Executive Overview

The quality of the search for DX Assets is terrible. The search is simple text search using an SQL Query. As a result of not using a more sophisticated text search such as Trigrams or Morphological Stemming, the search lacks the ability to:

• Find obvious hits. For example, if I'm looking for the "DataRobot" asset, the search "data robot" returns no results because simple text scan's can't deal with a simple blank.

• There is no ability to perform predictive search. That is, when you start typing the result "Data r" the search query will provide a list of potential queries with results such as "DataRobot" NOTE: This is a feature of both Lucene from 8+ years ago and Solr.

• Soundex was proposed as a potential solution, but the problem with Soundex is that it is optimized for American Surnames and thus would be of little use.

We need to have a high-quality, full text search that produces the results to help our customers find the Asset that they are looking for particularly as the number of Assets grow. The search should be sensitive to blanks in the way that tri-gram approaches are. The search should also predict by showing the user a set of possible search results such as the illustration below:

There are others who