WAVE: An Incremental Algorithm for Information Extraction
Jonathan Aseltine · 1999
This paper describes WAVE, a fully automatic, in-cremental induction algorithm for learning infor-mation extraction rules. Unlike traditionM batch learners, WAVE learns from a stream of training instances, not a set. WAVE overcomes the inher-ent problems of incremental operation by main-taining a generalization hierarchy of rules. Use of a hierarchy allows similar rules to be found efficiently, provides a natural bound on general-ization, enables recall/precision trade-offs without retraining, and speeds extraction since all rules need not be applied to an instance. Finally, be-cause the reliability of rule predictions are con-tinually updated throughout storage, the hierar-chy can be used for extraction at any time. Ex-periments show that WAVE performs as well as CRYSTAL, a related batch algorithm, in two very different extraction domains. WAVE is signifi-cantly faster in a simulated incremental applica-tion setting.