My students and I are currently exploring the task of detecting analogies in natural language text.

Analogy is the process of comparing a base case to a similar target case. For example, in an analogy comparing the solar system to an atom, the sun is like the nucleus, the planets are like the electrons, and gravity is like electromagnetic force. Analogical processing may support a variety of cognitive processes, and explicit cross-domain analogy is used to support learning.

A machine reading system that can process analogies it encounters can improve its understanding of the text and potentially allow for insight into how humans handle textual analogy. However, detecting when the author is using an analogy is not easy. Using modern language processing techniques, we look for ways to identify where in a text an author is using analogy. We also draw on what we know about how humans manage analogies. Previous work in this area suggests that semantics are likely critical in this endeavor, and we will focus our efforts on integrating semantics.

Students working on this research use a variety of text analysis tools and machine learning algorithms.