STEM与日常科技·英语30篇（3）

11 / 30

向量数据库与近似最近邻

Modern search engines convert text, images, or sounds into numerical vectors—lists of hundreds of numbers representing meaning or features.
Finding similar items means computing distances between vectors, but checking every pair in billion-item databases would take hours.
Approximate nearest neighbor (ANN) algorithms skip exhaustive checks, instead using clever indexing to find close matches in milliseconds.
Think of it like finding a restaurant in a city: ANN uses neighborhood maps and landmarks, not street-by-street walking.
Recommendation engines on streaming platforms rely on ANN to suggest films with ‘similar vibes’—not just same genre or actors.
These databases power real-time chatbots that retrieve relevant past answers without scanning every saved conversation.
Accuracy trades off slightly for speed—but for human-facing apps, ‘close enough’ is often faster and more useful than perfect.
ANN turns unstructured data—photos, voice notes, research papers—into searchable, relational knowledge with minimal latency.