Using a newly developed model to upgrade a legacy model is a common practice in machine learning applications. After the upgrade, it is expected that the new model should outperform the legacy model in the regions of interest. However, it is observed that the new model often makes incorrect decisions on some instances where the legacy model still performs well. For a binary classification model (e.g., click-through-rate/CTR prediction model), such undesirable behavior could even occur in the low false positive region of the receiver operating characteristic (ROC) curve. Finding the reasons behind this phenomenon can help business partners in an organization gain confidence in adopting the new model and help modelers to improve the new model in future releases. In this paper, we present the "Learning from Disagreement" framework to understand and improve the performance of a predictive model. Under the setting of a binary classification task, this proposed approach focuses on instances that lead to contradictory decisions between a pair of models at a given operating point. We perform feature importance analysis exclusively on these instances, gain insights into the pair of models without even knowing their inner operations, and offer actionable feedback for model improvement. We demonstrate the usefulness of this framework on two real-world event detection datasets.

Event Extraction (EE) is one of the fundamental tasks in Information Extraction (IE) that aims to recognize event mentions and their arguments (i.e., participants) from text.Due to its importance, extensive methods and resources have been developed for Event Extraction.However, one limitation of current research for EE involves the under-exploration for non-English languages in which the lack of high-quality multilingual EE datasets for model training and evaluation has been the main hindrance.To address this limitation, we propose a novel Multilingual Event Extraction dataset (MEE) that provides annotation for more than 50K event mentions in 8 typologically different languages.MEE comprehensively annotates data for entity mentions, event triggers and event arguments.We conduct extensive experiments on the proposed dataset to reveal challenges and opportunities for multilingual EE.

Graph neural networks (GNNs) are deep learning models designed specifically for graph data, and they typically rely on node features as the input to the first layer. When applying such a type of network on the graph without node features, one can extract simple graph-based node features (e.g., number of degrees) or learn the input node representations (i.e., embeddings) when training the network. While the latter approach, which trains node embeddings, more likely leads to better performance, the number of parameters associated with the embeddings grows linearly with the number of nodes. It is therefore impractical to train the input node embeddings together with GNNs within graphics processing unit (GPU) memory in an end-to-end fashion when dealing with industrial-scale graph data. Inspired by the embedding compression methods developed for natural language processing (NLP) tasks, we develop a node embedding compression method where each node is compactly represented with a bit vector instead of a floating-point vector. The parameters utilized in the compression method can be trained together with GNNs. We show that the proposed node embedding compression method achieves superior performance compared to the alternatives.

Predicting metrics associated with entities' transnational behavior within payment processing networks is essential for system monitoring. Multivariate time series, aggregated from the past transaction history, can provide valuable insights for such prediction. The general multivariate time series prediction problem has been well studied and applied across several domains, including manufacturing, medical, and entomology. However, new domain-related challenges associated with the data such as concept drift and multi-modality have surfaced in addition to the real-time requirements of handling the payment transaction data at scale. In this work, we study the problem of multivariate time series prediction for estimating transaction metrics associated with entities in the payment transaction database. We propose a model with five unique components to estimate the transaction metrics from multi-modality data. Four of these components capture interaction, temporal, scale, and shape perspectives, and the fifth component fuses these perspectives together. We also propose a hybrid offline/online training scheme to address concept drift in the data and fulfill the real-time requirements. Combining the estimation model with a graphical user interface, the prototype transaction metric estimation system has demonstrated its potential benefit as a tool for improving a payment processing company's system monitoring capability.

Evaluating on adversarial examples has become a standard procedure to measure robustness of deep learning models. Due to the difficulty of creating white-box adversarial examples for discrete text input, most analyses of the robustness of NLP models have been done through black-box adversarial examples. We investigate adversarial examples for character-level neural machine translation (NMT), and contrast black-box adversaries with a novel white-box adversary, which employs differentiable string-edit operations to rank adversarial changes. We propose two novel types of attacks which aim to remove or change a word in a translation, rather than simply break the NMT. We demonstrate that white-box adversarial examples are significantly stronger than their black-box counterparts in different attack scenarios, which show more serious vulnerabilities than previously known. In addition, after performing adversarial training, which takes only 3 times longer than regular training, we can improve the model’s robustness significantly.

Rumor stance classification is the task of determining the stance towards a rumor in text. This is the first step in effective rumor tracking on social media which is an increasingly important task. In this work, we analyze Twitter users' stance toward a rumorous tweet, in which users could support, deny, query, or comment upon the rumor. We propose a deep attentional CNN-LSTM approach, which takes the sequence of tweets in a thread of conversation as the input. We use neighboring tweets in the timeline as context vectors to capture the temporal dynamism in users' stance evolution. In addition, we use extra features such as friendship, to leverage useful relational features that are readily available in social media. Our model achieves the state-of-the-art results on rumor stance classification on a recent SemEval dataset, improving accuracy and F1 score by 3.6% and 4.2% respectively.

Distributed word representations are able to capture syntactic and semantic regularities in text. In this paper, we present a word representation scheme that incorporates authorship information. While maintaining similarity among related words in the induced distributed space, our word vectors can be effectively used for some text classification tasks too. We build on a log-bilinear document model (lbDm), which extracts document features, and word vectors based on word co-occurrence counts. First, we propose a log-bilinear author model (lbAm), which contains an additional author matrix. We show that by directly learning author feature vectors, as opposed to document vectors, we can learn better word representations for the authorship attribution task. Furthermore, authorship information has been found to be useful for sentiment classification. We enrich the author model with a sentiment tensor, and demonstrate the effectiveness of this hybrid model (lbHm) through our experiments on a movie review-classification dataset.

Classifying the stance expressed in online microblogging social media is an emerging problem in opinion mining. We propose a probabilistic approach to stance classification in tweets, which models stance, target of stance, and sentiment of tweet, jointly. Instead of simply conjoining the sentiment or target variables as extra variables to the feature space, we use a novel formulation to incorporate three-way interactions among sentiment-stance-input variables and three-way interactions among target-stance-input variables. The proposed specification intuitively aims to discriminate sentiment features from target features for stance classification. In addition, regularizing a single stance classifier, which handles all targets, acts as a soft weight-sharing among them. We demonstrate that discriminative training of this model achieves the state-of-the-art results in supervised stance classification, and its generative training obtains competitive results in the weakly supervised setting.

Supervised stance classification, in such domains as Congressional debates and online forums, has been a topic of interest in the past decade.Approaches have evolved from text classification to structured output prediction, including collective classification and sequence labeling.In this work, we investigate collective classification of stances on Twitter, using hinge-loss Markov random fields (HL-MRFs).Given the graph of all posts, users, and their relationships, we constrain the predicted post labels and latent user labels to correspond with the network structure.We focus on a weakly supervised setting, in which only a small set of hashtags or phrases is labeled.Using our relational approach, we are able to go beyond the stance-indicative patterns and harvest more stance-indicative tweets, which can also be used to train any linear text classifier when the network structure is not available or is costly.

Modeling physical activity propagation, such as activity level and intensity, is a key to preventing obesity from cascading through communities, and to helping spread wellness and healthy behavior in a social network. However, there have not been enough scientific and quantitative studies to elucidate how social communication may deliver physical activity interventions. In this work, we introduce a novel model named T opic-aware C ommunity-level P hysical Activity Propagation with T emporal Dynamics (TCPT) to analyze physical activity propagation and social influence at different granularities (i.e., individual level and community level). Given a social network, the TCPT model first integrates the correlations between the content of social communication, social influences, and temporal dynamics. Then, a hierarchical approach is utilized to detect a set of communities and their reciprocal influence strength of physical activities. The experimental evaluation shows not only the effectiveness of our approach but also the correlation of the detected communities with various health outcome measures. Our promising results pave a way for knowledge discovery in health social networks.

We present a novel approach for relation classification, using a recursive neural network (RNN), based on the shortest path between two entities in a dependency graph.Previous works on RNN are based on constituencybased parsing because phrasal nodes in a parse tree can capture compositionality in a sentence.Compared with constituency-based parse trees, dependency graphs can represent relations more compactly.This is particularly important in sentences with distant entities, where the parse tree spans words that are not relevant to the relation.In such cases RNN cannot be trained effectively in a timely manner.However, due to the lack of phrasal nodes in dependency graphs, application of RNN is not straightforward.In order to tackle this problem, we utilize dependency constituent units called chains.Our experiments on two relation classification datasets show that Chain based RNN provides a shallower network, which performs considerably faster and achieves better classification results.