START is a question-answering system built by Boris Katz at MIT’s Artificial Intelligence Laboratory, designed to answer questions posed in ordinary English rather than as keyword searches. According to MIT CSAIL’s own description, START “was first connected to the World Wide Web in December, 1993,” and has since “answered millions of questions from users around the world.” That early date makes it one of the first natural-language question-answering services to operate openly on the web, years before search engines or voice assistants made the idea familiar.
The system’s distinctive technique is what Katz called natural language annotation: pieces of information in its knowledge base are tagged with English-language descriptions, and when a user asks a question, START parses it and matches it against those annotations to retrieve a relevant answer, including text, images, video, or audio. This let the system respond to a real question with a real answer instead of a list of documents to read, which is the behavior people now expect from assistants and answer boxes.
START predates the large web-scale question-answering efforts that followed, and it influenced how researchers thought about turning a typed question directly into an answer, the same core problem that IBM’s DeepQA later tackled at championship scale for Jeopardy.
For a general reader, START is a reminder that the dream of just asking a computer a question in plain language and getting a direct answer is decades old, and that a working version was quietly serving the public on the early web.