Words as Numbers
Inside a word vector
A word vector is typically 50 to 300 numbers long. Each number represents some aspect of meaning, though these aspects aren't explicitly labeled. The patterns emerge from training on massive amounts of text.
Try selecting different words below to see their vector representations:
Explore Word Vectors
Green bars = positive values, Red bars = negative values. Each bar is one dimension.
Comparing similar words
The real magic happens when you compare vectors. Words with similar meanings have similar patterns. Try comparing two words:
Compare Two Words
Green = Word 1, Blue = Word 2. Notice how similar words show similar patterns.
What the dimensions mean
No one explicitly programmed what each dimension represents. The meanings emerge from patterns in text. However, researchers have found that some dimensions loosely correspond to concepts like:
- Gender (masculine ↔ feminine)
- Royalty (royal ↔ common)
- Size (big ↔ small)
- Sentiment (positive ↔ negative)
The key insight: Meaning emerges from patterns, not explicit programming. This is how AI learns to "understand" concepts without being told what they mean.
Key Takeaways
- Word vectors are lists of 50-300 numbers
- Each dimension captures some aspect of meaning
- Similar words have similar vector patterns