Concept Search
Contents
Concept Search#
Concept search is a way of searching without formulating an exact keyword query. Squirro offers two ways of running such concept searches: contextual concept search and Smart Filters.
Contextual concept search is exposed to end users in the form of a search icon when making a selection. When clicking on that search, a concept search is executed returning documents that match the key aspects of the selected text.
Introduction#
Concept search is a technology where a search engine can return documents matching a defined concept (see Wikipedia’s concept search article for further background).
As an example, compare normal full-text search with concept search. In standard full-text search, some sort of boolean search is used to return relevant documents. You can enter a query such as technology compan
to find any document that matches both technology
and company
. In concept search, this would be defined as a concept, where the system is taught to understand what “technology company” means.
It will then take into account additional terms around this area and search for those as well - for example the “technology company” concept may potentially include terms such as “IT” or “Silicon Valley”.
As a result, the search engine can return more and better results for this concept.
Smart Filters#
Smart Filters are a way of creating and modifying concepts for a power user or project creator. They can be used to quickly create document-level text classifiers. Once created, they are also useful as initial input into AI Studio Candidate Sets.
Reference: See Smart Filters for more information.
Contextual Concept Search#
When an end user is consuming content through the Squirro user interface, they may run concept searches by selecting parts of the content.
Whenever some content is selected, this shows a search icon around the selected text.
When clicking that button, the selection is used as search input to create a concept search. Query results are returned that conceptually match the selected text, without all of the words having to appear in the returned documents. To clearly show the difference, this is shown to the user as a separate text bubble in the search bar.
Tip: When hovering over the bubble, a tool-tip shows an extract of the text used for the concept search.
Limitations#
The contextual concept search functionality depends on a Global Search dashboard existing and will display the results using that dashboard.