How to use Synonym Suggestions

A step-by-step guide to using Synonym Suggestions when building Themes

C
Written by Cameron Parry
Updated over a week ago

Jump ahead to...


What are Synonym Suggestions?

Synonym Suggestions is a feature within the query builder that automatically suggests similar concepts from your analysis data to improve your query strength and save you time.


How to use Synonym Suggestions

Synonym suggestions are expected to help you build more comprehensive and accurate themes in a repeatable and systematic way.

  1. You can start as you would building any query either based on an initial concept from the Storyboard or one that you had in mind for a specific area of investigation.

  2. Then, when you click the “OR” button, to add more concepts to the same query row, you will see a list of suggested single and / or multi word phrases with their similarity and frequency.

  3. You then might select a single and / or multi word phrase that you think is relevant to the theme; If you do, then when you click the “OR” button again the suggestions will be updated to take into account both of the query phrases on the line.

  4. Every time you add a phrase to the query row, the next round of suggestions will be refined to to take into account the new phrase.

Eventually, usually after adding a few phrases, you’ll reach a point of diminishing returns where suggestions have low similarity, low frequency, or don’t seem relevant in your assessment.

Tip: You can add multiple Synonyms to your query at once by holding down the CTRL Key (CMD on Mac) and selecting the Synonyms you'd like to add to your Query!


How we Identify Suggested Synonyms

After a phrases is entered into a query and the user hits the "OR" button to add additional phrases to the line, the most similar concepts from the model are displayed. Suggestions are identified by similarity then ranked by frequency in the dataset to emphasise the concepts most relevant to that particular dataset.

  • Similarity is a rating on a scale from 0 to 1, with 0 meaning completely dissimilar and 1 meaning completely similar (identical).

  • Frequency is simply the total number of occurrences in that concept in the entire dataset (not the query).

When multiple concepts are in a query row, a centroid of all the phrases present on the line is used to identify the synonym suggestions.

  • The centroid represents a theoretical middle point between phrases and allows us to refine suggestions based on the context suggested by a combination of phrases.

Tip: Synonym Suggestions can detect potential spelling errors and alternative phrases, too!


Questions? 🤔

If you have any questions about Synonym Suggestions (or need some help!) you can get in touch with us any time by hitting the blue chat button to your right 👉

Did this answer your question?