EL-STEC: Shared Task Evaluation Campaigns with Endangered Language Data

实地语言学家的共同努力将不足以记录到本世纪末预计将消失的数千种语言。创建这些正在消失的语言的持久记录是至关重要的，因为所有语言都提供了关于人类如何学习和处理语言的线索。语言描述和分析可以通过在现有的和不断增长的数字音频和视频语言数据库上部署计算分析技术来促进。该项目将组织一次研讨会，汇集语音和语言技术研究人员、濒危语言研究人员、实地语言学家和语言存储库档案保管员，以帮助推动这一计算技术的发展。研讨会参与者将讨论共享任务评估活动如何促进记录濒危和资源匮乏语言的语音和语言技术的发展。共享任务评估活动旨在实现系统之间的可比性，并利用数据准备成本，将研究问题作为开放的任务，供研究人员参与，具有统一的数据、任务规范、评估指标以及公开的结果报告和系统方法。在语音和语言技术方面使用STEC的经验表明，它们帮助标准化了共享数据，创建了有用的评估工具，并提供了发布、比较和讨论结果的场所。与会者将评估可用于计算处理的语言档案数据和应采取的STEC类型。讲习班的成果可通过加速濒危语言文献、加强数字语言档案和开发新的低资源语言计算技术产生长期影响。

The combined efforts of field linguists will not be sufficient to document the thousands of languages that are expected to disappear by the end of this century. Creating a lasting record of these disappearing languages is essential since all languages provide clues about how humans learn and process language. Language description and analysis could be facilitated by deploying computational analytic techniques on existing and ever growing repositories of digital audio and video language data.This project will organize a workshop bringing together researchers in speech and language technology, researchers on endangered languages, field linguists, and language repository archivists to help drive the development of this computational technology. The workshop participants will discuss how Shared Task Evaluation Campaigns (STECs) can promote the development of speech and language technology for documenting endangered and low-resourced languages. Designed to enable comparability across systems and leverage the costs of data preparation, Shared Task Evaluation Campaigns operationalize research problems as tasks that are open to participation by researchers with uniform data, task specifications, evaluation metrics, and open reporting of results and system methodologies.Experience with STECs for speech and language technologies has shown that they have helped standardize shared data, create useful evaluation tools, and provide venues for publishing, comparing, and discussing results. Participants will evaluate both the language archive data available for computational processing and the types of STECs that should be undertaken. The results of the workshop can have long term impacts through acceleration of endangered language documentation, enhancement of digital language archives, and development of novel computational techniques for low-resource languages.