Event mining from the Blogosphere using topic words
Event mining from the Blogosphere using topic words
复制标题
DOI：
--
复制
发表时间：
2007-12
期刊：
影响因子：
--
通讯作者：
Yoshihiko Suhara;Hiroyuki Toda;A. Sakurai
中科院分区：
文献类型：
--
作者：
Yoshihiko Suhara;Hiroyuki Toda;A. Sakurai
本文提出了一种从日语博客中抽取相关主题词之间的"动作关系"的方法。动作关系是主体、目标和谓词的元组。我们的方法通过AND搜索获得包含两个关键词的博客文章，并输出由谓词和两个关键词构造的动作关系，后面跟着格助词。然而，由于单词经常在出现一次后被省略，因此包含这两个关键词的句子很少。我们的方法克服了这个问题，通过合并，ING句子，包含共同的谓词和附近的位置。因此，本研究的重点是从部落格中撷取相关主题词间的动作关系。在本文中，我们定义为“事件挖掘”的事件内容提取简单的表达式。为了提取动作关系，我们把注意力集中在谓词和它们的格元素上：名词后跟格助词。日语中格助词决定格成分的格，因此我们可以通过格助词来判断格成分的语义角色。虽然一个格助词根据其谓语和语境的不同可能具有多种意义，但我们把"ga"格看作动作的施事，把"ni"和"wo"格看作动作的目标，把谓语看作动作本身。然后，我们可以从这两个关键词作为格元素共现的句子中提取这两个关键词之间的“动作关系”。然而，在前一句中出现过的词常常被省略或用指示词代替。虽然有必要使用回指解析来适当地补充格元素，但回指解析还没有完全发展到实际应用中（2）。因此，我们考虑使用大量的博客文章来提取动作关系的简单方法来解决这个问题。
In this paper, we propose a method of extracting "action re- lations" between related topic words from Japanese weblogs (blogs). An action relation is a tuple of agent, target and predicate. Our method obtains blog articles that contain two keywords by AND search and outputs action relations con- structed from the predicate and the two keywords with case particles that follow. However, since words are often omitted after they appear once, sentences with both of the keywords are scarce. Our method overcomes this problem by combin- ing sentences that contain the common predicate and locate nearby. of this research, therefore, is to extract the action relation between related topic words from blogs. In this paper, we define "event mining" as the extraction of event contents as simple expressions. To extract action relations, we have focused attention on predicates and their case elements: nouns followed by a case particle. Since a case particle decides the case of the case element in Japanese, we can estimate the semantic role of the case element by its case particles. Although one case particle may have several meanings depending on its predicate and the context, we regard ga case as agent of the action, ni and wo case as target of the action, and the predicate as an action itself. Then we can extract the "action relation" between two keywords from the sentence in which these two keywords are co-occurred as case elements. However, a word which has appeared in a preceding sen- tence is often omitted or replaced with a demonstrative word. Although it is necessary to complement the case element properly using anaphora resolution, anaphora resolution is not yet fully developed for practical use (2). Therefore, we consider the simple way to solve this problem to extract ac- tion relations by using massive amount of blog articles.