Much of the work that products like Endeca focus on is related to the problem of how to most effectively get the user to tell the system what they’re looking for. It seems that an equally important aspect of a search interaction is for the system to be able to tell the user what it found.

Certainly the results standards of the title of a document/webpage may be instructive, as are highlighted text excerpts, etc. But this clearly doesn’t tell the whole story about an entity, especially within the context of a given search space. A number of aspects of data or metadata that largely exist already could be leveraged to provide a better answer to “what did you find?” such as:

• Embedded diagrams and images could be extracted and used as thumbnails
• Additional relevant metadata can be displayed along with the item
• Relationships between documents can be displayed

This last item is more subtle, but also, I think, potentially more impactful. By focusing on the relationships between searchable entities, one could construct display mechanisms much more effective than a list. When I’m not exactly sure what I’m looking for, or when I’m looking for several pieces of data all relevant to a central topic, this might become very valuable.

What if, when I search for “Harry Potter”, instead of getting this:

Figure 1

I get this:

Figure 2

In addition to the particular data or metadata, now I can potentially be presented with some understanding about how these things all fit together, which may be instrumental in telling me where to further refine my search. Placement, size, colors, shapes, connections, and virtually any aspect of this more flexible presentation space can all represent different, potentially useful aspects of what I’m searching for. This is clearly a simplistic example, but it illustrates the idea.

This sort of results richness may likely be more applicable to domain-specific searching rather than generalist searching – a pharmaceutical company has certain relationships between searchable entities within its enterprise than a financial services company doesn’t have, and vice versa. The richer the semantics around your metadata are, the more you could potentially leverage a system like this.

The tricky part seems like it would be finding the right level of richness. Too little, and what you present back to the user just doesn’t contain the information that would make it valuable. Too much, and you run the risk of completely overwhelming the user with data. Again, domain specificity may be where this sort of approach shines. When there are some underlying assumptions about relative importance, relationships, entity types, etc, the richer data can be presented in an easily consumable way.