Experimental Creative Writing with the Vectorized Word
summary
Uses pre-trained semantic vectors (GloVe, word2vec, fastText) and the CMU Pronouncing Dictionary to create continuous embeddings for meaning and sound, applies signal-processing ops (blends, blurs, resampling) to text via a prototype called Viktor Tightrope, and generates concrete artifacts: Frankenstein↔Genesis crossfades, phonetic tinting with “kiki/booba,” examples where frozen+water → ice and basketball in Dracula → tennis, published poem in vetch and a forthcoming book of phonetic-random-walk poems.
AI-written summary · about
chapters (14)
- 0:00Introduction
- 3:46Discrete units of language
- 5:09Color
- 9:20Examples
- 11:22Distributional Hypothesis
- 14:03Global Vectors
- 16:10Word Vectors
- 17:10Word Vectors as Waves
- 19:21Frankenstein Genesis
- 24:34Poetry and Sound
- 28:08Two Words Together
- 30:07Robert Frost
- 31:31Key Key
- 33:35Articulates
related talks
41mKeynote: computational stichography! (or: connecting lines)
51mPlayful Text
48mA fun and absurd introduction to Vector Databases
11mMutating text toward better nonsense?!
50mVector Ops: run vector embedding-powered apps in prod
15mSpeling werds egspressively with rrkurrent nuril nedwirques!
50mA fun and absurd introduction to Vector Databases
46mA fun and absurd introduction to Vector Databases