Machine Assistance in Collection Building: New Tools, Research, Issues, and Reflections

摘要

Digital tool making offers many challenges, involving much trial and error. Developing machine learning and assistance in automated and semi-automated Internet resource discovery, metadata generation, and rich-text identification provides opportunities for great discovery, innovation, and the potential for transformation of the library community. The areas of computer science involved, as applied to the library applications addressed, are among that discipline's leading edges. Making applied research practical and applicable, through placement within library/collection-management systems and services, involves equal parts computer scientist, research librarian, and legacy-systems archaeologist. Still, the early harvest is there for us now, with a large harvest pending. Data Fountains and iVia, the projects discussed, demonstrate this. Clearly, then, the present would be a good time for the library community to more proactively and significantly engage with this technology and research, to better plan for its impacts, to more proactively take up the challenges involved in its exploration, and to better and more comprehensively guide effort in this new territory. The alternative to doing this is that others will develop this territory for us, do it not as well, and sell it back to us at a premium. Awareness of this technology and its current capabilities, promises, limitations, and probable major impacts needs to be generalized throughout the library management, metadata, and systems communities. This article charts recent work, promising avenues for new research and development, and issues the library community needs to understand.