EMNLP 2018 Papers

Localizing moments in a longer video via natural language queries is a new, challenging task at the intersection of language and video understanding. Though moment localization with natural language is similar to other language and vision tasks like natural language object retrieval in images, moment localization offers an interesting opportunity to model temporal dependencies and reasoning in text. We propose a new model that explicitly reasons about different temporal segments in a video, and shows that temporal context is important for localizing phrases which include temporal language. To benchmark whether our model, and other recent video localization models, can effectively reason about temporal language, we collect the novel TEMPOral reasoning in video and language (TEMPO) dataset. Our dataset consists of two parts: a dataset with real videos and template sentences (TEMPO - Template Language) which allows for controlled studies on temporal language, and a human language dataset which consists of temporal sentences annotated by humans (TEMPO - Human Language).

Recent years have witnessed an increasing interest in image-based question-answering (QA) tasks. However, due to data limitations, there has been much less work on video-based QA. In this paper, we present TVQA, a large-scale video QA dataset based on 6 popular TV shows. TVQA consists of 152,545 QA pairs from 21,793 clips, spanning over 460 hours of video. Questions are designed to be compositional in nature, requiring systems to jointly localize relevant moments within a clip, comprehend subtitle-based dialogue, and recognize relevant visual concepts. We provide analyses of this new dataset as well as several baselines and a multi-stream end-to-end trainable neural network framework for the TVQA task. The dataset is publicly available at http://tvqa.cs.unc.edu.

Understanding and reasoning about cooking recipes is a fruitful research direction towards enabling machines to interpret procedural text. In this work, we introduce RecipeQA, a dataset for multimodal comprehension of cooking recipes. It comprises of approximately 20K instructional recipes with multiple modalities such as titles, descriptions and aligned set of images. With over 36K automatically generated question-answer pairs, we design a set of comprehension and reasoning tasks that require joint understanding of images and text, capturing the temporal flow of events and making sense of procedural knowledge. Our preliminary results indicate that RecipeQA will serve as a challenging test bed and an ideal benchmark for evaluating machine comprehension systems. The data and leaderboard are available at http://hucvl.github.io/recipeqa.

Active learning identifies data points to label that are expected to be the most useful in improving a supervised model. Opportunistic active learning incorporates active learning into interactive tasks that constrain possible queries during interactions. Prior work has shown that opportunistic active learning can be used to improve grounding of natural language descriptions in an interactive object retrieval task. In this work, we use reinforcement learning for such an object retrieval task, to learn a policy that effectively trades off task completion with model improvement that would benefit future tasks.

In Visual Question Answering, most existing approaches adopt the pipeline of representing an image via pre-trained CNNs, and then using the uninterpretable CNN features in conjunction with the question to predict the answer. Although such end-to-end models might report promising performance, they rarely provide any insight, apart from the answer, into the VQA process. In this work, we propose to break up the end-to-end VQA into two steps: explaining and reasoning, in an attempt towards a more explainable VQA by shedding light on the intermediate results between these two steps. To that end, we first extract attributes and generate descriptions as explanations for an image. Next, a reasoning module utilizes these explanations in place of the image to infer an answer. The advantages of such a breakdown include: (1) the attributes and captions can reflect what the system extracts from the image, thus can provide some insights for the predicted answer; (2) these intermediate results can help identify the inabilities of the image understanding or the answer inference part when the predicted answer is wrong. We conduct extensive experiments on a popular VQA dataset and our system achieves comparable performance with the baselines, yet with added benefits of explanability and the inherent ability to further improve with higher quality explanations.

To approximately parse an unfamiliar language, it helps to have a treebank of a similar language. But what if the closest available treebank still has the wrong word order? We show how to (stochastically) permute the constituents of an existing dependency treebank so that its surface part-of-speech statistics approximately match those of the target language. The parameters of the permutation model can be evaluated for quality by dynamic programming and tuned by gradient descent (up to a local optimum). This optimization procedure yields trees for a new artificial language that resembles the target language. We show that delexicalized parsers for the target language can be successfully trained using such “made to order” artificial languages.

We introduce a method to reduce constituent parsing to sequence labeling. For each word w_t, it generates a label that encodes: (1) the number of ancestors in the tree that the words w_t and w_{t+1} have in common, and (2) the nonterminal symbol at the lowest common ancestor. We first prove that the proposed encoding function is injective for any tree without unary branches. In practice, the approach is made extensible to all constituency trees by collapsing unary branches. We then use the PTB and CTB treebanks as testbeds and propose a set of fast baselines. We achieve 90% F-score on the PTB test set, outperforming the Vinyals et al. (2015) sequence-to-sequence parser. In addition, sacrificing some accuracy, our approach achieves the fastest constituent parsing speeds reported to date on PTB by a wide margin.

We introduce novel dynamic oracles for training two of the most accurate known shift-reduce algorithms for constituent parsing: the top-down and in-order transition-based parsers. In both cases, the dynamic oracles manage to notably increase their accuracy, in comparison to that obtained by performing classic static training. In addition, by improving the performance of the state-of-the-art in-order shift-reduce parser, we achieve the best accuracy to date (92.0 F1) obtained by a fully-supervised single-model greedy shift-reduce constituent parser on the WSJ benchmark.

Unsupervised learning of syntactic structure is typically performed using generative models with discrete latent variables and multinomial parameters. In most cases, these models have not leveraged continuous word representations. In this work, we propose a novel generative model that jointly learns discrete syntactic structure and continuous word representations in an unsupervised fashion by cascading an invertible neural network with a structured generative prior. We show that the invertibility condition allows for efficient exact inference and marginal likelihood computation in our model so long as the prior is well-behaved. In experiments we instantiate our approach with both Markov and tree-structured priors, evaluating on two tasks: part-of-speech (POS) induction, and unsupervised dependency parsing without gold POS annotation. On the Penn Treebank, our Markov-structured model surpasses state-of-the-art results on POS induction. Similarly, we find that our tree-structured model achieves state-of-the-art performance on unsupervised dependency parsing for the difficult training condition where neither gold POS annotation nor punctuation-based constraints are available.

We present a complete, automated, and efficient approach for utilizing valency analysis in making dependency parsing decisions. It includes extraction of valency patterns, a probabilistic model for tagging these patterns, and a joint decoding process that explicitly considers the number and types of each token’s syntactic dependents. On 53 treebanks representing 41 languages in the Universal Dependencies data, we find that incorporating valency information yields higher precision and F1 scores on the core arguments (subjects and complements) and functional relations (e.g., auxiliaries) that we employ for valency analysis. Precision on core arguments improves from 80.87 to 85.43. We further show that our approach can be applied to an ostensibly different formalism and dataset, Tree Adjoining Grammar as extracted from the Penn Treebank; there, we outperform the previous state-of-the-art labeled attachment score by 0.7. Finally, we explore the potential of extending valency patterns beyond their traditional domain by confirming their helpfulness in improving PP attachment decisions.

Traditional approaches to the task of ACE event detection primarily regard multiple events in one sentence as independent ones and recognize them separately by using sentence-level information. However, events in one sentence are usually interdependent and sentence-level information is often insufficient to resolve ambiguities for some types of events. This paper proposes a novel framework dubbed as Hierarchical and Bias Tagging Networks with Gated Multi-level Attention Mechanisms (HBTNGMA) to solve the two problems simultaneously. Firstly, we propose a hierachical and bias tagging networks to detect multiple events in one sentence collectively. Then, we devise a gated multi-level attention to automatically extract and dynamically fuse the sentence-level and document-level information. The experimental results on the widely used ACE 2005 dataset show that our approach significantly outperforms other state-of-the-art methods.

Distantly-supervised Relation Extraction (RE) methods train an extractor by automatically aligning relation instances in a Knowledge Base (KB) with unstructured text. In addition to relation instances, KBs often contain other relevant side information, such as aliases of relations (e.g., founded and co-founded are aliases for the relation founderOfCompany). RE models usually ignore such readily available side information. In this paper, we propose RESIDE, a distantly-supervised neural relation extraction method which utilizes additional side information from KBs for improved relation extraction. It uses entity type and relation alias information for imposing soft constraints while predicting relations. RESIDE employs Graph Convolution Networks (GCN) to encode syntactic information from text and improves performance even when limited side information is available. Through extensive experiments on benchmark datasets, we demonstrate RESIDE’s effectiveness. We have made RESIDE’s source code available to encourage reproducible research.

Event extraction is of practical utility in natural language processing. In the real world, it is a common phenomenon that multiple events existing in the same sentence, where extracting them are more difficult than extracting a single event. Previous works on modeling the associations between events by sequential modeling methods suffer a lot from the low efficiency in capturing very long-range dependencies. In this paper, we propose a novel Jointly Multiple Events Extraction (JMEE) framework to jointly extract multiple event triggers and arguments by introducing syntactic shortcut arcs to enhance information flow and attention-based graph convolution networks to model graph information. The experiment results demonstrate that our proposed framework achieves competitive results compared with state-of-the-art methods.

The current leading paradigm for temporal information extraction from text consists of three phases: (1) recognition of events and temporal expressions, (2) recognition of temporal relations among them, and (3) time-line construction from the temporal relations. In contrast to the first two phases, the last phase, time-line construction, received little attention and is the focus of this work. In this paper, we propose a new method to construct a linear time-line from a set of (extracted) temporal relations. But more importantly, we propose a novel paradigm in which we directly predict start and end-points for events from the text, constituting a time-line without going through the intermediate step of prediction of temporal relations as in earlier work. Within this paradigm, we propose two models that predict in linear complexity, and a new training loss using TimeML-style annotations, yielding promising results.

Identifying the salience (i.e. importance) of discourse units is an important task in language understanding. While events play important roles in text documents, little research exists on analyzing their saliency status. This paper empirically studies Event Salience and proposes two salience detection models based on discourse relations. The first is a feature based salience model that incorporates cohesion among discourse units. The second is a neural model that captures more complex interactions between discourse units. In our new large-scale event salience corpus, both methods significantly outperform the strong frequency baseline, while our neural model further improves the feature based one by a large margin. Our analyses demonstrate that our neural model captures interesting connections between salience and discourse unit relations (e.g., scripts and frame structures).

Many efforts have been made to facilitate natural language processing tasks with pre-trained language models (LMs), and brought significant improvements to various applications. To fully leverage the nearly unlimited corpora and capture linguistic information of multifarious levels, large-size LMs are required; but for a specific task, only parts of these information are useful. Such large-sized LMs, even in the inference stage, may cause heavy computation workloads, making them too time-consuming for large-scale applications. Here we propose to compress bulky LMs while preserving useful information with regard to a specific task. As different layers of the model keep different information, we develop a layer selection method for model pruning using sparsity-inducing regularization. By introducing the dense connectivity, we can detach any layer without affecting others, and stretch shallow and wide LMs to be deep and narrow. In model training, LMs are learned with layer-wise dropouts for better robustness. Experiments on two benchmark datasets demonstrate the effectiveness of our method.

Despite the tremendous empirical success of neural models in natural language processing, many of them lack the strong intuitions that accompany classical machine learning approaches. Recently, connections have been shown between convolutional neural networks (CNNs) and weighted finite state automata (WFSAs), leading to new interpretations and insights. In this work, we show that some recurrent neural networks also share this connection to WFSAs. We characterize this connection formally, defining rational recurrences to be recurrent hidden state update functions that can be written as the Forward calculation of a finite set of WFSAs. We show that several recent neural models use rational recurrences. Our analysis provides a fresh view of these models and facilitates devising new neural architectures that draw inspiration from WFSAs. We present one such model, which performs better than two recent baselines on language modeling and text classification. Our results demonstrate that transferring intuitions from classical models like WFSAs can be an effective approach to designing and understanding neural models.

We present a data set for evaluating the grammaticality of the predictions of a language model. We automatically construct a large number of minimally different pairs of English sentences, each consisting of a grammatical and an ungrammatical sentence. The sentence pairs represent different variations of structure-sensitive phenomena: subject-verb agreement, reflexive anaphora and negative polarity items. We expect a language model to assign a higher probability to the grammatical sentence than the ungrammatical one. In an experiment using this data set, an LSTM language model performed poorly on many of the constructions. Multi-task training with a syntactic objective (CCG supertagging) improved the LSTM’s accuracy, but a large gap remained between its performance and the accuracy of human participants recruited online. This suggests that there is considerable room for improvement over LSTMs in capturing syntax in a language model.

We propose a large margin criterion for training neural language models. Conventionally, neural language models are trained by minimizing perplexity (PPL) on grammatical sentences. However, we demonstrate that PPL may not be the best metric to optimize in some tasks, and further propose a large margin formulation. The proposed method aims to enlarge the margin between the “good” and “bad” sentences in a task-specific sense. It is trained end-to-end and can be widely applied to tasks that involve re-scoring of generated text. Compared with minimum-PPL training, our method gains up to 1.1 WER reduction for speech recognition and 1.0 BLEU increase for machine translation.