis Methods in Neural Language Processing: A Survey (2018, neutral): “, 2018) and speech (Wu and King, 2016; Nagamine et al., 2015; Wang et al., 2017b).”
What you can cram into a single $&!#* vector: Probing sentence embeddings for linguistic properties (2018, methodology): “…of recurrent networks and related models by using visualization, input/hidden representation deletion techniques or by looking at the word-by-word behaviour of the network (e.g., Nagamine et al., 2015; Hupkes et al., 2017; Li et al., 2016; Linzen et al., 2016; Kàdàr et al., 2017; Li et al., 2017).”
Analyzing Hidden Representations in End-to-End Automatic Speech Recognition Systems (2017, background): “In contrast, feed-forward neural networks trained on phoneme recognition were shown to learn increasingly better representations at higher layers [13, 14]; such networks do not need to model the full speech recognition task, different from end-to-end models.”
Building a Regular Decision Boundary with Deep Networks (2017, background): “As in [22], we chose instead to have an empirical analysis.”
On the Role of Nonlinear Transformations in Deep Neural Network Acoustic Models (2016, background): “Despite these advances, state-of-the-art ASR systems cannot match human-level performance in a variety of speech recognition tasks [6], motivating a number of recent studies aimed at better understanding deep learning in the hope of gaining intuitions that may lead to improved models [6, 7, 8, 9, 11”
On internal language representations in deep learning: an analysis of machine translation and speech recognition (2018, background): “For instance, previous work evaluated neural network acoustic models on phone recognition using different acoustic features [243] or investigated how such models learn invariant representations [352] and encode linguistic features in different layers [251, 252].”
Analyzing analytical methods: The case of phonology in neural models of spoken language (2020, background): “As an example of studies that track responses of neurons to controled input, Nagamine et al. (2015) analyze local representations acquired from a deep model of phoneme recognition and show that both individual and groups of nodes in the trained network are selective to various phonetic features,…”
How Does BERT Answer Questions?: A Layer-Wise Analysis of Transformer Representations (2019, background): “so been work which studies models not through probing tasks but through qualitative visual analysis. Zhang and Zhu [41] offer a survey of different approaches, though limited to CNNs. Nagamine et al. [25] explore phoneme recognition in DNNs by studying single node activations in the task of speech r”