EMNLP 2017 Papers

We present RACE, a new dataset for benchmark evaluation of methods in the reading comprehension task. Collected from the English exams for middle and high school Chinese students in the age range between 12 to 18, RACE consists of near 28,000 passages and near 100,000 questions generated by human experts (English instructors), and covers a variety of topics which are carefully designed for evaluating the students’ ability in understanding and reasoning. In particular, the proportion of questions that requires reasoning is much larger in RACE than that in other benchmark datasets for reading comprehension, and there is a significant gap between the performance of the state-of-the-art models (43%) and the ceiling human performance (95%). We hope this new dataset can serve as a valuable resource for research and evaluation in machine comprehension. The dataset is freely available at http://www.cs.cmu.edu/~glai1/data/race/ and the code is available at https://github.com/qizhex/RACE_AR_baselines.

Textbooks are rich sources of information. Harvesting structured knowledge from textbooks is a key challenge in many educational applications. As a case study, we present an approach for harvesting structured axiomatic knowledge from math textbooks. Our approach uses rich contextual and typographical features extracted from raw textbooks. It leverages the redundancy and shared ordering across multiple textbooks to further refine the harvested axioms. These axioms are then parsed into rules that are used to improve the state-of-the-art in solving geometry problems.

In this paper, we propose a new pipeline of word embedding for unsegmented languages, called segmentation-free word embedding, which does not require word segmentation as a preprocessing step. Unlike space-delimited languages, unsegmented languages, such as Chinese and Japanese, require word segmentation as a preprocessing step. However, word segmentation, that often requires manually annotated resources, is difficult and expensive, and unavoidable errors in word segmentation affect downstream tasks. To avoid these problems in learning word vectors of unsegmented languages, we consider word co-occurrence statistics over all possible candidates of segmentations based on frequent character n-grams instead of segmented sentences provided by conventional word segmenters. Our experiments of noun category prediction tasks on raw Twitter, Weibo, and Wikipedia corpora show that the proposed method outperforms the conventional approaches that require word segmenters.

Neural parsers have benefited from automatically labeled data via dependency-context word embeddings. We investigate training character embeddings on a word-based context in a similar way, showing that the simple method improves state-of-the-art neural word segmentation models significantly, beating tri-training baselines for leveraging auto-segmented data.

Even for common NLP tasks, sufficient supervision is not available in many languages – morphological tagging is no exception. In the work presented here, we explore a transfer learning scheme, whereby we train character-level recurrent neural taggers to predict morphological taggings for high-resource languages and low-resource languages together. Learning joint character representations among multiple related languages successfully enables knowledge transfer from the high-resource languages to the low-resource ones.

We introduce the task of book structure labeling: segmenting and assigning a fixed category (such as Table of Contents, Preface, Index) to the document structure of printed books. We manually annotate the page-level structural categories for a large dataset totaling 294,816 pages in 1,055 books evenly sampled from 1750-1922, and present empirical results comparing the performance of several classes of models. The best-performing model, a bidirectional LSTM with rich features, achieves an overall accuracy of 95.8 and a class-balanced macro F-score of 71.4.

A recent study by Plank et al. (2016) found that LSTM-based PoS taggers considerably improve over the current state-of-the-art when evaluated on the corpora of the Universal Dependencies project that use a coarse-grained tagset. We replicate this study using a fresh collection of 27 corpora of 21 languages that are annotated with fine-grained tagsets of varying size. Our replication confirms the result in general, and we additionally find that the advantage of LSTMs is even bigger for larger tagsets. However, we also find that for the very large tagsets of morphologically rich languages, hand-crafted morphological lexicons are still necessary to reach state-of-the-art performance.

We introduce a novel sub-character architecture that exploits a unique compositional structure of the Korean language. Our method decomposes each character into a small set of primitive phonetic units called jamo letters from which character- and word-level representations are induced. The jamo letters divulge syntactic and semantic information that is difficult to access with conventional character-level units. They greatly alleviate the data sparsity problem, reducing the observation space to 1.6% of the original while increasing accuracy in our experiments. We apply our architecture to dependency parsing and achieve dramatic improvement over strong lexical baselines.

The generation of complex derived word forms has been an overlooked problem in NLP; we fill this gap by applying neural sequence-to-sequence models to the task. We overview the theoretical motivation for a paradigmatic treatment of derivational morphology, and introduce the task of derivational paradigm completion as a parallel to inflectional paradigm completion. State-of-the-art neural models adapted from the inflection task are able to learn the range of derivation patterns, and outperform a non-neural baseline by 16.4%. However, due to semantic, historical, and lexical considerations involved in derivational morphology, future work will be needed to achieve performance parity with inflection-generating systems.

This paper presents a model for Arabic morphological disambiguation based on Recurrent Neural Networks (RNN). We train Long Short-Term Memory (LSTM) cells in several configurations and embedding levels to model the various morphological features. Our experiments show that these models outperform state-of-the-art systems without explicit use of feature engineering. However, adding learning features from a morphological analyzer to model the space of possible analyses provides additional improvement. We make use of the resulting morphological models for scoring and ranking the analyses of the morphological analyzer for morphological disambiguation. The results show significant gains in accuracy across several evaluation metrics. Our system results in 4.4% absolute increase over the state-of-the-art in full morphological analysis accuracy (30.6% relative error reduction), and 10.6% (31.5% relative error reduction) for out-of-vocabulary words.

Traditionally, word segmentation (WS) adopts the single-grained formalism, where a sentence corresponds to a single word sequence. However, Sproat et al. (1997) show that the inter-native-speaker consistency ratio over Chinese word boundaries is only 76%, indicating single-grained WS (SWS) imposes unnecessary challenges on both manual annotation and statistical modeling. Moreover, WS results of different granularities can be complementary and beneficial for high-level applications. This work proposes and addresses multi-grained WS (MWS). We build a large-scale pseudo MWS dataset for model training and tuning by leveraging the annotation heterogeneity of three SWS datasets. Then we manually annotate 1,500 test sentences with true MWS annotations. Finally, we propose three benchmark approaches by casting MWS as constituent parsing and sequence labeling. Experiments and analysis lead to many interesting findings.

Determining semantic textual similarity is a core research subject in natural language processing. Since vector-based models for sentence representation often use shallow information, capturing accurate semantics is difficult. By contrast, logical semantic representations capture deeper levels of sentence semantics, but their symbolic nature does not offer graded notions of textual similarity. We propose a method for determining semantic textual similarity by combining shallow features with features extracted from natural deduction proofs of bidirectional entailment relations between sentence pairs. For the natural deduction proofs, we use ccg2lambda, a higher-order automatic inference system, which converts Combinatory Categorial Grammar (CCG) derivation trees into semantic representations and conducts natural deduction proofs. Experiments show that our system was able to outperform other logic-based systems and that features derived from the proofs are effective for learning textual similarity.

Many modern NLP systems rely on word embeddings, previously trained in an unsupervised manner on large corpora, as base features. Efforts to obtain embeddings for larger chunks of text, such as sentences, have however not been so successful. Several attempts at learning unsupervised representations of sentences have not reached satisfactory enough performance to be widely adopted. In this paper, we show how universal sentence representations trained using the supervised data of the Stanford Natural Language Inference datasets can consistently outperform unsupervised methods like SkipThought vectors on a wide range of transfer tasks. Much like how computer vision uses ImageNet to obtain features, which can then be transferred to other tasks, our work tends to indicate the suitability of natural language inference for transfer learning to other NLP tasks. Our encoder is publicly available.

We present a feature vector formation technique for documents - Sparse Composite Document Vector (SCDV) - which overcomes several shortcomings of the current distributional paragraph vector representations that are widely used for text representation. In SCDV, word embeddings are clustered to capture multiple semantic contexts in which words occur. They are then chained together to form document topic-vectors that can express complex, multi-topic documents. Through extensive experiments on multi-class and multi-label classification tasks, we outperform the previous state-of-the-art method, NTSG. We also show that SCDV embeddings perform well on heterogeneous tasks like Topic Coherence, context-sensitive Learning and Information Retrieval. Moreover, we achieve a significant reduction in training and prediction times compared to other representation methods. SCDV achieves best of both worlds - better performance with lower time and space complexity.

In this paper, we introduce a new distributional method for modeling predicate-argument thematic fit judgments. We use a syntax-based DSM to build a prototypical representation of verb-specific roles: for every verb, we extract the most salient second order contexts for each of its roles (i.e. the most salient dimensions of typical role fillers), and then we compute thematic fit as a weighted overlap between the top features of candidate fillers and role prototypes. Our experiments show that our method consistently outperforms a baseline re-implementing a state-of-the-art system, and achieves better or comparable results to those reported in the literature for the other unsupervised systems. Moreover, it provides an explicit representation of the features characterizing verb-specific semantic roles.

Computerized generation of humor is a notoriously difficult AI problem. We develop an algorithm called Libitum that helps humans generate humor in a Mad Lib, which is a popular fill-in-the-blank game. The algorithm is based on a machine learned classifier that determines whether a potential fill-in word is funny in the context of the Mad Lib story. We use Amazon Mechanical Turk to create ground truth data and to judge humor for our classifier to mimic, and we make this data freely available. Our testing shows that Libitum successfully aids humans in filling in Mad Libs that are usually judged funnier than those filled in by humans with no computerized help. We go on to analyze why some words are better than others at making a Mad Lib funny.

In this paper we explore the effect of architectural choices on learning a variational autoencoder (VAE) for text generation. In contrast to the previously introduced VAE model for text where both the encoder and decoder are RNNs, we propose a novel hybrid architecture that blends fully feed-forward convolutional and deconvolutional components with a recurrent language model. Our architecture exhibits several attractive properties such as faster run time and convergence, ability to better handle long sequences and, more importantly, it helps to avoid the issue of the VAE collapsing to a deterministic model.

This paper presents a Generative Adversarial Network (GAN) to model single-turn short-text conversations, which trains a sequence-to-sequence (Seq2Seq) network for response generation simultaneously with a discriminative classifier that measures the differences between human-produced responses and machine-generated ones. In addition, the proposed method introduces an approximate embedding layer to solve the non-differentiable problem caused by the sampling-based output decoding procedure in the Seq2Seq generative model. The GAN setup provides an effective way to avoid noninformative responses (a.k.a “safe responses”), which are frequently observed in traditional neural response generators. The experimental results show that the proposed approach significantly outperforms existing neural response generation models in diversity metrics, with slight increases in relevance scores as well, when evaluated on both a Mandarin corpus and an English corpus.

We propose a new sentence simplification task (Split-and-Rephrase) where the aim is to split a complex sentence into a meaning preserving sequence of shorter sentences. Like sentence simplification, splitting-and-rephrasing has the potential of benefiting both natural language processing and societal applications. Because shorter sentences are generally better processed by NLP systems, it could be used as a preprocessing step which facilitates and improves the performance of parsers, semantic role labellers and machine translation systems. It should also be of use for people with reading disabilities because it allows the conversion of longer sentences into shorter ones. This paper makes two contributions towards this new task. First, we create and make available a benchmark consisting of 1,066,115 tuples mapping a single complex sentence to a sequence of sentences expressing the same meaning. Second, we propose five models (vanilla sequence-to-sequence to semantically-motivated models) to understand the difficulty of the proposed task.