Magellan: An adaptive ontology driven “breaking financial news” recommender

News recommendation has become an increasingly popular area of research for both the academic researcher and the commercial practitioner. Recently, a “semantic approach” to news recommendation has become popular. Semantic news recommendation normally relies upon a manually constructed domain Ontology and a user profile. The current semantic news recommenders have their flaws: 1. constructing and maintaining a manual Ontology is a manually intensive and laborious task, 2. user profiles may inhibit the selection of novel, but interesting stories which are atypical of the user´s previous selections and 3. the current approach semantic approaches ignores the “importance” of a story to the monitored domain. This paper describes Magellan (Magalhães), a semantic recommendation system which ranks “breaking financial news” by their “importance” to a monitored domain. In addition Magellan uses an automatically constructed and maintained adaptive Ontology which autonomously discovers new relations and entities whist removing outdated or erroneous information. Magellan applies a “temporal confidence” to each relation and once a relation´s “temporal confidence” falls below a preassigned constant it is removed. The ranking of stories is with a “weighted term frequency” (WTF) scheme. The domain Ontology is used to identify relevant terms in candidate stories. A term´s weight is determined by its: 1. Centrality and 2. Prestige (which incorporates a relation´s temporal confidence). An evaluation of Magellan with an inverted index and a semantic news recommender system with a static Ontology demonstrated a clear advantage for Magellan.