Improving Automatic Semantic Tag Recommendation through Fuzzy Ontologies

Semantic tagging of a textual document involves identifying and assigning to it appropriate entities that best summarize its content, i.e. entities that constitute a representative description of what the document is specifically about. The effective automation of this process requires from the system to be able to distinguish between the entities that play a central role to the documents´s meaning and those that are just complementary to it. For example, a news article might make reference to many politicians even when its primary subject is only one of them. To that end, various approaches have utilized ontologies as a means to narrow down the meaning of a document and infer appropriate tags, including a recent contribution of ours regarding a tagging framework that exploits ontological relations. In this work we revise and extend this framework so as to be able to exploit also fuzzy ontological information. Experiments in different domains show that this exploitation manages to improve the effectiveness of the tagging process.