Retrieving similar questions is very important in community-based question answering. A major challenge is the lexical gap in sentence matching. In this paper, we propose a convolutional neural tensor network architecture to encode the sentences in semantic space and model their interactions with a tensor layer. Our model integrates sentence modeling and semantic matching into a single model, which can not only capture the useful information with convolutional and pooling layers, but also learn the matching metrics between the question and its answer. Besides, our model is a general architecture, with no need for the other knowledge such as lexical or syntactic analysis. The experimental results shows that our method outperforms the other methods on two matching tasks.

We propose a graph-based method for extractive single-document summarization which considers importance, non-redundancy and local coherence simultaneously. We represent input documents by means of a bipartite graph consisting of sentence and entity nodes. We rank sentences on the basis of importance by applying a graph-based ranking algorithm to this graph and ensure non-redundancy and local coherence of the summary by means of an optimization step. Our graph based method is applied to scientific articles from the journal PLOS Medicine. We use human judgements to evaluate the coherence of our summaries. We compare ROUGE scores and human judgements for coherence of different systems on scientific articles. Our method performs considerably better than other systems on this data. Also, our graph-based summarization technique achieves state-of-the-art results on DUC 2002 data. Incorporating our local coherence measure always achieves the best results.

Aspect extraction aims to extract fine-grained opinion targets from opinion texts. Recent work has shown that the syntactical approach, which employs rules about grammar dependency relations between opinion words and aspects, performs quite well. This approach is highly desirable in practice because it is unsupervised and domain independent. However, the rules need to be carefully selected and tuned manually so as not to produce too many errors. Although it is easy to evaluate the accuracy of each rule automatically, it is not easy to select a set of rules that produces the best overall result due to the overlapping coverage of the rules. In this paper, we propose a novel method to select an effective set of rules. To our knowledge, this is the first work that selects rules automatically. Our experiment results show that the proposed method can select a subset of a given rule set to achieve significantly better results than the full rule set and the existing state-of-the-art CRF-based supervised method.

Distributed word representations have a rising interest in NLP community. Most of existing models assume only one vector for each individual word, which ignores polysemy and thus degrades their effectiveness for downstream tasks. To address this problem, some recent work adopts multi-prototype models to learn multiple embeddings per word type. In this paper, we distinguish the different senses of each word by their latent topics. We present a general architecture to learn the word and topic embeddings efficiently, which is an extension to the Skip-Gram model and can model the interaction between words and topics simultaneously. The experiments on the word similarity and text classification tasks show our model outperforms state-of-the-art methods.

Domain adaptation aims at learning robust classifiers across domains using labeled data from a source domain. Representation learning methods, which project the original features to a new feature space, have been proved to be quite effective for this task. However, these unsupervised methods neglect the domain information of the input and are not specialized for the classification task. In this work, we address two key factors to guide the representation learning process for domain adaptation of sentiment classification — one is domain supervision, enforcing the learned representation to better predict the domain of an input, and the other is sentiment supervision which utilizes the source domain sentiment labels to learn sentiment-favorable representations. Experimental results show that these two factors significantly improve the proposed models as expected.

Text normalization and part-of-speech (POS) tagging for social media data have been investigated recently, however, prior work has treated them separately. In this paper, we propose a joint Viterbi decoding process to determine each token’s POS tag and non-standard token’s correct form at the same time. In order to evaluate our approach, we create two new data sets with POS tag labels and non-standard tokens' correct forms. This is the first data set with such annotation. The experiment results demonstrate the effect of non-standard words on POS tagging, and also show that our proposed methods perform better than the state-of-the-art systems in both POS tagging and normalization

Most word embedding methods take a word as a basic unit and learn embeddings according to words' external contexts, ignoring the internal structures of words. However, in some languages such as Chinese, a word is usually composed of several  characters and contains rich internal information. The semantic meaning of a word is also related to the meanings of its composing characters. Hence, we take Chinese for example, and present a character-enhanced word embedding model (CWE). In order to address the issues of character ambiguity and non-compositional words, we propose multiple-prototype character embeddings and an effective word selection method. We evaluate the effectiveness of CWE on word relatedness computation and analogical reasoning. The results show that CWE outperforms other baseline methods which ignore internal character information.

We present a supervised framework for expanding an opinion lexicon for tweets. The lexicon contains part-of-speech (POS) disambiguated entries with a three-dimensional probability distribution for positive, negative, and neutral polarities. To obtain this distribution using machine learning, we propose word-level attributes based on POS tags and information calculated from streams of emoticon-annotated tweets. Our experimental results show that our method outperforms the three-dimensional word-level polarity classification performance obtained by semantic orientation, a state-of-the-art measure for establishing world-level sentiment.

During the last few years, the investigation of methodologies to automatically detect and characterise the figurative traits of textual contents has attracted a growing interest. Indeed, the capability to correctly deal with figurative language and more specifically with satire is fundamental to build robust approaches in several sub-fields of Artificial Intelligence including Sentiment Analysis and Affective Computing. In this paper we investigate the automatic detection of Tweets that advertise satirical news in English, Spanish and Italian. To this purpose we present a system that models Tweets from different languages by a set of language independent features that describe lexical, semantic and usage-related properties of the words of each Tweet. We approach the satire identification problem as binary classification of Tweets as satirical or not satirical messages. We test the performance of our system by performing experiments of both monolingual and cross-language classifications, evaluating the satire detection effectiveness of our features.Our system outperforms a word-based baseline and it is able to recognise if a news in Twitter is satirical or not with good accuracy. Moreover, we analyse the behaviour of the system across the different languages, obtaining interesting results.

Abstractive summarization is an ideal form of summarization since it can synthesize information from multiple documents to create concise informative summaries. In this work, we aim at developing an abstractive summarizer. First, our proposed approach identifies the most important document in the multi-document set. The sentences in the most important document are aligned to sentences in other documents to generate clusters of similar sentences. Second, we generate K-shortest paths from the sentences in each cluster using a word-graph structure. Finally, we select sentences from the set of shortest paths generated from all the clusters employing a novel integer linear programming (ILP) model with the objective of maximizing information content and readability of the final summary. Our ILP model represents the shortest paths as binary variables and considers the length of the path, information score and linguistic quality score in the objective function. Experimental results on the DUC 2004 and 2005 multi-document summarization datasets show that our proposed approach outperforms all the baselines and state-of-the-art extractive summarizers as measured by the ROUGE scores. Our method also outperforms a recent abstractive summarization technique. In manual evaluation, our approach also achieves promising results on informativeness and readability.