EMNLP 2015 Papers

We use character-based statistical machine translation in order to correct user search queries in the e-commerce domain. The training data is automatically extracted from event logs where users re-issue their search queries with potentially corrected spelling within the same session. We show results on a test set which was annotated by humans and compare against online autocorrection capabilities of three additional web sites. Overall, the methods presented in this paper outperform fully productized spellchecking and autocorrection services in terms of accuracy and F1 score. We also propose novel evaluation steps based on retrieved search results of the corrected queries in terms of quantity and relevance.

Engaging in a debate with oneself or others to take decisions is an integral part of our day-today life. A debate on a topic (say, use of performance enhancing drugs) typically proceeds by one party making an assertion/claim (say, PEDs are bad for health) and then providing an evidence to support the claim (say, a 2006 study shows that PEDs have psychiatric side effects). In this work, we propose the task of automatically detecting such evidences from unstructured text that support a given claim. This task has many practical applications in decision support and persuasion enhancement in a wide range of domains. We first introduce an extensive benchmark data set tailored for this task, which allows training statistical models and assessing their performance. Then, we suggest a system architecture based on supervised learning to address the evidence detection task. Finally, promising experimental results are reported.

Most of the current automated essay scoring (AES) systems are trained using manually graded essays from a specific prompt. These systems experience a drop in accuracy when used to grade an essay from a different prompt. Obtaining a large number of manually graded essays each time a new prompt is introduced is costly and not viable. We propose domain adaptation as a solution to adapt an AES system from an initial prompt to a new prompt. We also propose a novel domain adaptation technique that uses Bayesian linear ridge regression. We evaluate our domain adaptation technique on the publicly available Automated Student Assessment Prize (ASAP) dataset and show that our proposed technique is a competitive default domain adaptation algorithm for the AES task.

Social media represents a rich source of upto-date information about events such as incidents. The sheer amount of available information makes machine learning approaches a necessity for further processing. This learning problem is often concerned with regionally restricted datasets such as data from only one city. Because social media data such as tweets varies considerably across different cities, the training of efficient models requires labeling data from each city of interest, which is costly and time consuming. In this study, we investigate which features are most suitable for training generalizable models, i.e., models that show good performance across different datasets. We reimplemented the most popular features from the state of the art in addition to other novel approaches, and evaluated them on data from ten different cities. We show that many sophisticated features are not necessarily valuable for training a generalized model and are outperformed by classic features such as plain word-n-grams and character-n-grams.

This paper proposes a graph-based readability assessment method using word coupling. Compared to the state-of-theart methods such as the readability formulae, the word-based and feature-based methods, our method develops a coupled bag-of-words model which combines the merits of word frequencies and text features. Unlike the general bag-of-words model which assumes words are independent, our model correlates the words based on their similarities on readability. By applying TF-IDF (Term Frequency and Inverse Document Frequency), the coupled TF-IDF matrix is built, and used in the graph-based classification framework, which involves graph building, merging and label propagation. Experiments are conducted on both English and Chinese datasets. The results demonstrate both effectiveness and potential of the method.

In recent years, the task of recommending hashtags for microblogs has been given increasing attention. Various methods have been proposed to study the problem from different aspects. However, most of the recent studies have not considered the differences in the types or uses of hashtags. In this paper, we introduce a novel nonparametric Bayesian method for this task. Based on the Dirichlet Process Mixture Models (DPMM), we incorporate the type of hashtag as a hidden variable. The results of experiments on the data collected from a real world microblogging service demonstrate that the proposed method outperforms stateof-the-art methods that do not consider these aspects. By taking these aspects into consideration, the relative improvement of the proposed method over the state-of-theart methods is around 12.2% in F1- score.

We propose a summarization approach for scientific articles which takes advantage of citation-context and the document discourse model. While citations have been previously used in generating scientific summaries, they lack the related context from the referenced article and therefore do not accurately reflect the article’s content. Our method overcomes the problem of inconsistency between the citation summary and the article’s content by providing context for each citation. We also leverage the inherent scientific article’s discourse for producing better summaries. We show that our proposed method effectively improves over existing summarization approaches (greater than 30% improvement over the best performing baseline) in terms of ROUGE scores on TAC2014 scientific summarization dataset. While the dataset we use for evaluation is in the biomedical domain, most of our approaches are general and therefore adaptable to other domains.

Summarization based on text extraction is inherently limited, but generation-style abstractive methods have proven challenging to build. In this work, we propose a fully data-driven approach to abstractive sentence summarization. Our method utilizes a local attention-based model that generates each word of the summary conditioned on the input sentence. While the model is structurally simple, it can easily be trained end-to-end and scales to a large amount of training data. The model shows significant performance gains on the DUC-2004 shared task compared with several strong baselines.

Syntactic language models and N-gram language models have both been used in word ordering. In this paper, we give an empirical comparison between N-gram and syntactic language models on word order task. Our results show that the quality of automatically-parsed training data has a relatively small impact on syntactic models. Both of syntactic and N-gram models can benefit from large-scale raw text. Compared with N-gram models, syntactic models give overall better performance, but they require much more training time. In addition, the two models lead to different error distributions in word ordering. A combination of the two models integrates the advantages of each model, achieving the best result in a standard benchmark.

We present an LSTM approach to deletion-based sentence compression where the task is to translate a sentence into a sequence of zeros and ones, corresponding to token deletion decisions. We demonstrate that even the most basic version of the system, which is given no syntactic information (no PoS or NE tags, or dependencies) or desired compression length, performs surprisingly well: around 30% of the compressions from a large test set could be regenerated. We compare the LSTM system with a competitive baseline which is trained on the same amount of data but is additionally provided with all kinds of linguistic features. In an experiment with human raters the LSTMbased model outperforms the baseline achieving 4.5 in readability and 3.8 in informativeness.

We present extensions to a continuousstate dependency parsing method that makes it applicable to morphologically rich languages. Starting with a highperformance transition-based parser that uses long short-term memory (LSTM) recurrent neural networks to learn representations of the parser state, we replace lookup-based word representations with representations constructed from the orthographic representations of the words, also using LSTMs. This allows statistical sharing across word forms that are similar on the surface. Experiments for morphologically rich languages show that the parsing model benefits from incorporating the character-based encodings of words.

Accurate dependency parsing requires large treebanks, which are only available for a few languages. We propose a method that takes advantage of shared structure across languages to build a mature parser using less training data. We propose a model for learning a shared “universal” parser that operates over an interlingual continuous representation of language, along with language-specific mapping components. Compared with supervised learning, our methods give a consistent 8-10% improvement across several treebanks in low-resource simulations.

We present a novel method for the crosslingual transfer of dependency parsers. Our goal is to induce a dependency parser in a target language of interest without any direct supervision: instead we assume access to parallel translations between the target and one or more source languages, and to supervised parsers in the source language(s). Our key contributions are to show the utility of dense projected structures when training the target language parser, and to introduce a novel learning algorithm that makes use of dense structures. Results on several languages show an absolute improvement of 5.51% in average dependency accuracy over the state-of-the-art method of (Ma and Xia, 2014). Our average dependency accuracy of 82.18% compares favourably to the accuracy of fully supervised methods.

Path queries on a knowledge graph can be used to answer compositional questions such as "What languages are spoken by people living in Lisbon?". However, knowledge graphs often have missing facts (edges) which disrupts path queries. Recent models for knowledge base completion impute missing facts by embedding knowledge graphs in vector spaces. We show that these models can be recursively applied to answer path queries, but that they suffer from cascading errors. This motivates a new "compositional" training objective, which dramatically improves all models' ability to answer path queries, in some cases more than doubling accuracy. On a standard knowledge base completion task, we also demonstrate that compositional training acts as a novel form of structural regularization, reliably improving performance across all base models (reducing errors by up to 43%) and achieving new state-of-the-art results.

Latent Dirichlet allocation (LDA) is a popular topic modeling technique for exploring hidden topics in text corpora. Increasingly, topic modeling needs to scale to larger topic spaces and use richer forms of prior knowledge, such as word correlations or document labels. However, inference is cumbersome for LDA models with prior knowledge. As a result, LDA models that use prior knowledge only work in small-scale scenarios. In this work, we propose a factor graph framework, Sparse Constrained LDA (SC-LDA), for efficiently incorporating prior knowledge into LDA. We evaluate SC-LDA’s ability to incorporate word correlation knowledge and document label knowledge on three benchmark datasets. Compared to several baseline methods, SC-LDA achieves comparable performance but is significantly faster.

We present a comprehensive study of evaluation methods for unsupervised embedding techniques that obtain meaningful representations of words from text. Different evaluations result in different orderings of embedding methods, calling into question the common assumption that there is one single optimal vector representation. We present new evaluation techniques that directly compare embeddings with respect to specific queries. These methods reduce bias, provide greater insight, and allow us to solicit data-driven relevance judgments rapidly and accurately through crowdsourcing.

In order to reduce noise in training data, most natural language crowdsourcing annotation tasks gather redundant labels and aggregate them into an integrated label, which is provided to the classifier. However, aggregation discards potentially useful information from linguistically ambiguous instances. For five natural language tasks, we pass item agreement on to the task classifier via soft labeling and low-agreement filtering of the training dataset. We find a statistically significant benefit from low item agreement training filtering in four of our five tasks, and no systematic benefit from soft labeling.

Performing link prediction in Knowledge Bases (KBs) with embedding-based models, like with the model TransE (Bordes et al., 2013) which represents relationships as translations in the embedding space, have shown promising results in recent years. Most of these works are focused on modeling single relationships and hence do not take full advantage of the graph structure of KBs. In this paper, we propose an extension of TransE that learns to explicitly model composition of relationships via the addition of their corresponding translation vectors. We show empirically that this allows to improve performance for predicting single relationships as well as compositions of pairs of them.

There are two main types of word representations: low-dimensional embeddings and high-dimensional distributional vectors, in which each dimension corresponds to a context word. In this paper, we initialize an embedding-learning model with distributional vectors. Evaluation on word similarity shows that this initialization significantly increases the quality of embeddings for rare words.