Herman Kamper

Unsupervised neural network based feature extraction using weak top-down constraints

Deep neural networks (DNNs) have become a standard component in supervised ASR, used in both data-driven feature extraction and acoustic modelling. Supervision is typically obtained from a forced alignment that provides phone class targets, requiring transcriptions and pronunciations. We propose a novel unsupervised DNN-based feature extractor that can be trained without these resources in zero-resource settings. Using unsupervised term discovery, we find pairs of isolated word examples of the same unknown type; these provide weak top-down supervision. For each pair, dynamic programming is used to align the feature frames of the two words. Matching frames are presented as input-output pairs to a deep autoencoder (AE) neural network. Using this AE as feature extractor in a word discrimination task, we achieve 64% relative improvement over a previous state-of-the-art system, 57% improvement relative to a bottom-up trained deep AE, and come to within 23% of a supervised system.

Deep convolutional acoustic word embeddings using word-pair side information

Recent studies have been revisiting whole words as the basic modelling unit in speech recognition and query applications, instead of phonetic units. Such whole-word segmental systems rely on a function that maps a variable-length speech segment to a vector in a fixed-dimensional space; the resulting acoustic word embeddings need to allow for accurate discrimination between different word types, directly in the embedding space. We compare several old and new approaches in a word discrimination task. Our best approach uses side information in the form of known word pairs to train a Siamese convolutional neural network (CNN): a pair of tied networks that take two speech segments as input and produce their embeddings, trained with a hinge loss that separates same-word pairs and different-word pairs by some margin. A word classifier CNN performs similarly, but requires much stronger supervision. Both types of CNNs yield large improvements over the best previously published results on the word discrimination task.

Unsupervised word segmentation and lexicon discovery using acoustic word embeddings

In settings where only unlabeled speech data is available, speech technology needs to be developed without transcriptions, pronunciation dictionaries, or language modelling text. A similar problem is faced when modeling infant language acquisition. In these cases, categorical linguistic structure needs to be discovered directly from speech audio. We present a novel unsupervised Bayesian model that segments unlabeled speech and clusters the segments into hypothesized word groupings. The result is a complete unsupervised tokenization of the input speech in terms of discovered word types. In our approach, a potential word segment (of arbitrary length) is embedded in a fixed-dimensional acoustic vector space. The model, implemented as a Gibbs sampler, then builds a whole-word acoustic model in this space while jointly performing segmentation. We report word error rates in a small-vocabulary connected digit recognition task by mapping the unsupervised decoded output to ground truth transcriptions. The model achieves around 20% error rate, outperforming a previous HMM-based system by about 10% absolute. Moreover, in contrast to the baseline, our model does not require a pre-specified vocabulary size.

A segmental framework for fully-unsupervised large-vocabulary speech recognition

Zero-resource speech technology is a growing research area that aims to develop methods for speech processing in the absence of transcriptions, lexicons, or language modelling text. Early term discovery systems focused on identifying isolated recurring patterns in a corpus, while more recent full-coverage systems attempt to completely segment and cluster the audio into word-like units---effectively performing unsupervised speech recognition. This article presents the first attempt we are aware of to apply such a system to large-vocabulary multi-speaker data. Our system uses a Bayesian modelling framework with segmental word representations: each word segment is represented as a fixed-dimensional acoustic embedding obtained by mapping the sequence of feature frames to a single embedding vector. We compare our system on English and Xitsonga datasets to state-of-the-art baselines, using a variety of measures including word error rate (obtained by mapping the unsupervised output to ground truth transcriptions). Very high word error rates are reported---in the order of 70--80% for speaker-dependent and 80--95% for speaker-independent systems---highlighting the difficulty of this task. Nevertheless, in terms of cluster quality and word segmentation metrics, we show that by imposing a consistent top-down segmentation while also using bottom-up knowledge from detected syllable boundaries, both single-speaker and multi-speaker versions of our system outperform a purely bottom-up single-speaker syllable-based approach. We also show that the discovered clusters can be made less speaker- and gender-specific by using an unsupervised autoencoder-like feature extractor to learn better frame-level features (prior to embedding). Our systems discovered clusters are still less pure than those of unsupervised term discovery systems, but provide far greater coverage.

Unsupervised lexical clustering of speech segments using fixed-dimensional acoustic embeddings

Unsupervised speech processing methods are essential for applications ranging from zero-resource speech technology to modelling child language acquisition. One challenging problem is discovering the word inventory of the language: the lexicon. Lexical clustering is the task of grouping unlabelled acoustic word tokens according to type. We propose a novel lexical clustering model: variable-length word segments are embedded in a fixed-dimensional acoustic space in which clustering is then performed. We evaluate several clustering algorithms and find that the best methods produce clusters with wide variation in sizes, as observed in natural language. The best probabilistic approach is an infinite Gaussian mixture model (IGMM), which automatically chooses the number of clusters. Performance is comparable to that of non-probabilistic Chinese Whispers and average-linkage hierarchical clustering. We conclude that IGMM clustering of fixed-dimensional embeddings holds promise as the lexical clustering component in unsupervised speech processing systems.

Multi-accent acoustic modelling of South African English

Although English is spoken throughout South Africa it is most often used as a second or third language, resulting in several prevalent accents within the same population. When dealing with multiple accents in this under-resourced environment, automatic speech recognition (ASR) is complicated by the need to compile multiple, accent-specific speech corpora. We investigate how best to combine speech data from five South African accents of English in order to improve overall speech recognition performance. Three acoustic modelling approaches are considered: separate accent-specific models, accent-independent models obtained by pooling training data across accents, and multi-accent models. The latter approach extends the decision-tree clustering process normally used to construct tied-state hidden Markov models (HMMs) by allowing questions relating to accent. We find that multi-accent modelling outperforms accent-specific and accent-independent modelling in both phone and word recognition experiments, and that these improvements are statistically significant. Furthermore, we find that the relative merits of the accent-independent and accent-specific approaches depend on the particular accents involved. Multi-accent modelling therefore offers a mechanism by which speech recognition performance can be optimised automatically, and for hard decisions regarding which data to pool and which to separate to be avoided.

Capitalising on North American speech resources for the development of a South African English large vocabulary speech recognition system

South African English is currently considered an under-resourced variety of English. Extensive speech resources are, however, available for North American (US) English. In this paper we consider the use of these US resources in the development of a South African large vocabulary speech recognition system. Specifically we consider two research questions. Firstly, we determine the performance penalties that are incurred when using US instead of South African language models, pronunciation dictionaries and acoustic models. Secondly, we determine whether US acoustic and language modelling data can be used in addition to the much more limited South African resources to improve speech recognition performance. In the first case we find that using a US pronunciation dictionary or a US language model in a South African system results in fairly small penalties. However, a substantial penalty is incurred when using a US acoustic model. In the second investigation we find that small but consistent improvements over a baseline South African system can be obtained by the additional use of US acoustic data. Larger improvements are obtained when complementing the South African language modelling data with US and/or UK material. We conclude that, when developing resources for an under-resourced variety of English, the compilation of acoustic data should be prioritised, language modelling data has a weaker effect on performance and the pronunciation dictionary the smallest.

Weakly supervised spoken term discovery using cross-lingual side information

Recent work on unsupervised term discovery (UTD) aims to identify and cluster repeated word-like units from audio alone. These systems are promising for some very low-resource languages where transcribed audio is unavailable, or where no written form of the language exists. However, in some cases it may still be feasible (e.g., through crowdsourcing) to obtain (possibly noisy) text translations of the audio. If so, this information could be used as a source of side information to improve UTD. Here, we present a simple method for rescoring the output of a UTD system using text translations, and test it on a corpus of Spanish audio with English translations. We show that it greatly improves the average precision of the results over a wide range of system configurations and data preprocessing methods.