Concept Network: A Structure for Context Sensitive Document Representation
Abstract:
In this paper we propose a directed acyclic graphical structure called concept network (CNW) for context sensitive document representation for use in information filtering. Nodes of CNW represent concepts and links represent the relationships between the concepts. A concept can either be a phrase or a topic of discourse, or a mode of discourse. An important feature of the CNW based scheme (CNWBS) is context filters [Murthy and Keerthi, 1999] which are employed on the links of the graph to enable context sensitive analysis and representation of documents. Context filters are intended to filter the noise in the inputs of the concepts based on the context of appearance of their inputs. The CNWBS automatically finds the paragraphs related to all concepts in the document. It also provides good comprehensibility in representation; allows sharing of CNW among a group of users; and, reduces the credit-assignment problem during its construction. This representation scheme is used for document representation in Sharable Instructable Information Filtering Agent that we are developing. 1
Citations
No citations identified.
