Triaging Feedback Form Data
summary
Uses Python and R to clean and model Intercom feedback, starting from roughly 500 cleaned messages after deduplication and short-message removal. Builds TF-IDF document-term matrices, runs LDA with LDA posterior probabilities for predicting topics on unseen data, and tries k-means (three clusters) plus dendrograms and time-series sentiment smoothing. Shows which topics (image/colorize, video/search, account/docs) surface, what tooling struggles occur, and where a forecast or logistic model might triage effort.
AI-written summary · about
chapters (18)
- 0:00Introduction
- 2:09Triage customer feedback data
- 4:38Intercom and data sources
- 6:44Cleaning Intercom feedback
- 9:55Preparing text for NLP
- 11:05Frequent terms and word clouds
- 13:11Cluster dendrograms for topics
- 15:43Extracting documents from clusters
- 17:50Bringing R to platform
- 18:52Clustering feedback with k-means
- 21:28Refining clusters with verbs
- 22:38Actionable insights from clusters
- 24:51Topic modeling with LDA
- 27:22Evaluating LDA topics
- 28:50Visualizing LDA topics
- 29:31Sentiment analysis with time series
- 30:37Forecasting sentiment trends
- 31:01Conclusion
related talks
01h 01mWorkshop: Intro to topic modeling using Python
30mProductionizing an unsupervised machine learning model to...
01h 19mGenerating Customers Insights with Topic Modelling and HuggingFace SetFit Method
27mBuild Your Own ML Data Feedback Loop
31mWhen humans teach machines
46mAI-Driven Techniques for Noise Filtration in Software Validation Logs
01h 05mSocial Text, Sentiment, and Tone Analysis
37mData Driven Diagnosis of Alzheimer's Disease