Rumors spread through the Internet, especially on Twitter, have harmed social stability and residents' daily lives. Recently, in addition to utilizing text features of posts for rumor detection, structural information propagation trees has also been valued. Most rumors with salient can be quickly locked by graph models dominated cross entropy loss. However, these conventional may lead poor generalization, lack robustness face noise adversarial rumors, or even conversational structures that...

Due to its great importance in deep natural language understanding and various down-stream applications, text-level parsing of discourse rhetorical structure (DRS) has been drawing more attention recent years. However, all the previous studies on adopt bottom-up approaches, which much limit DRS determination local information fail well benefit from global overall discourse. In this paper, we justify both computational perceptive points-of-view that top-down architecture is suitable for...

With the increasing scale of mountain tunnel construction, control tunnelling quality is becoming a major concern. The efficient and accurate assessment overbreak underbreak vital to evaluation optimization quality, but remains challenge. Thus, this paper proposes an method for based on K-dimensional (KD) tree Iterative Closest Point (ICP) algorithm. Firstly, point clouds are acquired using laser scanning during 3D modeling performed. Secondly, as-designed models converted into registered...

This paper concentrates on Document-level Event Factuality Identification (DEFI) that predicts event factuality values from the viewpoint of document. At present, shortcomings previous studies are multi-fold, including data limitation and scarcity, coarsegrained interpretability without span-level clues, no unified model for different datasets. is devoted to address above problems by building Machine Reading Comprehension (MRC) frameworks comprised both span-extraction multiple-choice...

Event factuality identification is an important semantic task in NLP. Traditional research heavily relies on annotated texts. This paper proposes a two-step framework, first extracting essential factors related with event from raw texts as the input, and then identifying of events via Generative Adversarial Network Auxiliary Classification (AC-GAN). The use AC-GAN allows model to learn more syntactic information address imbalance among values. Experimental results FactBank show that our...

With more and messages in the form of text image being spread on Internet, multi-modal rumor detection has become focus recent research. However, most existing methods simply concatenate or fuse features with features, which can not fully explore interaction between modalities. Meanwhile, they ignore convergence inconsistency problem strong weak modalities, that is, dominant modality may inhibit optimization modality. In this paper, we investigate from a novel perspective, propose...

In the literature, most of previous studies on English implicit discourse relation recognition only use sentence-level representations, which cannot provide enough semantic information in Chinese due to its unique paratactic characteristics. this paper, we propose a topic tensor network recognize relations with both and topic-level representations. particular, besides encoding arguments (discourse units) using gated convolutional obtain train simplified model infer latent Moreover, feed two...

In this paper, we investigate the task scheduling in resource-limited mobile edge computing (MEC) network, where multiple base stations (BSs), each equipped with a MEC server, assist latency-sensitive user equipments (UEs) computing. We aim to jointly minimize system energy consumption and maximize number of offloaded tasks by optimizing between UEs BSs. A multiple-objective mix-integer problem is formulated, which difficult solve. To tackle problem, combine ant colony optimization (ACO)...

Few-shot remote sensing image classification entails identifying images using a limited set of labeled data within scenes, holding significant theoretical and practical implications. However, owing to the intricacy variety images, traditional methods usually struggle extract effective features learn robust classifiers. To address this issue, an end-to-end metric learning framework named Attention-based Contrastive Learning Network is introduced in paper. Specifically, Feature Optimization...

Event coreference resolution (ECR) aims to group event mentions referring the same real-world into clusters. Most previous studies adopt "encoding first, then scoring" framework, making judgment rely on encoding. Furthermore, current methods struggle leverage human-summarized ECR rules, e.g., coreferential events should have type, guide model. To address these two issues, we propose a prompt-based approach, CorefPrompt, transform cloze-style MLM (masked language model) task. This allows for...

Zhong Qian, Peifeng Li, Qiaoming Zhu, Guodong Zhou. Proceedings of the 2019 Conference North American Chapter Association for Computational Linguistics: Human Language Technologies, Volume 1 (Long and Short Papers). 2019.

Implicit discourse relation recognition (IDRR) is a critical task in analysis. Previous studies only regard it as classification and lack an in-depth understanding of the semantics different relations. Therefore, we first view IDRR generation further propose method joint modeling generation. Specifically, model, CG-T5, to recognize label generate target sentence containing meaning relations simultaneously. Furthermore, design three forms, including question form, for model incorporate prior...

Hierarchically constructing micro (i.e., intra-sentence or inter-sentence) discourse structure trees using explicit boundaries (e.g., sentence and paragraph boundaries) has been proved to be an effective strategy. However, it is difficult apply this strategy document-level macro inter-paragraph) parsing, the more challenging task, due lack of at higher level. To alleviate issue, we introduce a topic segmentation mechanism detect implicit then help parser construct better hierarchically. In...