Towards multi-granularity multi-facet e-book retrieval

Abstract

There are more than one million digitalized books (i.e. e-books) so far in China-US Million Book Digital Library Project (MBP for short). It is thus important to design effective and powerful tools that enable users to easily search the required information and appropriately access knowledge in the digital library. Towards this end, currently most digital libraries simply use the traditional metadata-based or fulltext-based retrieval technologies on the e-book collection. However, there are at least two limitations of such e-book retrieval systems. (1) The granularity of retrieval results is either too big or too small, and consequently the middle granularities such as chapters and paragraphs are ignored in the traditional e-book retrieval systems. (2) The mass of retrieval results are usually ill-organized so that users often need to pay more efforts to obtain the required items. Therefore, with the many complex data in MBP, new search models and algorithms need to be developed that can take advantage of the particularities of e-books, access them appropriately, and provide results efficiently. To tackle this challenge, this paper introduces our multigranularity and multi-aspect e-book retrieval approach for MBP. Firstly, a Multi-granularity Multi-facet Knowledge Network (MMKN) model is proposed to represent content from different granularities (e.g., books, chapters, pages, paragraphs and words) and different facets (e.g., time, space, etc.) to support retrieval of relevant items from an e-book collection. Then we implement a novel e-book retrieval system, called IQuery, to extract facetrelated information from e-books at several granularities and then support multi-granularity e-book retrieval with more retrievable units and multi-facet navigation. Experiments were conducted to validate the efficiency and effectiveness of the proposed MMKN model, as well as the performance of IQuery. The results are encouraging, demonstrating that IQuery can provide powerful capabilities for e-book retrieval in MBP.

iULib: where UDL and Wikipedia could meet

“…6 Naturally, the BookNet can be directly used to find similar e-books in iULib 1.0. For more details about the construction of BookNet, please refer to Ref.…”

“…In our previous work, 6 we presented a novel e-book retrieval system, called IQuery, which provides multigranularity and multi-facet e-book retrieval. IQuery extended its capability of retrieval and navigation with a pre-constructed TermNet.…”

“…In our previous work, 6 we have demonstrated that TermNet can be used in IQuery to extend the query terms by selecting related terms. However, there are some drawbacks of this query expansion approach: a) MARC or DC subject terms are defined with subject headings or thesauruses.…”

“…Similar books here can be obtained by using two methods: a) Index all books' metadata and find similar e-books with metadata (such as title); b) Select correlated e-books from BookNet constructed in our previous work 6. …”

Modularization and multi-granularity reuse of learning resources

“…Multimedia retrieval systems cope with this granularity issue by segmentation of contents at different levels of granularity; the segments are made individually retrievable [Sch05]. The same applies to the retrieval of e-books: the user does not always want to retrieve a whole book, which contains his search terms spread over all pages, but rather only few pages or a chapter [HTZH07]. …”