Zero-Shot Classification with OpenAI and...
summary
Scripts call OpenAI embeddings to turn survey comments into 372‑dim vectors, store them as CSVs, run K‑means via Accord.NET (square‑root heuristic chooses nine clusters for 68 lines), compute Euclidean distance to centroids, then ask ChatGPT to label the top one and top five exemplars. Output is a CSV of topics; cost and rate limits are handled with retries. Results broadly surface themes but sometimes mislabel (emoji responses, blanks).
AI-written summary · about
chapters (10)
- 0:00PowerShell data science intro
- 3:30Analyzing text data challenges
- 7:43Text analysis in surveys
- 10:52Data privacy disclaimer
- 11:12Understanding embeddings
- 14:22K-means clustering
- 20:23End-to-end process summary
- 22:09PowerShell implementation review
- 35:04Live demo survey responses
- 44:49Key takeaways and future
related talks
48mUnderstanding Embedding Vectors and the RAG Pattern in AI Solutions
31mText and Code Embeddings
45mScalable Open AI GPT Applications in the Cloud
45mAI Chats: Challenges in Business Integration
31mTaking AI/ML to the next level: Text!
33mBulk Labelling Techniques
44mBuild RAG from Scratch
44mContent Classification: NLP vs. AI in the Age of Data Vectorization