A naukri.com initiative
Medium
3w
347
Image Credit: Medium
Angles of Ambition: Mapping New Year’s Resolutions in Semantic Space
- Using over 5,000 New Year’s resolution tweets from 2015, the article explores semantic structures behind human expression by finding the underlying stories in the data.
- OpenAI embedding model creates a 1,536-dimensional vector for each tweet that can be used for exploring vast, multidimensional space.
- PCA is a dimensionality reduction technique that provides the most generic view of the data, whereas T-SNE and UMAP take a different approach by emphasizing local relationships.
- Supervised projections like Linear Discriminant Analysis explicitly align the projections with specific categories and reveal stunningly clear patterns. Cosmic graph visualization is used for the same.
- Validation is necessary to ensure that projections and patterns uncovered aren't arbitrary. Semantic embeddings aren't arbitrary and are usually structured by design.
- Visualization is about inviting curiosity, sparking insights, and uncovering truths hidden in the multidimensional echo of the data.
- Projecting 1,536 dimensions onto two is not just a technical challenge but a storytelling exercise, using machine learning to illuminate the human experience.
- Therefore, projecting the high-dimensional spaces can help people explore data from all possible angles.
Read Full Article
20 Likes
For uninterrupted reading, download the app