A comparability of two cutting-edge dynamic subject fashions fixing shopper complaints classification train
Buyer critiques about services and products present invaluable details about buyer satisfaction. They supply perception into what needs to be improved throughout the entire product improvement. Dynamic subject fashions in enterprise intelligence can establish key product qualities and different satisfaction components, cluster them into classes, and consider how enterprise selections materialized in buyer satisfaction over time. That is extremely invaluable info not just for product managers.
This text will examine two of the most recent subject fashions to categorise buyer complaints knowledge. BERTopic by Maarten Grootendorst (2022) and the current FASTopic by Xiaobao Wu et al. (2024) introduced finally 12 months’s NeurIPS, are the present main fashions for subject analytics of buyer knowledge. For these fashions, we’ll discover in Python code:
- methods to successfully preprocess knowledge
- methods to prepare a Bigram subject mannequin for buyer criticism evaluation
- methods to mannequin subject exercise over time.