STUDENT: Daniel G. Bobrow's Natural Language Algebra Problem Solver
In 1964, Daniel G. Bobrow, then a doctoral student at MIT, completed STUDENT, a LISP-based program that read English-language algebra word problems and solved them symbolically. It was among the earliest demonstrations that a computer could parse and act on natural-language mathematical statements.
Background
By the early 1960s, computers could solve algebra problems, but only if you wrote those problems in a formal language the machine could parse directly. A human operator had to translate the problem first. The gap between how people described a problem and how a machine could accept it was wide, and most researchers assumed it would stay that way.
Natural-language processing, the attempt to have computers read ordinary written or spoken language, was barely a field yet. The tools that existed were narrow. A program might recognise a small set of keywords or match a sentence against a rigid template, but anything more flexible was thought to require something close to a full account of how language worked. Nobody had that, and the difficulty of building it put most practical applications out of reach.
LISP, a programming language John McCarthy had designed at MIT in the late 1950s, gave researchers a way to handle symbolic expressions rather than just numbers. That made it easier to represent language structures in a form a computer could manipulate. Even so, nobody had yet shown that a program could take an ordinary English sentence, work out what mathematical relationship it described, and then solve it.
What happened
Daniel G. Bobrow, working as a doctoral student at MIT, built a program he called STUDENT and described it in his 1964 dissertation. The program accepted algebra word problems typed in plain English, the kind that appear in school textbooks, and solved them symbolically. A user could type something like “If the number of customers Tom gets is twice the square of 20 per cent of the number of advertisements he runs, and he runs 45 advertisements, what is the number of customers Tom gets?” and STUDENT would return the answer.
The approach was based on pattern matching. STUDENT scanned the input text for phrases that signalled mathematical relationships: words like “times”, “less than”, or “is”. It used a set of preprogrammed rules to convert those phrases into algebraic equations. Once it had the equations, it solved them using symbolic algebra. There was no deep grammatical analysis, and the program did not build anything like a general model of English. What it did was identify enough structure in a narrow class of sentences to make the translation reliable within that class.
Bobrow completed this work at MIT and later went on to Bolt Beranek and Newman. The dissertation was submitted in 1964, and the work was later republished in Marvin Minsky’s 1968 collection Semantic Information Processing. The program ran on real problems and produced correct answers, which was the point. It was not a theoretical sketch. It showed, concretely, that pattern-matching rules written in LISP could bridge at least some of the distance between English prose and symbolic mathematics, for a specific and well-defined type of problem.
Why it mattered
STUDENT demonstrated that natural language could serve as a practical interface to a symbolic reasoning system, an idea that had previously seemed intractable. By representing English sentences as algebraic equations through pattern-matching rules, Bobrow established a template for later natural-language understanding research. The program also showed that LISP was viable for building linguistically sophisticated AI systems, influencing subsequent work in both expert systems and natural-language processing.
People
Daniel G Bobrow
Organisations
Massachusetts Institute of Technology, Bolt Beranek and Newman
Sources
- Natural Language Input for a Computer Problem Solving System (MIT PhD Thesis).MIT DSpace institutional repository.Primary source
- STUDENT: A Program that Solves Algebra Word Problems (in Semantic Information Processing, ed. Minsky).ACM Digital Library / MIT Press.Primary source
- Paradigms of Artificial Intelligence Programming, Chapter 7, STUDENT.Peter Norvig / paip-lisp.Secondary
Cite this page
AI Achievements. (1964). STUDENT: Daniel G. Bobrow's Natural Language Algebra Problem Solver. Retrieved 2026-08-22, from https://achievements.ai/milestone/an-ai-student-program-designed-by-daniel-bobrow
@misc{achievements_an_ai_student_program_designed_by_daniel_bobrow,
title = {STUDENT: Daniel G. Bobrow's Natural Language Algebra Problem Solver},
author = {{AI Achievements}},
year = {1964},
url = {https://achievements.ai/milestone/an-ai-student-program-designed-by-daniel-bobrow}
}