Half II: Subject Extraction with BERTopic
and obtained from Pexels.com
Within the first a part of this collection, I launched you to my artificially created good friend John, who was good sufficient to offer us along with his chats with 5 of the closest individuals in his life. We used simply the metadata, resembling who despatched messages at what time, to visualise when John met his girlfriend, when he had fights with one among his finest buddies and which members of the family he ought to write to extra usually. If you happen to didn’t learn the primary a part of the collection, you will discover it right here.
What we didn’t cowl but however we’ll dive deeper into now could be an evaluation of precise messages. Subsequently, we’ll use the chat between John and Maria to determine the matters they talk about. And naturally, we won’t undergo the messages one after the other and classify them — no, we’ll use the Python library BERTopic to extract the matters that the chats revolve round.
What’s BERTopic?
BERTopic is a subject modeling approach launched by Maarten Grootendorst that makes use of transformer-based embeddings, particularly BERT embeddings, to generate coherent and interpretable matters from massive collections of paperwork. It was designed to beat the constraints of conventional subject modeling approaches like LDA (Latent Dirichlet Allocation), which frequently battle to deal with quick…
