Introducing Microsoft’s Azure Search Service

Microsoft has just announced the preview of its new search service.  You can provision the service using the standard Azure Portal.

image

The search service is a text indexing programmable engine for searching across any type of document or data that you can feed it.  One of the cool features is that the API looks to be quite open ended in that you can create indexes based on any kind of text data structure and with user defined indexes for metadata.  Populating the indexes is done using a REST API and results are returned in JSON format.

A couple key feature notes for the current version of the service:

  • Indexing is only on text input only, e.g. you cannot upload word documents and have the search engine parse them.  You would have to parse your own content and then send in the result text to the engine.
  • English is the only supported language at the moment, although additional languages will be available in future releases.
  • There is no built in concept of personalization or filtering by identity.  The query syntax supports a filter query but you would need to program the connection of identity to the filtering query.

The search service includes a concept of scoring profile that allows you adjust how search results are ranked.  In the current version of the service, the scoring profile allows you to boost results based on:

  • Freshness: the age of the document
  • Magnitude: scoring values stored in your metadata (e.g. star rating)
  • Distance: geographic distance (based on lon,lat data you store in the index) from a provided reference location (e.g. you could supply the current user’s position)

Stay tuned for some future experiments in using this service!