Link Discovery: A Comprehensive Analysis

Abstract

We present a comprehensive analysis of approaches for discovering links in document collections. We classify link discovery approaches with respect to the type of knowledge being used: the text of a document, its title, and already existing links. Using an evaluation dataset derived from Wikipedia, we find that link-based approaches outperform all other approaches if they can draw knowledge from a very large amount of already existing links. Simulating other document collections with fewer links, we show that text-based approaches yield better results. Furthermore, we argue that knowledge from Wikipedia cannot necessarily be applied to other domains, e.g. in corporate intranets. Thus, we conclude that text-based approaches are the best choice for reliable link discovery in arbitrary document collections.

An evaluation framework for cross-lingual link discovery

“…Link mining approaches have also been proposed, including the work of He & de Rijke (2010), and Sorg & Cimiano (2008) use Support Vector Machines (SVM). Erbs, Zesch, & Gurevych (2011) provide a comprehensive analysis of state-of-the-art mono-lingual link discovery approaches.…”

Comparative Evaluation of Link-Based Approaches for Candidate Ranking in Link-to-Wikipedia Systems

“…It is referred to as disambiguation by Hachey et al (2011). However, as selecting the best link target usually involves creating a ranking of all the candidates to choose the one with the highest rank, other authors refer to the process as target ranking (Erbs et al, 2011) or candidate ranking (Guo, Tang, Che, Liu, & Li, 2011;Ploch, Hennig, de Luca, & Albayrak, 2011;. In this article, we will also adopt the term candidate ranking.…”

“…The development of automatic link discovery systems (Erbs, Zesch, & Gurevych, 2011) is another area of research where Wikipedia has had an important impact. The task of discovering links to Wikipedia articles has been addressed, with slight variants and under different names, by different communities.…”

“…According to Erbs et al (2011), the task of discovering links can be divided into a series of steps. They include: identifying the anchors to be linked, searching for candidate link targets for each anchor, and selecting the best candidate from the results of the searching step.…”

“…These features are extracted from the document where the anchor is placed (the context document) and/or the different Wikipedia articles considered as candidates to become the link target. According to Erbs et al (2011) the features can be classified into three groups: (1) those extracted from the text of the document/articles, (2) those extracted from their titles; and (3) those based on existing links. The latter are the subject of study in this paper.…”