Knowing the particular context associated with a conversation can help improving the performance of an automatic speech recognition (ASR) system. For example, if we are provided with a list of in-context words or phrases — such as the speaker’s contacts or recent song playlists — during inference, we can bias the recognition process towards this list. There are many works addressing contextual ASR; however, there is few publicly available real benchmark for evaluation, making it difficult to compare different solutions. To this end, we provide a corpus (“ConEC”) and baselines to evaluate contextual ASR approaches, grounded on real-world applications. The ConEC corpus is based on public-domain earnings calls (ECs) and associated supplementary materials, such as presentation slides, earnings news release as well as a list of meeting participants’ names and affiliations. We demonstrate that such real contexts are noisier than artificially synthesized contexts that contain the ground truth, yet they still make great room for future improvement of contextual ASR technology

Keyword search (KWS) systems are commonly built on top of existing automatic speech recognition (ASR) systems. However, end-to-end (E2E) ASR models are not naturally equipped with word-level timing information or confidence. Existing methods for re-purposing E2E ASR systems for KWS are largely heuristic or model-specific. In this paper, we describe a general KWS pipeline, applicable to any ASR model that generates N-best lists. We extract timing information using either external word-aligners, or time-preserving weighted finite-state transducer-based decoders. We show that our light-weight, ASR-agnostic approach for confidence estimation based on N-best lists outperforms other commonly used heuristics, such as using the decoder’s softmax probability, and even a more complicated dedicated confidence estimation model (CEM). Finally, we compare our performance to hybrid ASR models, extensively evaluating the impact of word-level timing, confidence, and recall on KWS performance. Our KWS pipeline is available online1, suitable for evaluating the aforementioned ASR components as downstream tasks.

This paper introduces GigaSpeech, an evolving, multi-domain English speech recognition corpus with 10,000 hours of high quality labeled audio suitable for supervised training, and 40,000 hours of total audio suitable for semi-supervised and unsupervised training.Around 40,000 hours of transcribed audio is first collected from audiobooks, podcasts and YouTube, covering both read and spontaneous speaking styles, and a variety of topics, such as arts, science, sports, etc.A new forced alignment and segmentation pipeline is proposed to create sentence segments suitable for speech recognition training, and to filter out segments with low-quality transcription.For system training, GigaSpeech provides five subsets of different sizes, 10h, 250h, 1000h, 2500h, and 10000h.For our 10,000-hour XL training subset, we cap the word error rate at 4% during the filtering/validation stage, and for all our other smaller training subsets, we cap it at 0%.The DEV and TEST evaluation sets, on the other hand, are re-processed by professional human transcribers to ensure high transcription quality.Baseline systems are provided for popular speech recognition toolkits, namely Athena, ESPnet, Kaldi and Pika.

Despite the growing interest in unsupervised learning, extracting meaningful knowledge from unlabelled audio remains an open challenge. To take a step in this direction, we recently proposed a problem-agnostic speech encoder (PASE), that combines a convolutional encoder followed by multiple neural networks, called workers, tasked to solve self-supervised problems (i.e., ones that do not require manual annotations as ground truth). PASE was shown to capture relevant speech information, including speaker voice-print and phonemes. This paper proposes PASE+, an improved version of PASE for robust speech recognition in noisy and reverberant environments. To this end, we employ an online speech distortion module, that contaminates the input signals with a variety of random disturbances. We then propose a revised encoder that better learns short- and long-term speech dynamics with an efficient combination of recurrent and convolutional networks. Finally, we refine the set of workers used in self-supervision to encourage better cooperation. Results on TIMIT, DIRHA and CHiME-5 show that PASE+ significantly outperforms both the previous version of PASE as well as common acoustic features. Interestingly, PASE+ learns transferable representations suitable for highly mismatched acoustic conditions.

We present a speech data corpus that simulates a "dinner party" scenario taking place in an everyday home environment. The corpus was created by recording multiple groups of four Amazon employee volunteers having a natural conversation in English around a dining table. The participants were recorded by a single-channel close-talk microphone and by five far-field 7-microphone array devices positioned at different locations in the recording room. The dataset contains the audio recordings and human labeled transcripts of a total of 10 sessions with a duration between 15 and 45 minutes. The corpus was created to advance in the field of noise robust and distant speech processing and is intended to serve as a public research and benchmarking data set.

Existing acoustic models can be transferred to any language with a pronunciation lexicon (lexicon) that uses the same set of sub-word units as in training. Unfortunately such lexicons are not readily available in many low-resource languages. We bypass this requirement and create lexicons by training a grapheme-to-phoneme (G2P) transducer on a subset of words from other languages for which pronunciations are available. The subset of words is selected based on how representative it is of target language text. We find that cross-language acoustic model transfer using our selection strategy outperforms selection based on language similarity, and results in ASR performance approaching that of hand-crafted rule based lexicons in the majority of cases.

Hybrid deep neural network hidden Markov models (DNN-HMM) have achieved impressive results on large vocabulary continuous speech recognition (LVCSR) tasks. However, the recent approaches using DNN-HMM models are not explored much for text recognition. Inspired by the current work in automatic speech recognition (ASR) and machine translation, we present an open vocabulary sub-word text recognition system. The sub-word lexicon and sub-word language model (LM) helps in overcoming the challenge of recognizing out of vocabulary (OOV) words, and a time delay neural network (TDNN) and convolution neural network (CNN) based DNN-HMM optical model (OM) efficiently models the sequence dependency in the line image. We present results on 12 datasets with training data varying from 6k lines to 600k lines. The system is built for 8 languages, i.e., English, French, Arabic, Chinese, Farsi, Tamil, Russian, and Korean. We report competitive results on several commonly used handwritten and printed text datasets.

This paper explores the possibility to use grapheme-based word and sub-word models in the task of spoken term detection (STD). The usage of grapheme models eliminates the need for expert-prepared pronunciation lexicons (which are often far from complete) and/or trainable grapheme-to-phoneme (G2P) algorithms that are frequently rather inaccurate, especially for rare words (words coming from a different language). Moreover, the G2P conversion of the search terms that need to be performed on-line can substantially increase the response time of the STD system. Our results show that using various grapheme-based models, we can achieve STD performance (measured in terms of ATWV) comparable with phoneme-based models but without the additional burden of G2P conversion.

The CHiME challenge series aims to advance robust automatic speech recognition (ASR) technology by promoting research at the interface of speech and language processing, signal processing , and machine learning. This paper introduces the 5th CHiME Challenge, which considers the task of distant multi-microphone conversational ASR in real home environments. Speech material was elicited using a dinner party scenario with efforts taken to capture data that is representative of natural conversational speech and recorded by 6 Kinect microphone arrays and 4 binaural microphone pairs. The challenge features a single-array track and a multiple-array track and, for each track, distinct rankings will be produced for systems focusing on robustness with respect to distant-microphone capture vs. systems attempting to address all aspects of the task including conversational language modeling. We discuss the rationale for the challenge and provide a detailed description of the data collection procedure, the task, and the baseline systems for array synchronization, speech enhancement, and conventional and end-to-end ASR.

We present our work on improving the numerator graph for discriminative training using the lattice-free maximum mutual information (MMI) criterion. Specifically, we propose a scheme for creating unconstrained numerator graphs by removing time constraints from the baseline numerator graphs. This leads to much smaller graphs and therefore faster preparation of training supervisions. By testing the proposed un-constrained supervisions using factorized time-delay neural network (TDNN) models, we observe 0.5% to 2.6% relative improvement over the state-of-the-art word error rates on various large-vocabulary speech recognition databases.

In topic identification (topic ID) on real-world unstructured audio, an audio instance of variable topic shifts is first broken into sequential segments, and each segment is independently classified. We first present a general purpose method for topic ID on spoken segments in low-resource languages, using a cascade of universal acoustic modeling, translation lexicons to English, and English-language topic classification. Next, instead of classifying each segment independently, we demonstrate that exploring the contextual dependencies across sequential segments can provide large improvements. In particular, we propose an attention-based contextual model which is able to leverage the contexts in a selective manner. We test both our contextual and non-contextual models on four LORELEI languages, and on all but one our attention-based contextual model significantly outperforms the context-independent models.

We describe a fully Bayesian approach to grapheme-to-phoneme conversion based on the joint-sequence model (JSM). Usually, standard smoothed n-gram language models (LM, e.g. Kneser-Ney) are used with JSMs to model graphone sequences (joint grapheme-phoneme pairs). However, we take a Bayesian approach using a hierarchical Pitman-Yor-Process LM. This provides an elegant alternative to using smoothing techniques to avoid over-training. No held-out sets and complex parameter tuning is necessary, and several convergence problems encountered in the discounted Expectation-Maximization (as used in the smoothed JSMs) are avoided. Every step is modeled by weighted finite state transducers and implemented with standard operations from the OpenFST toolkit. We evaluate our model on a standard data set (CMUdict), where it gives comparable results to the previously reported smoothed JSMs in terms of phoneme-error rate while requiring a much smaller training/testing time. Most importantly, our model can be used in a Bayesian framework and for (partly) un-supervised training.

Modern topic identification (topic ID) systems for speech use automatic speech recognition (ASR) to produce speech transcripts, and perform supervised classification on such ASR outputs. However, under resource-limited conditions, the manually transcribed speech required to develop standard ASR systems can be severely limited or unavailable. In this paper, we investigate alternative unsupervised solutions to obtaining tokenizations of speech in terms of a vocabulary of automatically discovered word-like or phoneme-like units, without depending on the supervised training of ASR systems. Moreover, using automatic phoneme-like tokenizations, we demonstrate that a convolutional neural network based framework for learning spoken document representations provides competitive performance compared to a standard bag-of-words representation, as evidenced by comprehensive topic ID evaluations on both single-label and multi-label classification tasks.

Speech translation is conventionally carried out by cascading an automatic speech recognition (ASR) and a statistical machine translation (SMT) system.The hypotheses chosen for translation are based on the ASR system's acoustic and language model scores, and typically optimized for word error rate, ignoring the intended downstream use: automatic translation.In this paper, we present a coarseto-fine model that uses features from the ASR and SMT systems to optimize this coupling.We demonstrate that several standard features utilized by ASR and SMT systems can be used in such a model at the speech-translation interface, and we provide empirical results on the Fisher Spanish-English speech translation corpus.

In this paper we present an algorithm that produces pitch and probability-of-voicing estimates for use as features in automatic speech recognition systems. These features give large performance improvements on tonal languages for ASR systems, and even substantial improvements for non-tonal languages. Our method, which we are calling the Kaldi pitch tracker (because we are adding it to the Kaldi ASR toolkit), is a highly modified version of the getf0 (RAPT) algorithm. Unlike the original getf0 we do not make a hard decision whether any given frame is voiced or unvoiced; instead, we assign a pitch even to unvoiced frames while constraining the pitch trajectory to be continuous. Our algorithm also produces a quantity that can be used as a probability of voicing measure; it is based on the normalized autocorrelation measure that our pitch extractor uses. We present results on data from various languages in the BABEL project, and show a large improvement over systems without tonal features and systems where pitch and POV information was obtained from SAcC or getf0.

Recently, maxout networks have brought significant improvements to various speech recognition and computer vision tasks. In this paper we introduce two new types of generalized maxout units, which we call p-norm and soft-maxout. We investigate their performance in Large Vocabulary Continuous Speech Recognition (LVCSR) tasks in various languages with 10 hours and 60 hours of data, and find that the p-norm generalization of maxout consistently performs well. Because, in our training setup, we sometimes see instability during training when training unbounded-output nonlinearities such as these, we also present a method to control that instability. This is the “normalization layer”, which is a nonlinearity that scales down all dimensions of its input in order to stop the average squared output from exceeding one. The performance of our proposed nonlinearities are compared with maxout, rectified linear units (ReLU), tanh units, and also with a discriminatively trained SGMM/HMM system, and our p-norm units with p equal to 2 are found to perform best.

Spoken Term Detection (STD) focuses on finding instances of a particular spoken word or phrase in an audio corpus. Most STD systems have a two-step pipeline, ASR followed by search. Two approaches to search are common, Confusion Network (CN) based search and Finite State Transducer (FST) based search. In this paper, we examine combination of these two different search approaches, using the same ASR output. We find that the CN search performs better on shorter queries, and FST search performs better on longer queries. By combining the different search results from the same ASR decoding, we achieve better performance compared to either search approach on its own. We also find that this improvement is additive to the usual combination of decoder results using different modeling techniques.

The point process model (PPM) for keyword search is a wholeword parametric modeling framework based on the timing of phonetic events rather than the evolution of frame-level phonetic likelihoods. Recent progress in PPM training and decoding algorithms has yielded state-of-the-art phonetic search performance in high-resource settings, both in terms of accuracy and computational efficiency. In this paper, we consider PPM application to low-resource settings where the amount of transcribed speech is severely limited and the pronunciation dictionary is incomplete. By using (i) state-of-the-art deep neural network acoustic models to generate phonetic events and (ii) grapheme-to-phoneme conversion to generate pronunciations for out-of-vocabulary (OOV) keywords, we find the PPM system reaches state-of-the-art OOV search performance at a small computational cost. Moreover, due to their complementary methodologies, combining PPM outputs with the LVCSR baseline produces average relative ATWV improvements of 7% and 50% for in-vocabulary and OOV keywords, respectively (16% overall).

Provides an overview of a speech-to-text (STT) and keyword search (KWS) system architecture build primarily on the top of the Kaldi toolkit and expands on a few highlights. The system was developed as a part of the research efforts of the Radical team while participating in the IARPA Babel program. Our aim was to develop a general system pipeline which could be easily and rapidly deployed in any language, independently on the language script and phonological and linguistic features of the language.

We propose a simple but effective weighted finite state transducer (WFST) based framework for handling out-of-vocabulary (OOV) keywords in a speech search task. State-of-the-art large vocabulary continuous speech recognition (LVCSR) and keyword search (KWS) systems are developed for conversational telephone speech in Tagalog. Word-based and phone-based indexes are created from word lattices, the latter by using the LVCSR system's pronunciation lexicon. Pronunciations of OOV keywords are hypothesized via a standard grapheme-to-phoneme method. In-vocabulary proxies (word or phone sequences) are generated for each OOV keyword using WFST techniques that permit incorporation of a phone confusion matrix. Empirical results when searching for the Babel/NIST evaluation keywords in the Babel 10 hour development-test speech collection show that (i) searching for word proxies in the word index significantly outperforms searching for phonetic representations of OOV words in a phone index, and (ii) while phone confusion information yields minor improvement when searching a phone index, it yields up to 40% improvement in actual term weighted value when searching a word index with word proxies.

This paper quantifies the value of pronunciation lexicons in large vocabulary continuous speech recognition (LVCSR) systems that support keyword search (KWS) in low resource languages. State-of-the-art LVCSR and KWS systems are developed for conversational telephone speech in Tagalog, and the baseline lexicon is augmented via three different grapheme-to-phoneme models that yield increasing coverage of a large Tagalog word-list. It is demonstrated that while the increased lexical coverage - or reduced out-of-vocabulary (OOV) rate - leads to only modest (ca 1%-4%) improvements in word error rate, the concomitant improvements in actual term weighted value are as much as 60%. It is also shown that incorporating the augmented lexicons into the LVCSR system before indexing speech is superior to using them post facto, e.g., for approximate phonetic matching of OOV keywords in pre-indexed lattices. These results underscore the disproportionate importance of automatic lexicon augmentation for KWS in morphologically rich languages, and advocate for using them early in the LVCSR stage.

In this paper, we describe an optimized version of a Gaussian-mixture-based acoustic model likelihood evaluation algorithm for graphical processing units (GPUs). The evaluation of these likelihoods is one of the most computationally intensive parts of automatic speech recognizers, but it can be parallelized and offloaded to GPU devices. Our approach offers a significant speed-up over the recently published approaches, because it utilizes the GPU architecture in a more effective manner. All the recent implementations have been intended only for NVIDIA graphics processors, programmed either in CUDA or OpenCL GPU programming frameworks. We present results for both CUDA and OpenCL. Further, we have developed an OpenCL implementation optimized for ATI/AMD GPUs. Results suggest that even very large acoustic models can be used in real-time speech recognition engines on computers equipped with a low-end GPU or laptops. In addition, the completely asynchronous GPU management provides additional CPU resources for the decoder part of the LVCSR. The optimized implementation enables us to apply fusion techniques together with evaluating many (10 or even more) speaker-specific acoustic models. We apply this technique to a real-time parliamentary speech recognition system where the speaker changes frequently.

In this paper we present a highly optimized implementation of \nGaussian mixture acoustic model evaluation algorithm. Evaluation \nof these likelihoods is one of the most computationally \nintensive parts of automatics speech recognizers but it can be \nwell-parallelized and offloaded to GPU devices. Our approach \noffers significant speed-up compared to the recently published \napproaches, since it exploits the GPU architecture better. All \nthe recent implementations were programmed either in CUDA \nor OpenCL GPU programming frameworks. We present results \nfor both; CUDA as well as OpenCL. \nResults suggest that even very large acoustic models can \nbe utilized in real-time speech recognition engines on computers \nand laptops equipped with a low-end GPU. Optimization \nof acoustic likelihoods computation on GPU enables to use \nthe remaining GPU resources for offloading of other computeintensive \nparts of LVCSR decoder. \nOther possible use of the freed GPU resources is to evaluate \nseveral acoustic models at the same time and use fusion \ntechniques or model selection techniques to improve the quality \nof resulting conditional likelihoods under diverse conditions.

In the paper we present two techniques improving the recognition accuracy of multilayer perceptron neural networks (MLP ANN) by means of adopting Speaker Adaptive Training. The use of the MLP ANN, usually in combination with the TRAPS parametrization, includes applications in speech recognition tasks, discriminative features production for GMM-HMM and other. In the first SAT experiments, we used the VTLN as a speaker normalization technique. Moreover, we developed a novel speaker normalization technique called Minimum Error Linear Transform (MELT) that resembles the cMLLR/fMLLR method \\cite{gales96variance} with respect to the possible application either on the model or features. We tested these two methods extensively on telephone speech corpus SpeechDat-East. The results obtained in these experiments suggest that incorporation of SAT into MLP ANN training process is beneficial and depending on the setup leads to significant decrease of phoneme error rate (3% -- 8% absolute, 12% -- 25% relative).

In this paper we present three novel posterior transforms with the primary goal to achieve a high reduction of a feature vector size. The presented methods transform the posteriors to 1D or 2D space. For such a high reduction ratio the usually applied methods fail to keep the discriminative information. Contrary, the presented methods were specifically designed to retain most of the discriminative information. In our experiments, we used several different combinations of feature extraction methods nowadays commonly used, i.e. the PLP features (augmented with delta and acceleration coefficients) and two kinds of MLP-ANN features: the bottleneck (BN) and posterior estimates (PE). The experiments were designed with special attention to the assessment of possible improvements of the performance when the PLP features are combined either with the BN features or with the PE features whose dimensionality was reduced using the proposed feature transforms. The performance of the designed transforms was tested on two different speech corpora: a telephone speech SpeechDat-East corpus and multimodal Czech Audio-Visual corpus. Index Terms: speech recognition, posteriors, ANN, bottleneck

This paper presents initial experiments with the identification and automatic detection of parasitic sounds in speech signals. The main goal of this study is to identify such sounds in the source recordings for unit-selection-based speech synthesis systems and thus to avoid their unintended usage in synthesised speech. The first part of the paper describes the phonetic analysis and identification of parasitic phenomena in recordings of two Czech speakers. In the second part, experiments with the automatic detection of parasitic sounds using HMM-based and BVM classifiers are presented. The results are encouraging, especially those for glottalization phenomena.

In this paper we describe a HMM-based sign language recognition (SLR) system for isolated signs. In the first part we describe the image parametrization method producing features used for recognition. Our goal was to find the best combination of a feature space dimension reduction method and an HMM structure.

V článku prezentujeme komparativní studii nekolika dnes používaných metod pro extrakci a selekci příznaků. Porovnali jsme řadu ICA algoritmů oproti běžně používaným metodám DCT a PCA. Porovnání bylo provedeno na dvou úlohách. Jedná se jednak o úlohu verifikace řečníka a jednak o úlohu detekce řeči. Přesnost těchto systémů byla námi zvolenou metodikou významně pozitivně ovlivněna a to jen za cenu velmi mírně zvýšených nároků. Navíc, kde to bylo možné, jsme provedli stručné porovnání schopností extrahovat pro danou úlohu užitečnou informaci.