[OPEN-ILS-DEV] OPAC Search - Current and Future Directions

[Dan Wells]

Hello all,

I have been tracking Evergreen for several years now in the hopes of using it to replace our aging Sirsi system.  After giving it a serious amount of consideration, we have come to realize that the primary thing we dislike about our current system is the OPAC.  It basically stunk when we deployed it around eight years ago, and the smell has gotten much worse over time.

Since we need a new OPAC more or less immediately, we have been hoping to deploy Evergreen as an external catalog only, then move other library functions over piecemeal as time, logic, and system maturity allow.  This plan requires, however, that Evergreen be competitive when seen purely as a catalog, so I am hoping some people on this list can provide me with some support for this notion.  As an aid to organizing thoughts, here are what I see as my primary search considerations, in order of importance:

1. Relevancy Ranking - not only should relevancy ranking "work" out of the box, it should be customizable.  In regards to customization, as some like to say, easy things should be easy and hard things should be possible.

2. Facets/Attributes/Filters - easy and useful filtering of large result sets is becoming commonplace in modern search tools (e.g. Amazon).  This is less important, though, if the relevancy is top-notch (e.g. Google).

3. Speed/Size - all else being equal, an engine with faster searching, faster indexing, and smaller system requirements wins.  Where speed and size considerations are at odds (as is often the case), the best solutions are balanced by default yet easily configured.

4. Other Features - "Extras" like incremental indexing and built-in clustering, while in some situations critical, are for many search uses considered to be bonus or luxury items.

I have probably missed a few things, and perhaps you may disagree with my ordering, but if an OPAC is to function as a serious search interface, these are definite considerations.  Also note that I am not too concerned with non-search functionality like reviews, cover images, lists, etc.

So, from those who are most familiar with Evergreen's catalog, how does it stack up?  In my own research, it seems to me that tsearch2 may be falling behind other full-text indexes like Lucene/Solr and Sphinx.  Are there areas where tsearch2 clearly bests these other popular options?  Does the tight integration with PostgreSQL provide any exclusive yet also indispensable benefits?  Concerning facets, will Evergreen/tsearch2 provide for true facets or filters in the foreseeable future? (I would describe the current "related" links as pseudo-facets only.)  Finally, does Evergreen's catalog have any special features which are either not done in most other catalogs or not done well?

Thank you very much for any responses you can provide.  As my title suggests, feel free to frame your answers around where Evergreen's catalog is or where Evergreen's catalog is heading.

Sincerely,
Dan