Despite appearances, chatbots around the world have been a well-known and used marketing tool for many years. As early as 1959, the first weather chat application was created by L. Green, who, a year later, also co-authored a ‘baseball’ programme that chatted about baseball games. Shortly afterwards, a machine (Agile) was completed, which was more advanced in that it arranged the answers itself. Another small step towards an intelligent machine was Robert Lindsay’s Sad Sam programme (1961), which could answer questions on the degree of family collations.
In 1964, D. Bobrow wrote a programme, known as ‘student’, which led the conversation on algebra. Using logical formalisation, he substituted functors for conjunctions which enabled him to analyse appropriately arranged tasks.
In 1965, a group of researchers at MIT created Semantic Information Retrieval, or SIR. This was a program that gathered knowledge from the user’s utterances, creating a base along with simple relations (e.g. membership, quantity, position in space). It then built its responses on this basis.
However, a very important first step and the prototype of all chatbots was the creation of a bot named Eliza by Joseph Weizenbaum in 1966. She was also known as ‘the doctor’ because she assumed the role of a psychoanalyst in conversation. Her algorithm was based on a very simple principle – she would look for key words in a human utterance and respond in the form of questions. For example, if the bot gets a character’s utterance: “I’ve been feeling tired lately”, it would recognise the phrase ” I’ve been feeling” and, when responding, replace it with “why do you feel” and add a question mark at the end. In this trivial way, the phrase appears to have a logical connotation with the previous one, creating an impression of understanding. In 1986, Joseph Weintraub transferred Eliza to the PC. Today, dialogue conducted in a similar style is not well regarded and is referred to as the Eliza effect.
The research work undertaken in the following years was clearly directed towards the study of language syntax. In 1968, L. Gross and D. Walker used the so-called minimax method, previously used in work on computers playing chess, to analyse sentences. A year later, Dr Lee Mc Mahon (of Bell Telephone Laboratories) created a programming language known as Fase. It was, as a proposal for simplifying the English language, at the time used for grammatical dissection of little complex sentences. Edinburgh, on the other hand, developed a programme to study the structure of sentences written not necessarily quite orthographically correct.
A well-known successor to Eliza was the Parry bot written in the early 1970s by Dr Kenneth Mark Colby of Standford University and a group of his students. This time, the programme simulated the behaviour of a paranoid schizophrenic, based on the patient’s specific inference. In developing it, they borrowed the biography of a 28-year-old postal clerk who had spent his entire life in a state of irrational fear of retribution from the Mafia underworld. Perry was used to train students, learning how to deal with the mentally ill while talking to him. Technically, he was more advanced than Eliza, as he included strategies for conversation.
Despite its success in providing counselling based on large disease diagnosis databases and exchanging information with a human in diagnosing diseases, the first electronic assistant was not premiered until 1992 at the Consumer Electronic Show in Chicago. Since then, there has been rapid development of virtual humans, catalysed by their transfer from the laboratory to the commercial sector. Enhanced with the ability to answer questions, facial expressions, they are finding widespread use in the vast virtual world of the World Wide Web. In 1999. Oddcast Media Technology launched a product to facilitate the visualisation of chatterbots – SitePal, with an A.L.I.C.E.-based language engine and Flash-based visualisation.
In 2000, Dr Richard S. Wallace introduced the ALICE system , on which Alicebot is based, a programme that is currently considered one of the most capable simulators. It uses heuristic pattern-matching principles to process natural language. It does not, therefore, have the ability to learn on its own, as it needs to have built-in pattern-matching rules. ALICE is an open-source product on whose knowledge base a group of 500 volunteers worked to create approximately 40,000 categories alone as of today.
See what a conversation with our chatbot looks like: https://glivia.com/chatbot/