• Publications
  • Influence
Wrapper Induction for Information Extraction
This work introduces wrapper induction, a method for automatically constructing wrappers, and identifies hlrt, a wrapper class that is e ciently learnable, yet expressive enough to handle 48% of a recently surveyed sample of Internet resources. Expand
A scalable comparison-shopping agent for the World-Wide Web
ShopBot, a fully-implemented, domainindependent comparison-shopping agent that relies on a combination of heuristic search, pattern matching, and inductive learning techniques, enables users to both find superior prices and substantially reduce Web shopping time. Expand
Production Matching for Large Learning Systems
An improved match algorithm, Rete/UL, is presented, which is a general extension of the existing state-of-the-art Rete match algorithm and scales well on a significantly broader class of systems than existing match algorithms. Expand
Learning to Understand Information on the Internet: An Example-Based Approach
This work reports on ShopBot and ILA, two implemented agents that learn to interact with informationsources on the Internet and shows that ILA 's learning is fast and accurate, requiring only a small number of queries per information source. Expand
The Match Cost of Adding a New Rule: A Clash of Views
What is the match cost of adding a new rule to a production system (rule-based system)? Two conflicting views have emerged. Research in EBL indicates that learned rules add to the match cost of aExpand
Learning 10, 000 Chunks: What's It Like Out There?
An initial exploration into large learning systems, i.e., systems that learn a large number of rules, using a single problem-solving and learning system, Dispatcher-Soar, to begin to get answers to efficiency questions. Expand
Combining Left and Right Unlinking for Matching a Large Number of Learned Rules
A symmetric optimization, left unlinking, is introduced and it is shown that it makes Rete scale well on an even larger class of systems, and the interference is very small in practice. Expand
A Specification of the Soar Cognitive Architecture in Z
A formal specification of the sixth revision of the Soar architecture in the Z notation was constructed to elucidate and clarify the definition of Soar and to guide its implementation. Soar is aExpand
Matching 100,045 learned rules
Examination of large systems which learn a large number of rules reveals new phenomena and calls into question some common assumptions based on previous observations of smalkr systems, including one which learns 113,938 rules - the largest number ever learned by an AI system, and the largest in any production system in existence. Expand