EAGER: Medical Knowledge Graph Construction from Heterogeneous Sources

EAGER：异构来源的医学知识图构建

• 批准号: 1747614
• 负责人: Changyou Chen
• 金额: $ 19.96万
• 依托单位: SUNY at Buffalo
• 依托单位国家: 美国
• 项目类别: Standard Grant
• 财政年份: 2017
• 资助国家: 美国
• 起止时间: 2017-11-15 至 2024-06-30
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=1747614&HistoricalAwards=false
• 关键词: EAGER; Medical; Knowledge; Graph; Construction

The objective of this project is to construct comprehensive medical knowledge graphs. Such knowledge graphs can satisfy users' growing needs for reliable medical information, help the effective communication between patients and doctors, and potentially help reduce the high costs of health care. This project tackles a series of unique challenges observed in medical domains, and develops effective approaches that can extract knowledge from the deluge of crowdsourced data to augment medical knowledge graphs. The proposed research advances the fields of knowledge graph construction and information trustworthiness analysis by developing novel methods that mines knowledge from unstructured data in medical domains.This project investigates the problem of medical graph construction from the following perspectives: 1) Effective approaches are developed to take into account the semantic relations between medical terms during the extraction of reliable medical facts from noisy answers on healthcare question and answering websites; 2) A unified framework is designed to integrate information from heterogeneous data sources in the process of medical knowledge discovery; 3) The knowledge graph is completed by inferring new relations based on existing relations between medical terms in the graph. The proposed research is implemented into a system prototype that displays the extracted medical facts and the knowledge graph, which benefit online users who seek medical information. The proposed techniques are used to enhance educational methodologies. Research results of this project are integrated into course materials and projects that reinforce student training.

该项目的目标是构建全面的医学知识图谱。此类知识图谱可以满足用户对可靠医疗信息日益增长的需求，有助于患者和医生之间的有效沟通，并可能有助于降低高昂的医疗保健成本。该项目解决了医学领域中观察到的一系列独特挑战，并开发了有效的方法，可以从海量的众包数据中提取知识以增强医学知识图谱。 该研究通过开发从医学领域非结构化数据中挖掘知识的新方法，推进了知识图谱构建和信息可信度分析领域的发展。该项目从以下角度研究了医学图谱构建问题：1）在从医疗保健问答网站的嘈杂答案中提取可靠的医学事实时，开发了有效的方法来考虑医学术语之间的语义关系； 2）设计统一的框架，整合医学知识发现过程中异构数据源的信息； 3）根据图中医学术语之间的现有关系，推断出新的关系，完成知识图谱。所提出的研究被实施到一个系统原型中，该原型显示提取的医学事实和知识图，这有利于寻求医疗信息的在线用户。所提出的技术用于增强教育方法。该项目的研究成果被纳入课程材料和项目中，以加强学生培训。

项目成果

InterHG: an Interpretable and Accurate Model for Hypothesis Generation

InterHG：可解释且准确的假设生成模型

• DOI: 10.1109/bibm52615.2021.9669740
• 发表时间: 2021
• 期刊: IEEE International Conference on Bioinformatics and Biomedicine (BIBM
• 作者: Wang, Haoyu;Wang, Xuan;Wang, Yaqing;Xun, Guangxu;Jha, Kishlay;Gao, Jing
• 通讯作者: Gao, Jing

Data Poisoning Attack against Recommender System Using Incomplete and Perturbed Data

• DOI: 10.1145/3447548.3467233
• 发表时间: 2021-08
• 期刊: Proceedings of the 27th ACM SIGKDD Conference on Knowledge Discovery & Data Mining
• 作者: Hengtong Zhang;Changxin Tian;Yaliang Li;Lu Su;Nan Yang;Wayne Xin Zhao;Jing Gao

Knowledge-Guided Paraphrase Identification

• DOI: 10.18653/v1/2021.findings-emnlp.72
• 发表时间: 2021
• 期刊: ArXiv
• 作者: Haoyu Wang;Fenglong Ma;Yaqing Wang;Jing Gao

Learning from Language Description: Low-shot Named Entity Recognition via Decomposed Framework

• DOI: 10.18653/v1/2021.findings-emnlp.139
• 发表时间: 2021-09
• 期刊: ArXiv
• 作者: Yaqing Wang;Haoda Chu;Chao Zhang;Jing Gao

On the Estimation of Treatment Effect with Text Covariates

• DOI: 10.24963/ijcai.2019/570
• 发表时间: 2019-08
• 期刊: ArXiv
• 作者: Liuyi Yao;Sheng Li;Yaliang Li;Hongfei Xue;Jing Gao;Aidong Zhang