Search Interfaces for Specimen Data

While classic search interfaces often work beautifully for an expert user, our goal in defining the ideal interface for searching and browsing specimen metadata is to support users without expertise in the system, or even the collection. Although they may be experts in the subject area, interviews indicate that even museum scientists may use the interface only a few times a year, hardly qualifying as experts on the system. They may also need to use a variety of systems in order to accomplish their research, and so an easy-to-pick-up interface is most helpful. Finally, as museums seek to expand the value of their collections beyond traditional audiences, the tools with which they share their data must support a broader, non-expert audience as well.

How does an interface acheive ease-of-use by the non-expert? By supporting browsing & query refinement and applying principles from Usability a.k.a. Human Computer Interaction research.

In classic search interfaces, the user enters some representation of their information need, and gets zero results or tons of results. They then go back & try another representation, attempting to be more specific or more general and retrieve a manageable number of records.

A browsing system assists the user to express their information need in terms the system can understand, and supports expanding or reducing a result set by providing one-click access to logical refinements, and previews of the size of the found set under the new search conditions.

Principles from MaNIS Interface Design project

In our MaNIS search interface prototype, the browse features owe a direct intellectual debt to Flamenco, an image-browsing application by Professor Marti Hearst at SIMS. The particular application of those features to searching/browsing specimen metadata is highly shaped by the tasks real people do when using the these systems, as determined through interviews with emeritus MVZ scientists, graduate students, undergrads, teachers, curators, external scientists, and wildlife biologists. Finally, good usability means giving users feedback, preventing errors, and refining designs iteratively.

Notable features of the MaNIS search interface prototype:

Layout, vocabulary and defaults support common uses

  • Use of panels for common specifications avoids the need for specifying the various attributes that are necessary for a taxonomic search, a geographic search, when these are commonly performed.
  • Taxon panel as default panel follows our interview results - we had a hard time finding anyone who started their search with a place.
  • Avoid system-specific, implementation-oriented vocabulary - use words from common English & the domain, not database words.

Feedback! Found set is created and displayed on same page

The found set is displayed on the same page where the search conditions are specified, allowing for immediate feedback:

  • as you select a particular link, feature, or view
  • about what a typed-in word does for your search results. Bold text indicates how the system matched your input.

Help matching typed-in input with actual terms

"Did you mean..." feature suggests similar terms from among those actually found within the system.

One-click refinements of search conditions

One-click will expand or narrow the found set:

  1. intuitive, manipulatable representation of the search conditions on the same page.
  2. by showing current location on the taxonomic tree, and how you can go up & down
  3. provide dynamic preview counts of the found set if any particular term is added to the search conditions
  4. toggle between searching one specific museum or all museums

Finding AND colocating records

"Finding" helps you locate the thing you know the name of; "colocating" helps you to know what other names to look under, too, like a card catalogue saying "See also Samuel Longhorn Clemens" under Mark Twain. By checking taxonomic name servers for synonyms, and displaying those which have matches in the database, the system reduces the domain-knowledge required to retrieve a complete and meaningful dataset.

User-friendly niceties

  • One text box for searching multiple, non-independent fields in the database, such as those related to geography, or those related to taxonomy, thus preventing users from accidentally entering impossible, conflicting terms into different fields (like "Country:Brazil", "State:California").
  • Capitalization or case of the search terms doesn't matter
  • All records are returned; possibly provided as multiple pages if too many records would risk crashing a browser. Result sets are never truncated prematurely and without warning.

 

Heuristic evaluation of existing interfaces

  1. Suggestions for MaNIS Classic Data Portal Interface
  2. Review of BNHM's search interface
  3. Review of MVZ's data access application

March, 2005: This nice list of heuristics for general search interfaces is a great resource, too.