Topic Modeling

Unsupervised machine learning: it automatically identifies latent themes and semantic structure within a massive corpus of text.

Topic Modeling is a text mining method: it applies unsupervised learning (e.g., Latent Dirichlet Allocation (LDA) or Non-Negative Matrix Factorization (NMF)) to large document collections. The process treats each document as a 'bag of words,' analyzing word co-occurrence and frequency to deduce hidden themes. The output is a set of distinct topics, each defined by a probability distribution over the vocabulary. This technique is critical for scaling text analysis: it organizes and summarizes millions of customer reviews, legal briefs, or academic papers, making unstructured data immediately actionable for classification and information retrieval tasks.

https://www.ibm.com/topics/topic-modeling