Grid Parsing:Distributed Natural Language Analysis
Natural language analysis (aka parsing) with broad-coverage precision grammars has matured to a point that facilitates its application to large collections of running text. The LinGO English Resource Grammar (ERG) is a computational grammar couched in the HPSG framework. Combined with the open-source PET parser, processing times for the ERG average about one second per sentence. Although sufficient for interactive applications or smaller data sets, using the parser on large document collections (e.g. WikipediA articles) will require distribution over a high-performance computation grid.
Through collaboration with the UiO group for scientific computing (HPC), we have access to a cluster of thousands of compute nodes, each a standard x86 cpu running Linux. The dominant paradigm for distribution and parallelization in grid computation is the MPI message passing library. The goal of this project is to adapt the PET parser (which is implemented in C++) to support distributed and parallel execution on the HPC cluster, using MPI and the Grid Engine queueing system. Please contact Stephan Oepen to discuss this project further.
| Type: | 60 ECTS |
| Veileder(e): | Stephan Oepen |
| Tilknyttet forskningsgruppe(r): | LNS |