Detecting and aggregating sentiments toward people, organizations, and events expressed in unstructured social media have become critical text mining operations. Early systems detected sentiments over whole passages, whereas more recently, target-specific sentiments have been of greater interest. In this paper, we present MTTDSC, a multi-task target-dependent sentiment classification system that is informed by feature representation learnt for the related auxiliary task of passage-level sentiment classification. The auxiliary task uses a gated recurrent unit (GRU) and pools GRU states, followed by an auxiliary fully-connected layer that outputs passage-level predictions. In the main task, these GRUs contribute auxiliary per-token representations over and above word embeddings. The main task has its own, separate GRUs. The auxiliary and main GRUs send their states to a different fully connected layer, trained for the main task. Extensive experiments using two auxiliary datasets and three benchmark datasets (of which one is new, introduced by us) for the main task demonstrate that MTTDSC outperforms state-of-the-art baselines. Using word-level sensitivity analysis, we present anecdotal evidence that prior systems can make incorrect target-specific predictions because they miss sentiments expressed by words independent of target.
The quality of training data is one of the crucial problems when a learning-centered approach is employed. This paper proposes a new method to investigate the quality of a large corpus designed for the recognizing textual entailment (RTE) task. The proposed method, which is inspired by a statistical hypothesis test, consists of two phases: the first phase is to introduce the predictability of textual entailment labels as a null hypothesis which is extremely unacceptable if a target corpus has no hidden bias, and the second phase is to test the null hypothesis using a Naive Bayes model. The experimental result of the Stanford Natural Language Inference (SNLI) corpus does not reject the null hypothesis. Therefore, it indicates that the SNLI corpus has a hidden bias which allows prediction of textual entailment labels from hypothesis sentences even if no context information is given by a premise sentence. This paper also presents the performance impact of NN models for RTE caused by this hidden bias.
Multilingual parallel text corpora provide a powerful means for propagating linguistic knowledge across languages. We present a model which jointly learns linguistic structure for each language while inducing links between them. Our model supports fully symmetrical knowledge transfer, utilizing any combination of supervised and unsupervised data across language barriers. The proposed nonparametric Bayesian model effectively combines cross-lingual alignment with target language predictions. This architecture is a potent alternative to projection methods which decompose these decisions into two separate stages. We apply this approach to the task of morphological segmentation, where the goal is to separate a word into its individual morphemes. When tested on a parallel corpus of Hebrew and Arabic, our joint bilingual model effectively incorporates all available evidence from both languages, yielding significant performance gains.
Political discourse in the United States is getting increasingly polarized. This polarization frequently causes different communities to react very differently to the same news events. Political blogs as a form of social media provide an unique insight into this phenomenon. We present a multitarget, semisupervised latent variable model, MCR-LDA to model this process by analyzing political blogs posts and their comment sections from different political communities jointly to predict the degree of polarization that news topics cause. Inspecting the model after inference reveals topics and the degree to which it triggers polarization. In this approach, community responses to news topics are observed using sentiment polarity and comment volume which serves as a proxy for the level of interest in the topic. In this context, we also present computational methods to assign sentiment polarity to the comments which serve as targets for latent variable models that predict the polarity based on the topics in the blog content. Our results show that the joint modeling of communities with different political beliefs using MCR-LDA does not sacrifice accuracy in sentiment polarity prediction when compared to approaches that are tailored to specific communities and additionally provides a view of the polarization in responses from the different communities.
We present a hybrid algorithm for Bayesian topic models that combines the efficiency of sparse Gibbs sampling with the scalability of online stochastic inference. We used our algorithm to analyze a corpus of 1.2 million books (33 billion words) with thousands of topics. Our approach reduces the bias of variational inference and generalizes to many Bayesian hidden-variable models.