Back to top

Relevancy Ranked Query Completion Services

HEADLINES

  • Most business-critical search applications use automated query completion. Where offered, query completion is heavily used by visitors and highly influential on search results. 
  • Most off-the-shelf query completion functions are unsophisticated, and opportunities to improve user satisfaction and business performance are missed.


Search Technologies provides services and where appropriate, a search engine independent technology components to implement query completion best practices. Our technologies and services can be used with any of the leading search engines, including open source options such as Solr and Elasticsearch.

OVERVIEW
Query completion, or "type ahead” functionality, takes a few initial characters entered by the user and returns matching terms or phrases to auto-complete the search clue. 

Search Technologies approach differs from more simplistic methods in four important ways:

  • Suggestions can be sorted by popularity or any other metric identified by the customer, so that the most useful or advantageous suggestions are listed first
  • Suggestions can match with prefix-name fragments found on any token within the query phrase and not just on the first token
  • Name variations can be included as suggestions with no loss of accuracy
  • The approach taken is scalable and very fast, with response times typically less than 20 milliseconds, even for large information collections


TECHNOLOGY DESCRIPTION

  • Suggestions based on character strings representing fragments of tokens
  • Fragments can be as small as a single character (configurable)
  • Suggestions are returned in "real-time," with no noticeable delay to search execution
  • Suggestions are returned in JSON format so they can be used directly by the user interface for populating pull-down boxes
  • Suggestions can include:
    • The matching term from the suggestion database
    • A parent or normalized term (if appropriate)
    • Markup to indicate the match against the user's entry
  • Characters entered by the user can match either at the beginning of the first word, or at the beginning of any word in a phrase. For example, “man” can match "manager," "program manager" and "blue widget manufacturing"
  • Matches at the beginning of a phrase can be preferred over matches in other terms
  • Terms can be weighted by “popularity.” Suggestions can either be influenced or hard-sorted by this factor
  • If the characters entered by the user do not appear at the beginning of any word in any phrase in the input data, then no results will be returned
  • A parent term or a normalized term can optionally be returned
  • Stop words can be removed from the beginning of the phrase if desired. For example, a search on “th” would prefer “A Theory of Relativity” over “The Grapes of Wrath”
  • The service can optionally include data preparation, normalization and de-duplication. This includes accessing the data from raw sources (spreadsheets, databases, files, etc.)


"FUZZY" NAME SEARCHING
A full fuzzy name search capability is also available. For example, “Tchaikovsky” would be returned even if the user types “Chaikofsky.” Fuzzy results are sorted by closeness of match, optionally including relevancy boosts on factors such as "popularity."

SOFTWARE MAINTENANCE AND SUPPORT
All products or software tools used in the delivery are royalty-free or included within quoted prices.  A full maintenance and support service for the whole application is available under a commercial-grade Service Level Agreement.

0