Computer science researchers have developed a new computer system that has the capability of solving word problems automatically. The new system was developed by researchers from the Massachusetts Institute of Technology (MIT) with the help of other researchers from the University of Washington.
Most of the research to develop the new system was conducted at the Computer Science and Artificial Intelligence Laboratory in the MIT. Linguistic problems have always been a tricky subject for computer scientists. Unlike math, which is considered by many experts as a pure and accurate "language," computers often have difficulties in understanding the sometimes vague and confusing languages that humans use on a daily basis. However, the new computer system can actually be used to solve word problems often seen in basic math lessons at schools.
"In these algebra problems, you have to build these things up from many different sentences," said MIT electrical engineering and computer science graduate student Nate Kushman. "The fact that you're looking across multiple sentences to generate this semantic representation is really something new." Kushman is also the lead author of a paper on the new system to be presented at the next annual gathering of the Association for Computational Linguistics. The paper will be presented this June.
The new system is currently being eyed for a number of practical applications. The researchers behind the computer system say that it can soon be used in a variety of fields including education, math and even physics. Used as an educational aid, the system may be able to help students identify certain problems in reasoning. Moreover, the system can also be used to construct more efficient word problems by gauging the difficulty of current word problems used by teachers.
The MIT researchers also had the help of University of Washington's Regina Barzilay. Barzilay is a professor of computer science and engineering who is also acting as Kushman's thesis advisor.
To develop the system, the researcher used Macsyma and a sentence parser. These computational tools are used for research in math and linguistics. Macsyma is a tool that was also developed at the MIT back in the 60s, which may have potential applications in constructing an artificial intelligence. The tool is used to simplify mathematical equations by applying a template to make them appear for uniform and easier for computers to process. On the other hand, a sentence parser is used to represent certain segments of human words used in a sentence while showing the relationships between each part in a graph.
By utilizing these two tools, the team was able to come up with a system that can comprehend and analyze the content of a given word problem. Using a sentence parses, the system is able to understand the words used in the problem. The system can then start distilling the word problem into a mathematical equation using the templates used by Macsyma.
"The idea of this kind of supervision they have will be useful for lots of things," said University of Southern California professor of computer science Kevin Knight. "The approach of building a generative story of how people get from text to answers is a great idea."