Document recognition for a Digital Library

Abstract

Our work concerns three major processes which help to build a DL database for IR. Several components of the system are shown to be useful for creating a DL database. The design of these processes stresses the importance of robustness and ease of adaptation to the processing of different documents. Output generated by these processes facilitates the IR mechanism to produce intelligent response to user queries. An adaptive approach to document understanding was presented in this chapter. Its robustness was shown to be crucial to the success in processing varied library documents. This chapter also presented an adaptation of the vector space model for information retrieval to improving the performance of a word recognition algorithm. The neighborhoods of visually similar words determined by word recognition are matched to a database of documents and a subset of documents with topics that are similar to those of the input image are determined.