Text mining in a digital library

摘要

Text mining can be used to add value to documents in digital library collections; Greenstone already incorporates a few examples. What we have described is how a general text mining environment can be included within a digital library system. This provides a more satisfactory basis for putting future advances in text mining to work immediately for the benefit of digital library users. Our current implementation processes documents at display time only: in the future we plan to incorporate text mining into the collection-building process so that the data produced can be used for searching and browsing as well as for document display. It will also be necessary to design ways in which desired text mining operations can be specified in a collection's configuration file. The linkage of a general digital library system with a general text mining system presents many other possibilities. Tracking entities across documents leads to automatic hyperlinking of coreferences. Semantic indexing could be accomplished by annotating texts with ontology classes and allowing semantic searches rather than simple textual queries, providing some of the advantages of the Semantic Web. Document summaries could be generated automatically to serve as the "snippets" that are presented in lists of search results. Having GATE embedded in a digital library system will also benefit its own users by allowing them to experiment on real text collections in a digital library rather than with specially constructed text corpora. This spells good news for digital library users.