A Visual Approach for Interactive Keyterm-Based Clustering
summary
A term-first clustering algorithm (LDC) is implemented and exposed via the Vis-KT dashboard with cluster, term-cloud, graph and parallel-coordinate views supporting soft clustering and manual splits by rearranging top keyterms. Three interactive modes — term-supervised LDC, document-supervised, and dual — are evaluated: LDC matches LDA and beats NMF overall, outperforms LDA on two datasets, and term supervision wins; a user study and hour‑long sessions with materials scientists and clinicians report strong acceptance.
AI-written summary · about
chapters (6)
- 0:00Introduction
- 5:25Algorithms
- 9:43Interactive Visualization
- 12:06Real Clustering
- 13:59Future Work
- 16:21Questions
related talks
25mThe Art of Clustering: The Good, The Bad and The Beautiful
38mData Mapping for Data Exploration
01h 20mClustering Data
37mBuilding a Scientific Taxonomy at Scale
08mComparing Apples and Oranges: Human and Computer Clustered Affinity Diagrams Under the Microscope
41mBetter Data Exploration and Visualisation with Elasticsearch and Kibana
11mRethinking Dataset Discovery with DataScout
34mParallel Efficient Clustering with Graph-Based Approximate Nearest Neighbor Search