COLING 2016 Papers

WebQuestions and SimpleQuestions are two benchmark data-sets commonly used in recent knowledge-based question answering (KBQA) work. Most questions in them are ‘simple’ questions which can be answered based on a single relation in the knowledge base. Such data-sets lack the capability of evaluating KBQA systems on complicated questions. Motivated by this issue, we release a new data-set, namely ComplexQuestions, aiming to measure the quality of KBQA systems on ‘multi-constraint’ questions which require multiple knowledge base relations to get the answer. Beside, we propose a novel systematic KBQA approach to solve multi-constraint questions. Compared to state-of-the-art methods, our approach not only obtains comparable results on the two existing benchmark data-sets, but also achieves significant improvements on the ComplexQuestions.

Aspect phrase grouping is an important task in aspect-level sentiment analysis. It is a challenging problem due to polysemy and context dependency. We propose an Attention-based Deep Distance Metric Learning (ADDML) method, by considering aspect phrase representation as well as context representation. First, leveraging the characteristics of the review text, we automatically generate aspect phrase sample pairs for distant supervision. Second, we feed word embeddings of aspect phrases and their contexts into an attention-based neural network to learn feature representation of contexts. Both aspect phrase embedding and context embedding are used to learn a deep feature subspace for measure the distances between aspect phrases for K-means clustering. Experiments on four review datasets show that the proposed method outperforms state-of-the-art strong baseline methods.

Sentiment analysis (SA) using code-mixed data from social media has several applications in opinion mining ranging from customer satisfaction to social campaign analysis in multilingual societies. Advances in this area are impeded by the lack of a suitable annotated dataset. We introduce a Hindi-English (Hi-En) code-mixed dataset for sentiment analysis and perform empirical analysis comparing the suitability and performance of various state-of-the-art SA methods in social media. In this paper, we introduce learning sub-word level representations in our LSTM (Subword-LSTM) architecture instead of character-level or word-level representations. This linguistic prior in our architecture enables us to learn the information about sentiment value of important morphemes. This also seems to work well in highly noisy text containing misspellings as shown in our experiments which is demonstrated in morpheme-level feature maps learned by our model. Also, we hypothesize that encoding this linguistic prior in the Subword-LSTM architecture leads to the superior performance. Our system attains accuracy 4-5% greater than traditional approaches on our dataset, and also outperforms the available system for sentiment analysis in Hi-En code-mixed text by 18%.

We address the task of targeted sentiment as a means of understanding the sentiment that students hold toward courses and instructors, as expressed by students in their comments. We introduce a new dataset consisting of student comments annotated for targeted sentiment and describe a system that can both identify the courses and instructors mentioned in student comments, as well as label the students’ sentiment toward those entities. Through several comparative evaluations, we show that our system outperforms previous work on a similar task.

The automated comparison of points of view between two politicians is a very challenging task, due not only to the lack of annotated resources, but also to the different dimensions participating to the definition of agreement and disagreement. In order to shed light on this complex task, we first carry out a pilot study to manually annotate the components involved in detecting agreement and disagreement. Then, based on these findings, we implement different features to capture them automatically via supervised classification. We do not focus on debates in dialogical form, but we rather consider sets of documents, in which politicians may express their position with respect to different topics in an implicit or explicit way, like during an electoral campaign. We create and make available three different datasets.

Sarcasm detection has been modeled as a binary document classification task, with rich features being defined manually over input documents. Traditional models employ discrete manual features to address the task, with much research effect being devoted to the design of effective feature templates. We investigate the use of neural network for tweet sarcasm detection, and compare the effects of the continuous automatic features with discrete manual features. In particular, we use a bi-directional gated recurrent neural network to capture syntactic and semantic information over tweets locally, and a pooling neural network to extract contextual features automatically from history tweets. Results show that neural features give improved accuracies for sarcasm detection, with different error distributions compared with discrete manual features.

Rumour stance classification, the task that determines if each tweet in a collection discussing a rumour is supporting, denying, questioning or simply commenting on the rumour, has been attracting substantial interest. Here we introduce a novel approach that makes use of the sequence of transitions observed in tree-structured conversation threads in Twitter. The conversation threads are formed by harvesting users’ replies to one another, which results in a nested tree-like structure. Previous work addressing the stance classification task has treated each tweet as a separate unit. Here we analyse tweets by virtue of their position in a sequence and test two sequential classifiers, Linear-Chain CRF and Tree CRF, each of which makes different assumptions about the conversational structure. We experiment with eight Twitter datasets, collected during breaking news, and show that exploiting the sequential structure of Twitter conversations achieves significant improvements over the non-sequential methods. Our work is the first to model Twitter conversations as a tree structure in this manner, introducing a novel way of tackling NLP tasks on Twitter conversations.

Sentiment analysis of short texts is challenging because of the limited contextual information they usually contain. In recent years, deep learning models such as convolutional neural networks (CNNs) and recurrent neural networks (RNNs) have been applied to text sentiment analysis with comparatively remarkable results. In this paper, we describe a jointed CNN and RNN architecture, taking advantage of the coarse-grained local features generated by CNN and long-distance dependencies learned via RNN for sentiment analysis of short texts. Experimental results show an obvious improvement upon the state-of-the-art on three benchmark corpora, MR, SST1 and SST2, with 82.28%, 51.50% and 89.95% accuracy, respectively.

With the development and the advancement of social networks, forums, blogs and online sales, a growing number of Arabs are expressing their opinions on the web. In this paper, a scheme of Arabic sentiment classification, which evaluates and detects the sentiment polarity from Arabic reviews and Arabic social media, is studied. We investigated in several architectures to build a quality neural word embeddings using a 3.4 billion words corpus from a collected 10 billion words web-crawled corpus. Moreover, a convolutional neural network trained on top of pre-trained Arabic word embeddings is used for sentiment classification to evaluate the quality of these word embeddings. The simulation results show that the proposed scheme outperforms the existed methods on 4 out of 5 balanced and unbalanced datasets.

Distributed word representation is an efficient method for capturing semantic and syntactic word relations. In this work, we introduce an extension to the continuous bag-of-words model for learning word representations efficiently by using implicit structure information. Instead of relying on a syntactic parser which might be noisy and slow to build, we compute weights representing probabilities of syntactic relations based on the Huffman softmax tree in an efficient heuristic. The constructed “implicit graphs” from these weights show that these weights contain useful implicit structure information. Extensive experiments performed on several word similarity and word analogy tasks show gains compared to the basic continuous bag-of-words model.

Recent trend in question answering (QA) systems focuses on using structured knowledge bases (KBs) to find answers. While these systems are able to provide more precise answers than information retrieval (IR) based QA systems, the natural incompleteness of KB inevitably limits the question scope that the system can answer. In this paper, we present a hybrid question answering (hybrid-QA) system which exploits both structured knowledge base and free text to answer a question. The main challenge is to recognize the meaning of a question using these two resources, i.e., structured KB and free text. To address this, we map relational phrases to KB predicates and textual relations simultaneously, and further develop an integer linear program (ILP) model to infer on these candidates and provide a globally optimal solution. Experiments on benchmark datasets show that our system can benefit from both structured KB and free text, outperforming the state-of-the-art systems.

A spatial information extraction system retrieves spatial entities and their relationships for geological searches and reasoning. Spatial information systems have been developed mainly for English text, e.g., through the SpaceEval competition. Some of the techniques are useful but not directly applicable to Korean text, because of linguistic differences and the lack of language resources. In this paper, we propose a Korean spatial entity extraction model and a spatial relation extraction model; the spatial entity extraction model uses word vectors to alleviate the over generation and the spatial relation extraction mod-el uses dependency parse labels to find the proper arguments in relations. Experiments with Korean text show that the two models are effective for spatial information extraction.

Answer selection is a core component in any question-answering systems. It aims to select correct answer sentences for a given question from a pool of candidate sentences. In recent years, many deep learning methods have been proposed and shown excellent results for this task. However, these methods typically require extensive parameter (and hyper-parameter) tuning, which give rise to efficiency issues for large-scale datasets, and potentially make them less portable across new datasets and domains (as re-tuning is usually required). In this paper, we propose an extremely efficient hybrid model (FastHybrid) that tackles the problem from both an accuracy and scalability point of view. FastHybrid is a light-weight model that requires little tuning and adaptation across different domains. It combines a fast deep model (which will be introduced in the method section) with an initial information retrieval model to effectively and efficiently handle answer selection. We introduce a new efficient attention mechanism in the hybrid model and demonstrate its effectiveness on several QA datasets. Experimental results show that although the hybrid uses no training data, its accuracy is often on-par with supervised deep learning techniques, while significantly reducing training and tuning costs across different domains.

Relation classification is the task of classifying the semantic relations between entity pairs in text. Observing that existing work has not fully explored using different representations for relation instances, especially in order to better handle the asymmetry of relation types, in this paper, we propose a neural network based method for relation classification that combines the raw sequence and the shortest dependency path representations of relation instances and uses mirror instances to perform pairwise relation classification. We evaluate our proposed models on the SemEval-2010 Task 8 dataset. The empirical results show that with two additional features, our model achieves the state-of-the-art result of F1 score of 85.7.

Term co-occurrence in a sentence or paragraph is a powerful and often overlooked feature for text matching in document retrieval. In our experiments with matching email-style query messages to webpages, such term co-occurrence helped greatly to filter and rank documents, compared to matching document-size bags-of-words. The paper presents the results of the experiments as well as a text-matching model where the query shapes the vector space, a document is modelled by two or three vectors in this vector space, and the query-document similarity score depends on the length of the vectors and the relationships between them.

We present an approach to mathematical information retrieval (MIR) that exploits a special kind of technical terminology, referred to as a mathematical type. In this paper, we present and evaluate a type detection mechanism and show its positive effect on the retrieval of research-level mathematics. Our best model, which performs query expansion with a type-aware embedding space, strongly outperforms standard IR models with state-of-the-art query expansion (vector space-based and language modelling-based), on a relatively new corpus of research-level queries.

The conventional solution for handling sparsely labelled data is extensive feature engineering. This is time consuming and task and domain specific. We present a novel approach for learning embedded features that aims to alleviate this problem. Our approach jointly learns embeddings at different levels of granularity (word, sentence and document) along with the class labels. The intuition is that topic semantics represented by embeddings at multiple levels results in better classification. We evaluate this approach in unsupervised and semi-supervised settings on two sparsely labelled classification tasks, outperforming the handcrafted models and several embedding baselines.

Aspect extraction identifies relevant features from a textual description of an entity, e.g., a phone, and is typically targeted to product descriptions, reviews, and other short texts as an enabling task for, e.g., opinion mining and information retrieval. Current aspect extraction methods mostly focus on aspect terms and often neglect interesting modifiers of the term or embed them in the aspect term without proper distinction. Moreover, flat syntactic structures are often assumed, resulting in inaccurate extractions of complex aspects. This paper studies the problem of structured aspect extraction, a variant of traditional aspect extraction aiming at a fine-grained extraction of complex (i.e., hierarchical) aspects. We propose an unsupervised and scalable method for structured aspect extraction consisting of statistical noun phrase clustering, cPMI-based noun phrase segmentation, and hierarchical pattern induction. Our evaluation shows a substantial improvement over existing methods in terms of both quality and computational efficiency.

Previous studies have highlighted the necessity for entity linking systems to capture the local entity-mention similarities and the global topical coherence. We introduce a novel framework based on convolutional neural networks and recurrent neural networks to simultaneously model the local and global features for entity linking. The proposed model benefits from the capacity of convolutional neural networks to induce the underlying representations for local contexts and the advantage of recurrent neural networks to adaptively compress variable length sequences of predictions for global constraints. Our evaluation on multiple datasets demonstrates the effectiveness of the model and yields the state-of-the-art performance on such datasets. In addition, we examine the entity linking systems on the domain adaptation setting that further demonstrates the cross-domain robustness of the proposed model.