Description: This paper discusses Chinese information retrieval using Lemur. Abstract: This paper describes our participation in NTCIR-5 Chinese Information Retrieval (IR) evaluation. The main purpose is to evaluate Lemur, a freely available information retrieval toolkit. Our results showed that Lemur could provide above average performance on most of the runs. We also compared manual queries vs. automatic queries for Chinese IR. The results show that manually generated queries did not have much effect on IR performance. More analysis will be carried out to discover causes behind hard topics and ways to improve the overall retrieval performance.

Description: This paper describes the design and development of a learning object repository for a new statewide higher education initiative. The Texas Higher Education Coordinating Board is funding the redesign of large undergraduate courses; the redesigned courses are intended to improve student learning, retention and outcomes. The learning object repository stores and provides access to content from these courses. Content has been decomposed into discrete learning object varying in levels of granularity. The paper provides details on the proof-of-concept implementation developed in Phase I of a two-phase project. Special attention is given to key aspects such as the levels of granularity, metadata, technology, and user testing. ssues that emerged in Phase I are informing all facets of the next iteration of the repository.

Description: This paper discusses issues of semantic interoperability in digital libraries. The networked information environment comprising digital libraries, digital collections, and digital repositories increase people's expectations for information access. Specifically, users anticipate better search capabilities across these networked information resources and the metadata records associated with the resources.

Description: This paper discusses assessing descriptive substance in free-text collection-level metadata. Abstract: Collection-level metadata has the potential to provide important information about the features and purpose of individual collections. This paper reports on a content analysis of collection records in an aggregation of cultural heritage collections. The findings show that the free-text Description field often provides more accurate and complete representation of subjects and object types than the specified fields. Properties such as importance, uniqueness, comprehensiveness, provenance, and creator are articulated, as well as other vital contextual information about the intentions of a collector and the value of a collection, as a whole, for scholarly users. The results demonstrate that the semantically rich free-text Description field is essential to understanding the context of collections in large aggregations and can serve as a source of data for enhancing and customizing controlled vocabularies.

Description: This paper discusses building contextual mass in digital aggregations for scholarly use. Abstract: At present there are no established collection development methods for building large-scale digital aggregations. However, to realize the potential of the collective base of digital content and advance scholarship, aggregations must do more than provide search of sizable bodies of content. Informed by empirical understanding of scholarly information practices, the IMLS Digital Collections and Content project developed an aggregation strategy for building Opening History, one of the largest digital cultural heritage aggregations in the country. The strategy applied policy-driven collecting based on the principle of contextual mass, and conspectus-style evaluation of collection-level metadata to identify strong subject areas within the aggregation. Analysis of density, interconnectedness, diversity, and small/large collection complementary determined subject concentrations and thematic strengths to be prioritized for future collection development and used as organizational structures for browsing and visualization. The approach models how scholars build their own personal research collections, as they follow leads from collection to collection across institutions near and far, and adds value that cannot be achieved through conventional retrieval and browsing at the item-level.