Ticker: A System for Incremental ASP-based Stream Reasoning

摘要

In complex reasoning tasks, as expressible by Answer Set Programming (ASP),problems often permit for multiple solutions. In dynamic environments, whereknowledge is continuously changing, the question arises how a given model canbe incrementally adjusted relative to new and outdated information. This paperintroduces Ticker, a prototypical engine for well-defined logical reasoningover streaming data. Ticker builds on a practical fragment of the recentrule-based language LARS which extends Answer Set Programming for streams byproviding flexible expiration control and temporal modalities. We discussTicker's reasoning strategies: First, the repeated one-shot solving mode callsClingo on an ASP encoding. We show how this translation can be incrementallyupdated when new data is streaming in or time passes by. Based on this, webuild on Doyle's classic justification-based truth maintenance system (TMS) toupdate models of non-stratified programs. Finally, we empirically compare theobtained evaluation mechanisms. This paper is under consideration foracceptance in TPLP.