Tag Archives: Acoustic Modeling

AudioLM: a Language Modeling Approach to Audio Generation

Generating realistic audio requires modeling information represented at different scales. For example, just as music builds complex musical phrases from individual notes, speech combines temporally local structures, such as phonemes or syllables, into words and sentences. Creating well-structured and coherent audio sequences at all these scales is a challenge that has been addressed by coupling audio with transcriptions that can guide the generative process, be it text transcripts for speech synthesis or MIDI representations for piano. However, this approach breaks when trying to model untranscribed aspects of audio, such as speaker characteristics necessary to help people with speech impairments recover their voice, or stylistic components of a piano performance.

In “AudioLM: a Language Modeling Approach to Audio Generation”, we propose a new framework for audio generation that learns to generate realistic speech and piano music by listening to audio only. Audio generated by AudioLM demonstrates long-term consistency (e.g., syntax in speech, melody in music) and high fidelity, outperforming previous systems and pushing the frontiers of audio generation with applications in speech synthesis or computer-assisted music. Following our AI Principles, we've also developed a model to identify synthetic audio generated by AudioLM.

From Text to Audio Language Models
In recent years, language models trained on very large text corpora have demonstrated their exceptional generative abilities, from open-ended dialogue to machine translation or even common-sense reasoning. They have further shown their capacity to model other signals than texts, such as natural images. The key intuition behind AudioLM is to leverage such advances in language modeling to generate audio without being trained on annotated data.

However, some challenges need to be addressed when moving from text language models to audio language models. First, one must cope with the fact that the data rate for audio is significantly higher, thus leading to much longer sequences — while a written sentence can be represented by a few dozen characters, its audio waveform typically contains hundreds of thousands of values. Second, there is a one-to-many relationship between text and audio. This means that the same sentence can be rendered by different speakers with different speaking styles, emotional content and recording conditions.

To overcome both challenges, AudioLM leverages two kinds of audio tokens. First, semantic tokens are extracted from w2v-BERT, a self-supervised audio model. These tokens capture both local dependencies (e.g., phonetics in speech, local melody in piano music) and global long-term structure (e.g., language syntax and semantic content in speech, harmony and rhythm in piano music), while heavily downsampling the audio signal to allow for modeling long sequences.

However, audio reconstructed from these tokens demonstrates poor fidelity. To overcome this limitation, in addition to semantic tokens, we rely on acoustic tokens produced by a SoundStream neural codec, which capture the details of the audio waveform (such as speaker characteristics or recording conditions) and allow for high-quality synthesis. Training a system to generate both semantic and acoustic tokens leads simultaneously to high audio quality and long-term consistency.

Training an Audio-Only Language Model
AudioLM is a pure audio model that is trained without any text or symbolic representation of music. AudioLM models an audio sequence hierarchically, from semantic tokens up to fine acoustic tokens, by chaining several Transformer models, one for each stage. Each stage is trained for the next token prediction based on past tokens, as one would train a text language model. The first stage performs this task on semantic tokens to model the high-level structure of the audio sequence.

In the second stage, we concatenate the entire semantic token sequence, along with the past coarse acoustic tokens, and feed both as conditioning to the coarse acoustic model, which then predicts the future tokens. This step models acoustic properties such as speaker characteristics in speech or timbre in music.

In the third stage, we process the coarse acoustic tokens with the fine acoustic model, which adds even more detail to the final audio. Finally, we feed acoustic tokens to the SoundStream decoder to reconstruct a waveform.

After training, one can condition AudioLM on a few seconds of audio, which enables it to generate consistent continuation. In order to showcase the general applicability of the AudioLM framework, we consider two tasks from different audio domains:

  • Speech continuation, where the model is expected to retain the speaker characteristics, prosody and recording conditions of the prompt while producing new content that is syntactically correct and semantically consistent.
  • Piano continuation, where the model is expected to generate piano music that is coherent with the prompt in terms of melody, harmony and rhythm.

In the video below, you can listen to examples where the model is asked to continue either speech or music and generate new content that was not seen during training. As you listen, note that everything you hear after the gray vertical line was generated by AudioLM and that the model has never seen any text or musical transcription, but rather just learned from raw audio. We release more samples on this webpage.

To validate our results, we asked human raters to listen to short audio clips and decide whether it is an original recording of human speech or a synthetic continuation generated by AudioLM. Based on the ratings collected, we observed a 51.2% success rate, which is not statistically significantly different from the 50% success rate achieved when assigning labels at random. This means that speech generated by AudioLM is hard to distinguish from real speech for the average listener.

Our work on AudioLM is for research purposes and we have no plans to release it more broadly at this time. In alignment with our AI Principles, we sought to understand and mitigate the possibility that people could misinterpret the short speech samples synthesized by AudioLM as real speech. For this purpose, we trained a classifier that can detect synthetic speech generated by AudioLM with very high accuracy (98.6%). This shows that despite being (almost) indistinguishable to some listeners, continuations generated by AudioLM are very easy to detect with a simple audio classifier. This is a crucial first step to help protect against the potential misuse of AudioLM, with future efforts potentially exploring technologies such as audio “watermarking”.

We introduce AudioLM, a language modeling approach to audio generation that provides both long-term coherence and high audio quality. Experiments on speech generation show not only that AudioLM can generate syntactically and semantically coherent speech without any text, but also that continuations produced by the model are almost indistinguishable from real speech by humans. Moreover, AudioLM goes well beyond speech and can model arbitrary audio signals such as piano music. This encourages the future extensions to other types of audio (e.g., multilingual speech, polyphonic music, and audio events) as well as integrating AudioLM into an encoder-decoder framework for conditioned tasks such as text-to-speech or speech-to-speech translation.

The work described here was authored by Zalán Borsos, Raphaël Marinier, Damien Vincent, Eugene Kharitonov, Olivier Pietquin, Matt Sharifi, Olivier Teboul, David Grangier, Marco Tagliasacchi and Neil Zeghidour. We are grateful for all discussions and feedback on this work that we received from our colleagues at Google.

Source: Google AI Blog

Separating Birdsong in the Wild for Classification

Birds are all around us, and just by listening, we can learn many things about our environment. Ecologists use birds to understand food systems and forest health — for example, if there are more woodpeckers in a forest, that means there’s a lot of dead wood. Because birds communicate and mark territory with songs and calls, it’s most efficient to identify them by ear. In fact, experts may identify up to 10x as many birds by ear as by sight.

In recent years, autonomous recording units (ARUs) have made it easy to capture thousands of hours of audio in forests that could be used to better understand ecosystems and identify critical habitat. However, manually reviewing the audio data is very time consuming, and experts in birdsong are rare. But an approach based on machine learning (ML) has the potential to greatly reduce the amount of expert review needed for understanding a habitat.

However, ML-based audio classification of bird species can be challenging for several reasons. For one, birds often sing over one another, especially during the “dawn chorus” when many birds are most active. Also, there aren’t clear recordings of individual birds to learn from — almost all of the available training data is recorded in noisy outdoor conditions, where other sounds from the wind, insects, and other environmental sources are often present. As a result, existing birdsong classification models struggle to identify quiet, distant and overlapping vocalizations. Additionally, some of the most common species often appear unlabeled in the background of training recordings for less common species, leading models to discount the common species. These difficult cases are very important for ecologists who want to identify endangered or invasive species using automated systems.

To address the general challenge of training ML models to automatically separate audio recordings without access to examples of isolated sounds, we recently proposed a new unsupervised method called mixture invariant training (MixIT) in our paper, “Unsupervised Sound Separation Using Mixture Invariant Training”. Moreover, in our new paper, “Improving Bird Classification with Unsupervised Sound Separation,” we use MixIT training to separate birdsong and improve species classification. We found that including the separated audio in the classification improves precision and classification quality on three independent soundscape datasets. We are also happy to announce the open-source release of the birdsong separation models on GitHub.

Bird Song Audio Separation
MixIT learns to separate single-channel recordings into multiple individual tracks, and can be trained entirely with noisy, real-world recordings. To train the separation model, we create a “mixture of mixtures” (MoM) by mixing together two real-world recordings. The separation model then learns to take the MoM apart into many channels to minimize a loss function that uses the two original real-world recordings as ground-truth references. The loss function uses these references to group the separated channels such that they can be mixed back together to recreate the two original real-world recordings. Since there’s no way to know how the different sounds in the MoM were grouped together in the original recordings, the separation model has no choice but to separate the individual sounds themselves, and thus learns to place each singing bird in a different output audio channel, also separate from wind and other background noise.

We trained a new MixIT separation model using birdsong recordings from Xeno-Canto and the Macaulay Library. We found that for separating birdsong, this new model outperformed a MixIT separation model trained on a large amount of general audio from the AudioSet dataset. We measure the quality of the separation by mixing two recordings together, applying separation, and then remixing the separated audio channels such that they reconstruct the original two recordings. We measure the signal-to-noise ratio (SNR) of the remixed audio relative to the original recordings. We found that the model trained specifically for birds achieved 6.1 decibels (dB) better SNR than the model trained on AudioSet (10.5 dB vs 4.4 dB). Subjectively, we also found many examples where the system worked incredibly well, separating very difficult to distinguish calls in real-world data.

The following videos demonstrate separation of birdsong from two different regions (Caples and the High Sierras). The videos show the mel-spectrogram of the mixed audio (a 2D image that shows the frequency content of the audio over time) and highlight the audio separated into different tracks.

High Sierras

Classifying Bird Species
To classify birds in real-world audio captured with ARUs, we first split the audio into five-second segments and then create a mel-spectrogram of each segment. We then train an EfficientNet classifier to identify bird species from the mel-spectrogram images, training on audio from Xeno-Canto and the Macaulay Library. We trained two separate classifiers, one for species in the Sierra Nevada mountains and one for upstate New York. Note that these classifiers are not trained on separated audio; that’s an area for future improvement.

We also introduced some new techniques to improve classifier training. Taxonomic training asks the classifier to provide labels for each level of the species taxonomy (genus, family, and order), which allows the model to learn groupings of species before learning the sometimes-subtle differences between similar species. Taxonomic training also allows the model to benefit from expert information about the taxonomic relationships between different species. We also found that random low-pass filtering was helpful for simulating distant sounds during training: As an audio source gets further away, the high-frequency parts fade away before the low-frequency parts. This was particularly effective for identifying species from the High Sierras region, where bird songs cover very long distances, unimpeded by trees.

Classifying Separated Audio
We found that separating audio with the new MixIT model before classification improved the classifier performance on three independent real-world datasets. The separation was particularly successful for identification of quiet and background birds, and in many cases helped with overlapping vocalizations as well.

Top: A mel-spectrogram of two birds, an American pipit (amepip) and gray-crowned rosy finch (gcrfin), from the Sierra Nevadas. The legend shows the log-probabilities for the two species given by the pre-trained classifiers. Higher values indicate more confidence, and values greater than -1.0 are usually correct classifications. Bottom: A mel-spectrogram for the automatically separated audio, with the classifier log probabilities from the separated channels. Note that the classifier only identifies the gcrfin once the audio is separated.
Top: A complex mixture with three vocalizations: A golden-crowned kinglet (gockin), mountain chickadee (mouchi), and Steller’s jay (stejay). Bottom: Separation into three channels, with classifier log probabilities for the three species. We see good visual separation of the Steller’s jay (shown by the distinct pink marks), even though the classifier isn’t sure what it is.

The separation model does have some potential limitations. Occasionally we observe over-separation, where a single song is broken into multiple channels, which can cause misclassifications. We also notice that when multiple birds are vocalizing, the most prominent song often gets a lower score after separation. This may be due to loss of environmental context or other artifacts introduced by separation that do not appear during classifier training. For now, we get the best results by running the classifier on the separated channels and the original audio, and taking the maximum score for each species. We expect that further work will allow us to reduce over-separation and find better ways to combine separation and classification. You can see and hear more examples of the full system at our GitHub repo.

Future Directions
We are currently working with partners at the California Academy of Sciences to understand how habitat and species mix changes after prescribed fires and wildfires, applying these models to ARU audio collected over many years.

We also foresee many potential applications for the unsupervised separation models in ecology, beyond just birds. For example, the separated audio can be used to create better acoustic indices, which could measure ecosystem health by tracking the total activity of birds, insects, and amphibians without identifying particular species. Similar methods could also be adapted for use underwater to track coral reef health.

We would like to thank Mary Clapp, Jack Dumbacher, and Durrell Kapan from the California Academy of Sciences for providing extensive annotated soundscapes from the Sierra Nevadas. Stefan Kahl and Holger Klinck from the Cornell Lab of Ornithology provided soundscapes from Sapsucker Woods. Training data for both the separation and classification models came from Xeno-Canto and the Macaulay Library. Finally, we would like to thank Julie Cattiau, Lauren Harrell, Matt Harvey, and our co-author, John Hershey, from the Google Bioacoustics and Sound Separation teams.

Source: Google AI Blog

LEAF: A Learnable Frontend for Audio Classification

Developing machine learning (ML) models for audio understanding has seen tremendous progress over the past several years. Leveraging the ability to learn parameters from data, the field has progressively shifted from composite, handcrafted systems to today’s deep neural classifiers that are used to recognize speech, understand music, or classify animal vocalizations such as bird calls. However, unlike computer vision models, which can learn from raw pixels, deep neural networks for audio classification are rarely trained from raw audio waveforms. Instead, they rely on pre-processed data in the form of mel filterbanks — handcrafted mel-scaled spectrograms that have been designed to replicate some aspects of the human auditory response.

Although modeling mel filterbanks for ML tasks has been historically successful, it is limited by the inherent biases of fixed features: even though using a fixed mel-scale and a logarithmic compression works well in general, we have no guarantee that they provide the best representations for the task at hand. In particular, even though matching human perception provides good inductive biases for some application domains, e.g., speech recognition or music understanding, these biases may be detrimental to domains for which imitating the human ear is not important, such as recognizing whale calls. So, in order to achieve optimal performance, the mel filterbanks should be tailored to the task of interest, a tedious process that requires an iterative effort informed by expert domain knowledge. As a consequence, standard mel filterbanks are used for most audio classification tasks in practice, even though they are suboptimal. In addition, while researchers have proposed ML systems to address these problems, such as Time-Domain Filterbanks, SincNet and Wavegram, they have yet to match the performance of traditional mel filterbanks.

In “LEAF, A Fully Learnable Frontend for Audio Classification”, accepted at ICLR 2021, we present an alternative method for crafting learnable spectrograms for audio understanding tasks. LEarnable Audio Frontend (LEAF) is a neural network that can be initialized to approximate mel filterbanks, and then be trained jointly with any audio classifier to adapt to the task at hand, while only adding a handful of parameters to the full model. We show that over a wide range of audio signals and classification tasks, including speech, music and bird songs, LEAF spectrograms improve classification performance over fixed mel filterbanks and over previously proposed learnable systems. We have implemented the code in TensorFlow 2 and released it to the community through our GitHub repository.

Mel Filterbanks: Mimicking Human Perception of Sound
The first step in the traditional approach to creating a mel filterbank is to capture the sound’s time-variability by windowing, i.e., cutting the signal into short segments with fixed duration. Then, one performs filtering, by passing the windowed segments through a bank of fixed frequency filters, that replicate the human logarithmic sensitivity to pitch. Because we are more sensitive to variations in low frequencies than high frequencies, mel filterbanks give more importance to the low-frequency range of sounds. Finally, the audio signal is compressed to mimic the ear’s logarithmic sensitivity to loudness — a sound needs to double its power for a person to perceive an increase of 3 decibels.

LEAF loosely follows this traditional approach to mel filterbank generation, but replaces each of the fixed operations (i.e., the filtering layer, windowing layer, and compression function) by a learned counterpart. The output of LEAF is a time-frequency representation (a spectrogram) similar to mel filterbanks, but fully learnable. So, for example, while a mel filterbank uses a fixed scale for pitch, LEAF learns the scale that is best suited to the task of interest. Any model that can be trained using mel filterbanks as input features, can also be trained on LEAF spectrograms.

Diagram of computation of mel filterbanks compared to LEAF spectrograms.

While LEAF can be initialized randomly, it can also be initialized in a way that approximates mel filterbanks, which have been shown to be a better starting point. Then, LEAF can be trained with any classifier to adapt to the task of interest.

Left: Mel filterbanks for a person saying “wow”. Right: LEAF’s output for the same example, after training on a dataset of speech commands.

A Parameter-Efficient Alternative to Fixed Features
A potential downside of replacing fixed features that involve no learnable parameter with a trainable system is that it can significantly increase the number of parameters to optimize. To avoid this issue, LEAF uses Gabor convolution layers that have only two parameters per filter, instead of the ~400 parameters typical of a standard convolution layer. This way, even when paired with a small classifier, such as EfficientNetB0, the LEAF model only accounts for 0.01% of the total parameters.

Top: Unconstrained convolutional filters after training for audio event classification. Bottom: LEAF filters at convergence after training for the same task.

We apply LEAF to diverse audio classification tasks, including recognizing speech commands, speaker identification, acoustic scene recognition, identifying musical instruments, and finding birdsongs. On average, LEAF outperforms both mel filterbanks and previous learnable frontends, such as Time-Domain Filterbanks, SincNet and Wavegram. In particular, LEAF achieves a 76.9% average accuracy across the different tasks, compared to 73.9% for mel filterbanks. Moreover we show that LEAF can be trained in a multi-task setting, such that a single LEAF parametrization can work well across all these tasks. Finally, when combined with a large audio classifier, LEAF reaches state-of-the-art performance on the challenging AudioSet benchmark, with a 2.74 d-prime score.

D-prime score (the higher the better) of LEAF, mel filterbanks and previously proposed learnable spectrograms on the evaluation set of AudioSet.

The scope of audio understanding tasks keeps growing, from diagnosing dementia from speech to detecting humpback whale calls from underwater microphones. Adapting mel filterbanks to every new task can require a significant amount of hand-tuning and experimentation. In this context, LEAF provides a drop-in replacement for these fixed features, that can be trained to adapt to the task of interest, with minimal task-specific adjustments. Thus, we believe that LEAF can accelerate development of models for new audio understanding tasks.

We thank our co-authors, Olivier Teboul, Félix de Chaumont-Quitry and Marco Tagliasacchi. We also thank Dick Lyon, Vincent Lostanlen, Matt Harvey, and Alex Park for helpful discussions, and Julie Thomas for helping to design figures for this post.

Source: Google AI Blog

Improving End-to-End Models For Speech Recognition

Traditional automatic speech recognition (ASR) systems, used for a variety of voice search applications at Google, are comprised of an acoustic model (AM), a pronunciation model (PM) and a language model (LM), all of which are independently trained, and often manually designed, on different datasets [1]. AMs take acoustic features and predict a set of subword units, typically context-dependent or context-independent phonemes. Next, a hand-designed lexicon (the PM) maps a sequence of phonemes produced by the acoustic model to words. Finally, the LM assigns probabilities to word sequences. Training independent components creates added complexities and is suboptimal compared to training all components jointly. Over the last several years, there has been a growing popularity in developing end-to-end systems, which attempt to learn these separate components jointly as a single system. While these end-to-end models have shown promising results in the literature [2, 3], it is not yet clear if such approaches can improve on current state-of-the-art conventional systems.

Today we are excited to share “State-of-the-art Speech Recognition With Sequence-to-Sequence Models [4],” which describes a new end-to-end model that surpasses the performance of a conventional production system [1]. We show that our end-to-end system achieves a word error rate (WER) of 5.6%, which corresponds to a 16% relative improvement over a strong conventional system which achieves a 6.7% WER. Additionally, the end-to-end model used to output the initial word hypothesis, before any hypothesis rescoring, is 18 times smaller than the conventional model, as it contains no separate LM and PM.

Our system builds on the Listen-Attend-Spell (LAS) end-to-end architecture, first presented in [2]. The LAS architecture consists of 3 components. The listener encoder component, which is similar to a standard AM, takes the a time-frequency representation of the input speech signal, x, and uses a set of neural network layers to map the input to a higher-level feature representation, henc. The output of the encoder is passed to an attender, which uses henc to learn an alignment between input features x and predicted subword units {yn, … y0}, where each subword is typically a grapheme or wordpiece. Finally, the output of the attention module is passed to the speller (i.e., decoder), similar to an LM, that produces a probability distribution over a set of hypothesized words.
Components of the LAS End-to-End Model.
All components of the LAS model are trained jointly as a single end-to-end neural network, instead of as separate modules like conventional systems, making it much simpler.
Additionally, because the LAS model is fully neural, there is no need for external, manually designed components such as finite state transducers, a lexicon, or text normalization modules. Finally, unlike conventional models, training end-to-end models does not require bootstrapping from decision trees or time alignments generated from a separate system, and can be trained given pairs of text transcripts and the corresponding acoustics.

In [4], we introduce a variety of novel structural improvements, including improving the attention vectors passed to the decoder and training with longer subword units (i.e., wordpieces). In addition, we also introduce numerous optimization improvements for training, including the use of minimum word error rate training [5]. These structural and optimization improvements are what accounts for obtaining the 16% relative improvement over the conventional model.

Another exciting potential application for this research is multi-dialect and multi-lingual systems, where the simplicity of optimizing a single neural network makes such a model very attractive. Here data for all dialects/languages can be combined to train one network, without the need for a separate AM, PM and LM for each dialect/language. We find that these models work well on 7 english dialects [6] and 9 Indian languages [7], while outperforming a model trained separately on each individual language/dialect.

While we are excited by our results, our work is not done. Currently, these models cannot process speech in real time [8, 9], which is a strong requirement for latency-sensitive applications such as voice search. In addition, these models still compare negatively to production when evaluated on live production data. Furthermore, our end-to-end model is learned on 22,000 audio-text pair utterances compared to a conventional system that is typically trained on significantly larger corpora. In addition, our proposed model is not able to learn proper spellings for rarely used words such as proper nouns, which is normally performed with a hand-designed PM. Our ongoing efforts are focused now on addressing these challenges.

This work was done as a strong collaborative effort between Google Brain and Speech teams. Contributors include Tara Sainath, Rohit Prabhavalkar, Bo Li, Kanishka Rao, Shankar Kumar, Shubham Toshniwal, Michiel Bacchiani and Johan Schalkwyk from the Speech team; as well as Yonghui Wu, Patrick Nguyen, Zhifeng Chen, Chung-cheng Chiu, Anjuli Kannan, Ron Weiss and Navdeep Jaitly from the Google Brain team. The work is described in more detail in papers [4-11]

[1] G. Pundak and T. N. Sainath, “Lower Frame Rate Neural Network Acoustic Models," in Proc. Interspeech, 2016.

[2] W. Chan, N. Jaitly, Q. V. Le, and O. Vinyals, “Listen, attend and spell,” CoRR, vol. abs/1508.01211, 2015

[3] R. Prabhavalkar, K. Rao, T. N. Sainath, B. Li, L. Johnson, and N. Jaitly, “A Comparison of Sequence-to-sequence Models for Speech Recognition,” in Proc. Interspeech, 2017.

[4] C.C. Chiu, T.N. Sainath, Y. Wu, R. Prabhavalkar, P. Nguyen, Z. Chen, A. Kannan, R.J. Weiss, K. Rao, K. Gonina, N. Jaitly, B. Li, J. Chorowski and M. Bacchiani, “State-of-the-art Speech Recognition With Sequence-to-Sequence Models,” submitted to ICASSP 2018.

[5] R. Prabhavalkar, T.N. Sainath, Y. Wu, P. Nguyen, Z. Chen, C.C. Chiu and A. Kannan, “Minimum Word Error Rate Training for Attention-based Sequence-to-Sequence Models,” submitted to ICASSP 2018.

[6] B. Li, T.N. Sainath, K. Sim, M. Bacchiani, E. Weinstein, P. Nguyen, Z. Chen, Y. Wu and K. Rao, “Multi-Dialect Speech Recognition With a Single Sequence-to-Sequence Model” submitted to ICASSP 2018.

[7] S. Toshniwal, T.N. Sainath, R.J. Weiss, B. Li, P. Moreno, E. Weinstein and K. Rao, “End-to-End Multilingual Speech Recognition using Encoder-Decoder Models”, submitted to ICASSP 2018.

[8] T.N. Sainath, C.C. Chiu, R. Prabhavalkar, A. Kannan, Y. Wu, P. Nguyen and Z. Chen, “Improving the Performance of Online Neural Transducer Models”, submitted to ICASSP 2018.

[9] D. Lawson*, C.C. Chiu*, G. Tucker*, C. Raffel, K. Swersky, N. Jaitly. “Learning Hard Alignments with Variational Inference”, submitted to ICASSP 2018.

[10] T.N. Sainath, R. Prabhavalkar, S. Kumar, S. Lee, A. Kannan, D. Rybach, V. Schogol, P. Nguyen, B. Li, Y. Wu, Z. Chen and C.C. Chiu, “No Need for a Lexicon? Evaluating the Value of the Pronunciation Lexica in End-to-End Models,” submitted to ICASSP 2018.

[11] A. Kannan, Y. Wu, P. Nguyen, T.N. Sainath, Z. Chen and R. Prabhavalkar. “An Analysis of Incorporating an External Language Model into a Sequence-to-Sequence Model,” submitted to ICASSP 2018.