SHSarah Hänscheninquasiris.hashnode.dev·18h ago · 6 min readInside AI Search #1: Natural Language UnderstandingIn the “Inside AI Search” series, Quasiris shows what lies behind the individual features of modern AI-powered search. The series begins with Natural Language Understanding, or NLU for short. This tec00
MMMahesh Moholkarinmaheshmoholkar.hashnode.dev·5d ago · 70 min readText to Attention From ScratchThis is the second part of a series on how language models work, built from the ground up. The first part covered what a neural network is made of and how it learns: an office building of workers, man00
MAMuhammad Ariel Shakaramiroinshaka-ai.hashnode.dev·5d ago · 7 min readGPT-2 From Scratch vs Fine-Tuning: A Text Generation Experiment and Its Honest FindingsThere are two main ways to build a text generation model: train one from nothing, or take a model that already knows a lot and teach it something new. This post runs both, on the same small compute bu00
MMmohammad mohammadiinfixam.hashnode.dev·5d ago · 19 min readBuilding a Multilingual Location Parser for Local Service Pages with PythonLocal service pages become surprisingly difficult to process once a website supports multiple languages, multiple cities, and several physical locations. A human can open a page, read the heading, rec00
MAMuhammad Ariel Shakaramiroinshaka-ai.hashnode.dev·6d ago · 7 min readTransformer & Self-Attention: I Built an Encoder From Scratch, Then Caught It OverfittingThis week's topic is one of the most important building blocks of modern NLP: the Transformer and the self-attention mechanism behind it. Instead of just walking through the theory, I decided to build00
MAMuhammad Ariel Shakaramiroinshaka-ai.hashnode.dev·Sep 10 · 10 min readBuilding a Spam Classifier by Fine-tuning BERTA message like "Congratulations! You've won $10,000, click here to claim" reads as spam to pretty much anyone who gets it. What's worth digging into: a computer only sees a string of characters, so wh00
TTech-Aarvamintechaarvam.hashnode.dev·Sep 8 · 12 min readAttention is simpler than you thinkPart of the TechAarvam workshop support files — Build Your Own Model. A hand-crafted superhero transformer This notebook presents a hand-crafted example. The goal is to understand the intuition behind03T
MAMuhammad Ariel Shakaramiroinshaka-ai.hashnode.dev·Sep 7 · 9 min readUnderstanding Recurrent Neural Networks (RNN): Architecture for Sequential DataWhy does a model that reads sentences, recognizes speech, or forecasts stock prices need a different architecture than a plain image classifier? The answer comes down to one word: sequence. That's whe00
MAMuhammad Ariel Shakaramiroinshaka-ai.hashnode.dev·Sep 7 · 7 min readWord Embedding & Sentiment Analysis with LSTMIn the previous RNN post, I mentioned word embeddings would get their own post. Here it is — paired with a real use case: predicting whether a movie review is positive or negative, from scratch all th00
MAMuhammad Ariel Shakaramiroinshaka-ai.hashnode.dev·Sep 6 · 6 min readText Vectorizer Explained: BoW, N-gram, TF-IDF, and Word EmbeddingComputers can't read words the way humans do — everything has to be converted into numbers before a machine learning model can process it. That conversion is what's called a text vectorizer. In this n00
