We address the problem of unsupervised abstractive summarization of collections of user generated reviews with self-supervision and control. We propose a self-supervised setup that consider an individual document as a target summary for a set of similar documents. This setting makes training simpler than previous approaches by relying only on standard log-likelihood loss. We address the problem of hallucinations through the use of control codes, to steer the generation towards more coherent and relevant summaries.Finally, we extend the Transformer architecture to allow for multiple reviews as input. Our benchmarks on two datasets against graph-based and recent neural abstractive unsupervised models show that our proposed method generates summaries with a superior quality and relevance.This is confirmed in our human evaluation which focuses explicitly on the faithfulness of generated summaries We also provide an ablation study, which shows the importance of the control setup in controlling hallucinations and achieve high sentiment and topic alignment of the summaries with the input reviews.
Nowadays, listening music has been and will always be an indispensable part of our daily life. In recent years, sentiment analysis of music has been widely used in the information retrieval systems, personalized recommendation systems and so on. Due to the development of deep learning, this paper commits to find an effective approach for mood tagging of Chinese song lyrics. To achieve this goal, both machine-learning and deep-learning models have been studied and compared. Eventually, a CNN-based model with pre-trained word embedding has been demonstrated to effectively extract the distribution of emotional features of Chinese lyrics, with at least 15 percentage points higher than traditional machine-learning methods (i.e. TF-IDF+SVM and LIWC+SVM), and 7 percentage points higher than other deep-learning models (i.e. RNN, LSTM). In this paper, more than 160,000 lyrics corpus has been leveraged for pre-training word embedding for mood tagging boost.
Stance detection is a subproblem of sentiment analysis where the stance of the author of a piece of natural language text for a particular target (either explicitly stated in the text or not) is explored. The stance output is usually given as Favor, Against, or Neither. In this paper, we target at stance detection on sports-related tweets and present the performance results of our SVM-based stance classifiers on such tweets. First, we describe three versions of our proprietary tweet data set annotated with stance information, all of which are made publicly available for research purposes. Next, we evaluate SVM classifiers using different feature sets for stance detection on this data set. The employed features are based on unigrams, bigrams, hashtags, external links, emoticons, and lastly, named entities. The results indicate that joint use of the features based on unigrams, hashtags, and named entities by SVM classifiers is a plausible approach for stance detection problem on sports-related tweets.
Weibo, as the largest social media service in China, has billions of messages generated every day. The huge number of messages contain rich sentimental information. In order to analyze the emotional changes in accordance with time and space, this paper presents an Emotion Analysis Platform (EAP), which explores the emotional distribution of each province, so that can monitor the global pulse of each province in China. The massive data of Weibo and the real-time requirements make the building of EAP challenging. In order to solve the above problems, emoticons, emotion lexicon and emotion-shifting rules are adopted in EAP to analyze the emotion of each tweet. In order to verify the effectiveness of the platform, case study on the Sichuan earthquake is done, and the analysis result of the platform accords with the fact. In order to analyze from quantity, we manually annotate a test set and conduct experiment on it. The experimental results show that the macro-Precision of EAP reaches 80% and the EAP works effectively.
Event classification at sentence level is an important Information Extraction task with applications in several NLP, IR, and personalization systems. Multi-label binary relevance (BR) are the state-of-art methods. In this work, we explored new multi-label methods known for capturing relations between event types. These new methods, such as the ensemble Chain of Classifiers, improve the F1 on average across the 6 labels by 2.8% over the Binary Relevance. The low occurrence of multi-label sentences motivated the reduction of the hard imbalanced multi-label classification problem with low number of occurrences of multiple labels per instance to an more tractable imbalanced multiclass problem with better results (+ 4.6%). We report the results of adding new features, such as sentiment strength, rhetorical signals, domain-id (source-id and date), and key-phrases in both single-label and multi-label event classification scenarios.
Unlabeled data is often used to learn representations which can be used to supplement baseline features in a supervised learner. For example, for text applications where the words lie in a very high dimensional space (the size of the vocabulary), one can learn a low rank "dictionary" by an eigen-decomposition of the word co-occurrence matrix (e.g. using PCA or CCA). In this paper, we present a new spectral method based on CCA to learn an eigenword dictionary. Our improved procedure computes two set of CCAs, the first one between the left and right contexts of the given word and the second one between the projections resulting from this CCA and the word itself. We prove theoretically that this two-step procedure has lower sample complexity than the simple single step procedure and also illustrate the empirical efficacy of our approach and the richness of representations learned by our Two Step CCA (TSCCA) procedure on the tasks of POS tagging and sentiment classification.
*Content warning: This work displays examples of explicit and strongly offensive language. The COVID-19 pandemic has fueled a surge in anti-Asian xenophobia and prejudice. Many have taken to social media to express these negative sentiments, necessitating the development of reliable systems to detect hate speech against this often under-represented demographic. In this paper, we create and annotate a corpus of Twitter tweets using 2 experimental approaches to explore anti-Asian abusive and hate speech at finer granularity. Using the dataset with less biased annotation, we deploy multiple models and also examine the applicability of other relevant corpora to accomplish these multi-task classifications. In addition to demonstrating promising results, our experiments offer insights into the nuances of cultural and logistical factors in annotating hate speech for different demographics. Our analyses together aim to contribute to the understanding of the area of hate speech detection, particularly towards low-resource groups.
Deep neural network models have achieved state-of-the-art results in various tasks related to vision and/or language. Despite the use of large training data, most models are trained by iterating over single input-output pairs, discarding the remaining examples for the current prediction. In this work, we actively exploit the training data to improve the robustness and interpretability of deep neural networks, using the information from nearest training examples to aid the prediction both during training and testing. Specifically, the proposed approach uses the target of the nearest input example to initialize the memory state of an LSTM model or to guide attention mechanisms. We apply this approach to image captioning and sentiment analysis, conducting experiments with both image and text retrieval. Results show the effectiveness of the proposed models for the two tasks, on the widely used Flickr8 and IMDB datasets, respectively. Our code is publicly available http://github.com/RitaRamo/retrieval-augmentation-nn.
This study explores how people view and respond to the proposals of NYC congestion pricing evolve in time. To understand these responses, Twitter data is collected and analyzed. Critical groups in the recurrent process are detected by statistically analyzing the active users and the most mentioned accounts, and the trends of people's attitudes and concerns over the years are identified with text mining and hybrid Nature Language Processing techniques, including LDA topic modeling and LSTM sentiment classification. The result shows that multiple interest groups were involved and played crucial roles during the proposal, especially Mayor and Governor, MTA, and outer-borough representatives. The public shifted the concern of focus from the plan details to a wider city's sustainability and fairness. Furthermore, the plan's approval relies on several elements, the joint agreement reached in the political process, strong motivation in the real-world, the scheme based on balancing multiple interests, and groups' awareness of tolling's benefits and necessity.
The design of new products and services starts with the identification of needs of potential customers or users. Many existing methods like observations, surveys, and experiments draw upon specific efforts to elicit unsatisfied needs from individuals. At the same time, a huge amount of user-generated content in micro blogs is freely accessible at no cost. While this information is already analyzed to monitor sentiments towards existing offerings, it has not yet been tapped for the elicitation of needs. In this paper, we lay an important foundation for this endeavor: we propose a Machine Learning approach to identify those posts that do express needs. Our evaluation of tweets in the e-mobility domain demonstrates that the small share of relevant tweets can be identified with remarkable precision or recall results. Applied to huge data sets, the developed method should enable scalable need elicitation support for innovation managers - across thousands of users, and thus augment the service design tool set available to him.