Hristidis et al. extend the notion of ranking retrieved textual items from a database via Google’s PageRank by giving authority to the citing papers and to the citing authors. The basic data structure is a labeled directed graph. A demonstration Web site is available.
The system allows a user to input a set of keywords as a query, with optional Boolean operators. The user can also input a global object-rank importance parameter, a damping factor, and a specificity metric (inverse object rank). The database is an available set of documents that can be accessed on the Web. Hristidis et al. note that a query keyword provides a set of documents with that keyword, and then the database graph can be traversed to provide a keyword-specific ranking. Moreover, one can calibrate the specificity metric (inverse object rank) and quality metric (global object rank).
Finally, the paper offers an ontology graph, based on domain knowledge, to expand the search--only the inheritance of attributes, the “isa” relationship, is considered.
The paper provides an interesting approach that merits further consideration and testing.