Natural language processing gives web applications a way to process and understand human language in text or speech. A React and Node.js application can access pretrained models through OpenAI’s REST APIs instead of building a model from scratch. section 2.2.2
Transformer-based processing uses self-attention to evaluate relationships among words and preserve context across a sequence. This lets a model resolve references such as “it” meaning “the antelope” and generate contextually relevant continuations rather than merely grammatical ones. section 1.5
A conversational web application
Astra AI combines a React interface, Node.js backend, Tailwind CSS, Vite, OpenAI REST APIs, ephemeral conversation storage, input validation, sanitization, informative error handling, and Vitest unit tests. Its interface supports follow-up questions and scrollable conversation history. section 2.2.2
Figure 1.5 AI can be used to detect whether a picture of a cat is, in fact, a cat. It accepts an image as input and responds with yes or no (or 0 and 1).
Contrasts language processing with another AI capability by showing a model that accepts an image and returns a binary classification. section 1.5
