The TextLearner System: Reading Learning Comprehension

The goal of DARPA's Reading Learning Comprehension seedling was to determine the feasibility of autonomous knowledge acquisition through the analysis of text. This report describes the results of that effort by detailing the capabilities of the TextLearner prototype, a knowledge-acquisition program that represents the culmination of the year-long effort. Built atop the Cyc Knowledge Base and implemented almost entirely in the formal representation language of CycL, TextLearner is an anomaly in the way of Natural Language Understanding programs. The system operates by generating an information-rich model of its target document, and uses that model to explore learning opportunities. TextLearner uses this model to generate and evaluate hypotheses, not only about the possible contents of the target document, but about how to interpret unfamiliar natural language constructions it encounters. Thus TextLearner is able to do two important types of learning--content extraction and rule acquisition--that establish, the authors would argue, the value of knowledge acquisition from text as a rich and promising area of reasoning-based AI research. The original document contains color images.