Towards a Complete Event Type Taxonomy
Towards a Complete Event Type Taxonomy
复制标题
走向完整的事件类型分类
DOI：
10.1145/2740908.2742005
复制
发表时间：
2015
期刊：
影响因子：
--
通讯作者：
B. Fortuna
中科院分区：
文献类型：
--
作者：
Aljaz Kosmerlj;E. Belyaeva;Gregor Leban;M. Grobelnik;B. Fortuna
我们展示了我们努力建立广泛而完整的新闻文章中描述的事件分类的初步结果。通过抓取维基百科的当前事件门户，我们确定了九种顶级事件类型。利用门户引用的文章，我们使用词汇和语义特征为新闻文章构建了事件类型分类模型，并对其结果进行了小规模手动评估。结果表明，我们的模型可以准确地区分事件类型，但其覆盖范围仍然可以显着提高。
We present initial results of our effort to build an extensive and complete taxonomy of events described in news articles. By crawling Wikipedia's current events portal we identified nine top-level event types. Using articles referenced by the portal we built a event type classification model for news articles using lexical and semantic features and present a small-scale manual evaluation of its results. Results show that our model can accurately distinguish between event types but its coverage could still be significantly improved.