Elements: Data: Sustaining Modern Infrastructure For Political And Social Event Data

要素：数据：维持政治和社会事件数据的现代基础设施

• 批准号: 1931541
• 负责人: Patrick Brandt
• 金额: $ 58.8万
• 依托单位: University of Texas at Dallas
• 依托单位国家: 美国
• 项目类别: Standard Grant
• 财政年份: 2019
• 资助国家: 美国
• 起止时间: 2019-10-01 至 2023-09-30
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=1931541&HistoricalAwards=false
• 关键词: Elements; Data; Sustaining; Modern; Infrastructure

This project extracts quantitative summaries of political and international conflict events among national and non-state actors across the world by combing news reports across the internet in multiple languages. This generates event data, a machine-coded description of someone doing something to someone else as extracted from news reports. The project focuses on political and social events about conflict and cooperation between governments, individuals, non-governmental organizations, rebel groups, and others. The main goal of this project is to integrate and expand the end-to-end cyberinfrastructure for the robust creation, validation, access, and analysis of political event data by national security, government, academic, and non-governmental actors. A major component of this proposal is to continue to grow the project's engagement with the global event data community. This project extends, produces, and integrates a dynamic, robust system for event data to study sub-national and international conflict processes at a global scale, with applications to the needs of the national security and intelligence communities. Using natural language processing software tools to code event data by annotating the kinds of political events that are of interest to political scientists, international relations scholars, sociologists, and the national security community, the project analyzes contemporaneous news reports in English and Spanish, automatically encodes relevant political events for data analysts, and serves the data along with other open event data via the project websites. The technical challenges include: (1) additional extensions of the multilingual framework to more types of events; (2) smoother updates to political actor dictionaries; (3) robust data querying and linking mechanisms, and analytic tools for the broader research and user community; (4) improved methods for focus location extraction across languages and resolutions. This will improve event data quality and event detection through increased, multi-language comparisons. The multi-lingual extensions of the event encoding software and interface will produce novel methods for detecting and analyzing rare and local events. The proposed database integrations and query optimizations will streamline access to the many open access event datasets that exist, enabling researchers across diverse communities to analyze and compare conflict and political processes. The refinements of the geolocation modules will allow detection of locations from biased training samples, which is an important advancement since some political events, such as human rights violations, tend to occur in locations with low news coverage. The robust and innovative geolocation approaches can be carried over to other domain applications. Scaling the related software and data infrastructure aids the political science, national security and big data research communities. We also will provide robust data linkages across a diverse set of event data from multiple and multilingual news reporting services. The sustainable cyberinfrastructure not only includes the event data coding from news reports, but also analysis tools that include an R package and a thin-client browser-based analysis interface to the data. This sustains the cyberinfrastructure and creates a workforce that is able to work in both science, engineering, national security, and intelligence.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.

该项目通过梳理互联网上多种语言的新闻报道，提取世界各地国家和非国家行为者之间的政治和国际冲突事件的定量摘要。 这会生成事件数据，即从新闻报道中提取的某人对他人做某事的机器编码描述。该项目的重点是关于政府、个人、非政府组织、反叛团体和其他人之间的冲突与合作的政治和社会事件。 该项目的主要目标是整合和扩展端到端的网络基础设施，以便国家安全，政府，学术和非政府行为者强大地创建，验证，访问和分析政治事件数据。 该提议的一个主要组成部分是继续扩大该项目与全球活动数据界的接触。该项目扩展，生产，并集成了一个动态的，强大的事件数据系统，以研究在全球范围内的次国家和国际冲突过程，与应用程序的国家安全和情报界的需求。该项目使用自然语言处理软件工具对事件数据进行编码，对政治科学家、国际关系学者、社会学家和国家安全界感兴趣的政治事件进行注释，分析英语和西班牙语的同期新闻报道，为数据分析师自动编码相关政治事件，并通过项目网站将数据与其他开放事件数据一起提供沿着。 技术挑战包括：（1）将多语言框架进一步扩展到更多类型的事件;（2）更顺畅地更新政治行为者词典;（3）强大的数据查询和链接机制，以及更广泛的研究和用户社区的分析工具;（4）改进跨语言和分辨率的焦点位置提取方法。这将通过增加多语言比较来提高事件数据质量和事件检测。 事件编码软件和界面的多语言扩展将产生用于检测和分析罕见和本地事件的新方法。拟议的数据库集成和查询优化将简化对现有许多开放获取事件数据集的访问，使不同社区的研究人员能够分析和比较冲突和政治进程。地理定位模块的改进将允许从有偏见的训练样本中检测出地点，这是一个重要的进步，因为一些政治事件，如侵犯人权事件，往往发生在新闻报道较少的地点。强大而创新的地理定位方法可以应用于其他领域的应用。扩展相关的软件和数据基础设施有助于政治科学，国家安全和大数据研究社区。我们还将为来自多个和多语言新闻报道服务的各种事件数据提供强大的数据链接。 可持续的网络基础设施不仅包括来自新闻报道的事件数据编码，还包括分析工具，包括R包和基于瘦客户端浏览器的数据分析界面。该奖项反映了NSF的法定使命，并通过使用基金会的知识价值和更广泛的影响审查标准进行评估，被认为值得支持。

项目成果

Forecasting conflict in Africa with automated machine learning systems

使用自动化机器学习系统预测非洲冲突

• DOI: 10.1080/03050629.2022.2017290
• 发表时间: 2022
• 期刊: International Interactions
• 影响因子: 1.3
• 作者: D’Orazio, Vito;Lin, Yu
• 通讯作者: Lin, Yu

HANKE: Hierarchical Attention Networks for Knowledge Extraction in Political Science Domain

• DOI: 10.1109/dsaa49011.2020.00055
• 发表时间: 2020-10
• 期刊: 2020 IEEE 7th International Conference on Data Science and Advanced Analytics (DSAA)
• 作者: Eric Parolin;L. Khan;Javier Osorio;Vito D'Orazio;Patrick T. Brandt;J. Holmes

CoMe-KE: A New Transformers Based Approach for Knowledge Extraction in Conflict and Mediation Domain

CoMe-KE：一种基于 Transformers 的冲突与调解领域知识提取新方法

• DOI: 10.1109/bigdata52589.2021.9672080
• 发表时间: 2021
• 期刊: IEEE
• 作者: Parolin, Erick Skorupa;Hu, Yibo;Khan, Latifur;Osorio, Javier;Brandt, Patrick T.;D'Orazio, Vito
• 通讯作者: D'Orazio, Vito

Confli-T5: An AutoPrompt Pipeline for Conflict Related Text Augmentation

• DOI: 10.1109/bigdata55660.2022.10020509
• 发表时间: 2022-12
• 期刊: 2022 IEEE International Conference on Big Data (Big Data)
• 作者: Eric Parolin;Yibo Hu;Latif Khan;Patrick T. Brandt;Javier Osorio;Vito D'Orazio

Multi-CoPED: A Multilingual Multi-Task Approach for Coding Political Event Data on Conflict and Mediation Domain

• DOI: 10.1145/3514094.3534178
• 发表时间: 2022-07
• 期刊: Proceedings of the 2022 AAAI/ACM Conference on AI, Ethics, and Society
• 作者: Erick Skorupa Parolin;Mohammad Javad Hosseini;Yibo Hu;Latif Khan;Patrick T. Brandt;Javier Osorio;Vito D'Orazio