Topic Modeling

Also known as: LDA, BERTopic, Automated Thematic Analysis, Text Clustering

Unsupervised ML technique that automatically discovers themes or topic clusters in large text collections.

Topic Modeling is an unsupervised machine learning technique that analyzes large text collections to automatically discover the themes or topic clusters that emerge from the data, without the researcher having to define them a priori.

Common algorithms include LDA (Latent Dirichlet Allocation), NMF (Non-negative Matrix Factorization), and more recently transformer-based models like BERTopic or Top2Vec.

In research, topic modeling is extremely valuable for: analyzing thousands of open-ended survey responses, identifying emerging themes in social media conversations, discovering satisfaction or dissatisfaction drivers in customer feedback, and finding unexpected patterns in massive qualitative data.

Atlantia uses topic modeling as a complement to AI coding to ensure that emergent insights in verbatims are not missed.