Ruidong Wu, Yuan Yao, Xu Han, Ruobing Xie, Zhiyuan Liu, Fen Lin, Leyu Maosong Sun. Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and 9th International Joint (EMNLP-IJCNLP). 2019.

Recent studies show the effectiveness of interview chatbots for information elicitation. However, designing an effective chatbot is non-trivial. Few tools exist to help designers design, evaluate, and improve iteratively. Based on a formative study literature reviews, we propose computational framework quantifying performance chatbots. Incorporating framework, have developed iChatProfile, assistive design tool that can automatically generate profile with quantified metrics offer suggestions...

eHealth literacy is the ability to access, assess, and use digital health information. This study compared effects of a multimedia tutorial versus paper-based control in improving older adults’ from pre- posttest. A total 99 community-dwelling adults (63–90 years old; mean = 73.09) participated July 2019 February 2020. Overall, knowledge about computer/Internet terms, efficacy, quality information websites, procedural skills improved significantly No interaction effect was found between time...

Physical education, as a central component of educational systems, plays unique role in enhancing the psychological well-being university students. This study investigates impacts physical education on students' mental health and examines mediating roles social support exercise behavior. A cross-sectional survey was conducted using Education Satisfaction Scale, SCL-90, Social Support Questionnaire, International Activity Questionnaire. total 1,437 students were assessed. The research found...

Harmful text detection has become a crucial task in the development and deployment of large language models, especially as AI-generated content continues to expand across digital platforms. This study proposes joint retrieval framework that integrates pre-trained models with knowledge graphs improve accuracy robustness harmful detection. Experimental results demonstrate approach significantly outperforms single-model baselines, particularly low-resource training scenarios multilingual...

In this work, we propose a domain generalization (DG) approach to learn on several labeled source domains and transfer knowledge target that is inaccessible in training. Considering the inherent conditional label shifts, would expect alignment of p(x|y) p(y). However, widely used invariant feature learning (IFL) methods relies aligning marginal concept shift w.r.t. p(x), which rests an unrealistic assumption p(y) across domains. We thereby novel variational Bayesian inference framework...

Motivated by the need for multiple agents to collaborate in order solve a distributed resource allocation planning problem, this paper develops framework that combines each agent's information, expertise, responsibility, and asset ownership with goal of optimizing given mission objective. A is collection interdependent tasks be executed directed/sequential sequence. Each task modeled vector requirements, processing time, start time (release time). agent has subset which it responsible, owns...

Entity typing aims to classify semantic types of an entity mention in a specific context. Most existing models obtain training data using distant supervision, and inevitably suffer from the problem noisy labels. To address this issue, we propose with language model enhancement. It utilizes measure compatibility between context sentences labels, thereby automatically focuses more on context-dependent Experiments benchmark datasets demonstrate that our method is capable enhancing information...

Retrieval-augmented generation (RAG) is a powerful technique to facilitate language model with proprietary and private data, where data privacy pivotal concern. Whereas extensive research has demonstrated the risks of large models (LLMs), RAG could potentially reshape inherent behaviors LLM generation, posing new issues that are currently under-explored. In this work, we conduct empirical studies novel attack methods, which demonstrate vulnerability systems on leaking retrieval database....

Anti-submarine warfare (ASW) missions are the linchpin of maritime operations involving effective allocation and path planning scarce assets to search for, detect, classify, track, prosecute hostile submarines within a dynamic uncertain mission environment. Motivated by need assist ASW commanders make better decisions an evolving context, we investigate moving target problem with multiple searchers develop context-driven decision support tool for problem. Given spatial probability...

Text summarization is to generate a condensed version of the original document. The major issues for text are eliminating redundant information, identifying important difference among documents, and recovering informative content. This paper proposes Semantic Graph Model which exploits semantic information sentence using FSGM. FSGM treats sentences as vertexes while relationship edges. It uses FrameNet word embedding calculate similarity sentences. method assigns weight both nodes After all,...

Data center, as the core infrastructure of data storage and processing, calls for network security protection. Information has been addressed in a number specific ways. However, there are few studies that employ topology features to prevent transmission viruses. When virus spreads, different topologies display various properties. In this paper, we study three types center topologies, i.e., Fat-tree, Leaf-spine, Bcube, quantify propagation characteristics every through IC model. The...

A simplified magnetic resonance index of activity (MaRIAs) was recently proposed. Our aim to verify whether MaRIAs can accurately assess the degree CD.We retrospectively analyzed MRI, ileocolonoscopy, fecal calprotectin (FC), erythrocyte sedimentation rate (ESR) and C-reactive protein (CRP) data 93 CD patients. With SES-CD as gold standard, MaRIAs' accuracy, correlation SES-CD, FC, ESR, CRP, interevaluator reliability were assessed.MaRIAs ≥ 1 detected segments with active 90.80% specificity...

Motivated by the Navy's new emphasis on networked distributed planning capabilities in maritime operations centers (MOC), this paper presents an agent-based framework applicable to a range of missions. A mission is modeled as task graph, where each specified terms vector resource requirements, processing time, geographic location and opportunity window. Each agent has set mobile assets, asset it provides, its velocity, current location. Multiple agents need collaboratively solve resulting...

This study investigates coordinated behaviors and the underlying collective intelligence in biological groups, particularly those led by informed leaders. By establishing new convergence condition based on experiments involving real this research introduces concept of a volitional term heterogeneous networks, constructing coupled-force Cucker–Smale model with Incorporating leaders into leader-follower group enables more accurate representation behaviors. The paper then extracts Flock...

Abstract Separate-layer injection technology is a highly significant approach for enhancing oil recovery in the later stages of oilfield production. Both separate-layer and general information are crucial parameters multi-layer systems. However, significance usually overlooked during optimization process injection. Moreover, conventional schemes fail to meet immediate dynamic demands well Consequently, method based on artificial neural network residual (ANN-Res) model was proposed. Firstly,...

Retrieval-augmented generation (RAG) enhances the outputs of language models by integrating relevant information retrieved from external knowledge sources. However, when retrieval process involves private data, RAG systems may face severe privacy risks, potentially leading to leakage sensitive information. To address this issue, we propose using synthetic data as a privacy-preserving alternative for data. We SAGE, novel two-stage paradigm. In stage-1, employ an attribute-based extraction and...

Text summarization is to generate a condensed version of the original document. The major issues for text are eliminating redundant information, identifying important difference among documents and covering informative content. In this paper, we propose Sentence-Level Semantic Graph Model (SLSGM) which exploits semantic information sentence. SLSGM considers sentences as vertexes while relationship between edges. We calculate relevance values using analysis take weights edges sentences'...