Scenarios for Advanced Services in an ETD Digital Library

ETDs, typically in PDF, are a largely untapped international resource. Digital libraries (DLs) with advanced services can effectively address the broad needs to discover and utilize ETDs of interest. DLs support indexing, searching, and browsing. However, when only metadata is available, these capabilities are insufficient. Using full text to extend faceted searching provides improvement, but adds noise and reduces precision. Natural language processing (NLP), e.g., information extraction (IE), yields additional improvement, but results are like with Google Search. Google Scholar and CiteSeerX -- which extract, analyze, and link references in short publications -- provide additional capabilities, but do not work well with ETDs (due to length, complexity, and domain variations). We are working toward a tailored DL for English ETDs with special services -- including for processing references and citations, as well as for extraction from chapters, sections, and subsections -- that review the literature, state hypotheses, list research questions, explain the approach, describe methods, summarize results, discuss findings, draw conclusions, and provide insights about open problems. Such a domain independent DL can be prototyped now, using advanced NLP and IE techniques, coupled with machine learning and information retrieval methods. The result would enable stakeholders to engage in more advanced scenarios.