In task-oriented dialogue systems the dialogue state tracker (DST) component is responsible for predicting the state of the dialogue based on the dialogue history. Current DST approaches rely on a predefined domain ontology, a fact that limits their effective usage for large scale conversational agents, where the DST constantly needs to be interfaced with ever-increasing services and APIs. Focused towards overcoming this drawback, we propose a domain-aware dialogue state tracker, that is completely data-driven and it is modeled to predict for dynamic service schemas. The proposed model utilizes domain and slot information to extract both domain and slot specific representations for a given dialogue, and then uses such representations to predict the values of the corresponding slot. Integrating this mechanism with a pretrained language model (i.e. BERT), our approach can effectively learn semantic relations.
Pre-trained language models have been dominating the field of natural language processing in recent years, and have led to significant performance gains for various complex natural language tasks. One of the most prominent pre-trained language models is BERT, which was released as an English as well as a multilingual version. Although multilingual BERT performs well on many tasks, recent studies show that BERT models trained on a single language significantly outperform the multilingual version. Training a Dutch BERT model thus has a lot of potential for a wide range of Dutch NLP tasks. While previous approaches have used earlier implementations of BERT to train a Dutch version of BERT, we used RoBERTa, a robustly optimized BERT approach, to train a Dutch language model called RobBERT. We measured its performance on various tasks as well as the importance of the fine-tuning dataset size. We also evaluated the importance of language-specific tokenizers and the model's fairness. We found that RobBERT improves state-of-the-art results for various tasks, and especially significantly outperforms other models when dealing with smaller datasets. These results indicate that it is a powerful pre-trained model for a large variety of Dutch language tasks. The pre-trained and fine-tuned models are publicly available to support further downstream Dutch NLP applications.
Neural networks have achieved state of the art performance across a wide variety of machine learning tasks, often with large and computation-heavy models. Inducing sparseness as a way to reduce the memory and computation footprint of these models has seen significant research attention in recent years. In this paper, we present a new method for \emph{dynamic sparseness}, whereby part of the computations are omitted dynamically, based on the input. For efficiency, we combined the idea of dynamic sparseness with block-wise matrix-vector multiplications. In contrast to static sparseness, which permanently zeroes out selected positions in weight matrices, our method preserves the full network capabilities by potentially accessing any trained weights. Yet, matrix vector multiplications are accelerated by omitting a pre-defined fraction of weight blocks from the matrix, based on the input. Experimental results on the task of language modeling, using recurrent and quasi-recurrent models, show that the proposed method can outperform a magnitude-based static sparseness baseline. In addition, our method achieves similar language modeling perplexities as the dense baseline, at half the computational cost at inference time.
JavaScript (JS) engine vulnerabilities pose significant security threats affecting billions of web browsers. While fuzzing is a prevalent technique for finding such vulnerabilities, there have been few studies that leverage the recent advances in neural network language models (NNLMs). In this paper, we present Montage, the first NNLM-guided fuzzer for finding JS engine vulnerabilities. The key aspect of our technique is to transform a JS abstract syntax tree (AST) into a sequence of AST subtrees that can directly train prevailing NNLMs. We demonstrate that Montage is capable of generating valid JS tests, and show that it outperforms previous studies in terms of finding vulnerabilities. Montage found 37 real-world bugs, including three CVEs, in the latest JS engines, demonstrating its efficacy in finding JS engine bugs.
Large Transformer models routinely achieve state-of-the-art results on a number of tasks but training these models can be prohibitively costly, especially on long sequences. We introduce two techniques to improve the efficiency of Transformers. For one, we replace dot-product attention by one that uses locality-sensitive hashing, changing its complexity from O($L^2$) to O($L\log L$), where $L$ is the length of the sequence. Furthermore, we use reversible residual layers instead of the standard residuals, which allows storing activations only once in the training process instead of $N$ times, where $N$ is the number of layers. The resulting model, the Reformer, performs on par with Transformer models while being much more memory-efficient and much faster on long sequences.
Data-to-text generation models face challenges in ensuring data fidelity by referring to the correct input source. To inspire studies in this area, Wiseman et al. (2017) introduced the RotoWire corpus on generating NBA game summaries from the box- and line-score tables. However, limited attempts have been made in this direction and the challenges remain. We observe a prominent bottleneck in the corpus where only about 60% of the summary contents can be grounded to the boxscore records. Such information deficiency tends to misguide a conditioned language model to produce unconditioned random facts and thus leads to factual hallucinations. In this work, we restore the information balance and revamp this task to focus on fact-grounded data-to-text generation. We introduce a purified and larger-scale dataset, RotoWire-FG (Fact-Grounding), with 50% more data from the year 2017-19 and enriched input tables, hoping to attract more research focuses in this direction. Moreover, we achieve improved data fidelity over the state-of-the-art models by integrating a new form of table reconstruction as an auxiliary task to boost the generation quality.
Discrete structure rules for validating molecular structures are usually limited to fulfillment of the octet rule or similar simple deterministic heuristics. We propose a model, inspired by language modeling from natural language processing, with the ability to learn from a collection of undirected molecular graphs, enabling fitting of any underlying structure rule present in the collection. We introduce an adaption to the popular Transformer model, which can learn relationships between atoms and bonds. To our knowledge, the Transformer adaption is the first model that is trained to solve the unsupervised task of recovering partially observed molecules. In this work, we assess how different degrees of information impact performance w.r.t. to fitting the QM9 dataset, which conforms to the octet rule, and to fitting the ZINC dataset, which contains hypervalent molecules and ions requiring the model to learn a more complex structure rule. More specifically, we test a full discrete graph with bond order information, a full discrete graph with only connectivity, a bag-of-neighbors, a bag-of-atoms, and a count-based unigram statistics. These results provide encouraging evidence that neural networks, even when only connectivity is available, can learn arbitrary molecular structure rules specific to a dataset, as the Transformer adaption surpasses a strong octet rule baseline on the ZINC dataset.
Named entity recognition (NER) and relation extraction (RE) are two important tasks in information extraction and retrieval (IE \& IR). Recent work has demonstrated that it is beneficial to learn these tasks jointly, which avoids the propagation of error inherent in pipeline-based systems and improves performance. However, state-of-the-art joint models typically rely on external natural language processing (NLP) tools, such as dependency parsers, limiting their usefulness to domains (e.g. news) where those tools perform well. The few neural, end-to-end models that have been proposed are trained almost completely from scratch. In this paper, we propose a neural, end-to-end model for jointly extracting entities and their relations which does not rely on external NLP tools and which integrates a large, pre-trained language model. Because the bulk of our model's parameters are pre-trained and we eschew recurrence for self-attention, our model is fast to train. On 5 datasets across 3 domains, our model matches or exceeds state-of-the-art performance, sometimes by a large margin.
Sequential word order is important when processing text. Currently, neural networks (NNs) address this by modeling word position using position embeddings. The problem is that position embeddings capture the position of individual words, but not the ordered relationship (e.g., adjacency or precedence) between individual word positions. We present a novel and principled solution for modeling both the global absolute positions of words and their order relationships. Our solution generalizes word embeddings, previously defined as independent vectors, to continuous word functions over a variable (position). The benefit of continuous functions over variable positions is that word representations shift smoothly with increasing positions. Hence, word representations in different positions can correlate with each other in a continuous function. The general solution of these functions is extended to complex-valued domain due to richer representations. We extend CNN, RNN and Transformer NNs to complex-valued versions to incorporate our complex embedding (we make all code available). Experiments on text classification, machine translation and language modeling show gains over both classical word embeddings and position-enriched word embeddings. To our knowledge, this is the first work in NLP to link imaginary numbers in complex-valued representations to concrete meanings (i.e., word order).
Recent success of pre-trained language models (LMs) has spurred widespread interest in the language capabilities that they possess. However, efforts to understand whether LM representations are useful for symbolic reasoning tasks have been limited and scattered. In this work, we propose eight reasoning tasks, which conceptually require operations such as comparison, conjunction, and composition. A fundamental challenge is to understand whether the performance of a LM on a task should be attributed to the pre-trained representations or to the process of fine-tuning on the task data. To address this, we propose an evaluation protocol that includes both zero-shot evaluation (no fine-tuning), as well as comparing the learning curve of a fine-tuned LM to the learning curve of multiple controls, which paints a rich picture of the LM capabilities. Our main findings are that: (a) different LMs exhibit qualitatively different reasoning abilities, e.g., RoBERTa succeeds in reasoning tasks where BERT fails completely; (b) LMs do not reason in an abstract manner and are context-dependent, e.g., while RoBERTa can compare ages, it can do so only when the ages are in the typical range of human ages; (c) On half of our reasoning tasks all models fail completely. Our findings and infrastructure can help future work on designing new datasets, models and objective functions for pre-training.
Self-attention based Transformer has demonstrated the state-of-the-art performances in a number of natural language processing tasks. Self-attention is able to model long-term dependencies, but it may suffer from the extraction of irrelevant information in the context. To tackle the problem, we propose a novel model called \textbf{Explicit Sparse Transformer}. Explicit Sparse Transformer is able to improve the concentration of attention on the global context through an explicit selection of the most relevant segments. Extensive experimental results on a series of natural language processing and computer vision tasks, including neural machine translation, image captioning, and language modeling, all demonstrate the advantages of Explicit Sparse Transformer in model performance. We also show that our proposed sparse attention method achieves comparable or better results than the previous sparse attention method, but significantly reduces training and testing time. For example, the inference speed is twice that of sparsemax in Transformer model. Code will be available at \url{https://github.com/lancopku/Explicit-Sparse-Transformer}
The key to a Transformer model is the self-attention mechanism, which allows the model to analyze an entire sequence in a computationally efficient manner. Recent work has suggested the possibility that general attention mechanisms used by RNNs could be replaced by active-memory mechanisms. In this work, we evaluate whether various active-memory mechanisms could replace self-attention in a Transformer. Our experiments suggest that active-memory alone achieves comparable results to the self-attention mechanism for language modelling, but optimal results are mostly achieved by using both active-memory and self-attention mechanisms together. We also note that, for some specific algorithmic tasks, active-memory mechanisms alone outperform both self-attention and a combination of the two.
The Natural Language Processing (NLP) community has significantly contributed to the solutions for entity and relation recognition from the text, and possibly linking them to proper matches in Knowledge Graphs (KGs). Considering Wikidata as the background KG, still, there are limited tools to link knowledge within the text to Wikidata. In this paper, we present Falcon 2.0, first joint entity, and relation linking tool over Wikidata. It receives a short natural language text in the English language and outputs a ranked list of entities and relations annotated with the proper candidates in Wikidata. The candidates are represented by their Internationalized Resource Identifier (IRI) in Wikidata. Falcon 2.0 resorts to the English language model for the recognition task (e.g., N-Gram tiling and N-Gram splitting), and then an optimization approach for linking task. We have empirically studied the performance of Falcon 2.0 on Wikidata and concluded that it outperforms all the existing baselines. Falcon 2.0 is public and can be reused by the community; all the required instructions of Falcon 2.0 are well-documented at our GitHub repository. We also demonstrate an online API, which can be run without any technical expertise. Falcon 2.0 and its background knowledge bases are available as resources at https://labs.tib.eu/falcon/falcon2/.
To simultaneously capture syntax and global semantics from a text corpus, we propose a new larger-context recurrent neural network (RNN) based language model, which extracts recurrent hierarchical semantic structure via a dynamic deep topic model to guide natural language generation. Moving beyond a conventional RNN-based language model that ignores long-range word dependencies and sentence order, the proposed model captures not only intra-sentence word dependencies, but also temporal transitions between sentences and inter-sentence topic dependencies. For inference, we develop a hybrid of stochastic-gradient Markov chain Monte Carlo and recurrent autoencoding variational Bayes. Experimental results on a variety of real-world text corpora demonstrate that the proposed model not only outperforms larger-context RNN-based language models, but also learns interpretable recurrent multilayer topics and generates diverse sentences and paragraphs that are syntactically correct and semantically coherent.
The transformer-based pre-trained language model BERT has helped to improve state-of-the-art performance on many natural language processing (NLP) tasks. Using the same architecture and parameters, we developed and evaluated a monolingual Dutch BERT model called BERTje. Compared to the multilingual BERT model, which includes Dutch but is only based on Wikipedia text, BERTje is based on a large and diverse dataset of 2.4 billion tokens. BERTje consistently outperforms the equally-sized multilingual BERT model on downstream NLP tasks (part-of-speech tagging, named-entity recognition, semantic role labeling, and sentiment analysis). Our pre-trained Dutch BERT model is made available at https://github.com/wietsedv/bertje.
Logographs (Chinese characters) have recursive structures (i.e. hierarchies of sub-units in logographs) that contain phonological and semantic information, as developmental psychology literature suggests that native speakers leverage on the structures to learn how to read. Exploiting these structures could potentially lead to better embeddings that can benefit many downstream tasks. We propose building hierarchical logograph (character) embeddings from logograph recursive structures using treeLSTM, a recursive neural network. Using recursive neural network imposes a prior on the mapping from logographs to embeddings since the network must read in the sub-units in logographs according to the order specified by the recursive structures. Based on human behavior in language learning and reading, we hypothesize that modeling logographs' structures using recursive neural network should be beneficial. To verify this claim, we consider two tasks (1) predicting logographs' Cantonese pronunciation from logographic structures and (2) language modeling. Empirical results show that the proposed hierarchical embeddings outperform baseline approaches. Diagnostic analysis suggests that hierarchical embeddings constructed using treeLSTM is less sensitive to distractors, thus is more robust, especially on complex logographs.
Recent advances in text-to-speech (TTS) led to the development of flexible multi-speaker end-to-end TTS systems. We extend state-of-the-art attention-based automatic speech recognition (ASR) systems with synthetic audio generated by a TTS system trained only on the ASR corpora itself. ASR and TTS systems are built separately to show that text-only data can be used to enhance existing end-to-end ASR systems without the necessity of parameter or architecture changes. We compare our method with language model integration of the same text data and with simple data augmentation methods like SpecAugment and show that performance improvements are mostly independent. We achieve improvements of up to 33% relative in word-error-rate (WER) over a strong baseline with data-augmentation in a low-resource environment (LibriSpeech-100h), closing the gap to a comparable oracle experiment by more than 50\%. We also show improvements of up to 5% relative WER over our most recent ASR baseline on LibriSpeech-960h.
Bridging the exponentially growing gap between the numbers of unlabeled and labeled protein sequences, several studies adopted semi-supervised learning for protein sequence modeling. In these studies, models were pre-trained with a substantial amount of unlabeled data, and the representations were transferred to various downstream tasks. Most pre-training methods solely rely on language modeling and often exhibit limited performance. In this paper, we introduce a novel pre-training scheme called PLUS, which stands for Protein sequence representations Learned Using Structural information. PLUS consists of masked language modeling and a complementary protein-specific pre-training task, namely same-family prediction. PLUS can be used to pre-train various model architectures. In this work, we use PLUS to pre-train a bidirectional recurrent neural network and refer to the resulting model as PLUS-RNN. Our experiment results demonstrate that PLUS-RNN outperforms other models of similar size solely pre-trained with the language modeling in six out of seven widely used protein biology tasks. Furthermore, we present the results from our qualitative interpretation analyses to illustrate the strengths of PLUS-RNN. PLUS provides a novel way to exploit evolutionary relationships among unlabeled proteins and is broadly applicable across a variety of protein biology tasks. We expect that the gap between the numbers of unlabeled and labeled proteins will continue to grow exponentially, and the proposed pre-training method will play a larger role.
Deep learning based methods have achieved surprising progress in Scene Text Recognition (STR), one of classic problems in computer vision. In this paper, we propose a feasible framework for multi-lingual arbitrary-shaped STR, including instance segmentation based text detection and language model based attention mechanism for text recognition. Our STR algorithm not only recognizes Latin and Non-Latin characters, but also supports arbitrary-shaped text recognition. Our method wins the championship on Scene Text Spotting Task (Latin Only, Latin and Chinese) of ICDAR2019 Robust Reading Challenge on ArbitraryShaped Text Competition. Code is available at https://github.com/zhang0jhon/AttentionOCR.
Language models have become a key step to achieve state-of-the art results in many different Natural Language Processing (NLP) tasks. Leveraging the huge amount of unlabeled texts nowadays available, they provide an efficient way to pre-train continuous word representations that can be fine-tuned for a downstream task, along with their contextualization at the sentence level. This has been widely demonstrated for English using contextualized representations (Dai and Le, 2015; Peters et al., 2018; Howard and Ruder, 2018; Radford et al., 2018; Devlin et al., 2019; Yang et al., 2019b). In this paper, we introduce and share FlauBERT, a model learned on a very large and heterogeneous French corpus. Models of different sizes are trained using the new CNRS (French National Centre for Scientific Research) Jean Zay supercomputer. We apply our French language models to diverse NLP tasks (text classification, paraphrasing, natural language inference, parsing, word sense disambiguation) and show that most of the time they outperform other pre-training approaches. Different versions of FlauBERT as well as a unified evaluation protocol for the downstream tasks, called FLUE (French Language Understanding Evaluation), are shared to the research community for further reproducible experiments in French NLP.