Parallel Parsing: Different Distribution Schemata for Charts.
Jan W. Amtrup · 1995
Introduction We are going to present results from two experiments designed for parallel parsing within the chart paradigm. Parallel processing gains more relevance as applications become increasingly complex and nets of workstations as well as dedicated parallel computers are widely available. A chart-based parser is well suited for approaches to parallelism due to the identification of almost independent data objects that a chart is made of. A parallelization based on tasks of an agenda is only suitable for shared memory systems with a tight coupling, the choice of individual edges as autonomous agents may result in too many processes. But both nodes of a chart and rules of a grammar may provide sufficient possibilities for parallelization in a loosely coupled framework. Context-free parsing in a net of workstations These latter two aspects --- node-based and grammar-based approaches --- of a data-driven parallelization of chart parsing have been pursued in t