Searchable Metaspaces

The purpose of this presentation is to start a discussion about methodological and operational requirements for developing tools for internet browsing and/or querying of metadescriptions of language resources, in particular multimodal corpora. Among the most important requirements are: delimiting the relationship both between meta-descriptions and the resources they apply to, and between browsing and querying over the internet; establishing a standard for representing meta-descriptions; administering the web-based availability of language resource and their accessibility via meta-descriptions; and establishing user support for query editing and data interchange. We attempt to stake out positions regarding these requirements, addressing both their advantages and disadvantages. We base our positions on the EAGLES/ISLE proposal for a meta-description standard for language resources ([1]). Our views are also in uenced by work on the development of query languages for linguistic resources, such as CQP, the MATE query language Q4M, and the TIGER query language for syntactic tree annotations. 1 Overview of Objectives With the increasing development and use of multi-modal language resources, there is a growing need for suitable tools to access and query these resources. To facilitate these tasks, it is common and essential for the resources to be associated with meta-descriptions of their content (including object data annotations). Two perspectives are possible and relevant in this context: The local or site perspective: an institution has a (number of) multi-modal resource(s), and somebody wants to identify parts of these resources that satisfy certain meta-descriptions. Possibly one even wants to retrieve, from such resource(s), certain subsets (say turns, sentences, whole dialogues), according to a combination of metadata and linguistic (or other modality-speci c) criteria annotated in the resource. The resources are accessed locally and the search is also carried out on site. The global or web perspective: somebody wants to know about the existence of resources of a certain kind (i.e. satisfying certain conditions in terms of meta-descriptions); if a web search engine accepts the required meta-descriptions, then the resources can be located by browsing, and possibly even accessed and queried over the web. Although these perspectives make di erent demands on implementation, we will see that, from the point of view of the language resource user, they complement rather than compete with each other; hence, resource owners should accommodate both perspectives. Following the EAGLES/ISLE http://www.ims.uni-stuttgart.de/projekte/CorpusWorkbench/ http://www.cogsci.ed.ac.uk/ dmck/MateCode/ http://www.coli.uni-sb.de/cl/projects/tiger/