Table of Contents

How Searching Works

Niamh Ferns Updated by Niamh Ferns

How Searching Works in DeskDirector

In all products DeskDirector has created, we use full-text search for all list interfaces. Compared to traditional approaches that check whether an item contains or starts with a search term, full-text searching provide better flexibility and an overall cleaner user experience.

For more info regarding this technology, you can check out this Wikipedia page.

How do indexing and tokenisation work?

For each collection, we will mark a few properties as query fields (for example, the name, description, or content). The information within will be tokenised and indexed by the database. When searching, we search based on tokens instead of a full sentence and will search against the tokenised query fields in the DeskDirector database.

Basic tokenisation splits information by spaces, special characters, and new lines. All tokens are case-insensitive.

Example
Input string of text: The quick brown fox jumped over the lazy dogs, Bob@hotmail.com 123432.
Output tokens:[quick] [brown] [fox] [jumped] [over] [lazy] [dog] [bob@hotmail.com] [123432]
Do not confuse this tokenisation with that used in AI LLMs! AI LLMs will tokenise your input differently and will often split words into smaller chunks for further processing.

How does the system determine the most matched object?

When DeskDirector analyses search terms, it will tokenise the search string. It will match on two different criteria, and ones with the highest ratings will be returned near the top of your results.

  1. Any token that's exactly matched will give a rating of 2
  2. Any token that starts with the search term will give a rating of 1.
  3. A repeat match will increase its total rating.

For example, we have two users with the following names.

  1. Jamison Knight / jknight@example.com from Euro Health Center
  2. Jamison Coon / jamison@secondcompany.com from Europtimum

Scenario:

Search term: Euro Jamison will give a rating of 4 for Jamison Knight and a rating of 3 for Jamison Coon. Since Jamison Coon only has a starts-with match with Europtimum, while Jamison Knight has an exact match with Euro.

What kind of terms can we use for different collections?

DeskDirector has put a lot of effort into analysing the important fields that should be included in inside the index. I will summarise them below.

  • Company / Account: name, identifier
  • Contact: first name, last name, primary email, secondary emails, title, email domain.
  • Member / Resource: first name, last name, primary email, secondary emails, identifier
  • Ticket: title, initial description, identifier

Frequently Asked Questions

Same as many other full-text search implementations. You can include as many terms as you like. You can even type in a full sentence.

"Why can't we add sorting to the search results?"

Sorting of search results will often lead to confusion and unexpected results. For example, if sorting results by text, then higher-rated results could end up on page 2 or 4.

This is also why services like Google don't allow sorting.

"When I search contacts, there are results from different companies. How do I improve accuracy?"

You can add an email as an addition to an existing search term, or you can also add a company filter. That way, only contact from that company will be returned.

"How come there are many false positives?"

Full-text searching is likely to retrieve many documents that are not relevant to the intended search question. Such documents are called false positives. They often appear at the end of search results, where they are partially matched with search terms.

How did we do?

Office Hours

Embedding Help Page Media