In this chapter, we will have the chance to play with facets, text similarity and recommendations. We will again proceed back and forth from toy prototypes to the analysis of specific examples in order to give some direction and ideas for extended studies and improvements. I will also suggest some reading on math and information design theory for those who want to go further.
In order to explore the relation between these different approaches, we will see two different suggesters. We will start by using faceted search, which we have seen in action, without too much explanation, in the earlier examples. This is one of the most simple and yet complex features of Solr.
We will gradually shift our focus to findability and its implications for common users, introducing concepts such as similarity and recommendations.
Auto-completing terms will be a great improvement in user experience. We have already used an auto-suggester approach in Chapter 5, Extending Search. But an autocompletion approach could also be useful as a strategy to suggest navigation paths , because this could be adopted not only in the context of a traditional search but also for serendipity-based approaches. It gives the user the chance to discover some very different stuff while searching. These two directions can be implemented with several strategies, focusing on terms or some kind of data aggregation, from automatic clustering to faceting, that we will see in a while.
We can think about facets in analogy to categories, which can be used to collect many different documents sharing the same value. As we will see, however, facets do not need to be predefined of precomputed as categories generally are, but they can be requested on the fly, because they will be computed on term values that exist in the index. We can then think of facets as a way of narrowing down searches, while at the same time collecting suggestions useful to a broader exploration of the data.
Tip
A facet can also be seen as a sort of data clusterization on a specific term.
We will introduce Carrot2 in Chapter 7, Working with Multiple Entities, Multicores, and Distributed Search, which is a tool for document clustering and can be used with Solr. The clustering tools use documents' similarities or ranking metrics to compute data aggregation. This can involve complex processes, because every similar value can be computed using complex mathematical models, while the process of facet computation is simple. It's best to see a little introductory example to start.