• 主题
高级搜索  >
历史搜索 清空历史
首页> 外文会议>BioNLP shared task 2013 >Generalizing an Approximate Subgraph Matching-based System to Extract Events in Molecular Biology and Cancer Genetics
【24h】

Generalizing an Approximate Subgraph Matching-based System to Extract Events in Molecular Biology and Cancer Genetics

机译：概括基于近似子图匹配的系统以提取分子生物学和癌症遗传学中的事件

获取原文
获取原文并翻译 | 示例
页面导航
  • 摘要
  • 著录项
  • 相似文献
  • 相关主题

摘要

We participated in the BioNLP 2013 shared tasks, addressing the GENIA (GE) and the Cancer Genetics (CG) event extraction tasks. Our event extraction is based on the system we recently proposed for mining relations and events involving genes or proteins in the biomedical literature using a novel, approximate subgraph matching-based approach. In addition to handling the GE task involving 13 event types uniformly related to molecular biology, we generalized our system to address the CG task targeting a challenging set of 40 event types related to cancer biology with various arguments involving 18 kinds of biological entities. Moreover, we attempted to integrate a distributional similarity model into our system to extend the graph matching scheme for more events. In addition, we evaluated the impact of using paths of all possible lengths among event participants as key contextual dependencies to extract potential events as compared to using only the shortest paths within the framework of our system. We achieved a 46.38% F-score in the CG task and a 48.93% F-score in the GE task, ranking 3rd and 4th respectively. The consistent performance confirms that our system generalizes well to various event extraction tasks and scales to handle a large number of event and entity types.
机译：我们参加了BioNLP 2013共享任务，处理了GENIA（GE）和Cancer Genetics（CG）事件提取任务。我们的事件提取基于我们最近提出的一种系统，该系统使用一种新颖的基于近似子图匹配的方法来挖掘生物医学文献中涉及基因或蛋白质的关系和事件。除了处理涉及13个与分子生物学统一相关的事件类型的GE任务外，我们还对系统进行了通用化处理，以解决CG任务，该任务针对具有挑战性的40种与癌症生物学相关的事件类型，涉及18种生物学实体的各种论点。此外，我们尝试将分布相似性模型集成到我们的系统中，以扩展图匹配方案以用于更多事件。此外，与仅在系统框架内使用最短路径相比，我们评估了将事件参与者之间所有可能长度的路径用作关键上下文依存关系以提取潜在事件的影响。我们在CG任务中获得46.38％的F评分，在GE任务中获得48.93％的F评分，分别排名第三和第四。一致的性能证明我们的系统可以很好地概括各种事件提取任务，并且可以扩展以处理大量事件和实体类型。

著录项

相似文献

  • 外文文献
  • 中文文献
  • 专利
获取原文

客服邮箱：kefu@zhangqiaokeyan.com

京公网安备：11010802029741号 ICP备案号：京ICP备15016152号-6 六维联合信息科技 (北京) 有限公司©版权所有

  • 客服微信

  • 服务号