user@devops:~$ cat README.md
Embeddings & Semantic Search
# Description
Educational project on Embeddings and Semantic Search with Sentence Transformers. Uses the all-MiniLM-L6-v2 model (384 dimensions) to convert 30 Spanish documents (6 topics: ML, Climate Change, Space, Health, Technology, Art) into semantic vectors. Demonstrates four fundamental concepts: (1) embedding generation and cosine similarity analysis between documents, (2) semantic search by meaning vs keyword search, (3) embedding visualization with t-SNE showing topic-based clusters, and (4) basic RAG pipeline that retrieves relevant documents and generates contextualized answers. Includes 4 visualizations: t-SNE 2D projection, cosine similarity matrix, intra-topic vs inter-topic comparison, and semantic search results. Average intra-topic similarity: 0.324 vs inter-topic: 0.182 (Δ=0.142).
# Key features
$ Semantic embeddings with Sentence-BERT all-MiniLM-L6-v2 (384 dimensions)
$ Corpus of 30 Spanish documents across 6 thematic categories
$ Semantic search by cosine similarity: find documents by meaning
$ Keyword Search vs Semantic Search comparison with result analysis
$ 2D t-SNE visualization: clearly separated topic clusters
$ Complete 30×30 cosine similarity heatmap matrix
$ Basic RAG pipeline: context retrieval + answer generation
$ 4 professional visualizations saved in high resolution
# Gallery
# Technologies used
