Saturday, May 22 2004
Book indexes and tables of contents are useful, but of necessity general.
Researchers from Palo Alto Research Center (PARC) have given ebooks a more comprehensive index and table of contents tool that combines keyword searching and concept searching.
The system, dubbed ScentIndex, generates custom indexes and tables of contents organized according to a set of keywords that the user enters to describe the concepts she is interested in. The custom entries are drawn from the original manually-generated book index entries.
When a user enters a set of keywords, the system narrows down its index of thousands of entries to a single page's worth of a few dozen entries and displays them for the user. Keywords are also highlighted within the text of the ebook.
The researchers carried out a user study of the system using the non-fiction ebook Biohazard. The study showed that an ebook outfitted with the ability to custom generate tables of contents and indexes made for a faster and more accurate way to search through text than the physical book, largely by narrowing down the number of entries that the user searched through to find a correct answer.
The researchers' algorithm narrows down the entry using a word co-occurrence matrix to extract entries from the index that are conceptually relevant to the keywords, according to the researchers.
The work was presented at the Computer-Human Interaction (CHI) 2004 conference in Vienna, Austria, April 24 to 29.
