ACL 2016 Papers

We present a named entity recognition (NER) system for tagging fiction: LitNER . Relative to more traditional approaches, LitNER has two important properties: (1) it makes no use of hand-tagged data or gazetteers, instead it boot-straps a model from term clusters; and (2) it leverages multiple instances of the same name in a text. Our experiments show it to substantially outperform off-the-shelf supervised NER systems.

We present a new, structured approach to text simplification using conditional random fields over top-down traversals of dependency graphs that jointly predicts possible compressions and paraphrases. Our model reaches readability scores comparable to word-based compression approaches across a range of metrics and human judgements while maintaining more of the important information.

Microblogging sites have emerged as major platforms for bloggers to create and consume posts as well as to follow other bloggers and get informed of their updates. Due to the large number of users, and the huge amount of posts they create, it becomes extremely difficult to identify relevant and interesting blog posts. In this paper, we propose a novel convex collective matrix completion (CCMC) method that effectively utilizes user-item matrix and incorporates additional user activity and topic-based signals to recommend relevant content. The key advantage of CCMC over existing methods is that it can obtain a globally optimal solution and can easily scale to large-scale matrices us-ing Hazan’s algorithm. To the best of our knowledge, this is the first work which applies and studies CCMC as a recommendation method in social media. We conduct a large scale study and show significant improvement over existing state-of-the-art approaches.

Access to data is critical to any machine learning component aimed at training an accurate predictive model. In reality, data is often a subject of technical and legal constraints. Data may contain sensitive topics and data owners are often reluctant to share them. Instead of access to data, they make available decision making procedures to enable predictions on new data. Under the black box classifier constraint, we build an effective domain adaptation technique which adapts classifier predictions in a transductive setting. We run experiments on text categorization datasets and show that significant gains can be achieved, especially in the unsupervised case where no labels are available in the target domain.

Optimism is linked to various personality factors as well as both psychological and physical health, but how does it relate to the way a person tweets? We analyze the online activity of a set of Twitter users in order to determine how well machine learning algorithms can detect a person’s outlook on life by reading their tweets. A sample of tweets from each user is manually annotated in order to establish ground truth labels, and classifiers are trained to distinguish between optimistic and pessimistic users. Our results suggest that the words in people’s tweets provide ample evidence to identify them as optimists, pessimists, or somewhere in between. Additionally, several applications of these trained models are explored.

Writing style allows NLP tools to adjust to the traits of an author. In this paper, we explore the relation between stylistic and syntactic features and authors’ age and income. We confirm our hypothesis that for numerous feature types writing style is predictive of income even beyond age. We analyze the predictive power of writing style features in a regression task on two data sets of around 5,000 Twitter users each. Additionally, we use our validated features to study daily variations in writing style of users from distinct income groups. Temporal stylistic patterns not only provide novel psychological insight into user behavior, but are useful for future research and applications in social media.

People find it difficult to name odors and flavors. In blind tests with everyday smells and tastes like orange or chocolate, only 50% is recognized and described correctly. Certain experts like wine reviewers are trained in recognizing and reporting on odor and flavor on a daily basis and they have a much larger vocabulary than lay people. In this research, we want to examine whether expert wine tasters provide consistent descriptions in terms of perceived sensory attributes of wine, both across various wine types and colors. We collected a corpus of wine reviews and performed preliminary experiments to analyse the semantic fields of „flavor” and „odor” in wine reviews. To do so, we applied distributional methods as well as pattern-based approaches. In addition, we show the first results of automatically predicting „color” and „region” of a particular wine, solely based on the reviewer’s text. Our classifiers perform very well when predicting red and white wines, whereas it seems more challenging to distinguish rose wines.

We investigate the use of hierarchical phrase-based SMT lattices in end-to-end neural machine translation (NMT). Weight pushing transforms the Hiero scores for complete translation hypotheses, with the full translation grammar score and full n-gram language model score, into posteriors compatible with NMT predictive probabilities. With a slightly modified NMT beam-search decoder we find gains over both Hiero and NMT decoding alone, with practical advantages in extending NMT to very large input and output vocabularies.

Recently, neural network models have achieved consistent improvements in statistical machine translation. However, most networks only use one-hot encoded input vectors of words as their input. In this work, we investigated the exponentially decaying bag-of-words input features for feed-forward neural network translation models and proposed to train the decay rates along with other weight parameters. This novel bag-of-words model improved our phrase-based state-of-the-art system, which already includes a neural network translation model, by up to 0.5% B LEU and 0.6% T ER on three different translation tasks and even achieved a similar performance to the bidirectional LSTM translation model.

We address the problem of automatically cleaning a large-scale Translation Memory (TM) in a fully unsupervised fashion, i.e. without human-labelled data. We approach the task by: i) designing a set of features that capture the similarity between two text segments in different languages, ii) use them to induce reliable training labels for a subset of the translation units (TUs) contained in the TM, and iii) use the automatically labelled data to train an ensemble of binary classifiers. We apply our method to clean a test set composed of 1,000 TUs randomly extracted from the English-Italian version of MyMemory, the world’s largest public TM. Our results show competitive performance not only against a strong baseline that exploits machine translation, but also against a state-of-the-art method that relies on human-labelled data.

We present a neural network based automatic post-editing (APE) system to improve raw machine translation (MT) output. Our neural model of APE (NNAPE) is based on a bidirectional recurrent neural network (RNN) model and consists of an encoder that encodes an MT output into a fixed-length vector from which a de-coder provides a post-edited (PE) translation. APE translations produced by NNAPE show statistically significant improvements of 3.96, 2.68 and 1.35 BLEU points absolute over the original MT, phrase-based APE and hierarchical APE outputs, respectively. Furthermore, human evaluation shows that the NNAPE generated PE translations are much better than the original MT output.

Constrained translation has improved statistical machine translation (SMT) by combining it with translation memory (TM) at sentence-level. In this paper, we propose using a constrained word lattice, which encodes input phrases and TM constraints together, to combine SMT and TM at phrase-level. Experiments on English‐ Chinese and English‐French show that our approach is significantly better than previous combination methods, including sentence-level constrained translation and a recent phrase-level combination.

Text from social media provides a set of challenges that can cause traditional NLP approaches to fail. Informal language, spelling errors, abbreviations, and special characters are all commonplace in these posts, leading to a prohibitively large vocabulary size for word-level approaches. We propose a character composition model, tweet2vec, which finds vector-space representations of whole tweets by learning complex, non-local dependencies in character sequences. The proposed model outperforms a word-level baseline at predicting user-annotated hashtags associated with the posts, doing significantly better when the input contains many out-of-vocabulary words or unusual character sequences. Our tweet2vec encoder is publicly available.

Decision-making is often dependent on uncertain data, e.g. data associated with confidence scores or probabilities. We present a comparison of different information presentations for uncertain data and, for the first time, measure their effects on human decision-making. We show that the use of Natural Language Generation (NLG) improves decision-making under uncertainty, compared to state-of-the-art graphical-based representation methods. In a task-based study with 442 adults, we found that presentations using NLG lead to 24% better decision-making on average than the graphical presentations, and to 44% better decision-making when NLG is combined with graphics. We also show that women achieve significantly better results when presented with NLG output (an 87% increase on average compared to graphical presentations).

This paper addresses an automatic classification of preposition types in German, comparing hard and soft clustering approaches and various window- and syntax-based co-occurrence features. We show that (i) the semantically most salient preposition features (i.e., subcategorised nouns) are the most successful, and that (ii) soft clustering approaches are required for the task but reveal quite different attitudes towards predicting ambiguity.

We present a new annotation method for collecting data on relation inference in context. We convert the inference task to one of simple factoid question answering, allowing us to easily scale up to 16,000 high-quality examples. Our method corrects a major bias in previous evaluations, making our dataset much more realistic.

Finding quality descriptions on the web, such as those found in Wikipedia articles, of newer companies can be difficult: search engines show many pages with varying relevance, while multi-document summarization algorithms find it difficult to distinguish between core facts and other information such as news stories. In this paper, we propose an entity-focused, hybrid generation approach to automatically produce descriptions of previously unseen companies, and show that it outperforms a strong summarization baseline.

In this paper, we propose a method for referring to the real world to improve named entity recognition (NER) specialized for a domain. Our method adds a stacked autoencoder to a text-based deep neural network for NER. We first train the stacked auto-encoder only from the real world information, then the entire deep neural network from sentences annotated with NEs and accompanied by real world information. In our experiments, we took Japanese chess as the example. The dataset consists of pairs of a game state and commentary sentences about it annotated with gamespecific NE tags. We conducted NER experiments and showed that referring to the real world improves the NER accuracy.

In all previous work on deep multi-task learning we are aware of, all task supervisions are on the same (outermost) layer. We present a multi-task learning architecture with deep bi-directional RNNs, where different tasks supervision can happen at different layers. We present experiments in syntactic chunking and CCG supertagging, coupled with the additional task of POS-tagging. We show that it is consistently better to have POS supervision at the innermost rather than the outermost layer. We argue that this is because “lowlevel” tasks are better kept at the lower layers, enabling the higher-level tasks to make use of the shared representation of the lower-level tasks. Finally, we also show how this architecture can be used for domain adaptation.