Generating Customers Insights with Topic Modelling and HuggingFace SetFit Method
summary
Hands-on pipeline shows LDA topic modelling (gensim and sklearn) on a Kaggle British Airways reviews dataset, then fine-tunes a HuggingFace sentence-transformer with SetFit few-shot (8–16 examples/class), runs a Colab notebook, and pushes the fine-tuned model to the Hub. Reported demos hit ~83% accuracy (78% on private company data), and a keyword baseline comparison yields ~89% agreement. Code includes preprocessing, Bag‑of‑Words/TF‑IDF features, perplexity/coherence checks, and a simple production monitoring baseline.
AI-written summary · about
chapters (6)
- 0:00Customer insights with text analytics
- 17:27Topic modeling for reviews
- 30:29Topic modeling with LDA and Gensim
- 46:22SetFit for few-shot classification
- 1:00:38Fine-tuning SetFit and evaluation
- 1:18:03Q&A
related talks
48mTopic Modelling (and more) with NLP framework Gensim
01h 26mBagging to BERT- A tour of applied NLP
02h 10mIntro to Hugging Face: Fine-tuning BERT for NLP tasks
01h 01mWorkshop: Intro to topic modeling using Python
33mMachine Learning 2.0 with Hugging Face
35mLeveraging LLM for text augmentation in Topic Modeling
31mWhen humans teach machines
27mUnveiling Clustering in BERTopic Topic Modeling