// Commented out mbk Feb 2019 // //
Search this site:
Enterprise Search Blog
« The Technology of Search

Do You Need Clustering?

The answer to that question is "that depends"...

First off, what type of clustering has your vendor suggested? Many forms of results list navigators can loosely be called clustering:

  • Fully automatic clustering (based on statistical methods)
  • Unsupervised clustering (similar to above)
  • Noun phrase clustering
  • Automatic taxonomy generation / auto-categorization
  • Meta Data clustering (upgraded form of SQL group by)
  • Entity Extraction based clustering (related to above)

There are other forms of results list navigators not labeled as clustering. There's a handful of business drivers that help steer these decisions:

  • Is search a primary business requirement, or is it a supporting technology?
  • At what level is your data currently organized? Field level, corpus, vocabulary or completely organic?
  • What level of staffing will you allocate to this task? Beware of total automation if search is a primary business driver.
  • How sophisticated are your users?

Automation is appropriate in some applications, but in others it's better to focus on the source data and use more predictable methods. Another option is to upgrade the source data, to align it with a more appropriate navigation technology.