Mining Cross-document Relationships from Text
Petr Knoth, Zdeněk Zdráhal · Open Research Online (The Open University) · 2011
Abstract—The paper argues that automatic link generation and typing methods are needed to find and maintain crossdocument links in large and growing textual collections. Such links are important to organise information and to support search and navigation. We present an experimental study on mining cross-document links from a collection of 5000 documents. We identify a set of link types and show that the value of semantic similarity is a good distinguishing indicator. Keywords-text mining, automatic link generation and typing, semantic similarity, digital libraries I.