Proceedings of the BioNLP Shared Task 2011 Workshop
Proceedings of the BioNLP Shared Task 2011 Workshop
复制标题
BioNLP 共享任务 2011 研讨会论文集
DOI：
--
复制
发表时间：
2011
期刊：
影响因子：
--
通讯作者：
Sampo Pyysalo
中科院分区：
文献类型：
--
作者：
Junichi Tsujii;Jin;Sampo Pyysalo
在生物医学领域，通过语义搜索、辅助途径注释和自动识别特定生物分子反应以支持数据库管理等应用程序，提高对大量科学文献的访问要求，不断要求开发先进的生物医学信息提取和文本挖掘的方法和资源。BioNLP共享任务系列旨在通过更加关注提取信息的详细结构化表示，具有完全文本绑定注释的新语料库资源以及精确的任务定义，支持和评估来推动这一发展。
BioNLP共享任务2011是该系列的第二个，是2009年组织的第一个活动的后续活动。为了在上一次活动的成功基础上再接再厉，该任务由亚洲、欧洲和美国的几个小组合作组织，他们共定义了8项涉及不同挑战的具体任务，除了结构化事件提取外，还包括关系提取和支持任务，如共指消解。2011年活动的主题是泛化，主要任务在2009年的基础上进一步扩大了三个方面：文本类型，主题领域和新事件提取目标。
该任务吸引了社区的广泛兴趣，共收到来自24个团体的46份最终提交，保持了2009年任务参与人数，同时提交数量几乎翻了一番。除了生物医学文本挖掘社区的持续兴趣之外，我们很高兴欢迎来自学术界和工业界的许多新团体的参与。提交的材料表明，在既定的事件提取任务取得了实质性进展，并表明事件提取方法普遍适用于全文，新的主题领域，如传染病和细菌相互作用，以及新的事件集，如蛋白质翻译后修饰。
The requirements of improved access to the massive amount of scientific literature in biomedical domain - through applications such as semantic search, assisted pathway annotation, and the automatic identification of specific biomolecular reactions for database curation support - place continuing demands on the development of methods and resources for advanced biomedical information extraction and text mining. The BioNLP Shared Task series seeks to advance this development through an increased focus on detailed structured representations of extracted information, novel corpus resources with fully text-bound annotation, and precise task definitions, support and evaluation.
The BioNLP Shared Task 2011 is the second in the series, following up on the first event organized in 2009. Seeking to build on the success of the previous event, the task was organized as a collaboration between several groups in Asia, Europe and the US who defined in total eight specific tasks involving diverse challenges, including in addition to structured event extraction also relation extraction and supporting tasks such as coreference resolution. The main theme of the 2011 event was generalization, and the main tasks further broadened on the 2009 setup in three aspects: text types, subject domains, and novel event extraction targets.
The task attracted broad interest from the community, and a total of 46 final submissions were received from 24 groups, maintaining the 2009 task participation numbers while nearly doubling its number of submissions. In addition to the continued interest from the biomedical text mining community, we were glad to welcome the participation of many new groups from academia and industry. The submissions demonstrated substantial progress at the established event extraction task and showed that event extraction methods generalize well, among other aspects, to full papers, new subject domains such as infectious diseases and bacterial interactions, and new sets of events such as protein post-translational modifications.