MAG, a novel algorithm in AGDISTIS

This version contains the MAG approach which is included in the AGDISTIS framework. MAG consists of new capabilities.

• MAG is able to look for candidates considering the context from their neighbour entities. This context search is deterministic and does not rely on machine learning models.
• MAG sorts candidate entities based on their frequency in a given KB.
• MAG treats acronyms separately by performing normalization along with true-casing techniques (5 upper letters).
• MAG includes Page Rank and thus one more graph-based algorithm besides HITS.
• MAG can now return all possible candidates for a given mention.