Marek Lipczak
Arash Koushkestani
Evangelos Milios
This article presents Tulip, an ERD system submitted to the ERD 2014: Entity Recognition and Disambiguation Challenge. The objective of the proposed system is to spot mentions of entities in a document and link the mentions to corresponding Freebase articles. To achieve it, Tulip prunes the set of entity candidates focusing on a core subset of related entities capturing the context of the document. The relationship strength is measured as a similarity to a topic centroid generated from entity features. Each entity is represented by an accurate and compact feature vector extracted from a category graph built based on information from 120 language versions of Wikipedia. Given the core set of accepted entities Tulip uses the Wikipedia-based feature vectors to extract more related entities from the document text. Tulip received the first prize in the long document track with F1 score of 0:74, which confirms the effectiveness of our system. At the same, the system was faster than all other submissions with latency under 0:29 seconds.
http://dl.acm.org/citation.cfm?id=2634351&CFID=564807534&CFTOKEN=61327522