Order Retrieval

Abstract. Extensive work has been done in recent years on automatically grouping words into categories. For example, {Wednesday, Monday, Tuesday} could be grouped into a `days of week' category. However, not only grouping the words, but also ordering them is important, e.g. MondayTuesdayWednesday. The order relation is an important aspect that could be used to enrich existing ontologies, to determine the sequence of actions for planning tasks, and to determine the order of user's preferences for a set of items, etc. However, automatically determining the order relation seems to have been ignored. Pairwise similarity metric commonly used to cluster words may not be well suited for the ordering task. Therefore, we propose a new metric designed for the ordering task. We utilize statistical proximity features of the terms in the documents (in a large corpus) in order to determine the order relations between terms. The effectiveness of the proposed method is verified in experiment...

• Type: Conference
• Year: 2008
• Where: LKR
• Authors: Neil Rubens, Vera Sheinman, Takenobu Tokunaga, Masashi Sugiyama
• Keywords: Extensive Work, Important Aspect, Knowledge Management, LKR 2008, Order Relations