TACL 2019 Papers

Neural dependency parsing has proven very effective, achieving state-of-the-art results on numerous domains and languages. Unfortunately, it requires large amounts of labeled data, which is costly and laborious to create. In this paper we propose a self-training algorithm that alleviates this annotation bottleneck by training a parser on its own output. Our Deep Contextualized Self-training (DCST) algorithm utilizes representation models trained on sequence labeling tasks that are derived from the parser’s output when applied to unlabeled data, and integrates these models with the base parser through a gating mechanism. We conduct experiments across multiple languages, both in low resource in-domain and in cross-domain setups, and demonstrate that DCST substantially outperforms traditional self-training as well as recent semi-supervised training methods.1

We analyze human’s disagreements about the validity of natural language inferences. We show that, very often, disagreements are not dismissible as annotation “noise”, but rather persist as we collect more ratings and as we vary the amount of context provided to raters. We further show that the type of uncertainty captured by current state-of-the-art models for natural language inference is not reflective of the type of uncertainty present in human disagreements. We discuss implications of our results in relation to the recognizing textual entailment (RTE)/natural language inference (NLI) task. We argue for a refined evaluation objective that requires models to explicitly capture the full distribution of plausible human judgments.

Conventional neural autoregressive decoding commonly assumes a fixed left-to-right generation order, which may be sub-optimal. In this work, we propose a novel decoding algorithm— InDIGO—which supports flexible sequence generation in arbitrary orders through insertion operations. We extend Transformer, a state-of-the-art sequence generation model, to efficiently implement the proposed approach, enabling it to be trained with either a pre-defined generation order or adaptive orders obtained from beam-search. Experiments on four real-world tasks, including word order recovery, machine translation, image caption, and code generation, demonstrate that our algorithm can generate sequences following arbitrary orders, while achieving competitive or even better performance compared with the conventional left-to-right generation. The generated sequences show that InDIGO adopts adaptive generation orders based on input information.

The best solution of structured prediction models in NLP is often inaccurate because of limited expressive power of the model or to non-exact parameter estimation. One way to mitigate this problem is sampling candidate solutions from the model’s solution space, reasoning that effective exploration of this space should yield high-quality solutions. Unfortunately, sampling is often computationally hard and many works hence back-off to sub-optimal strategies, such as extraction of the best scoring solutions of the model, which are not as diverse as sampled solutions. In this paper we propose a perturbation-based approach where sampling from a probabilistic model is computationally efficient. We present a learning algorithm for the variance of the perturbations, and empirically demonstrate its importance. Moreover, while finding the argmax in our model is intractable, we propose an efficient and effective approximation. We apply our framework to cross-lingual dependency parsing across 72 corpora from 42 languages and to lightly supervised dependency parsing across 13 corpora from 12 languages, and demonstrate strong results in terms of both the quality of the entire solution list and of the final solution.1

This paper investigates the ability of artificial neural networks to judge the grammatical acceptability of a sentence, with the goal of testing their linguistic competence. We introduce the Corpus of Linguistic Acceptability (CoLA), a set of 10,657 English sentences labeled as grammatical or ungrammatical from published linguistics literature. As baselines, we train several recurrent neural network models on acceptability classification, and find that our models outperform unsupervised models by Lau et al. (2016) on CoLA. Error-analysis on specific grammatical phenomena reveals that both Lau et al.’s models and ours learn systematic generalizations like subject-verb-object order. However, all models we test perform far below human level on a wide range of grammatical constructions.

Contextual representation models have achieved great success in improving various downstream natural language processing tasks. However, these language-model-based encoders are difficult to train due to their large parameter size and high computational complexity. By carefully examining the training procedure, we observe that the softmax layer, which predicts a distribution of the target word, often induces significant overhead, especially when the vocabulary size is large. Therefore, we revisit the design of the output layer and consider directly predicting the pre-trained embedding of the target word for a given context. When applied to ELMo, the proposed approach achieves a 4-fold speedup and eliminates 80% trainable parameters while achieving competitive performance on downstream tasks. Further analysis shows that the approach maintains the speed advantage under various settings, even when the sentence encoder is scaled up.

We introduce an architecture to learn joint multilingual sentence representations for 93 languages, belonging to more than 30 different families and written in 28 different scripts. Our system uses a single BiLSTM encoder with a shared byte-pair encoding vocabulary for all languages, which is coupled with an auxiliary decoder and trained on publicly available parallel corpora. This enables us to learn a classifier on top of the resulting embeddings using English annotated data only, and transfer it to any of the 93 languages without any modification. Our experiments in cross-lingual natural language inference (XNLI data set), cross-lingual document classification (MLDoc data set), and parallel corpus mining (BUCC data set) show the effectiveness of our approach. We also introduce a new test set of aligned sentences in 112 languages, and show that our sentence embeddings obtain strong results in multilingual similarity search even for low- resource languages. Our implementation, the pre-trained encoder, and the multilingual test set are available at https://github.com/facebookresearch/LASER.

In this paper we introduce domain detection as a new natural language processing task. We argue that the ability to detect textual segments that are domain-heavy (i.e., sentences or phrases that are representative of and provide evidence for a given domain) could enhance the robustness and portability of various text classification applications. We propose an encoder-detector framework for domain detection and bootstrap classifiers with multiple instance learning. The model is hierarchically organized and suited to multilabel classification. We demonstrate that despite learning with minimal supervision, our model can be applied to text spans of different granularities, languages, and genres. We also showcase the potential of domain detection for text summarization.

We introduce Morse, a recurrent encoder-decoder model that produces morphological analyses of each word in a sentence. The encoder turns the relevant information about the word and its context into a fixed size vector representation and the decoder generates the sequence of characters for the lemma followed by a sequence of individual morphological features. We show that generating morphological features individually rather than as a combined tag allows the model to handle rare or unseen tags and to outperform whole-tag models. In addition, generating morphological features as a sequence rather than, for example, an unordered set allows our model to produce an arbitrary number of features that represent multiple inflectional groups in morphologically complex languages. We obtain state-of-the-art results in nine languages of different morphological complexity under low-resource, high-resource, and transfer learning settings. We also introduce TrMor2018, a new high-accuracy Turkish morphology data set. Our Morse implementation and the TrMor2018 data set are available online to support future research.1See https://github.com/ai-ku/Morse.jl for a Morse implementation in Julia/Knet (Yuret, 2016) and https://github.com/ai-ku/TrMor2018 for the new Turkish data set.

Until now, grammatical error correction (GEC) has been primarily evaluated on text written by non-native English speakers, with a focus on student essays. This paper enables GEC development on text written by native speakers by providing a new data set and metric. We present a multiple-reference test corpus for GEC that includes 4,000 sentences in two new domains (formal and informal writing by native English speakers) and 2,000 sentences from a diverse set of non-native student writing. We also collect human judgments of several GEC systems on this new test set and perform a meta-evaluation, assessing how reliable automatic metrics are across these domains. We find that commonly used GEC metrics have inconsistent performance across domains, and therefore we propose a new ensemble metric that is robust on all three domains of text.

Online debates allow people to express their persuasive abilities and provide exciting opportunities for understanding persuasion. Prior studies have focused on studying persuasion in debate content, but without accounting for each debater’s history or exploring the progression of a debater’s persuasive ability. We study debater skill by modeling how participants progress over time in a collection of debates from Debate.org. We build on a widely used model of skill in two-player games and augment it with linguistic features of a debater’s content. We show that online debaters’ skill levels do tend to improve over time. Incorporating linguistic profiles leads to more robust skill estimation than winning records alone. Notably, we find that an interaction feature combining uncertainty cues (hedging) with terms strongly associated with either side of a particular debate (fightin’ words) is more predictive than either feature on its own, indicating the importance of fine- grained linguistic features.

We provide the first computational treatment of fused-heads constructions (FHs), focusing on the numeric fused-heads (NFHs). FHs constructions are noun phrases in which the head noun is missing and is said to be “fused” with its dependent modifier. This missing information is implicit and is important for sentence understanding. The missing references are easily filled in by humans but pose a challenge for computational models. We formulate the handling of FHs as a two stages process: Identification of the FH construction and resolution of the missing head. We explore the NFH phenomena in large corpora of English text and create (1) a data set and a highly accurate method for NFH identification; (2) a 10k examples (1 M tokens) crowd-sourced data set of NFH resolution; and (3) a neural baseline for the NFH resolution task. We release our code and data set, to foster further research into this challenging problem.

We present a novel semantic framework for modeling linguistic expressions of generalization— generic, habitual, and episodic statements—as combinations of simple, real-valued referential properties of predicates and their arguments. We use this framework to construct a dataset covering the entirety of the Universal Dependencies English Web Treebank. We use this dataset to probe the efficacy of type-level and token-level information—including hand-engineered features and static (GloVe) and contextual (ELMo) word embeddings—for predicting expressions of generalization.

Domain-specific goal-oriented dialogue systems typically require modeling three types of inputs, namely, (i) the knowledge-base associated with the domain, (ii) the history of the conversation, which is a sequence of utterances, and (iii) the current utterance for which the response needs to be generated. While modeling these inputs, current state-of-the-art models such as Mem2Seq typically ignore the rich structure inherent in the knowledge graph and the sentences in the conversation context. Inspired by the recent success of structure-aware Graph Convolutional Networks (GCNs) for various NLP tasks such as machine translation, semantic role labeling, and document dating, we propose a memory-augmented GCN for goal-oriented dialogues. Our model exploits (i) the entity relation graph in a knowledge-base and (ii) the dependency graph associated with an utterance to compute richer representations for words and entities. Further, we take cognizance of the fact that in certain situations, such as when the conversation is in a code-mixed language, dependency parsers may not be available. We show that in such situations we could use the global word co-occurrence graph to enrich the representations of utterances. We experiment with four datasets: (i) the modified DSTC2 dataset, (ii) recently released code-mixed versions of DSTC2 dataset in four languages, (iii) Wizard-of-Oz style CAM676 dataset, and (iv) Wizard-of-Oz style MultiWOZ dataset. On all four datasets our method outperforms existing methods, on a wide range of evaluation metrics.

Recurrent neural networks (RNNs) have reached striking performance in many natural language processing tasks. This has renewed interest in whether these generic sequence processing devices are inducing genuine linguistic knowledge. Nearly all current analytical studies, however, initialize the RNNs with a vocabulary of known words, and feed them tokenized input during training. We present a multi-lingual study of the linguistic knowledge encoded in RNNs trained as character-level language models, on input data with word boundaries removed. These networks face a tougher and more cognitively realistic task, having to discover any useful linguistic unit from scratch based on input statistics. The results show that our “near tabula rasa” RNNs are mostly able to solve morphological, syntactic and semantic tasks that intuitively presuppose word-level knowledge, and indeed they learned, to some extent, to track word boundaries. Our study opens the door to speculations about the necessity of an explicit, rigid word lexicon in language learning and usage.

We present the Natural Questions corpus, a question answering data set. Questions consist of real anonymized, aggregated queries issued to the Google search engine. An annotator is presented with a question along with a Wikipedia page from the top 5 search results, and annotates a long answer (typically a paragraph) and a short answer (one or more entities) if present on the page, or marks null if no long/short answer is present. The public release consists of 307,373 training examples with single annotations; 7,830 examples with 5-way annotations for development data; and a further 7,842 examples with 5-way annotated sequestered as test data. We present experiments validating quality of the data. We also describe analysis of 25-way annotations on 302 examples, giving insights into human variability on the annotation task. We introduce robust metrics for the purposes of evaluating question answering systems; demonstrate high human upper bounds on these metrics; and establish baseline results using competitive methods drawn from related literature.

Composition models of distributional semantics are used to construct phrase representations from the representations of their words. Composition models are typically situated on two ends of a spectrum. They either have a small number of parameters but compose all phrases in the same way, or they perform word-specific compositions at the cost of a far larger number of parameters. In this paper we propose transformation weighting (TransWeight), a composition model that consistently outperforms existing models on nominal compounds, adjective-noun phrases, and adverb-adjective phrases in English, German, and Dutch. TransWeight drastically reduces the number of parameters needed compared with the best model in the literature by composing similar words in the same way.

We present a multiattentive recurrent neural network architecture for automatic multilingual readability assessment. This architecture considers raw words as its main input, but internally captures text structure and informs its word attention process using other syntax- and morphology-related datapoints, known to be of great importance to readability. This is achieved by a multiattentive strategy that allows the neural network to focus on specific parts of a text for predicting its reading level. We conducted an exhaustive evaluation using data sets targeting multiple languages and prediction task types, to compare the proposed model with traditional, state-of-the-art, and other neural network strategies.

Building meaningful phrase representations is challenging because phrase meanings are not simply the sum of their constituent meanings. Lexical composition can shift the meanings of the constituent words and introduce implicit information. We tested a broad range of textual representations for their capacity to address these issues. We found that, as expected, contextualized word representations perform better than static word embeddings, more so on detecting meaning shift than in recovering implicit information, in which their performance is still far from that of humans. Our evaluation suite, consisting of six tasks related to lexical composition effects, can serve future research aiming to improve representations.