Kyle Glandt, Sarthak Khanal, Yingjie Li, Doina Caragea, Cornelia Caragea. Proceedings of the 59th Annual Meeting Association for Computational Linguistics and 11th International Joint Conference on Natural Language Processing (Volume 1: Long Papers). 2021.

Yingjie Li, Cornelia Caragea. Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and 9th International Joint (EMNLP-IJCNLP). 2019.

Stance detection determines whether the author of a text is in favor of, against or neutral to specific target and provides valuable insights into important events such as presidential election.However, progress on stance has been hampered by absence large annotated datasets.In this paper, we present P-STANCE, dataset political domain, which contains 21,574 labeled tweets.We provide detailed description newly created develop deep learning models it.Our best model achieves macro-average...

The goal of zero-shot stance detection (ZSSD) is to identify the (in favor of, against, or neutral) a text towards an unseen target in inference stage. In this paper, we explore problem from novel angle by proposing Target-based Teacher-Student learning (TTS) framework. Specifically, first augment training set extracting diversified targets that are during with keyphrase generation model. Then, develop teacher-student framework which effectively utilizes augmented data. Extensive experiments...

The goal of stance detection is to identify whether the author a text in favor of, neutral or against specific target. Despite substantial progress on this task, one remaining challenges scarcity annotations. Data augmentation commonly used address annotation by generating more training samples. However, augmented sentences that are generated existing methods either less diversified inconsistent with given target and label. In paper, we formulate data as conditional masked language modeling...

Stance detection determines whether the author of a text is in favor of, against or neutral to specific target and provides valuable insights into important events such as legalization abortion. Despite significant progress on this task, one remaining challenges scarcity annotations. Besides, most previous works focused hard-label training which meaningful similarities among categories are discarded during training. To address these challenges, first, we evaluate multi-target multi-dataset...

Multi-target stance detection aims to identify the taken toward a pair of different targets from same text, and typically, there are multiple target pairs per dataset.Existing works generally train one model for each pair.However, they fail learn target-specific representations prone overfitting.In this paper, we propose new training strategy under multi-task learning setting by on all pairs, which helps more universal alleviate overfitting.Moreover, in order extract accurate...

Matchmaking systems are vital for creating fair matches in online multiplayer games, which directly affects players' satisfactions and game experience. Most of the matchmaking largely rely on precise estimation skills to construct equitable games. However, skill rating a novice is usually inaccurate, as current algorithms require considerable amount games learning true new player. Using these unreliable scores at early stages leads disparities terms team performance, causes negative This...

Zero-shot stance detection (ZSSD) aims to determine whether the author of a text is in favor of, against, or neutral toward target that unseen during training. Despite growing attention on ZSSD, most recent advances this task are limited English and do not pay much other languages such as Chinese. To support ZSSD research, paper, we present C-STANCE that, our knowledge, first Chinese dataset for zero-shot detection. We introduce two challenging subtasks ZSSD: target-based domain-based ZSSD....

Diffractive optical neural networks (DONNs) have attracted lots of attention as they bring significant advantages in terms power efficiency, parallelism, and computational speed compared with conventional deep (DNNs), which intrinsic limitations when implemented on digital platforms. However, inversely mapping algorithm-trained physical model parameters onto real-world devices discrete values is a non-trivial task existing non-unified levels non-monotonic properties. This work proposes novel...

The objective of this research was to develop a new computer-based system for psychomotor skill assessment. focus on the simulation Rey-Osterrieth Complex Figure (ROCF) reproduction test incorporating haptic interface. Various functions were created support customized testing protocols that are based specific user requirements, facilitate semiautomated scoring tests, and produce quantitative output. Advanced technologies pattern recognition reviewed adapted development. This approach yielded...

Image registration is a prerequisite for image fusion from multiple modalities, such as infrared (IR) and visible (VIS) images. Although there have been many various methods of registration, non-rigid IR VIS images still challenging due to large differences between In this work, point feature-based method proposed improve the performance on registration. Firstly, feature descriptor - Gaussian weighted shape context (GWSC) improved (SC) fast extract matching pairs edge maps in With set pairs,...

Choosing representative samples and removing data redundancy are two key issues in large-scale classification. This paper proposes a new model, named interval extreme learning machine (ELM), for big classification with continuous-valued attributes. The ELM model is built up based on techniques, i.e., discretization of conditional attributes fuzzification class labels. First, inspired by the traditional decision tree (DT) induction algorithm, each attribute discretized into number intervals...

Motion blur and out-of-focus are two main components of image in most cases. Blur classification parameter identification very important for processing such as restoration. In this paper, a novel method based on cepstrum peak detection is presented to classify identify three types including motion blur, mixed blur. First, frequency information detected utilized determine the Then we domain. Next parameters identified according distribution blurred Lastly, extensive experiments carried out...

Stance detection aims to detect the stance toward a corresponding target. Existing works use assumption that target is known in advance, which often not case wild. Given text from social media platforms, information unknown due implicit mentions source and it infeasible have manual annotations at large scale. Therefore, this paper, we propose new task Target-Stance Extraction (TSE) extract (target, stance) pair text. We benchmark by proposing two-stage framework first identifies relevant...

Description Logic possesses strong knowledge representation and reasoning capabilities offers logical foundation for Semantic Web ontology languages such as OWL OWL-S. However, the present implementations of OWL-S are deficient in semantic modeling dynamic service composition. They also do not consider user preferences. AI planning a better capability action state transformations provides an effective method solving problem task decomposition. it is limited capabilities. Based on merits...

The techniques of fuzzy measure and integral have been successfully applied in various real-world applications. determination measures is the most difficult part problem solving. Signed efficiency measure, which a special kind with best representation ability but highest complexity, even harder to determine. Some methodologies developed for solving this such as artificial neural networks (ANNs) genetic algorithms (GAs). However, none existing methods can outperform others unique advantages....