RAG 4 :The Evolution of Language Representation: Word Embeddings to Transformers
Details
### 🧠 What’s This Session About
Ever wondered how machines understand human language?
Atmaja Jape will take us on a fascinating journey — from the earliest symbolic representations to today’s powerful contextual embeddings.
We’ll start with one-hot encoding, bag-of-words, and TF-IDF, move through Word2Vec and GloVe, and finally reach Transformer-based models like BERT and Universal Sentence Encoder.
Through live coding demos, we’ll train a small embedding model to visualize how machines progressively capture semantic meaning — from symbols → context → understanding.
***
### ⚙️ Why It Matters
Understanding how language representations evolved helps you:
- Build stronger mental models for RAG (Retrieval-Augmented Generation) systems
- Design better LLM pipelines and embeddings
- Connect theory → code → applied GenAI workflows
***
### 💡 Bonus Peek
We’ll also connect this to practical automation — enriching CSV/LinkedIn data, finding HR contacts via APIs (like Hunter), personalizing outreach with LLMs, sending Gmail campaigns, and tracking responses — all as part of a “LLM + Workflow Automation” mini demo from the RAG toolkit on your own laptop.
About the topic/scope
The journey of how machines represent and understand language will be traced, starting from simple symbolic approaches and moving toward powerful contextual models. Early methods such as one-hot encoding, bag-of-words, and TF-IDF will be introduced, followed by the shift to dense word embeddings like Word2Vec and GloVe. The session will then highlight the breakthroughs brought by Transformer-based models such as BERT and the Universal Sentence Encoder, which enable deeper semantic understanding of text.
Concepts will be illustrated through coding demonstrations, and a small embedding model will be trained live to provide a hands-on view of how representations evolve. By the end, the progression from symbols to semantics will be clearly observed, showcasing how each stage of development has brought machines closer to capturing meaning in human language.
Date & Time: Saturday, 11th October | 2:00 PM – 5:30 PM
Venue: Harbinger Group, 102/A, ‘Global Port’, Mumbai-Bangalore Highway, Baner, Pune – 411045
RSVP: https://www.meetup.com/pune-women-in-machine-learning-and-data-science/events/311387609/