Link Analysis Approaches to Natural Language Processing and Text Mining

我们研究了如何有效地将链接分析技术应用于自然语言处理任务。具体来说，我们开发了一种方法，用于分析自然语言句子中的连接结构。该方法借鉴了链接分析中的相似度计算方法，即计算感兴趣节点之间的所有路径。我们提出的方法将这一思想应用于寻找句子中最相似的短语，这些短语可能会形成协调。我们还提出了一种图的构造方法，减少了“枢纽”的影响，这是位于附近的许多其他节点在一个图中的节点。

We investigated how to effectively apply link analysis techniques to the tasks of natural language processing. Specifically, we developed a method for analyzing conjunctive structures in natural language sentences. The method borrows an idea from the similarity computation method in link analysis, which counts all paths between the nodes of interest. Our proposed method applied the idea to finding the most similar phrases in a sentence, which are likely to form coordination. We also proposed a graph construction method that reduces the influence of "hubs", which are the nodes located close by to many other nodes in a graph.