Smart and smarter: Overcoming limitations of rule-based chatbots

Chatbots have come a long way since the first one was introduced in 1966. However, to become a truly helpful user support channel, they still have a long road ahead of them. Knowledge maps, user context and structured content can help make a difference.

Text by Alex Masycheff

Inhaltsübersicht

Smart and smarter: Overcoming limitations of rule-based chatbots

Image: © Sebastien Decoret/123rf.com

The first chatbot

The first chatbot was built by Joseph Weizenbaum, an American computer scientist from the Massachusetts Institute of Technology, in 1966. The bot was called Eliza, in reference to Eliza Doolittle of Bernard Show’s Pygmalion. To make Eliza understand the user’s questions and give answers, the bot used a script that consisted of pattern-answer pairs. If a user’s question matched one of the pre-defined patterns, Eliza would respond with the matching answer.

One of the most famous scripts was called Doctor. It emulated a conversation with a psychotherapist. Ironically, while Weizenbaum wanted to show the superficial nature of interaction between computers and humans, quite a few people believed they were talking to a human when chatting with Eliza. Weizenbaum’s own secretary asked him to go out of the room when she was talking to Eliza, saying that she was having a ...