Michael Picheny

Cal Peyser, Ronny Huang Andrew Rosenberg Tara N. Sainath, Michael Picheny et al.

The careful construction of audio representations has become a dominant feature in the design of approaches to many speech tasks. Increasingly, such approaches have emphasized "disentanglement", where a representation contains only parts of the speech signal relevant to transcription while discarding irrelevant information. In this paper, we construct a representation learning task based on joint modeling of ASR and TTS, and seek to learn a representation of audio that disentangles that part of the speech signal that is relevant to transcription from that part which is not. We present empirical evidence that successfully finding such a representation is tied to the randomness inherent in training. We then make the observation that these desired, disentangled solutions to the optimization problem possess unique statistical properties. Finally, we show that enforcing these properties during training improves WER by 24.5% relative on average for our joint modeling task. These observations motivate a novel approach to learning effective audio representations.

Cal Peyser, Zhong Meng, Ke Hu et al.

The last year has seen astonishing progress in text-prompted image generation premised on the idea of a cross-modal representation space in which the text and image domains are represented jointly. In ASR, this idea has found application as joint speech-text encoders that can scale to the capacities of very large parameter models by being trained on both unpaired speech and text. While these methods show promise, they have required special treatment of the sequence-length mismatch inherent in speech and text, either by up-sampling heuristics or an explicit alignment model. In this work, we offer evidence that joint speech-text encoders naturally achieve consistent representations across modalities by disregarding sequence length, and argue that consistency losses could forgive length differences and simply assume the best alignment. We show that such a loss improves downstream WER in both a large-parameter monolingual and multilingual system.

Cal Peyser, Michael Picheny, Kyunghyun Cho et al.

Unpaired text and audio injection have emerged as dominant methods for improving ASR performance in the absence of a large labeled corpus. However, little guidance exists on deploying these methods to improve production ASR systems that are trained on very large supervised corpora and with realistic requirements like a constrained model size and CPU budget, streaming capability, and a rich lattice for rescoring and for downstream NLU tasks. In this work, we compare three state-of-the-art semi-supervised methods encompassing both unpaired text and audio as well as several of their combinations in a controlled setting using joint training. We find that in our setting these methods offer many improvements beyond raw WER, including substantial gains in tail-word WER, decoder computation during inference, and lattice density.

Cal Peyser, Ronny Huang, Tara Sainath et al.

Dual learning is a paradigm for semi-supervised machine learning that seeks to leverage unsupervised data by solving two opposite tasks at once. In this scheme, each model is used to generate pseudo-labels for unlabeled examples that are used to train the other model. Dual learning has seen some use in speech processing by pairing ASR and TTS as dual tasks. However, these results mostly address only the case of using unpaired examples to compensate for very small supervised datasets, and mostly on large, non-streaming models. Dual learning has not yet been proven effective for using unsupervised data to improve realistic on-device streaming models that are already trained on large supervised corpora. We provide this missing piece though an analysis of an on-device-sized streaming conformer trained on the entirety of Librispeech, showing relative WER improvements of 10.7%/5.2% without an LM and 11.7%/16.4% with an LM.

Jeffrey Tumminia, Amanda Kuznecov, Sophia Tsilerides et al.

United States Courts make audio recordings of oral arguments available as public record, but these recordings rarely include speaker annotations. This paper addresses the Speech Audio Diarization problem, answering the question of "Who spoke when?" in the domain of judicial oral argument proceedings. We present a workflow for diarizing the speech of judges using audio recordings of oral arguments, a process we call Reference-Dependent Speaker Verification. We utilize a speech embedding network trained with the Generalized End-to-End Loss to encode speech into d-vectors and a pre-defined reference audio library based on annotated data. We find that by encoding reference audio for speakers and full arguments and computing similarity scores we achieve a 13.8% Diarization Error Rate for speakers covered by the reference audio library on a held-out test set. We evaluate our method on the Supreme Court of the United States oral arguments, accessed through the Oyez Project, and outline future work for diarizing legal proceedings. A code repository for this research is available at github.com/JeffT13/rd-diarization

Chunxi Liu, Michael Picheny, Leda Sarı et al.

It is well known that many machine learning systems demonstrate bias towards specific groups of individuals. This problem has been studied extensively in the Facial Recognition area, but much less so in Automatic Speech Recognition (ASR). This paper presents initial Speech Recognition results on "Casual Conversations" -- a publicly released 846 hour corpus designed to help researchers evaluate their computer vision and audio models for accuracy across a diverse set of metadata, including age, gender, and skin tone. The entire corpus has been manually transcribed, allowing for detailed ASR evaluations across these metadata. Multiple ASR models are evaluated, including models trained on LibriSpeech, 14,000 hour transcribed, and over 2 million hour untranscribed social media videos. Significant differences in word error rate across gender and skin tone are observed at times for all models. We are releasing human transcripts from the Casual Conversations dataset to encourage the community to develop a variety of techniques to reduce these statistical biases.

Andrew Rouditchenko, Angie Boggust, David Harwath et al.

In this paper, we explore self-supervised audio-visual models that learn from instructional videos. Prior work has shown that these models can relate spoken words and sounds to visual content after training on a large-scale dataset of videos, but they were only trained and evaluated on videos in English. To learn multilingual audio-visual representations, we propose a cascaded approach that leverages a model trained on English videos and applies it to audio-visual data in other languages, such as Japanese videos. With our cascaded approach, we show an improvement in retrieval performance of nearly 10x compared to training on the Japanese videos solely. We also apply the model trained on English videos to Japanese and Hindi spoken captions of images, achieving state-of-the-art performance.

Brian Chen, Andrew Rouditchenko, Kevin Duarte et al.

Multimodal self-supervised learning is getting more and more attention as it allows not only to train large networks without human supervision but also to search and retrieve data across various modalities. In this context, this paper proposes a self-supervised training framework that learns a common multimodal embedding space that, in addition to sharing representations across different modalities, enforces a grouping of semantically similar instances. To this end, we extend the concept of instance-level contrastive learning with a multimodal clustering step in the training pipeline to capture semantic similarities across modalities. The resulting embedding space enables retrieval of samples across all modalities, even from unseen datasets and different domains. To evaluate our approach, we train our model on the HowTo100M dataset and evaluate its zero-shot retrieval capabilities in two challenging domains, namely text-to-video retrieval, and temporal action localization, showing state-of-the-art results on four different datasets.

Xiangyun Chu, Elizabeth Combs, Amber Wang et al.

While improvements have been made in automatic speech recognition performance over the last several years, machines continue to have significantly lower performance on accented speech than humans. In addition, the most significant improvements on accented speech primarily arise by overwhelming the problem with hundreds or even thousands of hours of data. Humans typically require much less data to adapt to a new accent. This paper explores methods that are inspired by human perception to evaluate possible performance improvements for recognition of accented speech, with a specific focus on recognizing speech with a novel accent relative to that of the training data. Our experiments are run on small, accessible datasets that are available to the research community. We explore four methodologies: pre-exposure to multiple accents, grapheme and phoneme-based pronunciations, dropout (to improve generalization to a novel accent), and the identification of the layers in the neural network that can specifically be associated with accent modeling. Our results indicate that methods based on human perception are promising in reducing WER and understanding how accented speech is modeled in neural networks for novel accents.

Sujeong Cha, Wangrui Hou, Hyun Jung et al.

A major focus of recent research in spoken language understanding (SLU) has been on the end-to-end approach where a single model can predict intents directly from speech inputs without intermediate transcripts. However, this approach presents some challenges. First, since speech can be considered as personally identifiable information, in some cases only automatic speech recognition (ASR) transcripts are accessible. Second, intent-labeled speech data is scarce. To address the first challenge, we propose a novel system that can predict intents from flexible types of inputs: speech, ASR transcripts, or both. We demonstrate strong performance for either modality separately, and when both speech and ASR transcripts are available, through system combination, we achieve better results than using a single input modality. To address the second challenge, we leverage a semantically robust pre-trained BERT model and adopt a cross-modal system that co-trains text embeddings and acoustic embeddings in a shared latent space. We further enhance this system by utilizing an acoustic module pre-trained on LibriSpeech and domain-adapting the text module on our target datasets. Our experiments show significant advantages for these pre-training and fine-tuning strategies, resulting in a system that achieves competitive intent-classification performance on Snips SLU and Fluent Speech Commands datasets.

Yinghui Huang, Hong-Kwang Kuo, Samuel Thomas et al.

Training an end-to-end (E2E) neural network speech-to-intent (S2I) system that directly extracts intents from speech requires large amounts of intent-labeled speech data, which is time consuming and expensive to collect. Initializing the S2I model with an ASR model trained on copious speech data can alleviate data sparsity. In this paper, we attempt to leverage NLU text resources. We implemented a CTC-based S2I system that matches the performance of a state-of-the-art, traditional cascaded SLU system. We performed controlled experiments with varying amounts of speech and text training data. When only a tenth of the original data is available, intent classification accuracy degrades by 7.6% absolute. Assuming we have additional text-to-intent data (without speech) available, we investigated two techniques to improve the S2I system: (1) transfer learning, in which acoustic embeddings for intent classification are tied to fine-tuned BERT text embeddings; and (2) data augmentation, in which the text-to-intent data is converted into speech-to-intent data using a multi-speaker text-to-speech system. The proposed approaches recover 80% of performance lost due to using limited intent-labeled speech.

Andrew Rouditchenko, Angie Boggust, David Harwath et al.

Current methods for learning visually grounded language from videos often rely on text annotation, such as human generated captions or machine generated automatic speech recognition (ASR) transcripts. In this work, we introduce the Audio-Video Language Network (AVLnet), a self-supervised network that learns a shared audio-visual embedding space directly from raw video inputs. To circumvent the need for text annotation, we learn audio-visual representations from randomly segmented video clips and their raw audio waveforms. We train AVLnet on HowTo100M, a large corpus of publicly available instructional videos, and evaluate on image retrieval and video retrieval tasks, achieving state-of-the-art performance. We perform analysis of AVLnet's learned representations, showing our model utilizes speech and natural sounds to learn audio-visual concepts. Further, we propose a tri-modal model that jointly processes raw audio, video, and text captions from videos to learn a multi-modal semantic embedding space useful for text-video retrieval. Our code, data, and trained models will be released at avlnet.csail.mit.edu

Xiaodong Cui, Wei Zhang, Ulrich Finkler et al.

The past decade has witnessed great progress in Automatic Speech Recognition (ASR) due to advances in deep learning. The improvements in performance can be attributed to both improved models and large-scale training data. Key to training such models is the employment of efficient distributed learning techniques. In this article, we provide an overview of distributed training techniques for deep neural network acoustic models for ASR. Starting with the fundamentals of data parallel stochastic gradient descent (SGD) and ASR acoustic modeling, we will investigate various distributed training strategies and their realizations in high performance computing (HPC) environments with an emphasis on striking the balance between communication and computation. Experiments are carried out on a popular public benchmark to study the convergence, speedup and recognition performance of the investigated strategies.

Wei Zhang, Xiaodong Cui, Abdullah Kayi et al.

Decentralized Parallel SGD (D-PSGD) and its asynchronous variant Asynchronous Parallel SGD (AD-PSGD) is a family of distributed learning algorithms that have been demonstrated to perform well for large-scale deep learning tasks. One drawback of (A)D-PSGD is that the spectral gap of the mixing matrix decreases when the number of learners in the system increases, which hampers convergence. In this paper, we investigate techniques to accelerate (A)D-PSGD based training by improving the spectral gap while minimizing the communication cost. We demonstrate the effectiveness of our proposed techniques by running experiments on the 2000-hour Switchboard speech recognition task and the ImageNet computer vision task. On an IBM P9 supercomputer, our system is able to train an LSTM acoustic model in 2.28 hours with 7.5% WER on the Hub5-2000 Switchboard (SWB) test set and 13.3% WER on the CallHome (CH) test set using 64 V100 GPUs and in 1.98 hours with 7.7% WER on SWB and 13.3% WER on CH using 128 V100 GPUs, the fastest training time reported to date.

Zakaria Aldeneh, Mimansa Jaiswal, Michael Picheny et al.

Bipolar disorder, a severe chronic mental illness characterized by pathological mood swings from depression to mania, requires ongoing symptom severity tracking to both guide and measure treatments that are critical for maintaining long-term health. Mental health professionals assess symptom severity through semi-structured clinical interviews. During these interviews, they observe their patients' spoken behaviors, including both what the patients say and how they say it. In this work, we move beyond acoustic and lexical information, investigating how higher-level interactive patterns also change during mood episodes. We then perform a secondary analysis, asking if these interactive patterns, measured through dialogue features, can be used in conjunction with acoustic features to automatically recognize mood episodes. Our results show that it is beneficial to consider dialogue features when analyzing and building automated systems for predicting and monitoring mood.

Michael Picheny, Zóltan Tüske, Brian Kingsbury et al.

There has been huge progress in speech recognition over the last several years. Tasks once thought extremely difficult, such as SWITCHBOARD, now approach levels of human performance. The MALACH corpus (LDC catalog LDC2012S05), a 375-Hour subset of a large archive of Holocaust testimonies collected by the Survivors of the Shoah Visual History Foundation, presents significant challenges to the speech community. The collection consists of unconstrained, natural speech filled with disfluencies, heavy accents, age-related coarticulations, un-cued speaker and language switching, and emotional speech - all still open problems for speech recognition systems. Transcription is challenging even for skilled human annotators. This paper proposes that the community place focus on the MALACH corpus to develop speech recognition systems that are more robust with respect to accents, disfluencies and emotional speech. To reduce the barrier for entry, a lexicon and training and testing setups have been created and baseline results using current deep learning technologies are presented. The metadata has just been released by LDC (LDC2019S11). It is hoped that this resource will enable the community to build on top of these baselines so that the extremely important information in these and related oral histories becomes accessible to a wider audience.

Khoi-Nguyen C. Mac, Xiaodong Cui, Wei Zhang et al.

In automatic speech recognition (ASR), wideband (WB) and narrowband (NB) speech signals with different sampling rates typically use separate acoustic models. Therefore mixed-bandwidth (MB) acoustic modeling has important practical values for ASR system deployment. In this paper, we extensively investigate large-scale MB deep neural network acoustic modeling for ASR using 1,150 hours of WB data and 2,300 hours of NB data. We study various MB strategies including downsampling, upsampling and bandwidth extension for MB acoustic modeling and evaluate their performance on 8 diverse WB and NB test sets from various application domains. To deal with the large amounts of training data, distributed training is carried out on multiple GPUs using synchronous data parallelism.

Xiaodong Cui, Michael Picheny

Evolutionary stochastic gradient descent (ESGD) was proposed as a population-based approach that combines the merits of gradient-aware and gradient-free optimization algorithms for superior overall optimization performance. In this paper we investigate a variant of ESGD for optimization of acoustic models for automatic speech recognition (ASR). In this variant, we assume the existence of a well-trained acoustic model and use it as an anchor in the parent population whose good "gene" will propagate in the evolution to the offsprings. We propose an ESGD algorithm leveraging the anchor models such that it guarantees the best fitness of the population will never degrade from the anchor model. Experiments on 50-hour Broadcast News (BN50) and 300-hour Switchboard (SWB300) show that the ESGD with anchors can further improve the loss and ASR performance over the existing well-trained acoustic models.

Wei Zhang, Xiaodong Cui, Ulrich Finkler et al.

Modern Automatic Speech Recognition (ASR) systems rely on distributed deep learning to for quick training completion. To enable efficient distributed training, it is imperative that the training algorithms can converge with a large mini-batch size. In this work, we discovered that Asynchronous Decentralized Parallel Stochastic Gradient Descent (ADPSGD) can work with much larger batch size than commonly used Synchronous SGD (SSGD) algorithm. On commonly used public SWB-300 and SWB-2000 ASR datasets, ADPSGD can converge with a batch size 3X as large as the one used in SSGD, thus enable training at a much larger scale. Further, we proposed a Hierarchical-ADPSGD (H-ADPSGD) system in which learners on the same computing node construct a super learner via a fast allreduce implementation, and super learners deploy ADPSGD algorithm among themselves. On a 64 Nvidia V100 GPU cluster connected via a 100Gb/s Ethernet network, our system is able to train SWB-2000 to reach a 7.6% WER on the Hub5-2000 Switchboard (SWB) test-set and a 13.2% WER on the Call-home (CH) test-set in 5.2 hours. To the best of our knowledge, this is the fastest ASR training system that attains this level of model accuracy for SWB-2000 task to be ever reported in the literature.

Samuel Thomas, Masayuki Suzuki, Yinghui Huang et al.

With recent advances in deep learning, considerable attention has been given to achieving automatic speech recognition performance close to human performance on tasks like conversational telephone speech (CTS) recognition. In this paper we evaluate the usefulness of these proposed techniques on broadcast news (BN), a similar challenging task. We also perform a set of recognition measurements to understand how close the achieved automatic speech recognition results are to human performance on this task. On two publicly available BN test sets, DEV04F and RT04, our speech recognition system using LSTM and residual network based acoustic models with a combination of n-gram and neural network language models performs at 6.5% and 5.9% word error rate. By achieving new performance milestones on these test sets, our experiments show that techniques developed on other related tasks, like CTS, can be transferred to achieve similar performance. In contrast, the best measured human recognition performance on these test sets is much lower, at 3.6% and 2.8% respectively, indicating that there is still room for new techniques and improvements in this space, to reach human performance levels.

Wei Zhang, Xiaodong Cui, Ulrich Finkler et al.

In this paper, we propose and investigate a variety of distributed deep learning strategies for automatic speech recognition (ASR) and evaluate them with a state-of-the-art Long short-term memory (LSTM) acoustic model on the 2000-hour Switchboard (SWB2000), which is one of the most widely used datasets for ASR performance benchmark. We first investigate what are the proper hyper-parameters (e.g., learning rate) to enable the training with sufficiently large batch size without impairing the model accuracy. We then implement various distributed strategies, including Synchronous (SYNC), Asynchronous Decentralized Parallel SGD (ADPSGD) and the hybrid of the two HYBRID, to study their runtime/accuracy trade-off. We show that we can train the LSTM model using ADPSGD in 14 hours with 16 NVIDIA P100 GPUs to reach a 7.6% WER on the Hub5- 2000 Switchboard (SWB) test set and a 13.1% WER on the CallHome (CH) test set. Furthermore, we can train the model using HYBRID in 11.5 hours with 32 NVIDIA V100 GPUs without loss in accuracy.

Shane Settle, Kartik Audhkhasi, Karen Livescu et al.

Direct acoustics-to-word (A2W) systems for end-to-end automatic speech recognition are simpler to train, and more efficient to decode with, than sub-word systems. However, A2W systems can have difficulties at training time when data is limited, and at decoding time when recognizing words outside the training vocabulary. To address these shortcomings, we investigate the use of recently proposed acoustic and acoustically grounded word embedding techniques in A2W systems. The idea is based on treating the final pre-softmax weight matrix of an AWE recognizer as a matrix of word embedding vectors, and using an externally trained set of word embeddings to improve the quality of this matrix. In particular we introduce two ideas: (1) Enforcing similarity at training time between the external embeddings and the recognizer weights, and (2) using the word embeddings at test time for predicting out-of-vocabulary words. Our word embedding model is acoustically grounded, that is it is learned jointly with acoustic embeddings so as to encode the words' acoustic-phonetic content; and it is parametric, so that it can embed any arbitrary (potentially out-of-vocabulary) sequence of characters. We find that both techniques improve the performance of an A2W recognizer on conversational telephone speech.

Xiaodong Cui, Wei Zhang, Zoltán Tüske et al.

We propose a population-based Evolutionary Stochastic Gradient Descent (ESGD) framework for optimizing deep neural networks. ESGD combines SGD and gradient-free evolutionary algorithms as complementary algorithms in one framework in which the optimization alternates between the SGD step and evolution step to improve the average fitness of the population. With a back-off strategy in the SGD step and an elitist strategy in the evolution step, it guarantees that the best fitness in the population will never degrade. In addition, individuals in the population optimized with various SGD-based optimizers using distinct hyper-parameters in the SGD step are considered as competing species in a coevolution setting such that the complementarity of the optimizers is also taken into account. The effectiveness of ESGD is demonstrated across multiple applications including speech recognition, image recognition and language modeling, using networks with a variety of deep architectures.

Kartik Audhkhasi, Brian Kingsbury, Bhuvana Ramabhadran et al.

Direct acoustics-to-word (A2W) models in the end-to-end paradigm have received increasing attention compared to conventional sub-word based automatic speech recognition models using phones, characters, or context-dependent hidden Markov model states. This is because A2W models recognize words from speech without any decoder, pronunciation lexicon, or externally-trained language model, making training and decoding with such models simple. Prior work has shown that A2W models require orders of magnitude more training data in order to perform comparably to conventional models. Our work also showed this accuracy gap when using the English Switchboard-Fisher data set. This paper describes a recipe to train an A2W model that closes this gap and is at-par with state-of-the-art sub-word based models. We achieve a word error rate of 8.8%/13.9% on the Hub5-2000 Switchboard/CallHome test sets without any decoder or language model. We find that model initialization, training data order, and regularization have the most impact on the A2W model performance. Next, we present a joint word-character A2W model that learns to first spell the word and then recognize it. This model provides a rich output to the user instead of simple word hypotheses, making it especially useful in the case of words unseen or rarely-seen during training.

Kartik Audhkhasi, Bhuvana Ramabhadran, George Saon et al.

Recent work on end-to-end automatic speech recognition (ASR) has shown that the connectionist temporal classification (CTC) loss can be used to convert acoustics to phone or character sequences. Such systems are used with a dictionary and separately-trained Language Model (LM) to produce word sequences. However, they are not truly end-to-end in the sense of mapping acoustics directly to words without an intermediate phone representation. In this paper, we present the first results employing direct acoustics-to-word CTC models on two well-known public benchmark tasks: Switchboard and CallHome. These models do not require an LM or even a decoder at run-time and hence recognize speech with minimal complexity. However, due to the large number of word output units, CTC word models require orders of magnitude more data to train reliably compared to traditional systems. We present some techniques to mitigate this issue. Our CTC word model achieves a word error rate of 13.0%/18.8% on the Hub5-2000 Switchboard/CallHome test sets without any LM or decoder compared with 9.6%/16.0% for phone-based CTC with a 4-gram LM. We also present rescoring results on CTC word model lattices to quantify the performance benefits of a LM, and contrast the performance of word and phone CTC models.

George Saon, Gakuto Kurata, Tom Sercu et al.

One of the most difficult speech recognition tasks is accurate recognition of human to human communication. Advances in deep learning over the last few years have produced major speech recognition improvements on the representative Switchboard conversational corpus. Word error rates that just a few years ago were 14% have dropped to 8.0%, then 6.6% and most recently 5.8%, and are now believed to be within striking range of human performance. This then raises two issues - what IS human performance, and how far down can we still drive speech recognition error rates? A recent paper by Microsoft suggests that we have already achieved human performance. In trying to verify this statement, we performed an independent set of human performance measurements on two conversational tasks and found that human performance may be considerably better than what was earlier reported, giving the community a significantly harder goal to achieve. We also report on our own efforts in this area, presenting a set of acoustic and language modeling techniques that lowered the word error rate of our own English conversational telephone LVCSR system to the level of 5.5%/10.3% on the Switchboard/CallHome subsets of the Hub5 2000 evaluation, which - at least at the writing of this paper - is a new performance milestone (albeit not at what we measure to be human performance!). On the acoustic side, we use a score fusion of three models: one LSTM with multiple feature inputs, a second LSTM trained with speaker-adversarial multi-task learning and a third residual net (ResNet) with 25 convolutional layers and time-dilated convolutions. On the language modeling side, we use word and character LSTMs and convolutional WaveNet-style language models.

Avner May, Alireza Bagheri Garakani, Zhiyun Lu et al.

We study large-scale kernel methods for acoustic modeling in speech recognition and compare their performance to deep neural networks (DNNs). We perform experiments on four speech recognition datasets, including the TIMIT and Broadcast News benchmark tasks, and compare these two types of models on frame-level performance metrics (accuracy, cross-entropy), as well as on recognition metrics (word/character error rate). In order to scale kernel methods to these large datasets, we use the random Fourier feature method of Rahimi and Recht (2007). We propose two novel techniques for improving the performance of kernel acoustic models. First, in order to reduce the number of random features required by kernel models, we propose a simple but effective method for feature selection. The method is able to explore a large number of non-linear features while maintaining a compact model more efficiently than existing approaches. Second, we present a number of frame-level metrics which correlate very strongly with recognition performance when computed on the heldout set; we take advantage of these correlations by monitoring these metrics during training in order to decide when to stop learning. This technique can noticeably improve the recognition performance of both DNN and kernel models, while narrowing the gap between them. Additionally, we show that the linear bottleneck method of Sainath et al. (2013) improves the performance of our kernel models significantly, in addition to speeding up training and making the models more compact. Together, these three methods dramatically improve the performance of kernel acoustic models, making their performance comparable to DNNs on the tasks we explored.

Ewout van den Berg, Bhuvana Ramabhadran, Michael Picheny

In this work we study variance in the results of neural network training on a wide variety of configurations in automatic speech recognition. Although this variance itself is well known, this is, to the best of our knowledge, the first paper that performs an extensive empirical study on its effects in speech recognition. We view training as sampling from a distribution and show that these distributions can have a substantial variance. These results show the urgent need to rethink the way in which results in the literature are reported and interpreted.

Zhiyun Lu, Dong Guo, Alireza Bagheri Garakani et al.

We study large-scale kernel methods for acoustic modeling and compare to DNNs on performance metrics related to both acoustic modeling and recognition. Measuring perplexity and frame-level classification accuracy, kernel-based acoustic models are as effective as their DNN counterparts. However, on token-error-rates DNN models can be significantly better. We have discovered that this might be attributed to DNN's unique strength in reducing both the perplexity and the entropy of the predicted posterior probabilities. Motivated by our findings, we propose a new technique, entropy regularized perplexity, for model selection. This technique can noticeably improve the recognition performance of both types of models, and reduces the gap between them. While effective on Broadcast News, this technique could be also applicable to other tasks.

George Saon, Hong-Kwang J. Kuo, Steven Rennie et al.

We describe the latest improvements to the IBM English conversational telephone speech recognition system. Some of the techniques that were found beneficial are: maxout networks with annealed dropout rates; networks with a very large number of outputs trained on 2000 hours of data; joint modeling of partially unfolded recurrent neural networks and convolutional nets by combining the bottleneck and output layers and retraining the resulting model; and lastly, sophisticated language model rescoring with exponential and neural network LMs. These techniques result in an 8.0% word error rate on the Switchboard part of the Hub5-2000 evaluation test set which is 23% relative better than our previous best published result.

Zhiyun Lu, Avner May, Kuan Liu et al.

The computational complexity of kernel methods has often been a major barrier for applying them to large-scale learning problems. We argue that this barrier can be effectively overcome. In particular, we develop methods to scale up kernel models to successfully tackle large-scale learning problems that are so far only approachable by deep learning architectures. Based on the seminal work by Rahimi and Recht on approximating kernel functions with features derived from random projections, we advance the state-of-the-art by proposing methods that can efficiently train models with hundreds of millions of parameters, and learn optimal representations from multiple kernels. We conduct extensive empirical studies on problems from image recognition and automatic speech recognition, and show that the performance of our kernel models matches that of well-engineered deep neural nets (DNNs). To the best of our knowledge, this is the first time that a direct comparison between these two methods on large-scale problems is reported. Our kernel methods have several appealing properties: training with convex optimization, cost for training a single model comparable to DNNs, and significantly reduced total cost due to fewer hyperparameters to tune for model selection. Our contrastive study between these two very different but equally competitive models sheds light on fundamental questions such as how to learn good representations.