TL;DRword2vec was the breakthrough that learned word meaning from context.
word2vec (2013) trained shallow networks to predict a word from its neighbors (or vice versa), and discovered that the resulting vectors captured semantic relationships — famously king − man + woman ≈ queen. It showed meaning could be learned from co-occurrence statistics alone, kicking off the embedding era and foreshadowing how modern LLMs internalize meaning from context at massive scale.
Key points
Predict a word from its context (or vice versa)
Vectors captured analogies (king−man+woman≈queen)
Meaning learned from co-occurrence
Launched the embedding era
Common mistakes
Thinking word2vec understands like an LLM
Expecting context-sensitivity (one vector per word)
Overstating analogy reliability
Try it: Explain what the king−man+woman≈queen result demonstrated.
Example code
<!doctype html><html><head><meta charset="utf-8"></head>
<body style="background:#06040d;color:#e6e0ff;font-family:monospace;padding:20px"><pre>train: predict word ↔ context
result: vec(king) − vec(man) + vec(woman) ≈ vec(queen)
→ meaning from co-occurrence</pre></body></html>