Topic modeling

What is Topic Modeling?

Topic modeling consists in discovering abstract topics or themes within a collection of documents. It is a way to automatically identify and extract the underlying topics that exist in a large corpus of text without requiring any prior knowledge of the topics themselves.

• Input: A collection of documents (usually represented as a matrix of word frequencies or a document-term matrix).

Source: https://www.freecodecamp.org/news/content/images/2023/02/1.png

• Output: A set of topics, each represented as a distribution of words, and for each document, a distribution of topics.

In the rest of this section, we will assume that the document-term matrix is obtained using TF-IDF, i.e. $w_{ij} = \mathtt{tfidf}(Tj,Di, C)$.

Using non-negative matrix factorization (NMF)

Using Singular Value Decomposition (SVD)

SVD or NMF for topic modeling?

Previous Section

Home

Next Section