1	Dumbing Up or Dumbing Down?
2	Developing a flexible information architecture NYPL Digital Gallery Project Background Content vs. Context Building the Public Interface Challenges and lessons Defining categories (apples or oranges) ? New I.A. directions
3	Images at NYPL New York Public Library images have historically been given exhibition-level treatment online Single, distinct collections Curatorial oversight for images & descriptions Size = 150 to 8,000 images Web site(s) = stand-alone applications
4	Moving from an exhibition model to data-driven infrastructure Ingest data from disparate Library collections Oracle database & in-house Digital Assets Mgt. System Metadata for 500,000+ images Image data & storage Input by Metadata, Imaging staff Legacy-data loaded from other sources Multiple terabytes of archival TIFF, JPG and GIF files
5	APIs, hardware, software Back-end interface Oracle, Cold Fusion application server Public interface Sun (UNIX) presentation server Lucene (open source) search engine, XML data files CFMX custom tags, java, CFML for display
6	CONTENT VS. CONTEXT
7	Access points for Digital Library Program content Content can be accessed via: Descriptive Metadata Collection naming Categorization Randomly
8	Content description - examining a Real World Object
9	Assigning access points
10	Input metadata into Digital Asset Management System
11	Organization of a Real World collection
12	Real World Collection structure: for large collection “Photographic views of New York City, 1870's-1970's ” Over 50,000 images Flat structure, few levels Each directory holds thousands of images Each record contains image(s) Artifact has verso
13	Real World Collection structure: for small collection “The A.G. Spalding Baseball Card Collection Less than 1,000 images Flat structure, no levels Subjects might be leveraged for hierarchy? players, positions players, by teams
14	Metadata describes these collection/title “levels” ...
15	... while XML allows flattening of levels for search & retrieval
16	Example of earlier title-levels display
17	Searchable, expandable title-levels
18	Browsing a multi-level title
19	Functional Requirements for record-level access points
20	Interpreting access points
21	Providing context for NYPL Research Libraries’ material
22	Providing context: NYPL call number = 0 hits in library catalog
23	Providing context: NYPL title = 0 hits in library catalog
24	Providing context: Guide to the (NYPL) Research Collections
25	How can we provide sufficient Research Libraries’ context? In the previous example, a date or title is difficult to locate in the NYPL Research Libraries’ catalog Title is not available in online NYPL Catalog 18th Century date for Playing Cards: Mechanics is mentioned in the Guide to the Research Collections of the New York Public Library (published 1975) How do we link to this type of non-cataloged information from Digital Gallery metadata, especially if it is not in a standard reference format?
26	BUILDING THE PUBLIC INTERFACE
27	The new public (inter)face of NYPL’s Digital Gallery Where we’ve been Several earlier iterations in previous development Challenges: data issues, software development framework(s) Version 4.x : Dumbing up or down? simple = not necessarily elegant ? simpler = user-friendly ? more access paths = user confusion? better access paths = user confidence?
28	Where we’d been: hand-crafted collections
29	Collection example
30	Hand-building virtual collections Did not scale Necessitated endless revision of both content and labels Cognitive dissonance created by interface: when user clicked into a detail view or searched: the categories and titles in our data were often different than our curated title lists Content not managed by data
31	WHAT’S IN A CATEGORY (APPLES or ORANGES?)
32	2004 Digital Gallery development site overhaul In winter 2003-04, RazorFish, Inc., consulted by DLP Needs: Improve look-and-feel of overall navigation and site functions within the scope of existing back-end development (i.e. without re-writing our software) Outcomes: New: ‘topics/collections’ paradigm New: ‘find related images’ widget Improved visual language for user calls-to-action
33	New paradigm for categories & collections “Explore Topics” Contain topical featured selections Selections are hand-curated vs. cataloged Topics represent several broad categories (“big buckets”) Can feature topical searches or subjects
34	Example of “Explore Topic” feature Topical “Big Buckets”
35	Functional requirements for Topics & Collections Topics (or “big buckets”) Do not hold everything Virtually constructed container Are not exclusive: data can cross topics Only curated selections: “featured resources” Selected representation of online holdings Collections Do hold everything: a place for everything, and everything in .... Represent real-world objects User can scope search by collection Are exclusive (?) Comprehensive for online holdings
36	New Tool: “Search for Related Images”
37	Laser or a net?
38	Limit search
39	Expand search
40	Expanded search results
41	Improved visual language When selecting images for user’s portfolio ...
42	Improved visual language Portfolio selection is now highlighted.
43	Some of the lessons learned in Versions 1.0 > 4.0 If problems exist in data, try to fix them in data It is difficult, if not impossible, to curate your way out of data anomalies Rationalize look-and-feel of site Create visual language for interface Simplicity is hard to do well “Keep it simple, stupid”
44	Going forward ....
45	... more IA challenges …
46	... will present more opportunities for increased access.