With the prevalence of social media and online forum, opinion mining, aiming at analyzing discovering latent in user-generated reviews on Internet, has become a hot research topic. This survey focuses two important subtasks this field, stance detection product aspect both which can be formalized as problem triple (target, aspect, opinion) extraction. In paper, we first introduce general framework mining describe evaluation metrics. Then, methodologies for different sources, such forum are...

Jiasheng Si, Deyu Zhou, Tongzhe Li, Xingyu Shi, Yulan He. Proceedings of the 59th Annual Meeting Association for Computational Linguistics and 11th International Joint Conference on Natural Language Processing (Volume 1: Long Papers). 2021.

The opaqueness of the multi-hop fact verification model imposes imperative requirements for explainability. One feasible way is to extract rationales, a subset inputs, where performance prediction drops dramatically when being removed. Though explainable, most rationale extraction methods explore semantic information within each piece evidence individually, while ignoring topological interaction among different pieces evidence. Intuitively, faithful bears complementary able other rationales...

Biomedical argument mining aims to automatically identify and extract the argumentative structure in biomedical text. It helps determine not only what positions people adopt, but also why they hold such opinions, which provides valuable insights into medical decision making. Generally, consists of three subtasks: component identification, classification relation identification. Current approaches employ conventional multi-task learning framework for jointly addressing latter two subtasks,...

The success of deep learning models on multi-hop fact verification has prompted researchers to understand the behavior behind their veracity. One possible way is erasure search: obtaining rationale by entirely removing a subset input without compromising veracity prediction. Although extensively explored, existing approaches fall within scope single-granular (tokens or sentences) explanation, which inevitably leads explanation redundancy and inconsistency. To address such issues, this paper...

Automatic multi-hop fact verification task has gained significant attention in recent years. Despite impressive results, these well-designed models perform poorly on out-of-domain data. One possible solution is to augment the training data with counterfactuals, which are generated by minimally altering causal features of original However, current counterfactual augmentation techniques fail handle due their incapability preserve complex logical relationships within multiple correlated texts....

Slot filling and intent detection are two highly correlated tasks in spoken language understanding (SLU). Recent SLU research attempts to explore zero-shot prompting techniques large models alleviate the data scarcity problem. Nevertheless, existing work ignores cross-task interaction information for SLU, which leads sub-optimal performance. To solve this problem, we present pioneering of Cross-task Interactive Prompting (CroPrompt) enables model interactively leverage exchange across SLU....

Multi-modal sarcasm detection involves determining whether a given multi-modal input conveys sarcastic intent by analyzing the underlying sentiment. Recently, vision large language models have shown remarkable success on various of tasks. Inspired this, we systematically investigate impact in zero-shot task. Furthermore, to capture different perspectives expressions, propose multi-view agent framework, S 3 Agent, designed enhance leveraging three critical perspectives: superficial expression...

With the growing complexity of fact verification tasks, concern with "thoughtful" reasoning capabilities is increasing. However, recent benchmarks mainly focus on checking a narrow scope semantic factoids within claims and lack an explicit logical process. In this paper, we introduce CheckWhy, challenging dataset tailored to novel causal task: truthfulness relation through rigorous steps. CheckWhy consists over 19K "why" claim-evidence-argument structure triplets supports, refutes, not...

The inherent complexity of real-world time series data, combined with the cost and infeasibility manual labeling, presents considerable challenges to representation learning. Most existing studies tend utilize data augmentation techniques construct positive negative samples leverage a comparative learning framework generate representations. However, they typically employ simple techniques, such as jitter cropping, while randomly selecting irrelevant ones, which are easily distinguished...

Chain-of-Thought (CoT) has become a vital technique for enhancing the performance of Large Language Models (LLMs), attracting increasing attention from researchers. One stream approaches focuses on iterative enhancement LLMs by continuously verifying and refining their reasoning outputs desired quality. Despite its impressive results, this paradigm faces two critical issues: (1) Simple verification methods: The current relies solely single method. (2) Wrong Information Ignorance: Traditional...

The rise of large language models (LLMs) has driven significant progress in medical applications, including traditional Chinese medicine (TCM). However, current LLMs struggle with TCM diagnosis and syndrome differentiation due to substantial differences between modern theory, the scarcity specialized, high-quality corpora. This paper addresses these challenges by proposing BianCang, a TCM-specific LLM, using two-stage training process that first injects domain-specific knowledge then aligns...

The steam turbine is one of the major pieces equipment in thermal power plants. It crucial to predict its output accurately. However, because complex coupling relationships with other equipment, it still a challenging task. Previous methods mainly focus on operation individually while ignoring relationship condenser, which we believe for prediction. Therefore, this paper, explore between and propose novel approach prediction based encode-decoder framework guided by condenser vacuum degree...

Storyline extraction aims to generate concise summaries of related events unfolding over time from a collection temporally-ordered news articles. Some existing approaches storyline are typically built on probabilistic graphical models that jointly model the and storylines published in different periods. However, their parameter inference procedures often complex require long converge, which hinders use practical applications. More recently, neural network-based approach has been proposed...

Given a controversial target such as ``nuclear energy'', argument mining aims to identify the argumentative text from heterogeneous sources. Current approaches focus on exploring better ways of integrating target-associated semantic information with text. Despite their empirical successes, two issues remain unsolved: (i) is represented by word or phrase, which insufficient cover diverse set target-related subtopics; (ii) sentence-level topic within an argument, we believe crucial for mining,...

Automatic multi-hop fact verification task has gained significant attention in recent years. Despite impressive results, these well-designed models perform poorly on out-of-domain data. One possible solution is to augment the training data with counterfactuals, which are generated by minimally altering causal features of original However, current counterfactual augmentation techniques fail handle due their incapability preserve complex logical relationships within multiple correlated texts....

The opaqueness of the multi-hop fact verification model imposes imperative requirements for explainability. One feasible way is to extract rationales, a subset inputs, where performance prediction drops dramatically when being removed. Though explainable, most rationale extraction methods explore semantic information within each piece evidence individually, while ignoring topological interaction among different pieces evidence. Intuitively, faithful bears complementary able other rationales...

Fact verification is a challenging task that requires simultaneously reasoning and aggregating over multiple retrieved pieces of evidence to evaluate the truthfulness claim. Existing approaches typically (i) explore semantic interaction between claim at different granularity levels but fail capture their topical consistency during process, which we believe crucial for verification; (ii) aggregate equally without considering implicit stances claim, thereby introducing spurious information. To...