SIR (Semantic Information Retrieval) by Bertram Raphael, MIT
In 1964, Bertram Raphael at MIT completed his doctoral thesis describing SIR (Semantic Information Retrieval), a LISP program that answered English-language questions about simple relational facts by constructing an internal semantic model. It was an early demonstration that a computer could parse and reason over a restricted subset of natural language.
Background
By the early 1960s, computers were useful tools for people who could speak their language: precise, formal, unforgiving notation. Everyone else was locked out. Researchers were beginning to ask whether a machine could instead meet people partway, accepting something closer to ordinary English as input.
The problem was harder than it looked. Natural language is full of ambiguity. Words shift meaning depending on context, and even a simple sentence like “every boy has a father” carries an implicit logical structure that a program has to reconstruct before it can do anything useful with it. The programs of the time mostly skipped that reconstruction step. They matched patterns of text against stored patterns of text, which worked in narrow cases but broke down the moment a question arrived in a form the designer had not anticipated.
What was missing was an internal model. Without one, a program could not combine two facts it had been told to produce a third. It could retrieve, but it could not reason.
What happened
Bertram Raphael, working at MIT, completed his doctoral thesis in 1964 describing a program called SIR, for Semantic Information Retrieval. Written in LISP, SIR took simple English sentences as input and built an internal representation of the facts they contained, using property lists and word associations to capture relationships between concepts. When asked a question, it worked from that internal model rather than searching its input for a matching string of words.
The relationships SIR could handle were deliberately limited. It understood set membership and inclusion, ownership, and part-whole relationships. Tell it that a boy is a person, and that every person has a heart, and it could answer a question about whether a boy has a heart by chaining those two facts together. That kind of inference, modest as it sounds, was not something programs had reliably done from natural-language input before.
The system also knew when it lacked enough information to answer, and would say so rather than guess. It handled some ambiguous sentences by asking the user to clarify. These were small things, but they pointed toward something the field would spend decades building on: the idea that a question-answering program needs an explicit model of what it knows, and an honest account of what it does not.
SIR was a demonstration, not a product. Its coverage of English was narrow and its reasoning was shallow by any later standard. Raphael himself was clear about that. But it showed, concretely, that a computer could parse a natural-language sentence, extract relational meaning from it, and use that meaning to answer a question it had not seen before.
Why it mattered
SIR was among the earliest programs to demonstrate that a computer could accept natural-language input, build an internal representation of the stated facts, and draw inferences from that representation to answer queries. It helped establish the foundational problems of natural-language understanding and knowledge representation that would preoccupy the field for decades. The work directly influenced subsequent question-answering and expert-system research by showing both the promise and the strict limits of semantically shallow, rule-governed approaches.
People
Organisations
Massachusetts Institute of Technology
Sources
- SIR: A Computer Program for Semantic Information Retrieval (MIT Master's Thesis).MIT DSpace Institutional Repository.Primary source
- A computer program for semantic information retrieval.ACM Digital Library (Computers and Thought, 1965).Secondary
Cite this page
AI Achievements. (1964). SIR (Semantic Information Retrieval) by Bertram Raphael, MIT. Retrieved 2026-08-22, from https://achievements.ai/milestone/sir-computer-program-bert-raphaels-mit
@misc{achievements_sir_computer_program_bert_raphaels_mit,
title = {SIR (Semantic Information Retrieval) by Bertram Raphael, MIT},
author = {{AI Achievements}},
year = {1964},
url = {https://achievements.ai/milestone/sir-computer-program-bert-raphaels-mit}
}