Embedding is the most rational and efficient general-purpose way we found to transform natural language into numbers that machines can understand and work with. Why? 

1. Embeddings preserve meaning and semantic relationships

Natural language is full of ambiguity, nuance, and context. Embeddings convert words, phrases and whole documents into dense numerical vectors that preserve meaning and semantic relationships.

2. Context

Modern embeddings go beyond static meanings of words. For example, the word “bank” in:

• "I sat by the river bank"
• "I deposited money in the bank"

…will be mapped to different vectors depending on the context. This is known as contextual emb…