Word Embeddings (வார்த்தைகளின் வரைபடம் / Meaning Map)

Let's go shopping in Koyambedu! (கோயம்பேடு மார்க்கெட் போவோமா!)

Technical Meaning: சொல் பொதிவுகள் (Sol Pothivugal) - Representing words as numbers so computers can understand their meaning.

The Core Idea

Word Embeddings are how an AI understands language. Because computers can only calculate numbers, they assign every word a set of numerical coordinates in a giant, multi-dimensional space. Words with similar meanings are given coordinates that place them physically close together.

The Origin Story

Early NLP (Natural Language Processing) used something called "One-Hot Encoding", treating every word in the dictionary as completely unrelated. "Cat" and "Dog" were considered as mathematically different as "Cat" and "Helicopter." Word2Vec changed everything by realizing that words appearing in similar contexts (e.g., "I walked my ___") usually have related meanings, allowing us to map meaning into physical distance.

The Tamil Analogy

Imagine the sprawling vegetable market at Koyambedu (கோயம்பேடு மார்க்கெட்).

If you are a total stranger looking for a carrot, you don't wander randomly. You know that vendors are grouped by meaning.

• All the root vegetables (கேரட், பீட்ரூட், உருளைக்கிழங்கு) are clustered in one corner.
• The leafy greens (கீரை வகைகள்) are in a different aisle.
• The fruits (பழங்கள்) are in a completely different block.

If you are standing next to the carrot vendor, the physical distance to the beetroot vendor is very short. The distance to the banana vendor is very far. Word Embeddings do exactly this: they map the English language into a giant digital Koyambedu market, where related concepts are mathematically clustered next to each other!

Try It Yourself

Click the concepts below to drop them into the AI's semantic space. Notice how related things cluster together!