Summary: Proc. Int'l Workshop on Mixed Language Explanations in Learning, Environments
(XLANG), in conj. with AIED2005,
Amsterdam, July, 2005, Copyright 2005.
Text Classification Rule Induction in the
Presence of Domain-Specific Expression
Adam Carlson and Steven Tanimoto 2
Dept. of Computer Science and Engineering
Univ. of Washington, Seattle, WA, 98195, USA.
Abstract. We describe a new method for learning text-classification rules from ex-
amples. The text consists of messages written by students in an online learning en-
vironment, and it may contain ungrammatical expressions as well as specialized
expressions such as formulas. The method is based on the version-space machine
learning technique. Experiments show that our method successfully generalizes
over certain classes of embedded numerical expressions involving ranges of values
in RGB triples that represent colors in an image processing system.
Keywords. text classification, learning, numerical expressions, domain-specific
terms, rule induction