Information retrieval from a large-scale document network

Takashi Sonoda System Technology Laboratory,
Research & Technology Group
Hiroshi Okamoto same as above
Yukihiro Tsuboshita same as above

In the average office, a variety of documents are accumulated daily. A large-scale network is gained when these documents are linked by their attributes (for instance, similarities or citation relations). Here, we propose a method to extract relevant information from such document networks. This method is based on the algorithm inferred from the brain mechanism for retrieval of associative memory. We demonstrated the validity of this method by applying the algorithm to a citation network of patents. A set of patents retrieved by this method was in good agreement with those arranged by human experts. These results suggest that the proposed method might be useful for finding valuable information in a pile of office documents.

To Subsequent page

Return to Top page