A5017024092- Topic Modeling
- Diabetes Management and Research
- Natural Language Processing Techniques
- Diabetes and associated disorders
- Gut microbiota and health
- Data Quality and Management
- Text and Document Classification Technologies
- Machine Learning in Healthcare
- Pancreatic function and diabetes
- Hyperglycemia and glycemic control in critically ill and hospitalized patients
- Growth Hormone and Insulin-like Growth Factors
- Diet and metabolism studies
- Advanced Text Analysis Techniques
- Birth, Development, and Health
- Child and Adolescent Health
- Chronic Disease Management Strategies
- Clostridium difficile and Clostridium perfringens research
- Metabolism and Genetic Disorders
- Metabolism, Diabetes, and Cancer
- 3D Shape Modeling and Analysis
- Tryptophan and brain disorders
- Infant Nutrition and Health
- Diet, Metabolism, and Disease
- Diabetes Treatment and Management
- Diabetes, Cardiovascular Risks, and Lipoproteins
Children's Hospital of Fudan University
2019-2024
Harbin Institute of Technology
2023
Fudan University
2022
Abstract Gut dysbiosis has been linked to type 1 diabetes (T1D); however, microbial capacity in T1D remains unclear. Here, we integratively profiled gut functional and metabolic alterations children with new-onset independent cohorts investigated the underlying mechanisms. In T1D, microbiota was characterized by decreased butyrate production bile acid metabolism increased lipopolysaccharide biosynthesis at species, gene, metabolite levels. The combination of 18 bacterial species fecal...
Blood microbiome signatures in patients with type 1 diabetes (T1D) remain unclear. We profile blood using 16S rRNA gene sequencing 77 controls and 64 children new-onset T1D, compared it the gut oral microbiomes. The of T1D is characterized by increased diversity perturbed microbial features, a significant increase potentially pathogenic bacteria controls. Thirty-six representative genera were identified random forest analysis, providing strong discriminatory power for an AUC 0.82. PICRUSt...
Virtual film production requires intricate decision-making processes, including scriptwriting, virtual cinematography, and precise actor positioning actions. Motivated by recent advances in automated with language agent-based societies, this paper introduces FilmAgent, a novel LLM-based multi-agent collaborative framework for end-to-end automation our constructed 3D spaces. FilmAgent simulates various crew roles, directors, screenwriters, actors, cinematographers, covers key stages of...
Abstract Background The real-world exposure levels of non-therapeutic antibiotics and neonicotinoids in type 1 diabetes (T1D) children their associations as environmental triggers through gut microbiota shifts remained unknown. We thus investigated the neonicotinoids’ with pediatric T1D. Methods Fifty-one newly onset T1D along 67 age-matched healthy controls were recruited. Urine concentrations 28 12 measured by mass spectrometry. Children grouped according to kinds antibiotics’ exposures,...
Lifestyle changes including COVID-19 lockdown cause weight gain and may change obesity trends; however, timely are largely unknown monitoring measures usually lack. This first large-scale study aimed to analyze the real-world national trends of prevalence Chinese children in past five years, impact pandemic on pediatric development through both mobile- hospital-based data. included aged 3 19 years old all over China from January 2017 April 2021. Hospital-measured parent-reported cases XIGAO...
In the field of Large Language Models (LLMs), researchers are increasingly exploring their effectiveness across a wide range tasks. However, critical area that requires further investigation is interpretability these models, particularly ability to generate rational explanations for decisions. Most existing explanation datasets limited English language and general domain, which leads scarcity linguistic diversity lack resources in specialized domains, such as medical. To mitigate this, we...
BACKGROUND In addition to insulin resistance, impaired secretion has recently been identified as a crucial factor in the pathogenesis of type 2 diabetes mellitus (T2DM). Scarce clinical data exist for pediatric T2DM. AIM To investigate association β-cell function and resistance with T2DM first Chinese multicenter study. METHODS This cross-sectional study included 161 newly diagnosed children adolescents between January 2017 October 2019. Children normal glycemic levels (n = 1935) were...
Multimodal Entity Linking (MEL) is the task of mapping mentions with multimodal contexts to referent entities from a knowledge base. Existing MEL methods mainly focus on designing complex interaction mechanisms and require fine-tuning all model parameters, which can be prohibitively costly difficult scale in era Large Language Models (LLMs). In this work, we propose GEMEL, Generative framework based LLMs, directly generates target entity names. We keep vision language frozen only train...
Large Language Models (LLMs) have shown remarkable capabilities in general natural language processing tasks but often fall short complex reasoning tasks. Recent studies explored human-like problem-solving strategies, such as self-correct, to push further the boundary of single-model ability. In this work, we let a single model "step outside box" by engaging multiple models correct each other. We introduce multi-agent collaboration strategy that emulates academic peer review process. Each...
The coronavirus disease 2019 (COVID-19) pandemic has been linked to an increased incidence of diabetes and diabetic ketoacidosis (DKA). However, the relationship between COVID-19 infection progression type 1 (T1D) in children not well defined.
Background: Congenital hyperinsulinemic hypoglycemia (HH) is a rare metabolic disease. Suppressed free fat acids, β-hydroxybutyrate levels and altered serum branched-chain amino acid were found during the acute course of HH. While scarce studies have described overall secondary spectrum changes. We thus investigate profiles in HH ketotic children analyze their distinguished features. Methods: A total 97 children, 74 with 23 hypoglycemia, 170 euglycemia control subjects studied...
Event Causality Identification (ECI) aims to identify the causality between a pair of event mentions in document, which is composed sentence-level ECI (SECI) and document-level (DECI). Previous work applies various reasoning models implicit causality. However, they indiscriminately reason all same way, ignoring that most inter-sentence depends on intra-sentence infer. In this paper, we propose progressive graph pairwise attention network (PPAT) consider above dependence. PPAT strategy, as it...
As ChatGPT and GPT-4 spearhead the development of Large Language Models (LLMs), more researchers are investigating their performance across various tasks. But research needs to be done on interpretability capabilities LLMs, that is, ability generate reasons after an answer has been given. Existing explanation datasets mostly English-language general knowledge questions, which leads insufficient thematic linguistic diversity. To address language bias lack medical resources in generating...
Zero-shot entity linking (EL) aims at aligning mentions to unseen entities challenge the generalization ability. Previous methods largely focus on candidate retrieval stage and ignore essential ranking stage, which disambiguates among makes final prediction. In this paper, we propose a read-and-select (ReS) framework by modeling main components of disambiguation, i.e., mention-entity matching cross-entity comparison. First, for each candidate, reading module leverages mention context output...
Entity linking aims to link ambiguous mentions their corresponding entities in a knowledge base. One of the key challenges comes from insufficient labeled data for specific domains. Although dense retrievers have achieved excellent performance on several benchmarks, decreases significantly when only limited amount in-domain is available. In such few-shot setting, we revisit sparse retrieval method, and propose an ELECTRA-based keyword extractor denoise mention context construct better query...
Zero-shot entity linking (EL) aims at aligning mentions to unseen entities challenge the generalization ability. Previous methods largely focus on candidate retrieval stage and ignore essential ranking stage, which disambiguates among makes final prediction. In this paper, we propose a read-and-select (ReS) framework by modeling main components of disambiguation, i.e., mention-entity matching cross-entity comparison. First, for each candidate, reading module leverages mention context output...
Entity linking aims to link ambiguous mentions their corresponding entities in a knowledge base. One of the key challenges comes from insufficient labeled data for specific domains. Although dense retrievers have achieved excellent performance on several benchmarks, decreases significantly when only limited amount in-domain is available. In such few-shot setting, we revisit sparse retrieval method, and propose an ELECTRA-based keyword extractor denoise mention context construct better query...
Modern Entity Linking (EL) systems entrench a popularity bias, yet there is no dataset focusing on tail and emerging entities in languages other than English. We present Hansel, new benchmark Chinese that fills the vacancy of non-English few-shot zero-shot EL challenges. The test set Hansel human annotated reviewed, created with novel method for collecting datasets. It covers 10K diverse documents news, social media posts web articles, Wikidata as its target Knowledge Base. demonstrate...