Topic Modeling vs Text Classification
People think topic modeling is just sorting text by topic. It is not. Text classification uses groups chosen ahead of time and learns them from labeled examples. Topic modeling has no groups and no labels. It finds themes from words that show up together, and a person names them after.
Topic Modeling
Finding the main themes in many texts. Words that show up together form a theme.
Topic modeling finds the main themes in a large set of texts without being told what to look for. It looks at which words tend to show up together. Words like goal, match and team form one topic, and vote, party and law form another. Each text then gets a mix of topics. No one labels anything first, so it is a kind of unsupervised learning. People still have to read the word lists and name each topic.
Text Classification
Sorting text into groups that were picked first, like spam or not spam.
Text classification puts a piece of text into one or more set groups. The groups are chosen first, such as spam or not spam, or billing, tech and sales. A model learns from examples that people have labeled. Then it labels new text by itself. Sentiment analysis and spam filters are both kinds of text classification. It is one of the most common jobs in NLP.
In one line
- Topic Modeling
- Finding the main themes in many texts. Words that show up together form a theme.
- Text Classification
- Sorting text into groups that were picked first, like spam or not spam.