A Parallel Chart-Based Parser for Analyzing Ill-Formed Inputs

概要

論文の詳細を見る
When a natural language processing system encounters unparsable inputs, the analysis should not be rejected. Instead, the system should attempt to detect the cause of ill-formedness and generate a set of possible interpretations. However, parsing ill-formed inputs suffers from large computation time due to extra mechanisms for detecting the existing ill-formedness. This indicates the importance of developing a parallel robust parsing algorithm. The goal of this research is to develop an effective parallel algorithm for parsing an ill-formed input under a loosely-coupled hardware environment. The parallel parser is implemented on PIM/m with 256 processors. Through the experiments, we point out that parsing ill-formed inputs with the proposed parser can acquire a satisfactory result in its performance.
社団法人人工知能学会の論文
1995-07-01