1. Why are there so many catalogs and what can we do about it? Robin Wendler (and Dale Flecker) November 2, 2000 Tufts Metadata Conference

2. Catalogs galore Traditional library materials Social science data sets Art and Cultural Images Archives Botanical specimens Biomedical Images Geo-spatial Data Networked Resources Robin: Research resources do not reside only in libraries. The visibility made possible by the web is breaking down the barriers between different kinds of research organizations, not least in the minds of researchers. the greater availability of automation options and, most importantly, the easy distribution made possible by the web have influenced more departments to make their information available online and has led to rising expectations about online access to collections and services At Harvard, this manifests itself in the development of multiple systems to provide public access to different parts of the collections. These parallel public access systems augment HOLLIS as ways to make Harvard resources known to the community. These systems are tailored to a particular kind of intellectual access, which is often closely but not exclusively associated with a particular type of material In each of these systems, some form of cataloging is created, and internal consistency in the data is necessary if searches of the system are to yield predictable results. The information is created in different departments, by staff with different kinds of training, with reference to different standards and vocabularies Robin: Research resources do not reside only in libraries. The visibility made possible by the web is breaking down the barriers between different kinds of research organizations, not least in the minds of researchers. the greater availability of automation options and, most importantly, the easy distribution made possible by the web have influenced more departments to make their information available online and has led to rising expectations about online access to collections and services At Harvard, this manifests itself in the development of multiple systems to provide public access to different parts of the collections. These parallel public access systems augment HOLLIS as ways to make Harvard resources known to the community. These systems are tailored to a particular kind of intellectual access, which is often closely but not exclusively associated with a particular type of material In each of these systems, some form of cataloging is created, and internal consistency in the data is necessary if searches of the system are to yield predictable results. The information is created in different departments, by staff with different kinds of training, with reference to different standards and vocabularies

3. REASONS FOR MULTIPLE CATALOGS Desire for autonomy Varying functional requirements Community-specific conventions, terminology Different metadata formats appropriate for different materials or in different contexts

4. DESIRE FOR AUTONOMY –libraries –museums –archives –herbaria –academic departments –research labs –hospitals –... Catalogs operated by different administrative units such as …units which may have more interest in interoperating with their fellows across institutional boundaries than with other kinds of organizations within the institution

5. FUNCTIONAL NEEDS DIFFER Library catalogs: –support circulation; placing holds, recalls, or requests from remote storage –optimized for searching a large database and browsing large result sets –draw a line between finding and using material –use standards to support large scale exchange of metadata –standard metadata lends itself to automated processing (e.g., authority control, identifying duplicates, merging records, creating well-ordered result lists

6. FUNCTIONAL NEEDS DIFFER Image catalogs: integrate display of images with the catalog; “light table”, image comparison tools Geospatial catalogs: search via “bounding polygon” interface and determine relevance based on proportion of overlap, support “preview” rendering of data Statistical data catalogs: order datasets from ICPSR, exploratory statistical modeling Biomedical image catalogs: link between research projects, supporting images and resulting publications

7. TERMINOLOGY AND CONVENTIONS DIFFER For people, organizations, places, topics... –Libraries use Library of Congress and Medical Subject Headings –VIA uses the Art and Architecture Thesaurus and Union List of Artists Names –Herbaria use standardized botanical names and form personal names according to centuries-old practice –Geodesy uses conventional notations for geographic coordinates

8. METADATA DIFFERS... Because of historically different practices –Library standards require describing the object in hand –Photo collection standards describe the object pictured –Archives describe collective materials as they are organized And these differences are reflected in the formats used to record the descriptions

9. METADATA DIFFERS... With the structure of what is being described –Image cataloging is often hierarchic, with many pictures of a single described object, site, etc. –The cataloging for an archival collection is structured to replicate the logical arrangement of the collection –Dataset descriptions include variables and their locations

10. METADATA DIFFERS... With community schemes and standards –Libraries use MARC and AACR2 –The GIS community uses FGDC –The archival community uses EAD –The survey data community will be using DDI –The image community will use VRA Core –The text encoding community uses TEI Start with elements, move toward rules

11. MORE REASONS FOR MULTIPLE CATALOGS Smaller catalogs are easier to use –1.8% of all HOLLIS searches exceed maximum result set limit (126,659 searches of 7 mill. in FY99) –fewer functions to learn, but those used more often Specific catalogs can be tailored to targeted audiences –increasing precision of search results –providing richer (or more frequently needed) functionality

12. BUT... Multiple catalogs are confusing –How does a user know where to look? Multiple catalogs are inconvenient –Need to repeat a search multiple times

13. SOME POSSIBLE SOLUTIONS Replicated descriptions Distributed search Super-catalog Links

14. REPLICATED DESCRIPTIONS Same material described in more than one catalog –MARC AMC records and EAD finding aids –MARC and the library Portal –MARC for ICPSR datasets and Harvard/MIT Data Center records Geodesy to experiment with single point of metadata creation/maintenance feeding two catalogs (HOLLIS and Geodesy)

15. REPLICATED DESCRIPTIONS Issues –Can be labor intensive –Added maintenance burden –Mapping between metadata standards doesn’t work well ALWAYS involves some loss (of data, of meaning, of specificity, and/or of accuracy) may be extremely difficult, e.g., Hierarchical VIA records or EAD finding aids would not map well into MARC

16. DISTRIBUTED SEARCH SEARCH FRONT END 1. QUERY SYSTEM 1 SYSTEM 2 SYSTEM 3 2. QUERY 3. RESPONSE 4. SUMMARY OR CONSOLIDATED RESPONSE

17. DISTRIBUTED SEARCH Front-end query interface –Reformats user query as appropriate for each target system May allow user to choose which target(s) to query –Sends queries in parallel –Handles search results May consolidate results into single set May simply summarize number of hits, and pass user to specific target system to display results

18. DISTRIBUTED SEARCH -- ISSUES Front-end system is complex –Need to understand each target system Search syntax Results responses and formats Easier if all targets support Z39.50 –Constant maintenance is required as target systems are modified Performance sensitive to weakest link

19. DISTRIBUTED SEARCH -- ISSUES Target systems frequently have non-parallel functions or use different terminology “find author” vs “find person” “cancer” vs “neoplasms” Consolidating results into a single set is difficult –How to de-duplicate when same item is described in more than one system –How to order heterogeneous result sets –How to display heterogeneous data formats

20. SUPER-CATALOG SUPER- CATALOG 2. QUERY SYSTEM 1 SYSTEM 2 SYSTEM 3 3. RESPONSE 1. CONTRIBUTE METADATA 1. CONTRIBUTE METADATA 1. CONTRIBUTE METADATA

21. SUPER-CATALOG Union catalog of data from separate systems –Data collected through contribution or via “harvesting” Data may require homogenizing –Format –Data elements –Terminology

22. SUPER-CATALOG -- ISSUES Homogenizing can be complex –Terminology particularly difficult Homogenizing tends towards least-common- denominator –If one contributor only labels “person”, cannot offer “author” search Likely to produce a catalog of “apples and oranges” –Single photographs/whole archival collections

23. RELATED IDEA: “ACADEMIC LYCOS” Super catalog built from data in many academic research catalogs across institutions Built on Internet search engine technology –Based on familiar concepts and interfaces Being explored by DLF with Mellon foundation encouragement

24. LINKS Supports navigation and assistance for sequential searching of multiple systems After searching one catalog, user given options of pursuing same query in other sources Primary exemplar is SFX system

25. SFX SYSTEM SYSTEM 1 1. QUERY 2. RESULTS WITH “LINKS?” BUTTON SFX SYSTEM 3. USER CLICKS “LINKS?” BUTTON 4. PAGE WITH MULTIPLE LINK OPTION BUTTONS SYSTEM 2 5. BUTTON GENERATES PRE-FORMATTED SEARCH 6. RESULTS

26. LINKS -- ISSUES Each source system must be modified to provide appropriate “LINKS?” button Links server must understand data formats and search syntax for each linked system Does not address problems of non-parallel terminology and search functionality Potential user frustration, as many links will be dead ends

27. THEREFORE…. Many approaches, no ideal solution –Fundamental problem in digital libraries –Problem and solutions being widely analyzed today