2019 CharacterLevelLanguageModelingw

From GM-RKB
Jump to navigation Jump to search

Subject Headings: Transformer-based Character-Level Language Model, Character-Level Language Modeling.

Notes

Cited By

2019

2019b

Quotes

Abstract

LSTMs and other RNN variants have shown strong performance on character-level language modeling. These models are typically trained using truncated backpropagation through time, and it is common to assume that their success stems from their ability to remember long-term contexts. In this paper, we show that a deep (64-layer) transformer model with fixed context outperforms RNN variants by a large margin, achieving state of the art on two popular benchmarks: 1.13 bits per character on text8 and 1.06 on enwik8. To get good results at this depth, we show that it is important to add auxiliary losses, both at intermediate network layers and intermediate sequence positions.

Introduction

Character-level modeling of natural language text is challenging, for several reasons. First, the model must learn a large vocabulary of words “from scratch”. Second, natural text exhibits dependencies over long distances of hundreds or thousands of time steps. Third, character sequences are longer than word sequences and thus require significantly more steps of computation.

In recent years, strong character-level language models typically follow a common template (Mikolov et al. 2010; 2011; Sundermeyer, Schl¨uter, and Ney 2012). A recurrent neural net (RNN) is trained over mini-batches of text sequences, using a relatively short sequence length (e.g. 200 tokens). To capture context longer than the batch sequence length, training batches are provided in sequential order, and the hidden states from the previous batch are passed forward to the current batch. This procedure is known as “truncated backpropagation through time” (TBTT), because the gradient computation doesn’t proceed further than a single batch (Werbos 1990). A range of methods have arisen for unbiasing and improving TBTT (Tallec and Ollivier 2017; Ke et al. 2017).

While this technique gets good results, it adds complexity to the training procedure, and recent work suggests that models trained in this manner don’t actually make “strong” use of long-term context. For example Khandelwal et al. (2018) find that a word-based LSTM language model only effectively uses around 200 tokens of context (even if more is provided), and that word order only has an effect within approximately the last 50 tokens.

In this paper, we show that a non-recurrent model can achieve strong results on character-level language modeling.

Specifically, we use a deep network of transformer self-attention layers (Vaswani et al. 2017) with causal (backward-looking) attention to process fixed-length inputs and predict upcoming characters. The model is trained on mini-batches of sequences from random positions in the training corpus, with no information passed from one batch to the next.

Our primary finding is that the transformer architecture is well-suited to language modeling over long sequences and could replace RNNs in this domain. We speculate that the transformer’s success here is due to its ability to “quickly” propagate information over arbitrary distances; by comparison, RNNs need to learn to pass relevant information forward step by step.

We also find that some modifications to the basic transformer architecture are beneficial in this domain. Most importantly, we add three auxiliary losses, requiring the model to predict upcoming characters (i) at intermediate sequence positions, (ii) from intermediate hidden representations, and (iii) at target positions multiple steps in the future. These losses speed up convergence, and make it possible to train deeper networks.

Character Transformer Model

Language models assign a probability distribution over token sequences t0:L by factoring out the joint probability as follows, where L is the sequence length:

Pr(t0:L) = P(t0) YL i=1 Pr(tijt0:i􀀀1); (1)

To model the conditional probability Pr(tijt0:i􀀀1), we train a transformer network to process the character sequence t0:i􀀀1. Transformer networks have recently showed significant gains in tasks that require processing sequences accurately and efficiently.

Our character-level transformer architecture has 64 transformer layers. Following Vaswani et al. (2017), by “transformer layer” we mean a block containing a multihead self-attention sub-layer followed by a feed-forward network of two fully connected sub-layers. For more details on the transformer architecture, refer to Vaswani et al. (2017) and the tensor2tensor library1. To ensure that the model’s predictions are only conditioned on past characters, we mask our attention layers with a causal attention, so each position can only attend leftward. This is the same as the “masked attention” in the decoder component of the original transformer architecture used for sequence-to-sequence problems (Vaswani et al. 2017).

Figure 1 shows our initial model with the causal attention mask limiting information flow from left to right. Each character prediction is conditioned only on the characters that appeared earlier.

t0 t1 t2 t3 t4 Transformer Layers

Figure 1: Character transformer network of two layers processing a four character sequence to predict t4. The causal attention mask limits information to left-to-right flow. Red arrows highlight the prediction task the network has to learn.

Auxiliary Losses

Our network is, to our knowledge, deeper than any transformer network discussed in previous work. In initial experiments, we found training a network deeper than ten layers to be challenging, with slow convergence and poor accuracy. We were able to deepen the network to better effect through the addition auxiliary losses, which sped up convergence of the training significantly.

We add several types of auxiliary losses, corresponding to intermediate positions, intermediate layers, and nonadjacent targets. We hypothesize that these losses not only speed up convergence but also serve as an additional regularizer. During training, the auxiliary losses get added to the total loss of the network with discounted weights. Each type of auxiliary loss has its own schedule of decay. During evaluation and inference time, only the prediction of the final position at the final layer is used.

One consequence of this approach is that a number of the network parameters are only used during training— specifically, the parameters in the output classification layers associated with predictions made from intermediate layers and predictions over non-adjacent targets. Thus, when

1 https://github.com/tensorflow/tensor2tensor

listing the number of parameters in our models, we distinguish between “training parameters” and “inference parameters”. Multiple Positions First, we add prediction tasks for each position in the final layer, extending our predictions from one per example to jLj (sequence length). Note, predicting over all sequence positions is standard practice in RNNbased approaches. However in our case, since no information is passed forward across batches, this is forcing the model to predict given smaller contexts — sometimes just one or two characters. It is not obvious whether these secondary training tasks should help on the primary task of predicting with full context. However, we find that adding this auxiliary loss speeds up training and gives better results (see Ablation Experiments below). Figure 2 illustrates the task of predicting across all sequence positions. We add these losses during training without decaying their weights.

t0 t1 t2 t3
t1 t2 t3 t4
Transformer Layers

Figure 2: Adding the intermediate positions prediction tasks to our network. Now, we predict the final character t4 and all intermediate characters t0:3. t3 has access only to t0:2 because of the causal attention masks. All of these losses contribute equally during training.

Intermediate Layer Losses In addition to the final prediction layer, we add predictions made from the output of each intermediate transformer layer. As with the final layer, we add predictions for all intermediate positions in the sequence (see Figure 3). Lower layers are weighted to contribute less and less to the loss as training progresses. If there are n layers total, then the lth intermediate layer stops contributing any loss after finishing l=2n of the training. This schedule drops all intermediate losses after half of the training is done.

Multiple Targets At each position in the sequence, the model makes two (or more) predictions of future characters. For each new target we introduce a separate classifier. The losses of the extra targets get weighted by a multiplier of 0.5 before being added to their corresponding layer loss.

Positional Embeddings

In the basic transformer network described in Vaswani et al. (2017), a sinusoidal timing signal is added to the input sequence prior to the first transformer layer. However, as our network is deeper (64 layers), we hypothesize that the timing information may get lost during the propagation through the layers. To address this, we replace the timing signal with a learned per-layer positional embedding added to the input sequence before each transformer layer. Specifically, the model learns a unique 512-dimensional embedding vector for each of the L context positions within each of N layers, giving a total of L�N �512 additional parameters. We are able to safely use positional embeddings for our task, as we don’t require the model to generalize to longer contexts than those seen during training.

Transformer Layers
t0 t1 t2 t3
t1 t2 t3 t4
t1 t2 t3 t4

Figure 3: Our network after adding prediction tasks for the intermediate layers. For this example of two layers, the losses of the intermediate layer prediction tasks will be absent after finishing 25% of the training.

t0 t1 t2 t3 t1 t2 t2 t3 t3 t4 t4 t5 t1 t2 t2 t3 t3 t4 t4 t5

Transformer Layers

Figure 4: Our example network after adding two predictions per position.

Experimental Setup

Datasets

For evaluation we focus mainly on text8 (Mahoney 2009). This dataset consists of English Wikipedia articles, with superfluous content removed (tables, links to foreign language versions, citations, footnotes, markup, punctuation). The remaining text is processed to use a minimal character vocabulary of 27 unique characters — lowercase letters a through z, and space. Digits are replaced by their spelled-out equivalents, so “20” becomes “two zero”. Character sequences not in the range [a-zA-Z] are converted to a single space. Finally, the text is lowercased. The size of the corpus is 100M characters. Following Mikolov et al. (2012) and Zhang et al. (2016), we split the data into 90M characters for train, 5M characters for dev, and 5M characters for test.

To aid in comparison with other recent approaches, we also evaluate our model on enwik8 (Mahoney 2009) which is 100M bytes of unprocessed Wikipedia text, including markup and non-Latin characters. There are 205 unique bytes in the dataset. Following Chung et al. (2015), and as in text8, we split the data into 90M, 5M and 5M for training, dev and test respectively.

Training

Compared to most models based on transformers (Vaswani et al. 2017; Salimans et al. 2018), our model is very deep, with 64 transformer layers and each layer using two attention heads. Each transformer layer has a hidden size of 512 and a filter size of 2048. We feed our model sequences of length 512. Each item in the sequence represents a single byte (or equivalently, one character in text8) which gets replaced by its embedding, a vector of size 512. We add to the byte embeddings a separate learned positional embedding for each of the 512 token positions, as described in the Positional Embeddings section above. We do the same addition at each layer activation throughout the network. The positional embeddings are not shared across the layers.With two predictions per position, each layer learns to predict 1024 characters. Because we are primarily interested in predicting the immediately following character (one step away), we halve the loss of predicting characters two steps away. The prediction layers are logistic regression layers over the full 256 outputs (the number of unique bytes). To demonstrate the generality of the model, we always train and predict over all 256 labels, even on datasets that cover a smaller vocabulary. Despite this, we found that in practice the model never predicted a byte value outside of the ones observed in the training dataset.

The model has approximately 235 million parameters, which is larger than the number of characters in the text8 training corpus. To regularize the model, we apply dropout in the attention and ReLU layers with a probability of 0.55. We use the momentum optimizer with 0.99 momentum. The learning rate is fixed during training to 0.003. We train our model for 4 million steps, with each step processing a batch of 16 randomly selected sequences. We drop the intermediate layer losses consecutively, as described in the Intermediate Layer Losses section above. Starting from the first layer, after every 62:5K (= 4M� 1 2�64 ) steps, we drop the losses introduced by the next layer. According to this schedule, after training is halfway complete, only the final layer losses are present.

Evaluation

At inference time, we use the model’s prediction at the final position of the final layer to compute the probability of a character given a context of 512 characters. There is no state passed between predictions as would be the case with RNN models, so for each character predicted we have to process the context from scratch. Because there is no reused Parameters (�106)

Model train inference bpc LSTM (Cooijmans et al. 2016) - - 1.43 BN-LSTM (Cooijmans et al. 2016) - - 1.36 HM-LSTM (Chung, Ahn, and Bengio 2016) 35 35 1.29 Recurrent Highway (Zilly et al. 2016) 45 45 1.27 mLSTM (Krause et al. 2016) 45 45 1.27 T12 (ours) 44 41 1.18 T64 (ours) 235 219 1.13 mLSTM + dynamic eval (Krause et al. 2017) 45 - 1.19

Table 1: Comparison of various models on text8 test.

bpc Accuracy (%) Context dev test dev test 32 1.25 1.34 72.8 71.1 64 1.17 1.26 74.8 73.0 128 1.12 1.20 76.1 74.4 256 1.09 1.16 76.9 75.3 512 1.06 1.13 77.3 75.9

Table 2: Bits per character (bpc) and accuracy of our best model on text8 dev and test, for different context lengths.

computation from previous steps, our model requires expensive computational resources for evaluation and inference. We measure the performance of training checkpoints (roughly every 10,000 steps) by evaluating bits per character (bpc) over the entire the validation set, and save the parameters that perform the best. Our best model is achieved after around 2.5 million steps of training, which takes 175 hours on a single Google Cloud TPU v2.

Results

We report the performance of our best model (T64) on the validation and test sets. Table 1 compares our models against several recent results. On the test set, we achieve a new state of the art, 1.13 bpc. This model is 5x larger than previous models, which necessitated aggressive dropout rates of 0.55. For better comparison with smaller models, we also train a smaller model (T12) with 41M parameters. This model consists of 12 layers, and trained for 8M steps, with a reduced dropout rate of 0.2. All other settings were left the same as T64. Our smaller model still outperforms previous models, achieving 1.18 bpc on the test dataset. Increasing the depth of the network from 12 layers to 64 improved the results significantly, with the auxiliary losses enabling the training to better utilize the depth of the network. Note, our models do not use dynamic evaluation (Krause et al. 2017), a technique that adjusts model weights at test time by training on test Parameters (�106)

Model train inference bpb FS-LSTM-4 (Mujika, Meier, and Steger 2017) 47 - 1.25 mLSTM (Krause et al. 2016) 46 - 1.24 cmix v13 (Knol 2017) - - 1.23 T12 (ours) 44 41 1.11 T64 (ours) 235 219 1.06 mLSTM + dynamic eval (Krause et al. 2017) 46 - 1.08

Table 3: Comparison of various models on enwik8 test. data.

Table 2 shows the performance of our model given different context sizes. We are able to achieve state-of-the-art results once the context increases beyond 128 characters, with the best performance of 1.06 bpc at 512 characters. As expected, the model performs better when it is given more context. However this trend levels off after 512 characters; we do not see better results using a context of 1024. Using the same hyperparameters and training procedure for text8, we also train and evaluate the T12 and T64 architectures on enwik8 (see Table 3). Note, several previous authors discuss “bits per character” on enwik8 but are in fact reporting bits per byte. Without retuning for this dataset, our models still achieve state-of-the-art performance.

Ablation Experiments

To better understand the relative importance of the several modifications we proposed, we run an ablation analysis. We start from our best model T64 and then remove one modification at a time. For example, when we disable Multiple Positions, the model gets trained with only the last position loss for each layer. This corresponds to calculating fL(t4 j t0:3), L(t5 j t0:3)g in the example shown in Figure 4 for both the first and the second layers. When disabling Positional Embeddings, we add the default transformer sinusoidal timing signal before the first layer.

Description bpc �bpc T64 (Baseline) 1.062 - T64 w/out Multiple Positions 2.482 1.420 T64 w/out Intermediate Layer Losses 1.158 0.096 T64 w/out Positional Embeddings 1.069 0.007 T64 w/out Multiple Targets 1.068 0.006 T64 w/ SGD Optimizer 1.065 0.003

Table 4: Evaluation of T64 on text8 dev with context set to 512. Disabling each feature or loss lowers the quality of the model. The biggest win comes from adding multiple positions and intermediate layers losses.

For the ablation experiments, we reuse the hyperparameters from our best model to avoid a prohibitively expensive parameter search for each ablation. The only exception is the SGD experiment, where we vary the learning rate. The analysis shows that the biggest advantage comes from multiple positions and intermediate layers losses. Predicting all the intermediate positions leads to significant speed up in convergence, since the model sees more effective training examples per batch. Adding losses at the intermediate layers acts in the same spirit by forcing more predictions per training step. Finally, we replace momentum with SGD as our optimizer, using a range of learning rates (0.3, 0.1, 0.03, 0.01, 0.003, 0.001). This ablation shows that SGD produces competitive models, with learning rate 0.1 giving the best performance. Despite the depth of our network, SGD is able to train the network efficiently with the help of our auxiliary losses.

Type Model bpb ppl Word J´ozefowicz et al. (2016) - 23.7 Byte T64 1.03 40.6

Table 5: Performance of T64 on the lm1b test set.

Comparison with Word-Level Models To understand how byte-level language models perform in comparison to word-level language models, we train T64 on the lm1b corpus (Chelba et al. 2013). For lm1b, we use the standard train/test split of the preprocessed corpus, where out-of-vocab words have been replaced with UNK, to allow comparison to previous work on word and word-piece models. We report word perplexity (ppl) by converting bitsper- byte (bpb) into ppl2. During training we use the second shard (01) of the heldout dataset as a dev set, as the first shard (00) is the test. Given this is a significantly larger dataset than text8, we set all dropouts to zero. Table 5 shows a gap in performance between the two classes of language models. This comparison can serve as a starting point for researching possible ways to bridge the gap.

Qualitative Analysis

To probe the strengths and weaknesses of our best model (T64), we run the model forward, starting with the seed sequence of 512 characters in Figure 5, taken from the text8 test set. Figure 6 shows several per-character metrics for the model’s predictions over the true continuation of this seed text. At each position, we measure i) the model’s prediction entropy in bits across all 256 output classes, ii) its loss — the negative log probability of the target label, i.e. the “bits per character” for this position, and iii) the rank of the target in the list of output classes sorted by likelihood. Unsurprisingly, the model is least certain when predicting the first character of a word, and becomes progressively more confident and correct as subsequent characters are seen.

To investigate the degree to which our model prefers actual English words over non-existent words, we compute the likelihood the model assigns to all continuations after the seed. We cut off continuations when they reach a space character, or when the total probability of the continuation falls below 0.001. Figure 5 shows the entire set of word completions, in order of probability, where the initial pr- from the seed is repeated for readability. Note that these are all real or plausible (proofed) English words, and that even short but bad continuations like prz are assigned a lower cumulative probability than long realistic word completions like predictable. We expect that the transformer self-attention should make it easy for our model to copy sequences observed in the context over long distances (up to the context size of 512 characters). To test this expectation, we corrupt the seed and continuation from above by introducing a fake name zjakdmu bmijwxn. Specifically, we change the first occurrence of elizabeth in the seed to zjakdmu bmijwxn, and the 2For this test set, ppl = 2bpb�826189=159658, where 826,189 is the number of bytes and 159,658 is the number of tokens.

Seed

mary was not permitted to see them or to speak in her own defence at the tribunal she refused to offer a written defence unless elizabeth would guarantee a verdict of not guilty which elizabeth would not do although the casket letters were accepted by the inquiry as genuine after a study of the handwriting and of the information contained therein and were generally held to be certain proof of guilt if authentic the inquiry reached the conclusion that nothing was proven from the start this could have been pr Word Completions

proven, proved, proof, prevented, presented, problematic, probably, provided, practical, provoked, preceded, predicted, previously, presumed, praised, proposed, practicable, produced, present, preserved, precisely, prior, protected, probable, prompted, proofed, properly, practiced, prohibited, profound, preferable, proceeded, precise, predictable, practically, prevalent Figure 5: A seed sequence of 512 characters taken from the text8 test set, and all word completions assigned cumulative probability above 0.001 to follow the seed, in order from most likely (0.529) to least likely (0.001). second occurrence to she. Similarly, in the continuation, we change elizabeth to zjakdmu bmijwxn. The resulting distance between the two occurrences of the fake name is 434 characters.

Figure 7a confirms that the model can successfully copy over this long distance. While the initial z in zjakdmu is unexpected, the model immediately chooses to copy the remainder of this word from the context, as opposed to predicting any real z- words learned during training. Similarly, while the model is somewhat unsure whether the fake surname bmijwxn will appear (assigning the initial b a rank of two), it immediately picks up on the correspondence after the b is observed, correctly predicting the remainder of the fake surname.

For comparison, Figure 7b shows how the model would rank the targets in our fake continuation if the original seed with elizabeth were used. This confirms that the fake name is not predictable based on knowledge gained through training, and is indeed being copied from the preceding context. Generation For generating samples using our language model, we train on a larger and less processed dataset, enwik9 (Mahoney 2009). We split enwik9 into 900M, 50M and 50M for training, dev and test. Using the dev dataset to tune our dropout, we find that dropout=0.1 performs the best. On the test dataset, T64 achieves 0.85 bpb. Table 6 shows different generated samples following the seed text, using a sampling temperature of 1.0.

Related Work

Character-level modeling has shown promise in many areas such as sentiment analysis (Radford, J´ozefowicz, and

Figure 6: Per-character entropy, loss and rank assigned by T64 after seeding on the 512 character sequence from Figure 5.

Seed Computational neuroscience is an interdisciplinary field which draws on neuroscience, computer science, and applied mathematics. It most often uses mathematical and computational techniques such as computer simulations and mathematical models to understand the function of the nervous system. The field of computational neuroscience began with the work of Andrew Huxley, Alan Hodgkin, and David Marr. The results of Hodgkin and Huxley's pioneering work in developing Sample 1 computational neuroscience were chronicled in Is Mathematics Anything I Could Learn?. (ISBN 0826412246). Computational Sample 2 neuroscience concerned neurological auraria and the inherited ability to communicate and respond to environmental destruction - Sample 3 the model were published in 1982 and 1983 respectively, and the subsequent work on the field began its graduate program with [[M Truth the voltage clamp allowed them to develop the first mathematical model of the action potential. David Marr's work focuses on

Table 6: Samples generated by T64, seeded with text from the enwik9 dev set, using a sampling temperature of 1.0.

Sutskever 2017), question answering (Kenter, Jones, and Hewlett 2018) and classification (Zhang, Zhao, and LeCun 2015), and is an exciting area due to its simplicity and the ability to easily adapt to other languages. Neural network based language modeling has been heavily researched since its effectiveness was shown by Bengio et al. (2003). By far, the most popular architecture in this area is the RNN and variants, first studied in Mikolov et al. (2010). Much of the progress in this area has been made by mitigating the vanishing gradients problem (Hochreiter et al. 2001) by architectures such as LSTMs (Hochreiter and Schmidhuber 1997), GRU (Cho et al. 2014), Recurrent Highway Networks (Zilly et al. 2016), Unitary RNNs (Arjovsky, Shah, and Bengio 2015) and others. This is an issue that transformers do not have, due to attention allowing short paths to all inputs. Methods of normalizing activation functions, such as Batch Normalization (Ioffe and Szegedy 2015; Merity, Keskar, and Socher 2017) and Layer Normalization (Lei Ba, Kiros, and Hinton 2016) have also demonstrated improvements on language modeling tasks. As with this work, progress has been made with discovering ways to regularize sequential architectures, with techniques such as Recurrent Dropout (Zaremba, Sutskever, and Vinyals 2014; Gal and Ghahramani 2015) and Zoneout (Krueger et al. 2016; Rocki 2016).

A closely related architecture is the Neural Cache Model (Grave, Joulin, and Usunier 2016), where the RNN is allowed to attend to all of its previous hidden states at each step. Another similar model is used in (Daniluk et al. 2017) where a key-value attention mechanism similar to transformers is used. Both approaches show improvements on word level language modeling. Memory Networks (Weston, Chopra, and Bordes 2014) have a similarity to the transformer model in design as it also has layers of attention for processing a fix memory representing the input doc( a) Continuing after the modified seed (including the fake name 434 characters away). (b) Continuing after the original seed from Figure 5. Figure 7: Per-character rank assigned by T64 to a fake continuation, after being seeded on either (a) the fake context where elizabeth is replaced with zjakdmu bmijwxn, or (b) the original context. ument and has been shown to be effective for language modeling in (Sukhbaatar et al. 2015). ByteNet (Kalchbrenner et al. 2016), which is related but uses layers of dilated convolutions rather than attention, showed promising results on byte level language modeling. Gated Convolutional Networks (Dauphin et al. 2016) was an early non-recurrent model to show superior performance on word level language modeling.

Language models are not usually very deep due to computational constraints of training RNNs, and this also limits the number of parameters. The transformer architecture allowed us to build very deep (64 layer) models with a large number of parameters. A recent CNN model for text classification (Conneau et al. 2016) at 29 layers is considered deep in the NLP community. A Sparsely-Gated Mixture-of- Experts Layer (Shazeer et al. 2017) allowed language modeling experiments with a greatly increased number of parameters by only accessing a small portion of parameters every time step, showing a reduction in bits per word. In Exploring the Limits of Language Modeling (J´ozefowicz et al. 2016), an increase in the number of parameters was achieved by mixing character-level and word level models, using specialized softmaxes and using a large amount of computational resources to train. IndRNN (Li et al. 2018) uses a simplified RNN architecture that allows deeper stacking with 21-layers, achieving near SOTA character-level language modeling. Fast-Slow Recurrent Neural Networks (Mujika, Meier, and Steger 2017) also achieved near SOTA by increasing the number of recurrent steps for each character processed.

Conclusion

Character language modeling has been dominated by recurrent network approaches. In this paper, we show that a network of 12 stacked transformer layers achieves state-ofthe- art results on this task. We gain further improvements in quality by deepening the network to 64 layers, utilizing capacity and depth efficiently. The use of auxiliary losses at intermediate layers and positions is critical for reaching this performance, and these losses allow us to train much deeper transformer networks. Finally, we analyze the behavior of our network and find that it is able to exploit dependencies in structure and content over long distances, over 400 characters apart.

References

  • Arjovsky, M.; Shah, A.; and Bengio, Y. 2015. Unitary evolution recurrent neural networks. CoRR abs/1511.06464.
  • Bengio, Y.; Ducharme, R.; Vincent, P.; and Janvin, C. 2003. A neural probabilistic language model. J. Mach. Learn. Res. 3:1137–1155.

Chelba, C.; Mikolov, T.; Schuster, M.; Ge, Q.; Brants, T.; Koehn, P.; and Robinson, T. 2013. One billion word benchmark for measuring progress in statistical language modeling. arXiv preprint arXiv:1312.3005. Cho, K.; van Merrienboer, B.; G¨ulc¸ehre, C¸ .; Bougares, F.; Schwenk, H.; and Bengio, Y. 2014. Learning phrase representations using RNN encoder-decoder for statistical machine translation. CoRR abs/1406.1078. Chung, J.; Ahn, S.; and Bengio, Y. 2016. Hierarchical multiscale recurrent neural networks. arXiv preprint arXiv:1609.01704. Chung, J.; Gulcehre, C.; Cho, K.; and Bengio, Y. 2015. Gated feedback recurrent neural networks. In: Proceedings of The International Conference on Machine Learning, 2067–2075. Conneau, A.; Schwenk, H.; Barrault, L.; and LeCun, Y. 2016. Very deep convolutional networks for natural language processing. CoRR abs/1606.01781. Cooijmans, T.; Ballas, N.; Laurent, C.; and Courville, A. C. 2016. Recurrent batch normalization. CoRR abs/1603.09025. Daniluk, M.; Rockt¨aschel, T.; Welbl, J.; and Riedel, S. 2017. Frustratingly short attention spans in neural language modeling. CoRR abs/1702.04521. Dauphin, Y. N.; Fan, A.; Auli, M.; and Grangier, D. 2016. Language modeling with gated convolutional networks. CoRR abs/1612.08083. Gal, Y., and Ghahramani, Z. 2015. A Theoretically Grounded Application of Dropout in Recurrent Neural Networks. ArXiv e-prints. Grave, E.; Joulin, A.; and Usunier, N. 2016. Improving neural language models with a continuous cache. CoRR abs/1612.04426. Hochreiter, S., and Schmidhuber, J. 1997. Long short-term memory. 9:1735–80. Hochreiter, S.; Bengio, Y.; Frasconi, P.; Schmidhuber, J.; et al. 2001. Gradient flow in recurrent nets: the difficulty of learning long-term dependencies. Ioffe, S., and Szegedy, C. 2015. Batch normalization: Accelerating deep network training by reducing internal covariate shift. CoRR abs/1502.03167. J´ozefowicz, R.; Vinyals, O.; Schuster, M.; Shazeer, N.; andWu, Y. 2016. Exploring the limits of language modeling. CoRR abs/1602.02410. Kalchbrenner, N.; Espeholt, L.; Simonyan, K.; Oord, A. v. d.; Graves, A.; and Kavukcuoglu, K. 2016. Neural machine translation in linear time. arXiv preprint arXiv:1610.10099. Ke, N. R.; Goyal, A.; Bilaniuk, O.; Binas, J.; Charlin, L.; Pal, C.; and Bengio, Y. 2017. Sparse attentive backtracking: Longrange credit assignment in recurrent networks. arXiv preprint arXiv:1711.02326. Kenter, T.; Jones, L.; and Hewlett, D. 2018. Byte-level machine reading across morphologically varied languages. Khandelwal, U.; He, H.; Qi, P.; and Jurafsky, D. 2018. Sharp nearby, fuzzy far away: How neural language models use context. In Association for Computational Linguistics (ACL). Knol, B. 2017. cmix - http://www.byronknoll.com/cmix.html. Krause, B.; Lu, L.; Murray, I.; and Renals, S. 2016. Multiplicative lstm for sequence modelling. arXiv preprint arXiv:1609.07959. Krause, B.; Kahembwe, E.; Murray, I.; and Renals, S. 2017. Dynamic evaluation of neural sequence models. arXiv preprint arXiv:1709.07432. Krueger, D.; Maharaj, T.; Kram´ar, J.; Pezeshki, M.; Ballas, N.; Ke, N. R.; Goyal, A.; Bengio, Y.; Courville, A.; and Pal, C. 2016. Zoneout: Regularizing rnns by randomly preserving hidden activations. arXiv preprint arXiv:1606.01305. Lei Ba, J.; Kiros, J. R.; and Hinton, G. E. 2016. Layer Normalization. ArXiv e-prints. Li, S.; Li, W.; Cook, C.; Zhu, C.; and Gao, Y. 2018. Independently recurrent neural network (indrnn): Building A longer and deeper RNN. CoRR abs/1803.04831. Mahoney, M. 2009. Large text compression benchmark. http://www.mattmahoney.net/text/text.html. Merity, S.; Keskar, N. S.; and Socher, R. 2017. Regularizing and optimizing LSTM language models. CoRR abs/1708.02182. Mikolov, T.; Karafit, M.; Burget, L.; Cernock, J.; and Khudanpur, S. 2010. Recurrent neural network based language model. In Kobayashi, T.; Hirose, K.; and Nakamura, S., eds., INTERSPEECH, 1045–1048. ISCA. Mikolov, T.; Kombrink, S.; Burget, L.; ernock, J.; and Khudanpur, S. 2011. Extensions of recurrent neural network language model. In 2011 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), 5528–5531. Mikolov, T.; Sutskever, I.; Deoras, A.; Le, H.-S.; Kombrink, S.; and Cernocky, J. 2012. Subword language modeling with neural networks. preprint (http://www. fit. vutbr. cz/imikolov/rnnlm/char. pdf) 8. Mujika, A.; Meier, F.; and Steger, A. 2017. Fast-slow recurrent neural networks. In Advances in Neural Information Processing Systems, 5915–5924. Radford, A.; J´ozefowicz, R.; and Sutskever, I. 2017. Learning to generate reviews and discovering sentiment. CoRR abs/1704.01444. Rocki, K. M. 2016. Surprisal-driven feedback in recurrent networks. arXiv preprint arXiv:1608.06027. Salimans, T.; Zhang, H.; Radford, A.; and Metaxas, D. N. 2018. Improving gans using optimal transport. CoRR abs/1803.05573. Shazeer, N.; Mirhoseini, A.; Maziarz, K.; Davis, A.; Le, Q. V.; Hinton, G. E.; and Dean, J. 2017. Outrageously large neural networks: The sparsely-gated mixture-of-experts layer. CoRR abs/1701.06538. Sukhbaatar, S.; Weston, J.; Fergus, R.; et al. 2015. End-to-end memory networks. In Advances in neural information processing systems, 2440–2448. Sundermeyer, M.; Schl¨uter, R.; and Ney, H. 2012. Lstm neural networks for language modeling. In Thirteenth annual conference of the international speech communication association. Tallec, C., and Ollivier, Y. 2017. Unbiasing truncated backpropagation through time. arXiv preprint arXiv:1705.08209. Vaswani, A.; Shazeer, N.; Parmar, N.; Uszkoreit, J.; Jones, L.; Gomez, A. N.; Kaiser, L.; and Polosukhin, I. 2017. Attention is all you need. In Advances in Neural Information Processing Systems 30. Curran Associates, Inc. 5998–6008. Werbos, P. J. 1990. Backpropagation through time: what it does and how to do it. Proceedings of the IEEE 78(10):1550–1560. Weston, J.; Chopra, S.; and Bordes, A. 2014. Memory networks. CoRR abs/1410.3916. Zaremba, W.; Sutskever, I.; and Vinyals, O. 2014. Recurrent neural network regularization. CoRR abs/1409.2329. Zhang, S.; Wu, Y.; Che, T.; Lin, Z.; Memisevic, R.; Salakhutdinov, R. R.; and Bengio, Y. 2016. Architectural complexity measures of recurrent neural networks. In Advances in Neural Information Processing Systems, 1822–1830. Zhang, X.; Zhao, J. J.; and LeCun, Y. 2015. Characterlevel convolutional networks for text classification. CoRR abs/1509.01626.

  • Zilly, J. G.; Srivastava, R. K.; Koutn´ık, J.; and Schmidhuber, J. 2016. Recurrent highway networks. CoRR abs/1607.03474.

BibTeX

@inproceedings{2019_CharacterLevelLanguageModelingw,
  author    = {Rami Al-Rfou and
               Dokook Choe and
               Noah Constant and
               Mandy Guo and
               Llion Jones},
  title     = {Character-Level Language Modeling with Deeper Self-Attention},
  booktitle = {Proceedings of the Thirty-Third AAAI Conference on Artificial Intelligence (AAAI
               2019), The Thirty-First Innovative Applications of Artificial Intelligence
               Conference (IAAI 2019), The Ninth AAAI Symposium on Educational
               Advances in Artificial Intelligence (EAAI 2019)},
  pages     = {3159--3166},
  publisher = {AAAI Press},
  year      = {2019},
  url       = {https://doi.org/10.1609/aaai.v33i01.33013159},
  doi       = {10.1609/aaai.v33i01.33013159},
}


 AuthorvolumeDate ValuetitletypejournaltitleUrldoinoteyear
2019 CharacterLevelLanguageModelingwRami Al-Rfou
Llion Jones
Dokook Choe
Noah Constant
Mandy Guo
Character-level Language Modeling with Deeper Self-attention2019