Improved topic modeling and analysis in Relative Insight

What we like to call topical analysis, you might know as thematic analysis, topic modeling or qualitative data coding. Regardless of your preferred terminology, topical text analytics is a powerful tool for gaining insight from your qualitative data. It provides an efficient alternative to having teams of analysts manually code themes in surveys, customer reviews and other text-based data sources.

Over the next month, we’ll be rolling out a series of improvements to the topic modeling capabilities in Relative Insight Explore.

Here’s a rundown of what’s coming…

Intuitive labels for easier topical analysis

Relative Insight was born out of academic research – and we’ll admit that our topic names sometimes include overly complex wording that reflects our academic roots.

Topic tags such as “speech acts”, “generally kinds, groups, examples” and “discourse bin” often lead to more questions than answers and examples of words that fall under these topics don’t always immediately spring to mind.

We heard from you that these confusing labels were, at times, a barrier to getting the insights you needed from the platform. Recognizing the challenge, our all-star team of data scientists set out to give topics a makeover in hopes of making topical insights more accessible and useful.

Over the next month, you’ll see over 250 topics getting fresh new names in the platform. Some of the most requested changes include:

Relationship: Intimacy and sex becoming Love
Discourse bin becoming Slang
Time: Present; simultaneous becoming Current
Warfare, defence and the army; weapons becoming Military
Vehicles and transport on land becoming Transportation

These updates will be applied to new and existing projects. Where you’ve previously renamed a topic yourself, you’ll continue to see your custom label.

New topics to reflect our ever-changing vocabulary

The topic modeling makeover doesn’t stop there…

The words we use, and the way we use them change over time. A person living in the 1950s would be terribly confused by talk of computers, cyber security, mobile phones and dating apps. A member of gen Z would be similarly perplexed by talk of 8-track tapes, VCRs or telegrams. These changes in language can give birth to brand new topics and render others obsolete.

In other cases, as time goes on topics become too broad to be useful. In the early days of computing, for example, it may have made sense to track the topic of computers and electronics as a single entity. Today, it’s likely more useful to break this topic into parts such as software, hardware, consumer electronics, mobile apps and cybersecurity.

To ensure we’re keeping up with the times, in April we’ll be adding over 125 new topics that better reflect the nature of today’s economy and the realities of modern communication.

Among the new additions are:

Mental health
Reproductive health
Athleisure
Cybersecurity
Streaming

Once released, new topics will be included in the analysis for new projects.

These updates to our topic modeling will enable insight teams with a more nuanced view into their qualitative data, helping them efficiently gain a view into what audiences and competitors are saying – all without having to read a word!